id
stringlengths
36
36
text
stringlengths
114
429k
url
stringlengths
54
121
d9a37339-4bad-4cf3-9fb5-1dd03debfa69
Models LangChain provides interfaces and integrations for a number of different types of models. LLMs Chat Models
https://api.python.langchain.com/en/latest/models.html
8ff91ce1-7d62-4479-9049-f8f70b82ef58
Model I/O LangChain provides interfaces and integrations for working with language models. Prompts Models Output Parsers
https://api.python.langchain.com/en/latest/model_io.html
2412f67f-00bb-42fb-9975-249ace261d8e
Prompts The reference guides here all relate to objects for working with Prompts. Prompt Templates Example Selector
https://api.python.langchain.com/en/latest/prompts.html
f7f922b9-f340-4163-bd4d-a15304000f03
Data connection LangChain has a number of modules that help you load, structure, store, and retrieve documents. Document Loaders Document Transformers Embeddings Vector Stores Retrievers
https://api.python.langchain.com/en/latest/data_connection.html
583cb37c-049b-4ce1-bba8-5e96effe2285
Embeddings Wrappers around embedding modules. class langchain.embeddings.OpenAIEmbeddings(*, client=None, model='text-embedding-ada-002', deployment='text-embedding-ada-002', openai_api_version=None, openai_api_base=None, openai_api_type=None, openai_proxy=None, embedding_ctx_length=8191, openai_api_key=None, openai_organization=None, allowed_special={}, disallowed_special='all', chunk_size=1000, max_retries=6, request_timeout=None, headers=None, tiktoken_model_name=None)[source] Bases: pydantic.main.BaseModel, langchain.embeddings.base.Embeddings Wrapper around OpenAI embedding models. To use, you should have the openai python package installed, and the environment variable OPENAI_API_KEY set with your API key or pass it as a named parameter to the constructor. Example from langchain.embeddings import OpenAIEmbeddings openai = OpenAIEmbeddings(openai_api_key="my-api-key") In order to use the library with Microsoft Azure endpoints, you need to set the OPENAI_API_TYPE, OPENAI_API_BASE, OPENAI_API_KEY and OPENAI_API_VERSION. The OPENAI_API_TYPE must be set to β€˜azure’ and the others correspond to the properties of your endpoint. In addition, the deployment name must be passed as the model parameter. Example import os os.environ["OPENAI_API_TYPE"] = "azure" os.environ["OPENAI_API_BASE"] = "https://<your-endpoint.openai.azure.com/" os.environ["OPENAI_API_KEY"] = "your AzureOpenAI key" os.environ["OPENAI_API_VERSION"] = "2023-03-15-preview" os.environ["OPENAI_PROXY"] = "http://your-corporate-proxy:8080" from langchain.embeddings.openai import OpenAIEmbeddings embeddings = OpenAIEmbeddings( deployment="your-embeddings-deployment-name", model="your-embeddings-model-name", openai_api_base="https://your-endpoint.openai.azure.com/", openai_api_type="azure", ) text = "This is a test query." query_result = embeddings.embed_query(text) Parameters client (Any) – model (str) – deployment (str) – openai_api_version (Optional[str]) – openai_api_base (Optional[str]) – openai_api_type (Optional[str]) – openai_proxy (Optional[str]) – embedding_ctx_length (int) – openai_api_key (Optional[str]) – openai_organization (Optional[str]) – allowed_special (Union[Literal['all'], typing.Set[str]]) – disallowed_special (Union[Literal['all'], typing.Set[str], typing.Sequence[str]]) – chunk_size (int) – max_retries (int) – request_timeout (Optional[Union[float, Tuple[float, float]]]) – headers (Any) – tiktoken_model_name (Optional[str]) – Return type None attribute chunk_size: int = 1000 Maximum number of texts to embed in each batch attribute max_retries: int = 6 Maximum number of retries to make when generating. attribute request_timeout: Optional[Union[float, Tuple[float, float]]] = None Timeout in seconds for the OpenAPI request. attribute tiktoken_model_name: Optional[str] = None The model name to pass to tiktoken when using this class. Tiktoken is used to count the number of tokens in documents to constrain them to be under a certain limit. By default, when set to None, this will be the same as the embedding model name. However, there are some cases where you may want to use this Embedding class with a model name not supported by tiktoken. This can include when using Azure embeddings or when using one of the many model providers that expose an OpenAI-like API but with different models. In those cases, in order to avoid erroring when tiktoken is called, you can specify a model name to use here. async aembed_documents(texts, chunk_size=0)[source] Call out to OpenAI’s embedding endpoint async for embedding search docs. Parameters texts (List[str]) – The list of texts to embed. chunk_size (Optional[int]) – The chunk size of embeddings. If None, will use the chunk size specified by the class. Returns List of embeddings, one for each text. Return type List[List[float]] async aembed_query(text)[source] Call out to OpenAI’s embedding endpoint async for embedding query text. Parameters text (str) – The text to embed. Returns Embedding for the text. Return type List[float] embed_documents(texts, chunk_size=0)[source] Call out to OpenAI’s embedding endpoint for embedding search docs. Parameters texts (List[str]) – The list of texts to embed. chunk_size (Optional[int]) – The chunk size of embeddings. If None, will use the chunk size specified by the class. Returns List of embeddings, one for each text. Return type List[List[float]] embed_query(text)[source] Call out to OpenAI’s embedding endpoint for embedding query text. Parameters text (str) – The text to embed. Returns Embedding for the text. Return type List[float] class langchain.embeddings.HuggingFaceEmbeddings(*, client=None, model_name='sentence-transformers/all-mpnet-base-v2', cache_folder=None, model_kwargs=None, encode_kwargs=None)[source] Bases: pydantic.main.BaseModel, langchain.embeddings.base.Embeddings Wrapper around sentence_transformers embedding models. To use, you should have the sentence_transformers python package installed. Example from langchain.embeddings import HuggingFaceEmbeddings model_name = "sentence-transformers/all-mpnet-base-v2" model_kwargs = {'device': 'cpu'} encode_kwargs = {'normalize_embeddings': False} hf = HuggingFaceEmbeddings( model_name=model_name, model_kwargs=model_kwargs, encode_kwargs=encode_kwargs ) Parameters client (Any) – model_name (str) – cache_folder (Optional[str]) – model_kwargs (Dict[str, Any]) – encode_kwargs (Dict[str, Any]) – Return type None attribute cache_folder: Optional[str] = None Path to store models. Can be also set by SENTENCE_TRANSFORMERS_HOME environment variable. attribute encode_kwargs: Dict[str, Any] [Optional] Key word arguments to pass when calling the encode method of the model. attribute model_kwargs: Dict[str, Any] [Optional] Key word arguments to pass to the model. attribute model_name: str = 'sentence-transformers/all-mpnet-base-v2' Model name to use. embed_documents(texts)[source] Compute doc embeddings using a HuggingFace transformer model. Parameters texts (List[str]) – The list of texts to embed. Returns List of embeddings, one for each text. Return type List[List[float]] embed_query(text)[source] Compute query embeddings using a HuggingFace transformer model. Parameters text (str) – The text to embed. Returns Embeddings for the text. Return type List[float] class langchain.embeddings.CohereEmbeddings(*, client=None, model='embed-english-v2.0', truncate=None, cohere_api_key=None)[source] Bases: pydantic.main.BaseModel, langchain.embeddings.base.Embeddings Wrapper around Cohere embedding models. To use, you should have the cohere python package installed, and the environment variable COHERE_API_KEY set with your API key or pass it as a named parameter to the constructor. Example from langchain.embeddings import CohereEmbeddings cohere = CohereEmbeddings( model="embed-english-light-v2.0", cohere_api_key="my-api-key" ) Parameters client (Any) – model (str) – truncate (Optional[str]) – cohere_api_key (Optional[str]) – Return type None attribute model: str = 'embed-english-v2.0' Model name to use. attribute truncate: Optional[str] = None Truncate embeddings that are too long from start or end (β€œNONE”|”START”|”END”) embed_documents(texts)[source] Call out to Cohere’s embedding endpoint. Parameters texts (List[str]) – The list of texts to embed. Returns List of embeddings, one for each text. Return type List[List[float]] embed_query(text)[source] Call out to Cohere’s embedding endpoint. Parameters text (str) – The text to embed. Returns Embeddings for the text. Return type List[float] class langchain.embeddings.ElasticsearchEmbeddings(client, model_id, *, input_field='text_field')[source] Bases: langchain.embeddings.base.Embeddings Wrapper around Elasticsearch embedding models. This class provides an interface to generate embeddings using a model deployed in an Elasticsearch cluster. It requires an Elasticsearch connection object and the model_id of the model deployed in the cluster. In Elasticsearch you need to have an embedding model loaded and deployed. - https://www.elastic.co/guide/en/elasticsearch/reference/current/infer-trained-model.html - https://www.elastic.co/guide/en/machine-learning/current/ml-nlp-deploy-models.html Parameters client (MlClient) – model_id (str) – input_field (str) – classmethod from_credentials(model_id, *, es_cloud_id=None, es_user=None, es_password=None, input_field='text_field')[source] Instantiate embeddings from Elasticsearch credentials. Parameters model_id (str) – The model_id of the model deployed in the Elasticsearch cluster. input_field (str) – The name of the key for the input text field in the document. Defaults to β€˜text_field’. es_cloud_id (Optional[str]) – (str, optional): The Elasticsearch cloud ID to connect to. es_user (Optional[str]) – (str, optional): Elasticsearch username. es_password (Optional[str]) – (str, optional): Elasticsearch password. Return type langchain.embeddings.elasticsearch.ElasticsearchEmbeddings Example from langchain.embeddings import ElasticsearchEmbeddings # Define the model ID and input field name (if different from default) model_id = "your_model_id" # Optional, only if different from 'text_field' input_field = "your_input_field" # Credentials can be passed in two ways. Either set the env vars # ES_CLOUD_ID, ES_USER, ES_PASSWORD and they will be automatically # pulled in, or pass them in directly as kwargs. embeddings = ElasticsearchEmbeddings.from_credentials( model_id, input_field=input_field, # es_cloud_id="foo", # es_user="bar", # es_password="baz", ) documents = [ "This is an example document.", "Another example document to generate embeddings for.", ] embeddings_generator.embed_documents(documents) classmethod from_es_connection(model_id, es_connection, input_field='text_field')[source] Instantiate embeddings from an existing Elasticsearch connection. This method provides a way to create an instance of the ElasticsearchEmbeddings class using an existing Elasticsearch connection. The connection object is used to create an MlClient, which is then used to initialize the ElasticsearchEmbeddings instance. Args: model_id (str): The model_id of the model deployed in the Elasticsearch cluster. es_connection (elasticsearch.Elasticsearch): An existing Elasticsearch connection object. input_field (str, optional): The name of the key for the input text field in the document. Defaults to β€˜text_field’. Returns: ElasticsearchEmbeddings: An instance of the ElasticsearchEmbeddings class. Example from elasticsearch import Elasticsearch from langchain.embeddings import ElasticsearchEmbeddings # Define the model ID and input field name (if different from default) model_id = "your_model_id" # Optional, only if different from 'text_field' input_field = "your_input_field" # Create Elasticsearch connection es_connection = Elasticsearch( hosts=["localhost:9200"], http_auth=("user", "password") ) # Instantiate ElasticsearchEmbeddings using the existing connection embeddings = ElasticsearchEmbeddings.from_es_connection( model_id, es_connection, input_field=input_field, ) documents = [ "This is an example document.", "Another example document to generate embeddings for.", ] embeddings_generator.embed_documents(documents) Parameters model_id (str) – es_connection (Elasticsearch) – input_field (str) – Return type ElasticsearchEmbeddings embed_documents(texts)[source] Generate embeddings for a list of documents. Parameters texts (List[str]) – A list of document text strings to generate embeddings for. Returns A list of embeddings, one for each document in the inputlist. Return type List[List[float]] embed_query(text)[source] Generate an embedding for a single query text. Parameters text (str) – The query text to generate an embedding for. Returns The embedding for the input query text. Return type List[float] class langchain.embeddings.LlamaCppEmbeddings(*, client=None, model_path, n_ctx=512, n_parts=- 1, seed=- 1, f16_kv=False, logits_all=False, vocab_only=False, use_mlock=False, n_threads=None, n_batch=8, n_gpu_layers=None)[source] Bases: pydantic.main.BaseModel, langchain.embeddings.base.Embeddings Wrapper around llama.cpp embedding models. To use, you should have the llama-cpp-python library installed, and provide the path to the Llama model as a named parameter to the constructor. Check out: https://github.com/abetlen/llama-cpp-python Example from langchain.embeddings import LlamaCppEmbeddings llama = LlamaCppEmbeddings(model_path="/path/to/model.bin") Parameters client (Any) – model_path (str) – n_ctx (int) – n_parts (int) – seed (int) – f16_kv (bool) – logits_all (bool) – vocab_only (bool) – use_mlock (bool) – n_threads (Optional[int]) – n_batch (Optional[int]) – n_gpu_layers (Optional[int]) – Return type None attribute f16_kv: bool = False Use half-precision for key/value cache. attribute logits_all: bool = False Return logits for all tokens, not just the last token. attribute n_batch: Optional[int] = 8 Number of tokens to process in parallel. Should be a number between 1 and n_ctx. attribute n_ctx: int = 512 Token context window. attribute n_gpu_layers: Optional[int] = None Number of layers to be loaded into gpu memory. Default None. attribute n_parts: int = -1 Number of parts to split the model into. If -1, the number of parts is automatically determined. attribute n_threads: Optional[int] = None Number of threads to use. If None, the number of threads is automatically determined. attribute seed: int = -1 Seed. If -1, a random seed is used. attribute use_mlock: bool = False Force system to keep model in RAM. attribute vocab_only: bool = False Only load the vocabulary, no weights. embed_documents(texts)[source] Embed a list of documents using the Llama model. Parameters texts (List[str]) – The list of texts to embed. Returns List of embeddings, one for each text. Return type List[List[float]] embed_query(text)[source] Embed a query using the Llama model. Parameters text (str) – The text to embed. Returns Embeddings for the text. Return type List[float] class langchain.embeddings.HuggingFaceHubEmbeddings(*, client=None, repo_id='sentence-transformers/all-mpnet-base-v2', task='feature-extraction', model_kwargs=None, huggingfacehub_api_token=None)[source] Bases: pydantic.main.BaseModel, langchain.embeddings.base.Embeddings Wrapper around HuggingFaceHub embedding models. To use, you should have the huggingface_hub python package installed, and the environment variable HUGGINGFACEHUB_API_TOKEN set with your API token, or pass it as a named parameter to the constructor. Example from langchain.embeddings import HuggingFaceHubEmbeddings repo_id = "sentence-transformers/all-mpnet-base-v2" hf = HuggingFaceHubEmbeddings( repo_id=repo_id, task="feature-extraction", huggingfacehub_api_token="my-api-key", ) Parameters client (Any) – repo_id (str) – task (Optional[str]) – model_kwargs (Optional[dict]) – huggingfacehub_api_token (Optional[str]) – Return type None attribute model_kwargs: Optional[dict] = None Key word arguments to pass to the model. attribute repo_id: str = 'sentence-transformers/all-mpnet-base-v2' Model name to use. attribute task: Optional[str] = 'feature-extraction' Task to call the model with. embed_documents(texts)[source] Call out to HuggingFaceHub’s embedding endpoint for embedding search docs. Parameters texts (List[str]) – The list of texts to embed. Returns List of embeddings, one for each text. Return type List[List[float]] embed_query(text)[source] Call out to HuggingFaceHub’s embedding endpoint for embedding query text. Parameters text (str) – The text to embed. Returns Embeddings for the text. Return type List[float] class langchain.embeddings.ModelScopeEmbeddings(*, embed=None, model_id='damo/nlp_corom_sentence-embedding_english-base')[source] Bases: pydantic.main.BaseModel, langchain.embeddings.base.Embeddings Wrapper around modelscope_hub embedding models. To use, you should have the modelscope python package installed. Example from langchain.embeddings import ModelScopeEmbeddings model_id = "damo/nlp_corom_sentence-embedding_english-base" embed = ModelScopeEmbeddings(model_id=model_id) Parameters embed (Any) – model_id (str) – Return type None attribute model_id: str = 'damo/nlp_corom_sentence-embedding_english-base' Model name to use. embed_documents(texts)[source] Compute doc embeddings using a modelscope embedding model. Parameters texts (List[str]) – The list of texts to embed. Returns List of embeddings, one for each text. Return type List[List[float]] embed_query(text)[source] Compute query embeddings using a modelscope embedding model. Parameters text (str) – The text to embed. Returns Embeddings for the text. Return type List[float] class langchain.embeddings.TensorflowHubEmbeddings(*, embed=None, model_url='https://tfhub.dev/google/universal-sentence-encoder-multilingual/3')[source] Bases: pydantic.main.BaseModel, langchain.embeddings.base.Embeddings Wrapper around tensorflow_hub embedding models. To use, you should have the tensorflow_text python package installed. Example from langchain.embeddings import TensorflowHubEmbeddings url = "https://tfhub.dev/google/universal-sentence-encoder-multilingual/3" tf = TensorflowHubEmbeddings(model_url=url) Parameters embed (Any) – model_url (str) – Return type None attribute model_url: str = 'https://tfhub.dev/google/universal-sentence-encoder-multilingual/3' Model name to use. embed_documents(texts)[source] Compute doc embeddings using a TensorflowHub embedding model. Parameters texts (List[str]) – The list of texts to embed. Returns List of embeddings, one for each text. Return type List[List[float]] embed_query(text)[source] Compute query embeddings using a TensorflowHub embedding model. Parameters text (str) – The text to embed. Returns Embeddings for the text. Return type List[float] class langchain.embeddings.SagemakerEndpointEmbeddings(*, client=None, endpoint_name='', region_name='', credentials_profile_name=None, content_handler, model_kwargs=None, endpoint_kwargs=None)[source] Bases: pydantic.main.BaseModel, langchain.embeddings.base.Embeddings Wrapper around custom Sagemaker Inference Endpoints. To use, you must supply the endpoint name from your deployed Sagemaker model & the region where it is deployed. To authenticate, the AWS client uses the following methods to automatically load credentials: https://boto3.amazonaws.com/v1/documentation/api/latest/guide/credentials.html If a specific credential profile should be used, you must pass the name of the profile from the ~/.aws/credentials file that is to be used. Make sure the credentials / roles used have the required policies to access the Sagemaker endpoint. See: https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies.html Parameters client (Any) – endpoint_name (str) – region_name (str) – credentials_profile_name (Optional[str]) – content_handler (langchain.embeddings.sagemaker_endpoint.EmbeddingsContentHandler) – model_kwargs (Optional[Dict]) – endpoint_kwargs (Optional[Dict]) – Return type None attribute content_handler: langchain.embeddings.sagemaker_endpoint.EmbeddingsContentHandler [Required] The content handler class that provides an input and output transform functions to handle formats between LLM and the endpoint. attribute credentials_profile_name: Optional[str] = None The name of the profile in the ~/.aws/credentials or ~/.aws/config files, which has either access keys or role information specified. If not specified, the default credential profile or, if on an EC2 instance, credentials from IMDS will be used. See: https://boto3.amazonaws.com/v1/documentation/api/latest/guide/credentials.html attribute endpoint_kwargs: Optional[Dict] = None Optional attributes passed to the invoke_endpoint function. See `boto3`_. docs for more info. .. _boto3: <https://boto3.amazonaws.com/v1/documentation/api/latest/index.html> attribute endpoint_name: str = '' The name of the endpoint from the deployed Sagemaker model. Must be unique within an AWS Region. attribute model_kwargs: Optional[Dict] = None Key word arguments to pass to the model. attribute region_name: str = '' The aws region where the Sagemaker model is deployed, eg. us-west-2. embed_documents(texts, chunk_size=64)[source] Compute doc embeddings using a SageMaker Inference Endpoint. Parameters texts (List[str]) – The list of texts to embed. chunk_size (int) – The chunk size defines how many input texts will be grouped together as request. If None, will use the chunk size specified by the class. Returns List of embeddings, one for each text. Return type List[List[float]] embed_query(text)[source] Compute query embeddings using a SageMaker inference endpoint. Parameters text (str) – The text to embed. Returns Embeddings for the text. Return type List[float] class langchain.embeddings.HuggingFaceInstructEmbeddings(*, client=None, model_name='hkunlp/instructor-large', cache_folder=None, model_kwargs=None, encode_kwargs=None, embed_instruction='Represent the document for retrieval: ', query_instruction='Represent the question for retrieving supporting documents: ')[source] Bases: pydantic.main.BaseModel, langchain.embeddings.base.Embeddings Wrapper around sentence_transformers embedding models. To use, you should have the sentence_transformers and InstructorEmbedding python packages installed. Example from langchain.embeddings import HuggingFaceInstructEmbeddings model_name = "hkunlp/instructor-large" model_kwargs = {'device': 'cpu'} encode_kwargs = {'normalize_embeddings': True} hf = HuggingFaceInstructEmbeddings( model_name=model_name, model_kwargs=model_kwargs, encode_kwargs=encode_kwargs ) Parameters client (Any) – model_name (str) – cache_folder (Optional[str]) – model_kwargs (Dict[str, Any]) – encode_kwargs (Dict[str, Any]) – embed_instruction (str) – query_instruction (str) – Return type None attribute cache_folder: Optional[str] = None Path to store models. Can be also set by SENTENCE_TRANSFORMERS_HOME environment variable. attribute embed_instruction: str = 'Represent the document for retrieval: ' Instruction to use for embedding documents. attribute encode_kwargs: Dict[str, Any] [Optional] Key word arguments to pass when calling the encode method of the model. attribute model_kwargs: Dict[str, Any] [Optional] Key word arguments to pass to the model. attribute model_name: str = 'hkunlp/instructor-large' Model name to use. attribute query_instruction: str = 'Represent the question for retrieving supporting documents: ' Instruction to use for embedding query. embed_documents(texts)[source] Compute doc embeddings using a HuggingFace instruct model. Parameters texts (List[str]) – The list of texts to embed. Returns List of embeddings, one for each text. Return type List[List[float]] embed_query(text)[source] Compute query embeddings using a HuggingFace instruct model. Parameters text (str) – The text to embed. Returns Embeddings for the text. Return type List[float] class langchain.embeddings.MosaicMLInstructorEmbeddings(*, endpoint_url='https://models.hosted-on.mosaicml.hosting/instructor-xl/v1/predict', embed_instruction='Represent the document for retrieval: ', query_instruction='Represent the question for retrieving supporting documents: ', retry_sleep=1.0, mosaicml_api_token=None)[source] Bases: pydantic.main.BaseModel, langchain.embeddings.base.Embeddings Wrapper around MosaicML’s embedding inference service. To use, you should have the environment variable MOSAICML_API_TOKEN set with your API token, or pass it as a named parameter to the constructor. Example from langchain.llms import MosaicMLInstructorEmbeddings endpoint_url = ( "https://models.hosted-on.mosaicml.hosting/instructor-large/v1/predict" ) mosaic_llm = MosaicMLInstructorEmbeddings( endpoint_url=endpoint_url, mosaicml_api_token="my-api-key" ) Parameters endpoint_url (str) – embed_instruction (str) – query_instruction (str) – retry_sleep (float) – mosaicml_api_token (Optional[str]) – Return type None attribute embed_instruction: str = 'Represent the document for retrieval: ' Instruction used to embed documents. attribute endpoint_url: str = 'https://models.hosted-on.mosaicml.hosting/instructor-xl/v1/predict' Endpoint URL to use. attribute query_instruction: str = 'Represent the question for retrieving supporting documents: ' Instruction used to embed the query. attribute retry_sleep: float = 1.0 How long to try sleeping for if a rate limit is encountered embed_documents(texts)[source] Embed documents using a MosaicML deployed instructor embedding model. Parameters texts (List[str]) – The list of texts to embed. Returns List of embeddings, one for each text. Return type List[List[float]] embed_query(text)[source] Embed a query using a MosaicML deployed instructor embedding model. Parameters text (str) – The text to embed. Returns Embeddings for the text. Return type List[float] class langchain.embeddings.SelfHostedEmbeddings(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, pipeline_ref=None, client=None, inference_fn=<function _embed_documents>, hardware=None, model_load_fn, load_fn_kwargs=None, model_reqs=['./', 'torch'], inference_kwargs=None)[source] Bases: langchain.llms.self_hosted.SelfHostedPipeline, langchain.embeddings.base.Embeddings Runs custom embedding models on self-hosted remote hardware. Supported hardware includes auto-launched instances on AWS, GCP, Azure, and Lambda, as well as servers specified by IP address and SSH credentials (such as on-prem, or another cloud like Paperspace, Coreweave, etc.). To use, you should have the runhouse python package installed. Example using a model load function:from langchain.embeddings import SelfHostedEmbeddings from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline import runhouse as rh gpu = rh.cluster(name="rh-a10x", instance_type="A100:1") def get_pipeline(): model_id = "facebook/bart-large" tokenizer = AutoTokenizer.from_pretrained(model_id) model = AutoModelForCausalLM.from_pretrained(model_id) return pipeline("feature-extraction", model=model, tokenizer=tokenizer) embeddings = SelfHostedEmbeddings( model_load_fn=get_pipeline, hardware=gpu model_reqs=["./", "torch", "transformers"], ) Example passing in a pipeline path:from langchain.embeddings import SelfHostedHFEmbeddings import runhouse as rh from transformers import pipeline gpu = rh.cluster(name="rh-a10x", instance_type="A100:1") pipeline = pipeline(model="bert-base-uncased", task="feature-extraction") rh.blob(pickle.dumps(pipeline), path="models/pipeline.pkl").save().to(gpu, path="models") embeddings = SelfHostedHFEmbeddings.from_pipeline( pipeline="models/pipeline.pkl", hardware=gpu, model_reqs=["./", "torch", "transformers"], ) Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – pipeline_ref (Any) – client (Any) – inference_fn (Callable) – hardware (Any) – model_load_fn (Callable) – load_fn_kwargs (Optional[dict]) – model_reqs (List[str]) – inference_kwargs (Any) – Return type None attribute inference_fn: Callable = <function _embed_documents> Inference function to extract the embeddings on the remote hardware. attribute inference_kwargs: Any = None Any kwargs to pass to the model’s inference function. embed_documents(texts)[source] Compute doc embeddings using a HuggingFace transformer model. Parameters texts (List[str]) – The list of texts to embed.s Returns List of embeddings, one for each text. Return type List[List[float]] embed_query(text)[source] Compute query embeddings using a HuggingFace transformer model. Parameters text (str) – The text to embed. Returns Embeddings for the text. Return type List[float] class langchain.embeddings.SelfHostedHuggingFaceEmbeddings(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, pipeline_ref=None, client=None, inference_fn=<function _embed_documents>, hardware=None, model_load_fn=<function load_embedding_model>, load_fn_kwargs=None, model_reqs=['./', 'sentence_transformers', 'torch'], inference_kwargs=None, model_id='sentence-transformers/all-mpnet-base-v2')[source] Bases: langchain.embeddings.self_hosted.SelfHostedEmbeddings Runs sentence_transformers embedding models on self-hosted remote hardware. Supported hardware includes auto-launched instances on AWS, GCP, Azure, and Lambda, as well as servers specified by IP address and SSH credentials (such as on-prem, or another cloud like Paperspace, Coreweave, etc.). To use, you should have the runhouse python package installed. Example from langchain.embeddings import SelfHostedHuggingFaceEmbeddings import runhouse as rh model_name = "sentence-transformers/all-mpnet-base-v2" gpu = rh.cluster(name="rh-a10x", instance_type="A100:1") hf = SelfHostedHuggingFaceEmbeddings(model_name=model_name, hardware=gpu) Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – pipeline_ref (Any) – client (Any) – inference_fn (Callable) – hardware (Any) – model_load_fn (Callable) – load_fn_kwargs (Optional[dict]) – model_reqs (List[str]) – inference_kwargs (Any) – model_id (str) – Return type None attribute hardware: Any = None Remote hardware to send the inference function to. attribute inference_fn: Callable = <function _embed_documents> Inference function to extract the embeddings. attribute load_fn_kwargs: Optional[dict] = None Key word arguments to pass to the model load function. attribute model_id: str = 'sentence-transformers/all-mpnet-base-v2' Model name to use. attribute model_load_fn: Callable = <function load_embedding_model> Function to load the model remotely on the server. attribute model_reqs: List[str] = ['./', 'sentence_transformers', 'torch'] Requirements to install on hardware to inference the model. class langchain.embeddings.SelfHostedHuggingFaceInstructEmbeddings(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, pipeline_ref=None, client=None, inference_fn=<function _embed_documents>, hardware=None, model_load_fn=<function load_embedding_model>, load_fn_kwargs=None, model_reqs=['./', 'InstructorEmbedding', 'torch'], inference_kwargs=None, model_id='hkunlp/instructor-large', embed_instruction='Represent the document for retrieval: ', query_instruction='Represent the question for retrieving supporting documents: ')[source] Bases: langchain.embeddings.self_hosted_hugging_face.SelfHostedHuggingFaceEmbeddings Runs InstructorEmbedding embedding models on self-hosted remote hardware. Supported hardware includes auto-launched instances on AWS, GCP, Azure, and Lambda, as well as servers specified by IP address and SSH credentials (such as on-prem, or another cloud like Paperspace, Coreweave, etc.). To use, you should have the runhouse python package installed. Example from langchain.embeddings import SelfHostedHuggingFaceInstructEmbeddings import runhouse as rh model_name = "hkunlp/instructor-large" gpu = rh.cluster(name='rh-a10x', instance_type='A100:1') hf = SelfHostedHuggingFaceInstructEmbeddings( model_name=model_name, hardware=gpu) Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – pipeline_ref (Any) – client (Any) – inference_fn (Callable) – hardware (Any) – model_load_fn (Callable) – load_fn_kwargs (Optional[dict]) – model_reqs (List[str]) – inference_kwargs (Any) – model_id (str) – embed_instruction (str) – query_instruction (str) – Return type None attribute embed_instruction: str = 'Represent the document for retrieval: ' Instruction to use for embedding documents. attribute model_id: str = 'hkunlp/instructor-large' Model name to use. attribute model_reqs: List[str] = ['./', 'InstructorEmbedding', 'torch'] Requirements to install on hardware to inference the model. attribute query_instruction: str = 'Represent the question for retrieving supporting documents: ' Instruction to use for embedding query. embed_documents(texts)[source] Compute doc embeddings using a HuggingFace instruct model. Parameters texts (List[str]) – The list of texts to embed. Returns List of embeddings, one for each text. Return type List[List[float]] embed_query(text)[source] Compute query embeddings using a HuggingFace instruct model. Parameters text (str) – The text to embed. Returns Embeddings for the text. Return type List[float] class langchain.embeddings.FakeEmbeddings(*, size)[source] Bases: langchain.embeddings.base.Embeddings, pydantic.main.BaseModel Parameters size (int) – Return type None embed_documents(texts)[source] Embed search docs. Parameters texts (List[str]) – Return type List[List[float]] embed_query(text)[source] Embed query text. Parameters text (str) – Return type List[float] class langchain.embeddings.AlephAlphaAsymmetricSemanticEmbedding(*, client=None, model='luminous-base', hosting='https://api.aleph-alpha.com', normalize=True, compress_to_size=128, contextual_control_threshold=None, control_log_additive=True, aleph_alpha_api_key=None)[source] Bases: pydantic.main.BaseModel, langchain.embeddings.base.Embeddings Wrapper for Aleph Alpha’s Asymmetric Embeddings AA provides you with an endpoint to embed a document and a query. The models were optimized to make the embeddings of documents and the query for a document as similar as possible. To learn more, check out: https://docs.aleph-alpha.com/docs/tasks/semantic_embed/ Example from aleph_alpha import AlephAlphaAsymmetricSemanticEmbedding embeddings = AlephAlphaSymmetricSemanticEmbedding() document = "This is a content of the document" query = "What is the content of the document?" doc_result = embeddings.embed_documents([document]) query_result = embeddings.embed_query(query) Parameters client (Any) – model (Optional[str]) – hosting (Optional[str]) – normalize (Optional[bool]) – compress_to_size (Optional[int]) – contextual_control_threshold (Optional[int]) – control_log_additive (Optional[bool]) – aleph_alpha_api_key (Optional[str]) – Return type None attribute aleph_alpha_api_key: Optional[str] = None API key for Aleph Alpha API. attribute compress_to_size: Optional[int] = 128 Should the returned embeddings come back as an original 5120-dim vector, or should it be compressed to 128-dim. attribute contextual_control_threshold: Optional[int] = None Attention control parameters only apply to those tokens that have explicitly been set in the request. attribute control_log_additive: Optional[bool] = True Apply controls on prompt items by adding the log(control_factor) to attention scores. attribute hosting: Optional[str] = 'https://api.aleph-alpha.com' Optional parameter that specifies which datacenters may process the request. attribute model: Optional[str] = 'luminous-base' Model name to use. attribute normalize: Optional[bool] = True Should returned embeddings be normalized embed_documents(texts)[source] Call out to Aleph Alpha’s asymmetric Document endpoint. Parameters texts (List[str]) – The list of texts to embed. Returns List of embeddings, one for each text. Return type List[List[float]] embed_query(text)[source] Call out to Aleph Alpha’s asymmetric, query embedding endpoint :param text: The text to embed. Returns Embeddings for the text. Parameters text (str) – Return type List[float] class langchain.embeddings.AlephAlphaSymmetricSemanticEmbedding(*, client=None, model='luminous-base', hosting='https://api.aleph-alpha.com', normalize=True, compress_to_size=128, contextual_control_threshold=None, control_log_additive=True, aleph_alpha_api_key=None)[source] Bases: langchain.embeddings.aleph_alpha.AlephAlphaAsymmetricSemanticEmbedding The symmetric version of the Aleph Alpha’s semantic embeddings. The main difference is that here, both the documents and queries are embedded with a SemanticRepresentation.Symmetric .. rubric:: Example from aleph_alpha import AlephAlphaSymmetricSemanticEmbedding embeddings = AlephAlphaAsymmetricSemanticEmbedding() text = "This is a test text" doc_result = embeddings.embed_documents([text]) query_result = embeddings.embed_query(text) Parameters client (Any) – model (Optional[str]) – hosting (Optional[str]) – normalize (Optional[bool]) – compress_to_size (Optional[int]) – contextual_control_threshold (Optional[int]) – control_log_additive (Optional[bool]) – aleph_alpha_api_key (Optional[str]) – Return type None embed_documents(texts)[source] Call out to Aleph Alpha’s Document endpoint. Parameters texts (List[str]) – The list of texts to embed. Returns List of embeddings, one for each text. Return type List[List[float]] embed_query(text)[source] Call out to Aleph Alpha’s asymmetric, query embedding endpoint :param text: The text to embed. Returns Embeddings for the text. Parameters text (str) – Return type List[float] langchain.embeddings.SentenceTransformerEmbeddings alias of langchain.embeddings.huggingface.HuggingFaceEmbeddings class langchain.embeddings.MiniMaxEmbeddings(*, endpoint_url='https://api.minimax.chat/v1/embeddings', model='embo-01', embed_type_db='db', embed_type_query='query', minimax_group_id=None, minimax_api_key=None)[source] Bases: pydantic.main.BaseModel, langchain.embeddings.base.Embeddings Wrapper around MiniMax’s embedding inference service. To use, you should have the environment variable MINIMAX_GROUP_ID and MINIMAX_API_KEY set with your API token, or pass it as a named parameter to the constructor. Example from langchain.embeddings import MiniMaxEmbeddings embeddings = MiniMaxEmbeddings() query_text = "This is a test query." query_result = embeddings.embed_query(query_text) document_text = "This is a test document." document_result = embeddings.embed_documents([document_text]) Parameters endpoint_url (str) – model (str) – embed_type_db (str) – embed_type_query (str) – minimax_group_id (Optional[str]) – minimax_api_key (Optional[str]) – Return type None attribute embed_type_db: str = 'db' For embed_documents attribute embed_type_query: str = 'query' For embed_query attribute endpoint_url: str = 'https://api.minimax.chat/v1/embeddings' Endpoint URL to use. attribute minimax_api_key: Optional[str] = None API Key for MiniMax API. attribute minimax_group_id: Optional[str] = None Group ID for MiniMax API. attribute model: str = 'embo-01' Embeddings model name to use. embed_documents(texts)[source] Embed documents using a MiniMax embedding endpoint. Parameters texts (List[str]) – The list of texts to embed. Returns List of embeddings, one for each text. Return type List[List[float]] embed_query(text)[source] Embed a query using a MiniMax embedding endpoint. Parameters text (str) – The text to embed. Returns Embeddings for the text. Return type List[float] class langchain.embeddings.BedrockEmbeddings(*, client=None, region_name=None, credentials_profile_name=None, model_id='amazon.titan-e1t-medium', model_kwargs=None)[source] Bases: pydantic.main.BaseModel, langchain.embeddings.base.Embeddings Embeddings provider to invoke Bedrock embedding models. To authenticate, the AWS client uses the following methods to automatically load credentials: https://boto3.amazonaws.com/v1/documentation/api/latest/guide/credentials.html If a specific credential profile should be used, you must pass the name of the profile from the ~/.aws/credentials file that is to be used. Make sure the credentials / roles used have the required policies to access the Bedrock service. Parameters client (Any) – region_name (Optional[str]) – credentials_profile_name (Optional[str]) – model_id (str) – model_kwargs (Optional[Dict]) – Return type None attribute credentials_profile_name: Optional[str] = None The name of the profile in the ~/.aws/credentials or ~/.aws/config files, which has either access keys or role information specified. If not specified, the default credential profile or, if on an EC2 instance, credentials from IMDS will be used. See: https://boto3.amazonaws.com/v1/documentation/api/latest/guide/credentials.html attribute model_id: str = 'amazon.titan-e1t-medium' Id of the model to call, e.g., amazon.titan-e1t-medium, this is equivalent to the modelId property in the list-foundation-models api attribute model_kwargs: Optional[Dict] = None Key word arguments to pass to the model. attribute region_name: Optional[str] = None The aws region e.g., us-west-2. Fallsback to AWS_DEFAULT_REGION env variable or region specified in ~/.aws/config in case it is not provided here. embed_documents(texts, chunk_size=1)[source] Compute doc embeddings using a Bedrock model. Parameters texts (List[str]) – The list of texts to embed. chunk_size (int) – Bedrock currently only allows single string inputs, so chunk size is always 1. This input is here only for compatibility with the embeddings interface. Returns List of embeddings, one for each text. Return type List[List[float]] embed_query(text)[source] Compute query embeddings using a Bedrock model. Parameters text (str) – The text to embed. Returns Embeddings for the text. Return type List[float] class langchain.embeddings.DeepInfraEmbeddings(*, model_id='sentence-transformers/clip-ViT-B-32', normalize=False, embed_instruction='passage: ', query_instruction='query: ', model_kwargs=None, deepinfra_api_token=None)[source] Bases: pydantic.main.BaseModel, langchain.embeddings.base.Embeddings Wrapper around Deep Infra’s embedding inference service. To use, you should have the environment variable DEEPINFRA_API_TOKEN set with your API token, or pass it as a named parameter to the constructor. There are multiple embeddings models available, see https://deepinfra.com/models?type=embeddings. Example from langchain.embeddings import DeepInfraEmbeddings deepinfra_emb = DeepInfraEmbeddings( model_id="sentence-transformers/clip-ViT-B-32", deepinfra_api_token="my-api-key" ) r1 = deepinfra_emb.embed_documents( [ "Alpha is the first letter of Greek alphabet", "Beta is the second letter of Greek alphabet", ] ) r2 = deepinfra_emb.embed_query( "What is the second letter of Greek alphabet" ) Parameters model_id (str) – normalize (bool) – embed_instruction (str) – query_instruction (str) – model_kwargs (Optional[dict]) – deepinfra_api_token (Optional[str]) – Return type None attribute embed_instruction: str = 'passage: ' Instruction used to embed documents. attribute model_id: str = 'sentence-transformers/clip-ViT-B-32' Embeddings model to use. attribute model_kwargs: Optional[dict] = None Other model keyword args attribute normalize: bool = False whether to normalize the computed embeddings attribute query_instruction: str = 'query: ' Instruction used to embed the query. embed_documents(texts)[source] Embed documents using a Deep Infra deployed embedding model. Parameters texts (List[str]) – The list of texts to embed. Returns List of embeddings, one for each text. Return type List[List[float]] embed_query(text)[source] Embed a query using a Deep Infra deployed embedding model. Parameters text (str) – The text to embed. Returns Embeddings for the text. Return type List[float] class langchain.embeddings.DashScopeEmbeddings(*, client=None, model='text-embedding-v1', dashscope_api_key=None, max_retries=5)[source] Bases: pydantic.main.BaseModel, langchain.embeddings.base.Embeddings Wrapper around DashScope embedding models. To use, you should have the dashscope python package installed, and the environment variable DASHSCOPE_API_KEY set with your API key or pass it as a named parameter to the constructor. Example from langchain.embeddings import DashScopeEmbeddings embeddings = DashScopeEmbeddings(dashscope_api_key="my-api-key") Example import os os.environ["DASHSCOPE_API_KEY"] = "your DashScope API KEY" from langchain.embeddings.dashscope import DashScopeEmbeddings embeddings = DashScopeEmbeddings( model="text-embedding-v1", ) text = "This is a test query." query_result = embeddings.embed_query(text) Parameters client (Any) – model (str) – dashscope_api_key (Optional[str]) – max_retries (int) – Return type None attribute dashscope_api_key: Optional[str] = None Maximum number of retries to make when generating. embed_documents(texts)[source] Call out to DashScope’s embedding endpoint for embedding search docs. Parameters texts (List[str]) – The list of texts to embed. chunk_size – The chunk size of embeddings. If None, will use the chunk size specified by the class. Returns List of embeddings, one for each text. Return type List[List[float]] embed_query(text)[source] Call out to DashScope’s embedding endpoint for embedding query text. Parameters text (str) – The text to embed. Returns Embedding for the text. Return type List[float] class langchain.embeddings.EmbaasEmbeddings(*, model='e5-large-v2', instruction=None, api_url='https://api.embaas.io/v1/embeddings/', embaas_api_key=None)[source] Bases: pydantic.main.BaseModel, langchain.embeddings.base.Embeddings Wrapper around embaas’s embedding service. To use, you should have the environment variable EMBAAS_API_KEY set with your API key, or pass it as a named parameter to the constructor. Example # Initialise with default model and instruction from langchain.embeddings import EmbaasEmbeddings emb = EmbaasEmbeddings() # Initialise with custom model and instruction from langchain.embeddings import EmbaasEmbeddings emb_model = "instructor-large" emb_inst = "Represent the Wikipedia document for retrieval" emb = EmbaasEmbeddings( model=emb_model, instruction=emb_inst ) Parameters model (str) – instruction (Optional[str]) – api_url (str) – embaas_api_key (Optional[str]) – Return type None attribute api_url: str = 'https://api.embaas.io/v1/embeddings/' The URL for the embaas embeddings API. attribute instruction: Optional[str] = None Instruction used for domain-specific embeddings. attribute model: str = 'e5-large-v2' The model used for embeddings. embed_documents(texts)[source] Get embeddings for a list of texts. Parameters texts (List[str]) – The list of texts to get embeddings for. Returns List of embeddings, one for each text. Return type List[List[float]] embed_query(text)[source] Get embeddings for a single text. Parameters text (str) – The text to get embeddings for. Returns List of embeddings. Return type List[float]
https://api.python.langchain.com/en/latest/modules/embeddings.html
c3d05344-a68a-42c1-a6e5-0cf256519de1
Memory class langchain.memory.CassandraChatMessageHistory(contact_points, session_id, port=9042, username='cassandra', password='cassandra', keyspace_name='chat_history', table_name='message_store')[source] Bases: langchain.schema.BaseChatMessageHistory Chat message history that stores history in Cassandra. Parameters contact_points (List[str]) – list of ips to connect to Cassandra cluster session_id (str) – arbitrary key that is used to store the messages of a single chat session. port (int) – port to connect to Cassandra cluster username (str) – username to connect to Cassandra cluster password (str) – password to connect to Cassandra cluster keyspace_name (str) – name of the keyspace to use table_name (str) – name of the table to use property messages: List[langchain.schema.BaseMessage] Retrieve the messages from Cassandra add_message(message)[source] Append the message to the record in Cassandra Parameters message (langchain.schema.BaseMessage) – Return type None clear()[source] Clear session memory from Cassandra Return type None class langchain.memory.ChatMessageHistory(*, messages=[])[source] Bases: langchain.schema.BaseChatMessageHistory, pydantic.main.BaseModel Parameters messages (List[langchain.schema.BaseMessage]) – Return type None attribute messages: List[langchain.schema.BaseMessage] = [] add_message(message)[source] Add a self-created message to the store Parameters message (langchain.schema.BaseMessage) – Return type None clear()[source] Remove all messages from the store Return type None class langchain.memory.CombinedMemory(*, memories)[source] Bases: langchain.schema.BaseMemory Class for combining multiple memories’ data together. Parameters memories (List[langchain.schema.BaseMemory]) – Return type None attribute memories: List[langchain.schema.BaseMemory] [Required] For tracking all the memories that should be accessed. clear()[source] Clear context from this session for every memory. Return type None load_memory_variables(inputs)[source] Load all vars from sub-memories. Parameters inputs (Dict[str, Any]) – Return type Dict[str, str] save_context(inputs, outputs)[source] Save context from this session for every memory. Parameters inputs (Dict[str, Any]) – outputs (Dict[str, str]) – Return type None property memory_variables: List[str] All the memory variables that this instance provides. class langchain.memory.ConversationBufferMemory(*, chat_memory=None, output_key=None, input_key=None, return_messages=False, human_prefix='Human', ai_prefix='AI', memory_key='history')[source] Bases: langchain.memory.chat_memory.BaseChatMemory Buffer for storing conversation memory. Parameters chat_memory (langchain.schema.BaseChatMessageHistory) – output_key (Optional[str]) – input_key (Optional[str]) – return_messages (bool) – human_prefix (str) – ai_prefix (str) – memory_key (str) – Return type None attribute ai_prefix: str = 'AI' attribute human_prefix: str = 'Human' load_memory_variables(inputs)[source] Return history buffer. Parameters inputs (Dict[str, Any]) – Return type Dict[str, Any] property buffer: Any String buffer of memory. class langchain.memory.ConversationBufferWindowMemory(*, chat_memory=None, output_key=None, input_key=None, return_messages=False, human_prefix='Human', ai_prefix='AI', memory_key='history', k=5)[source] Bases: langchain.memory.chat_memory.BaseChatMemory Buffer for storing conversation memory. Parameters chat_memory (langchain.schema.BaseChatMessageHistory) – output_key (Optional[str]) – input_key (Optional[str]) – return_messages (bool) – human_prefix (str) – ai_prefix (str) – memory_key (str) – k (int) – Return type None attribute ai_prefix: str = 'AI' attribute human_prefix: str = 'Human' attribute k: int = 5 load_memory_variables(inputs)[source] Return history buffer. Parameters inputs (Dict[str, Any]) – Return type Dict[str, str] property buffer: List[langchain.schema.BaseMessage] String buffer of memory. class langchain.memory.ConversationEntityMemory(*, chat_memory=None, output_key=None, input_key=None, return_messages=False, human_prefix='Human', ai_prefix='AI', llm, entity_extraction_prompt=PromptTemplate(input_variables=['history', 'input'], output_parser=None, partial_variables={}, template='You are an AI assistant reading the transcript of a conversation between an AI and a human. Extract all of the proper nouns from the last line of conversation. As a guideline, a proper noun is generally capitalized. You should definitely extract all names and places.\n\nThe conversation history is provided just in case of a coreference (e.g. "What do you know about him" where "him" is defined in a previous line) -- ignore items mentioned there that are not in the last line.\n\nReturn the output as a single comma-separated list, or NONE if there is nothing of note to return (e.g. the user is just issuing a greeting or having a simple conversation).\n\nEXAMPLE\nConversation history:\nPerson #1: how\'s it going today?\nAI: "It\'s going great! How about you?"\nPerson #1: good! busy working on Langchain. lots to do.\nAI: "That sounds like a lot of work! What kind of things are you doing to make Langchain better?"\nLast line:\nPerson #1: i\'m trying to improve Langchain\'s interfaces, the UX, its integrations with various products the user might want ... a lot of stuff.\nOutput: Langchain\nEND OF EXAMPLE\n\nEXAMPLE\nConversation history:\nPerson #1: how\'s it going today?\nAI: "It\'s going great! How about you?"\nPerson #1: good! busy working on Langchain. lots to do.\nAI: "That sounds like a lot of work! What kind of things are you doing to make Langchain better?"\nLast line:\nPerson #1: i\'m trying to improve Langchain\'s interfaces, the UX, its integrations with various products the user might want ... a lot of stuff. I\'m working with Person #2.\nOutput: Langchain, Person #2\nEND OF EXAMPLE\n\nConversation history (for reference only):\n{history}\nLast line of conversation (for extraction):\nHuman: {input}\n\nOutput:', template_format='f-string', validate_template=True), entity_summarization_prompt=PromptTemplate(input_variables=['entity', 'summary', 'history', 'input'], output_parser=None, partial_variables={}, template='You are an AI assistant helping a human keep track of facts about relevant people, places, and concepts in their life. Update the summary of the provided entity in the "Entity" section based on the last line of your conversation with the human. If you are writing the summary for the first time, return a single sentence.\nThe update should only include facts that are relayed in the last line of conversation about the provided entity, and should only contain facts about the provided entity.\n\nIf there is no new information about the provided entity or the information is not worth noting (not an important or relevant fact to remember long-term), return the existing summary unchanged.\n\nFull conversation history (for context):\n{history}\n\nEntity to summarize:\n{entity}\n\nExisting summary of {entity}:\n{summary}\n\nLast line of conversation:\nHuman: {input}\nUpdated summary:', template_format='f-string', validate_template=True), entity_cache=[], k=3, chat_history_key='history', entity_store=None)[source] Bases: langchain.memory.chat_memory.BaseChatMemory Entity extractor & summarizer memory. Extracts named entities from the recent chat history and generates summaries. With a swapable entity store, persisting entities across conversations. Defaults to an in-memory entity store, and can be swapped out for a Redis, SQLite, or other entity store. Parameters chat_memory (langchain.schema.BaseChatMessageHistory) – output_key (Optional[str]) – input_key (Optional[str]) – return_messages (bool) – human_prefix (str) – ai_prefix (str) – llm (langchain.base_language.BaseLanguageModel) – entity_extraction_prompt (langchain.prompts.base.BasePromptTemplate) – entity_summarization_prompt (langchain.prompts.base.BasePromptTemplate) – entity_cache (List[str]) – k (int) – chat_history_key (str) – entity_store (langchain.memory.entity.BaseEntityStore) – Return type None attribute ai_prefix: str = 'AI' attribute chat_history_key: str = 'history' attribute entity_cache: List[str] = [] attribute entity_extraction_prompt: langchain.prompts.base.BasePromptTemplate = PromptTemplate(input_variables=['history', 'input'], output_parser=None, partial_variables={}, template='You are an AI assistant reading the transcript of a conversation between an AI and a human. Extract all of the proper nouns from the last line of conversation. As a guideline, a proper noun is generally capitalized. You should definitely extract all names and places.\n\nThe conversation history is provided just in case of a coreference (e.g. "What do you know about him" where "him" is defined in a previous line) -- ignore items mentioned there that are not in the last line.\n\nReturn the output as a single comma-separated list, or NONE if there is nothing of note to return (e.g. the user is just issuing a greeting or having a simple conversation).\n\nEXAMPLE\nConversation history:\nPerson #1: how\'s it going today?\nAI: "It\'s going great! How about you?"\nPerson #1: good! busy working on Langchain. lots to do.\nAI: "That sounds like a lot of work! What kind of things are you doing to make Langchain better?"\nLast line:\nPerson #1: i\'m trying to improve Langchain\'s interfaces, the UX, its integrations with various products the user might want ... a lot of stuff.\nOutput: Langchain\nEND OF EXAMPLE\n\nEXAMPLE\nConversation history:\nPerson #1: how\'s it going today?\nAI: "It\'s going great! How about you?"\nPerson #1: good! busy working on Langchain. lots to do.\nAI: "That sounds like a lot of work! What kind of things are you doing to make Langchain better?"\nLast line:\nPerson #1: i\'m trying to improve Langchain\'s interfaces, the UX, its integrations with various products the user might want ... a lot of stuff. I\'m working with Person #2.\nOutput: Langchain, Person #2\nEND OF EXAMPLE\n\nConversation history (for reference only):\n{history}\nLast line of conversation (for extraction):\nHuman: {input}\n\nOutput:', template_format='f-string', validate_template=True) attribute entity_store: langchain.memory.entity.BaseEntityStore [Optional] attribute entity_summarization_prompt: langchain.prompts.base.BasePromptTemplate = PromptTemplate(input_variables=['entity', 'summary', 'history', 'input'], output_parser=None, partial_variables={}, template='You are an AI assistant helping a human keep track of facts about relevant people, places, and concepts in their life. Update the summary of the provided entity in the "Entity" section based on the last line of your conversation with the human. If you are writing the summary for the first time, return a single sentence.\nThe update should only include facts that are relayed in the last line of conversation about the provided entity, and should only contain facts about the provided entity.\n\nIf there is no new information about the provided entity or the information is not worth noting (not an important or relevant fact to remember long-term), return the existing summary unchanged.\n\nFull conversation history (for context):\n{history}\n\nEntity to summarize:\n{entity}\n\nExisting summary of {entity}:\n{summary}\n\nLast line of conversation:\nHuman: {input}\nUpdated summary:', template_format='f-string', validate_template=True) attribute human_prefix: str = 'Human' attribute k: int = 3 attribute llm: langchain.base_language.BaseLanguageModel [Required] clear()[source] Clear memory contents. Return type None load_memory_variables(inputs)[source] Returns chat history and all generated entities with summaries if available, and updates or clears the recent entity cache. New entity name can be found when calling this method, before the entity summaries are generated, so the entity cache values may be empty if no entity descriptions are generated yet. Parameters inputs (Dict[str, Any]) – Return type Dict[str, Any] save_context(inputs, outputs)[source] Save context from this conversation history to the entity store. Generates a summary for each entity in the entity cache by prompting the model, and saves these summaries to the entity store. Parameters inputs (Dict[str, Any]) – outputs (Dict[str, str]) – Return type None property buffer: List[langchain.schema.BaseMessage] Access chat memory messages. class langchain.memory.ConversationKGMemory(*, chat_memory=None, output_key=None, input_key=None, return_messages=False, k=2, human_prefix='Human', ai_prefix='AI', kg=None, knowledge_extraction_prompt=PromptTemplate(input_variables=['history', 'input'], output_parser=None, partial_variables={}, template="You are a networked intelligence helping a human track knowledge triples about all relevant people, things, concepts, etc. and integrating them with your knowledge stored within your weights as well as that stored in a knowledge graph. Extract all of the knowledge triples from the last line of conversation. A knowledge triple is a clause that contains a subject, a predicate, and an object. The subject is the entity being described, the predicate is the property of the subject that is being described, and the object is the value of the property.\n\nEXAMPLE\nConversation history:\nPerson #1: Did you hear aliens landed in Area 51?\nAI: No, I didn't hear that. What do you know about Area 51?\nPerson #1: It's a secret military base in Nevada.\nAI: What do you know about Nevada?\nLast line of conversation:\nPerson #1: It's a state in the US. It's also the number 1 producer of gold in the US.\n\nOutput: (Nevada, is a, state)<|>(Nevada, is in, US)<|>(Nevada, is the number 1 producer of, gold)\nEND OF EXAMPLE\n\nEXAMPLE\nConversation history:\nPerson #1: Hello.\nAI: Hi! How are you?\nPerson #1: I'm good. How are you?\nAI: I'm good too.\nLast line of conversation:\nPerson #1: I'm going to the store.\n\nOutput: NONE\nEND OF EXAMPLE\n\nEXAMPLE\nConversation history:\nPerson #1: What do you know about Descartes?\nAI: Descartes was a French philosopher, mathematician, and scientist who lived in the 17th century.\nPerson #1: The Descartes I'm referring to is a standup comedian and interior designer from Montreal.\nAI: Oh yes, He is a comedian and an interior designer. He has been in the industry for 30 years. His favorite food is baked bean pie.\nLast line of conversation:\nPerson #1: Oh huh. I know Descartes likes to drive antique scooters and play the mandolin.\nOutput: (Descartes, likes to drive, antique scooters)<|>(Descartes, plays, mandolin)\nEND OF EXAMPLE\n\nConversation history (for reference only):\n{history}\nLast line of conversation (for extraction):\nHuman: {input}\n\nOutput:", template_format='f-string', validate_template=True), entity_extraction_prompt=PromptTemplate(input_variables=['history', 'input'], output_parser=None, partial_variables={}, template='You are an AI assistant reading the transcript of a conversation between an AI and a human. Extract all of the proper nouns from the last line of conversation. As a guideline, a proper noun is generally capitalized. You should definitely extract all names and places.\n\nThe conversation history is provided just in case of a coreference (e.g. "What do you know about him" where "him" is defined in a previous line) -- ignore items mentioned there that are not in the last line.\n\nReturn the output as a single comma-separated list, or NONE if there is nothing of note to return (e.g. the user is just issuing a greeting or having a simple conversation).\n\nEXAMPLE\nConversation history:\nPerson #1: how\'s it going today?\nAI: "It\'s going great! How about you?"\nPerson #1: good! busy working on Langchain. lots to do.\nAI: "That sounds like a lot of work! What kind of things are you doing to make Langchain better?"\nLast line:\nPerson #1: i\'m trying to improve Langchain\'s interfaces, the UX, its integrations with various products the user might want ... a lot of stuff.\nOutput: Langchain\nEND OF EXAMPLE\n\nEXAMPLE\nConversation history:\nPerson #1: how\'s it going today?\nAI: "It\'s going great! How about you?"\nPerson #1: good! busy working on Langchain. lots to do.\nAI: "That sounds like a lot of work! What kind of things are you doing to make Langchain better?"\nLast line:\nPerson #1: i\'m trying to improve Langchain\'s interfaces, the UX, its integrations with various products the user might want ... a lot of stuff. I\'m working with Person #2.\nOutput: Langchain, Person #2\nEND OF EXAMPLE\n\nConversation history (for reference only):\n{history}\nLast line of conversation (for extraction):\nHuman: {input}\n\nOutput:', template_format='f-string', validate_template=True), llm, summary_message_cls=<class 'langchain.schema.SystemMessage'>, memory_key='history')[source] Bases: langchain.memory.chat_memory.BaseChatMemory Knowledge graph memory for storing conversation memory. Integrates with external knowledge graph to store and retrieve information about knowledge triples in the conversation. Parameters chat_memory (langchain.schema.BaseChatMessageHistory) – output_key (Optional[str]) – input_key (Optional[str]) – return_messages (bool) – k (int) – human_prefix (str) – ai_prefix (str) – kg (langchain.graphs.networkx_graph.NetworkxEntityGraph) – knowledge_extraction_prompt (langchain.prompts.base.BasePromptTemplate) – entity_extraction_prompt (langchain.prompts.base.BasePromptTemplate) – llm (langchain.base_language.BaseLanguageModel) – summary_message_cls (Type[langchain.schema.BaseMessage]) – memory_key (str) – Return type None attribute ai_prefix: str = 'AI' attribute entity_extraction_prompt: langchain.prompts.base.BasePromptTemplate = PromptTemplate(input_variables=['history', 'input'], output_parser=None, partial_variables={}, template='You are an AI assistant reading the transcript of a conversation between an AI and a human. Extract all of the proper nouns from the last line of conversation. As a guideline, a proper noun is generally capitalized. You should definitely extract all names and places.\n\nThe conversation history is provided just in case of a coreference (e.g. "What do you know about him" where "him" is defined in a previous line) -- ignore items mentioned there that are not in the last line.\n\nReturn the output as a single comma-separated list, or NONE if there is nothing of note to return (e.g. the user is just issuing a greeting or having a simple conversation).\n\nEXAMPLE\nConversation history:\nPerson #1: how\'s it going today?\nAI: "It\'s going great! How about you?"\nPerson #1: good! busy working on Langchain. lots to do.\nAI: "That sounds like a lot of work! What kind of things are you doing to make Langchain better?"\nLast line:\nPerson #1: i\'m trying to improve Langchain\'s interfaces, the UX, its integrations with various products the user might want ... a lot of stuff.\nOutput: Langchain\nEND OF EXAMPLE\n\nEXAMPLE\nConversation history:\nPerson #1: how\'s it going today?\nAI: "It\'s going great! How about you?"\nPerson #1: good! busy working on Langchain. lots to do.\nAI: "That sounds like a lot of work! What kind of things are you doing to make Langchain better?"\nLast line:\nPerson #1: i\'m trying to improve Langchain\'s interfaces, the UX, its integrations with various products the user might want ... a lot of stuff. I\'m working with Person #2.\nOutput: Langchain, Person #2\nEND OF EXAMPLE\n\nConversation history (for reference only):\n{history}\nLast line of conversation (for extraction):\nHuman: {input}\n\nOutput:', template_format='f-string', validate_template=True) attribute human_prefix: str = 'Human' attribute k: int = 2 attribute kg: langchain.graphs.networkx_graph.NetworkxEntityGraph [Optional] attribute knowledge_extraction_prompt: langchain.prompts.base.BasePromptTemplate = PromptTemplate(input_variables=['history', 'input'], output_parser=None, partial_variables={}, template="You are a networked intelligence helping a human track knowledge triples about all relevant people, things, concepts, etc. and integrating them with your knowledge stored within your weights as well as that stored in a knowledge graph. Extract all of the knowledge triples from the last line of conversation. A knowledge triple is a clause that contains a subject, a predicate, and an object. The subject is the entity being described, the predicate is the property of the subject that is being described, and the object is the value of the property.\n\nEXAMPLE\nConversation history:\nPerson #1: Did you hear aliens landed in Area 51?\nAI: No, I didn't hear that. What do you know about Area 51?\nPerson #1: It's a secret military base in Nevada.\nAI: What do you know about Nevada?\nLast line of conversation:\nPerson #1: It's a state in the US. It's also the number 1 producer of gold in the US.\n\nOutput: (Nevada, is a, state)<|>(Nevada, is in, US)<|>(Nevada, is the number 1 producer of, gold)\nEND OF EXAMPLE\n\nEXAMPLE\nConversation history:\nPerson #1: Hello.\nAI: Hi! How are you?\nPerson #1: I'm good. How are you?\nAI: I'm good too.\nLast line of conversation:\nPerson #1: I'm going to the store.\n\nOutput: NONE\nEND OF EXAMPLE\n\nEXAMPLE\nConversation history:\nPerson #1: What do you know about Descartes?\nAI: Descartes was a French philosopher, mathematician, and scientist who lived in the 17th century.\nPerson #1: The Descartes I'm referring to is a standup comedian and interior designer from Montreal.\nAI: Oh yes, He is a comedian and an interior designer. He has been in the industry for 30 years. His favorite food is baked bean pie.\nLast line of conversation:\nPerson #1: Oh huh. I know Descartes likes to drive antique scooters and play the mandolin.\nOutput: (Descartes, likes to drive, antique scooters)<|>(Descartes, plays, mandolin)\nEND OF EXAMPLE\n\nConversation history (for reference only):\n{history}\nLast line of conversation (for extraction):\nHuman: {input}\n\nOutput:", template_format='f-string', validate_template=True) attribute llm: langchain.base_language.BaseLanguageModel [Required] attribute summary_message_cls: Type[langchain.schema.BaseMessage] = <class 'langchain.schema.SystemMessage'> Number of previous utterances to include in the context. clear()[source] Clear memory contents. Return type None get_current_entities(input_string)[source] Parameters input_string (str) – Return type List[str] get_knowledge_triplets(input_string)[source] Parameters input_string (str) – Return type List[langchain.graphs.networkx_graph.KnowledgeTriple] load_memory_variables(inputs)[source] Return history buffer. Parameters inputs (Dict[str, Any]) – Return type Dict[str, Any] save_context(inputs, outputs)[source] Save context from this conversation to buffer. Parameters inputs (Dict[str, Any]) – outputs (Dict[str, str]) – Return type None class langchain.memory.ConversationStringBufferMemory(*, human_prefix='Human', ai_prefix='AI', buffer='', output_key=None, input_key=None, memory_key='history')[source] Bases: langchain.schema.BaseMemory Buffer for storing conversation memory. Parameters human_prefix (str) – ai_prefix (str) – buffer (str) – output_key (Optional[str]) – input_key (Optional[str]) – memory_key (str) – Return type None attribute ai_prefix: str = 'AI' Prefix to use for AI generated responses. attribute buffer: str = '' attribute human_prefix: str = 'Human' attribute input_key: Optional[str] = None attribute output_key: Optional[str] = None clear()[source] Clear memory contents. Return type None load_memory_variables(inputs)[source] Return history buffer. Parameters inputs (Dict[str, Any]) – Return type Dict[str, str] save_context(inputs, outputs)[source] Save context from this conversation to buffer. Parameters inputs (Dict[str, Any]) – outputs (Dict[str, str]) – Return type None property memory_variables: List[str] Will always return list of memory variables. :meta private: class langchain.memory.ConversationSummaryBufferMemory(*, human_prefix='Human', ai_prefix='AI', llm, prompt=PromptTemplate(input_variables=['summary', 'new_lines'], output_parser=None, partial_variables={}, template='Progressively summarize the lines of conversation provided, adding onto the previous summary returning a new summary.\n\nEXAMPLE\nCurrent summary:\nThe human asks what the AI thinks of artificial intelligence. The AI thinks artificial intelligence is a force for good.\n\nNew lines of conversation:\nHuman: Why do you think artificial intelligence is a force for good?\nAI: Because artificial intelligence will help humans reach their full potential.\n\nNew summary:\nThe human asks what the AI thinks of artificial intelligence. The AI thinks artificial intelligence is a force for good because it will help humans reach their full potential.\nEND OF EXAMPLE\n\nCurrent summary:\n{summary}\n\nNew lines of conversation:\n{new_lines}\n\nNew summary:', template_format='f-string', validate_template=True), summary_message_cls=<class 'langchain.schema.SystemMessage'>, chat_memory=None, output_key=None, input_key=None, return_messages=False, max_token_limit=2000, moving_summary_buffer='', memory_key='history')[source] Bases: langchain.memory.chat_memory.BaseChatMemory, langchain.memory.summary.SummarizerMixin Buffer with summarizer for storing conversation memory. Parameters human_prefix (str) – ai_prefix (str) – llm (langchain.base_language.BaseLanguageModel) – prompt (langchain.prompts.base.BasePromptTemplate) – summary_message_cls (Type[langchain.schema.BaseMessage]) – chat_memory (langchain.schema.BaseChatMessageHistory) – output_key (Optional[str]) – input_key (Optional[str]) – return_messages (bool) – max_token_limit (int) – moving_summary_buffer (str) – memory_key (str) – Return type None attribute max_token_limit: int = 2000 attribute memory_key: str = 'history' attribute moving_summary_buffer: str = '' clear()[source] Clear memory contents. Return type None load_memory_variables(inputs)[source] Return history buffer. Parameters inputs (Dict[str, Any]) – Return type Dict[str, Any] prune()[source] Prune buffer if it exceeds max token limit Return type None save_context(inputs, outputs)[source] Save context from this conversation to buffer. Parameters inputs (Dict[str, Any]) – outputs (Dict[str, str]) – Return type None property buffer: List[langchain.schema.BaseMessage] class langchain.memory.ConversationSummaryMemory(*, human_prefix='Human', ai_prefix='AI', llm, prompt=PromptTemplate(input_variables=['summary', 'new_lines'], output_parser=None, partial_variables={}, template='Progressively summarize the lines of conversation provided, adding onto the previous summary returning a new summary.\n\nEXAMPLE\nCurrent summary:\nThe human asks what the AI thinks of artificial intelligence. The AI thinks artificial intelligence is a force for good.\n\nNew lines of conversation:\nHuman: Why do you think artificial intelligence is a force for good?\nAI: Because artificial intelligence will help humans reach their full potential.\n\nNew summary:\nThe human asks what the AI thinks of artificial intelligence. The AI thinks artificial intelligence is a force for good because it will help humans reach their full potential.\nEND OF EXAMPLE\n\nCurrent summary:\n{summary}\n\nNew lines of conversation:\n{new_lines}\n\nNew summary:', template_format='f-string', validate_template=True), summary_message_cls=<class 'langchain.schema.SystemMessage'>, chat_memory=None, output_key=None, input_key=None, return_messages=False, buffer='', memory_key='history')[source] Bases: langchain.memory.chat_memory.BaseChatMemory, langchain.memory.summary.SummarizerMixin Conversation summarizer to memory. Parameters human_prefix (str) – ai_prefix (str) – llm (langchain.base_language.BaseLanguageModel) – prompt (langchain.prompts.base.BasePromptTemplate) – summary_message_cls (Type[langchain.schema.BaseMessage]) – chat_memory (langchain.schema.BaseChatMessageHistory) – output_key (Optional[str]) – input_key (Optional[str]) – return_messages (bool) – buffer (str) – memory_key (str) – Return type None attribute buffer: str = '' clear()[source] Clear memory contents. Return type None classmethod from_messages(llm, chat_memory, *, summarize_step=2, **kwargs)[source] Parameters llm (langchain.base_language.BaseLanguageModel) – chat_memory (langchain.schema.BaseChatMessageHistory) – summarize_step (int) – kwargs (Any) – Return type langchain.memory.summary.ConversationSummaryMemory load_memory_variables(inputs)[source] Return history buffer. Parameters inputs (Dict[str, Any]) – Return type Dict[str, Any] save_context(inputs, outputs)[source] Save context from this conversation to buffer. Parameters inputs (Dict[str, Any]) – outputs (Dict[str, str]) – Return type None class langchain.memory.ConversationTokenBufferMemory(*, chat_memory=None, output_key=None, input_key=None, return_messages=False, human_prefix='Human', ai_prefix='AI', llm, memory_key='history', max_token_limit=2000)[source] Bases: langchain.memory.chat_memory.BaseChatMemory Buffer for storing conversation memory. Parameters chat_memory (langchain.schema.BaseChatMessageHistory) – output_key (Optional[str]) – input_key (Optional[str]) – return_messages (bool) – human_prefix (str) – ai_prefix (str) – llm (langchain.base_language.BaseLanguageModel) – memory_key (str) – max_token_limit (int) – Return type None attribute ai_prefix: str = 'AI' attribute human_prefix: str = 'Human' attribute llm: langchain.base_language.BaseLanguageModel [Required] attribute max_token_limit: int = 2000 attribute memory_key: str = 'history' load_memory_variables(inputs)[source] Return history buffer. Parameters inputs (Dict[str, Any]) – Return type Dict[str, Any] save_context(inputs, outputs)[source] Save context from this conversation to buffer. Pruned. Parameters inputs (Dict[str, Any]) – outputs (Dict[str, str]) – Return type None property buffer: List[langchain.schema.BaseMessage] String buffer of memory. class langchain.memory.CosmosDBChatMessageHistory(cosmos_endpoint, cosmos_database, cosmos_container, session_id, user_id, credential=None, connection_string=None, ttl=None, cosmos_client_kwargs=None)[source] Bases: langchain.schema.BaseChatMessageHistory Chat history backed by Azure CosmosDB. Parameters cosmos_endpoint (str) – cosmos_database (str) – cosmos_container (str) – session_id (str) – user_id (str) – credential (Any) – connection_string (Optional[str]) – ttl (Optional[int]) – cosmos_client_kwargs (Optional[dict]) – prepare_cosmos()[source] Prepare the CosmosDB client. Use this function or the context manager to make sure your database is ready. Return type None load_messages()[source] Retrieve the messages from Cosmos Return type None add_message(message)[source] Add a self-created message to the store Parameters message (langchain.schema.BaseMessage) – Return type None upsert_messages()[source] Update the cosmosdb item. Return type None clear()[source] Clear session memory from this memory and cosmos. Return type None class langchain.memory.DynamoDBChatMessageHistory(table_name, session_id, endpoint_url=None)[source] Bases: langchain.schema.BaseChatMessageHistory Chat message history that stores history in AWS DynamoDB. This class expects that a DynamoDB table with name table_name and a partition Key of SessionId is present. Parameters table_name (str) – name of the DynamoDB table session_id (str) – arbitrary key that is used to store the messages of a single chat session. endpoint_url (Optional[str]) – URL of the AWS endpoint to connect to. This argument is optional and useful for test purposes, like using Localstack. If you plan to use AWS cloud service, you normally don’t have to worry about setting the endpoint_url. property messages: List[langchain.schema.BaseMessage] Retrieve the messages from DynamoDB add_message(message)[source] Append the message to the record in DynamoDB Parameters message (langchain.schema.BaseMessage) – Return type None clear()[source] Clear session memory from DynamoDB Return type None class langchain.memory.FileChatMessageHistory(file_path)[source] Bases: langchain.schema.BaseChatMessageHistory Chat message history that stores history in a local file. Parameters file_path (str) – path of the local file to store the messages. property messages: List[langchain.schema.BaseMessage] Retrieve the messages from the local file add_message(message)[source] Append the message to the record in the local file Parameters message (langchain.schema.BaseMessage) – Return type None clear()[source] Clear session memory from the local file Return type None class langchain.memory.InMemoryEntityStore(*, store={})[source] Bases: langchain.memory.entity.BaseEntityStore Basic in-memory entity store. Parameters store (Dict[str, Optional[str]]) – Return type None attribute store: Dict[str, Optional[str]] = {} clear()[source] Delete all entities from store. Return type None delete(key)[source] Delete entity value from store. Parameters key (str) – Return type None exists(key)[source] Check if entity exists in store. Parameters key (str) – Return type bool get(key, default=None)[source] Get entity value from store. Parameters key (str) – default (Optional[str]) – Return type Optional[str] set(key, value)[source] Set entity value in store. Parameters key (str) – value (Optional[str]) – Return type None class langchain.memory.MomentoChatMessageHistory(session_id, cache_client, cache_name, *, key_prefix='message_store:', ttl=None, ensure_cache_exists=True)[source] Bases: langchain.schema.BaseChatMessageHistory Chat message history cache that uses Momento as a backend. See https://gomomento.com/ Parameters session_id (str) – cache_client (momento.CacheClient) – cache_name (str) – key_prefix (str) – ttl (Optional[timedelta]) – ensure_cache_exists (bool) – classmethod from_client_params(session_id, cache_name, ttl, *, configuration=None, auth_token=None, **kwargs)[source] Construct cache from CacheClient parameters. Parameters session_id (str) – cache_name (str) – ttl (timedelta) – configuration (Optional[momento.config.Configuration]) – auth_token (Optional[str]) – kwargs (Any) – Return type MomentoChatMessageHistory property messages: list[langchain.schema.BaseMessage] Retrieve the messages from Momento. Raises SdkException – Momento service or network error Exception – Unexpected response Returns List of cached messages Return type list[BaseMessage] add_message(message)[source] Store a message in the cache. Parameters message (BaseMessage) – The message object to store. Raises SdkException – Momento service or network error. Exception – Unexpected response. Return type None clear()[source] Remove the session’s messages from the cache. Raises SdkException – Momento service or network error. Exception – Unexpected response. Return type None class langchain.memory.MongoDBChatMessageHistory(connection_string, session_id, database_name='chat_history', collection_name='message_store')[source] Bases: langchain.schema.BaseChatMessageHistory Chat message history that stores history in MongoDB. Parameters connection_string (str) – connection string to connect to MongoDB session_id (str) – arbitrary key that is used to store the messages of a single chat session. database_name (str) – name of the database to use collection_name (str) – name of the collection to use property messages: List[langchain.schema.BaseMessage] Retrieve the messages from MongoDB add_message(message)[source] Append the message to the record in MongoDB Parameters message (langchain.schema.BaseMessage) – Return type None clear()[source] Clear session memory from MongoDB Return type None class langchain.memory.MotorheadMemory(*, chat_memory=None, output_key=None, input_key=None, return_messages=False, url='https://api.getmetal.io/v1/motorhead', session_id, context=None, api_key=None, client_id=None, timeout=3000, memory_key='history')[source] Bases: langchain.memory.chat_memory.BaseChatMemory Parameters chat_memory (langchain.schema.BaseChatMessageHistory) – output_key (Optional[str]) – input_key (Optional[str]) – return_messages (bool) – url (str) – session_id (str) – context (Optional[str]) – api_key (Optional[str]) – client_id (Optional[str]) – timeout (int) – memory_key (str) – Return type None attribute api_key: Optional[str] = None attribute client_id: Optional[str] = None attribute context: Optional[str] = None attribute session_id: str [Required] attribute url: str = 'https://api.getmetal.io/v1/motorhead' delete_session()[source] Delete a session Return type None async init()[source] Return type None load_memory_variables(values)[source] Return key-value pairs given the text input to the chain. If None, return all memories Parameters values (Dict[str, Any]) – Return type Dict[str, Any] save_context(inputs, outputs)[source] Save context from this conversation to buffer. Parameters inputs (Dict[str, Any]) – outputs (Dict[str, str]) – Return type None property memory_variables: List[str] Input keys this memory class will load dynamically. class langchain.memory.PostgresChatMessageHistory(session_id, connection_string='postgresql://postgres:mypassword@localhost/chat_history', table_name='message_store')[source] Bases: langchain.schema.BaseChatMessageHistory Chat message history stored in a Postgres database. Parameters session_id (str) – connection_string (str) – table_name (str) – property messages: List[langchain.schema.BaseMessage] Retrieve the messages from PostgreSQL add_message(message)[source] Append the message to the record in PostgreSQL Parameters message (langchain.schema.BaseMessage) – Return type None clear()[source] Clear session memory from PostgreSQL Return type None class langchain.memory.ReadOnlySharedMemory(*, memory)[source] Bases: langchain.schema.BaseMemory A memory wrapper that is read-only and cannot be changed. Parameters memory (langchain.schema.BaseMemory) – Return type None attribute memory: langchain.schema.BaseMemory [Required] clear()[source] Nothing to clear, got a memory like a vault. Return type None load_memory_variables(inputs)[source] Load memory variables from memory. Parameters inputs (Dict[str, Any]) – Return type Dict[str, str] save_context(inputs, outputs)[source] Nothing should be saved or changed Parameters inputs (Dict[str, Any]) – outputs (Dict[str, str]) – Return type None property memory_variables: List[str] Return memory variables. class langchain.memory.RedisChatMessageHistory(session_id, url='redis://localhost:6379/0', key_prefix='message_store:', ttl=None)[source] Bases: langchain.schema.BaseChatMessageHistory Chat message history stored in a Redis database. Parameters session_id (str) – url (str) – key_prefix (str) – ttl (Optional[int]) – property key: str Construct the record key to use property messages: List[langchain.schema.BaseMessage] Retrieve the messages from Redis add_message(message)[source] Append the message to the record in Redis Parameters message (langchain.schema.BaseMessage) – Return type None clear()[source] Clear session memory from Redis Return type None class langchain.memory.RedisEntityStore(session_id='default', url='redis://localhost:6379/0', key_prefix='memory_store', ttl=86400, recall_ttl=259200, *args, redis_client=None)[source] Bases: langchain.memory.entity.BaseEntityStore Redis-backed Entity store. Entities get a TTL of 1 day by default, and that TTL is extended by 3 days every time the entity is read back. Parameters session_id (str) – url (str) – key_prefix (str) – ttl (Optional[int]) – recall_ttl (Optional[int]) – args (Any) – redis_client (Any) – Return type None attribute key_prefix: str = 'memory_store' attribute recall_ttl: Optional[int] = 259200 attribute redis_client: Any = None attribute session_id: str = 'default' attribute ttl: Optional[int] = 86400 clear()[source] Delete all entities from store. Return type None delete(key)[source] Delete entity value from store. Parameters key (str) – Return type None exists(key)[source] Check if entity exists in store. Parameters key (str) – Return type bool get(key, default=None)[source] Get entity value from store. Parameters key (str) – default (Optional[str]) – Return type Optional[str] set(key, value)[source] Set entity value in store. Parameters key (str) – value (Optional[str]) – Return type None property full_key_prefix: str class langchain.memory.SQLChatMessageHistory(session_id, connection_string, table_name='message_store')[source] Bases: langchain.schema.BaseChatMessageHistory Chat message history stored in an SQL database. Parameters session_id (str) – connection_string (str) – table_name (str) – property messages: List[langchain.schema.BaseMessage] Retrieve all messages from db add_message(message)[source] Append the message to the record in db Parameters message (langchain.schema.BaseMessage) – Return type None clear()[source] Clear session memory from db Return type None class langchain.memory.SQLiteEntityStore(session_id='default', db_file='entities.db', table_name='memory_store', *args)[source] Bases: langchain.memory.entity.BaseEntityStore SQLite-backed Entity store Parameters session_id (str) – db_file (str) – table_name (str) – args (Any) – Return type None attribute session_id: str = 'default' attribute table_name: str = 'memory_store' clear()[source] Delete all entities from store. Return type None delete(key)[source] Delete entity value from store. Parameters key (str) – Return type None exists(key)[source] Check if entity exists in store. Parameters key (str) – Return type bool get(key, default=None)[source] Get entity value from store. Parameters key (str) – default (Optional[str]) – Return type Optional[str] set(key, value)[source] Set entity value in store. Parameters key (str) – value (Optional[str]) – Return type None property full_table_name: str class langchain.memory.SimpleMemory(*, memories={})[source] Bases: langchain.schema.BaseMemory Simple memory for storing context or other bits of information that shouldn’t ever change between prompts. Parameters memories (Dict[str, Any]) – Return type None attribute memories: Dict[str, Any] = {} clear()[source] Nothing to clear, got a memory like a vault. Return type None load_memory_variables(inputs)[source] Return key-value pairs given the text input to the chain. If None, return all memories Parameters inputs (Dict[str, Any]) – Return type Dict[str, str] save_context(inputs, outputs)[source] Nothing should be saved or changed, my memory is set in stone. Parameters inputs (Dict[str, Any]) – outputs (Dict[str, str]) – Return type None property memory_variables: List[str] Input keys this memory class will load dynamically. class langchain.memory.VectorStoreRetrieverMemory(*, retriever, memory_key='history', input_key=None, return_docs=False)[source] Bases: langchain.schema.BaseMemory Class for a VectorStore-backed memory object. Parameters retriever (langchain.vectorstores.base.VectorStoreRetriever) – memory_key (str) – input_key (Optional[str]) – return_docs (bool) – Return type None attribute input_key: Optional[str] = None Key name to index the inputs to load_memory_variables. attribute memory_key: str = 'history' Key name to locate the memories in the result of load_memory_variables. attribute retriever: langchain.vectorstores.base.VectorStoreRetriever [Required] VectorStoreRetriever object to connect to. attribute return_docs: bool = False Whether or not to return the result of querying the database directly. clear()[source] Nothing to clear. Return type None load_memory_variables(inputs)[source] Return history buffer. Parameters inputs (Dict[str, Any]) – Return type Dict[str, Union[List[langchain.schema.Document], str]] save_context(inputs, outputs)[source] Save context from this conversation to buffer. Parameters inputs (Dict[str, Any]) – outputs (Dict[str, str]) – Return type None property memory_variables: List[str] The list of keys emitted from the load_memory_variables method. class langchain.memory.ZepChatMessageHistory(session_id, url='http://localhost:8000')[source] Bases: langchain.schema.BaseChatMessageHistory A ChatMessageHistory implementation that uses Zep as a backend. Recommended usage: # Set up Zep Chat History zep_chat_history = ZepChatMessageHistory( session_id=session_id, url=ZEP_API_URL, ) # Use a standard ConversationBufferMemory to encapsulate the Zep chat history memory = ConversationBufferMemory( memory_key="chat_history", chat_memory=zep_chat_history ) Zep provides long-term conversation storage for LLM apps. The server stores, summarizes, embeds, indexes, and enriches conversational AI chat histories, and exposes them via simple, low-latency APIs. For server installation instructions and more, see: https://getzep.github.io/ This class is a thin wrapper around the zep-python package. Additional Zep functionality is exposed via the zep_summary and zep_messages properties. For more information on the zep-python package, see: https://github.com/getzep/zep-python Parameters session_id (str) – url (str) – Return type None property messages: List[langchain.schema.BaseMessage] Retrieve messages from Zep memory property zep_messages: List[Message] Retrieve summary from Zep memory property zep_summary: Optional[str] Retrieve summary from Zep memory add_message(message)[source] Append the message to the Zep memory history Parameters message (langchain.schema.BaseMessage) – Return type None search(query, metadata=None, limit=None)[source] Search Zep memory for messages matching the query Parameters query (str) – metadata (Optional[Dict]) – limit (Optional[int]) – Return type List[MemorySearchResult] clear()[source] Clear session memory from Zep. Note that Zep is long-term storage for memory and this is not advised unless you have specific data retention requirements. Return type None
https://api.python.langchain.com/en/latest/modules/memory.html
7702ed66-abcf-4761-ad44-02001b146132
Output Parsers class langchain.output_parsers.BooleanOutputParser(*, true_val='YES', false_val='NO')[source] Bases: langchain.schema.BaseOutputParser[bool] Parameters true_val (str) – false_val (str) – Return type None attribute false_val: str = 'NO' attribute true_val: str = 'YES' parse(text)[source] Parse the output of an LLM call to a boolean. Parameters text (str) – output of language model Returns boolean Return type bool class langchain.output_parsers.CombiningOutputParser(*, parsers)[source] Bases: langchain.schema.BaseOutputParser Class to combine multiple output parsers into one. Parameters parsers (List[langchain.schema.BaseOutputParser]) – Return type None attribute parsers: List[langchain.schema.BaseOutputParser] [Required] get_format_instructions()[source] Instructions on how the LLM output should be formatted. Return type str parse(text)[source] Parse the output of an LLM call. Parameters text (str) – Return type Dict[str, Any] class langchain.output_parsers.CommaSeparatedListOutputParser[source] Bases: langchain.output_parsers.list.ListOutputParser Parse out comma separated lists. Return type None get_format_instructions()[source] Instructions on how the LLM output should be formatted. Return type str parse(text)[source] Parse the output of an LLM call. Parameters text (str) – Return type List[str] class langchain.output_parsers.DatetimeOutputParser(*, format='%Y-%m-%dT%H:%M:%S.%fZ')[source] Bases: langchain.schema.BaseOutputParser[datetime.datetime] Parameters format (str) – Return type None attribute format: str = '%Y-%m-%dT%H:%M:%S.%fZ' get_format_instructions()[source] Instructions on how the LLM output should be formatted. Return type str parse(response)[source] Parse the output of an LLM call. A method which takes in a string (assumed output of a language model ) and parses it into some structure. Parameters text – output of language model response (str) – Returns structured output Return type datetime.datetime class langchain.output_parsers.EnumOutputParser(*, enum)[source] Bases: langchain.schema.BaseOutputParser Parameters enum (Type[enum.Enum]) – Return type None attribute enum: Type[enum.Enum] [Required] get_format_instructions()[source] Instructions on how the LLM output should be formatted. Return type str parse(response)[source] Parse the output of an LLM call. A method which takes in a string (assumed output of a language model ) and parses it into some structure. Parameters text – output of language model response (str) – Returns structured output Return type Any class langchain.output_parsers.GuardrailsOutputParser(*, guard=None, api=None, args=None, kwargs=None)[source] Bases: langchain.schema.BaseOutputParser Parameters guard (Any) – api (Optional[Callable]) – args (Any) – kwargs (Any) – Return type None attribute api: Optional[Callable] = None attribute args: Any = None attribute guard: Any = None attribute kwargs: Any = None classmethod from_rail(rail_file, num_reasks=1, api=None, *args, **kwargs)[source] Parameters rail_file (str) – num_reasks (int) – api (Optional[Callable]) – args (Any) – kwargs (Any) – Return type langchain.output_parsers.rail_parser.GuardrailsOutputParser classmethod from_rail_string(rail_str, num_reasks=1, api=None, *args, **kwargs)[source] Parameters rail_str (str) – num_reasks (int) – api (Optional[Callable]) – args (Any) – kwargs (Any) – Return type langchain.output_parsers.rail_parser.GuardrailsOutputParser get_format_instructions()[source] Instructions on how the LLM output should be formatted. Return type str parse(text)[source] Parse the output of an LLM call. A method which takes in a string (assumed output of a language model ) and parses it into some structure. Parameters text (str) – output of language model Returns structured output Return type Dict class langchain.output_parsers.ListOutputParser[source] Bases: langchain.schema.BaseOutputParser Class to parse the output of an LLM call to a list. Return type None abstract parse(text)[source] Parse the output of an LLM call. Parameters text (str) – Return type List[str] class langchain.output_parsers.OutputFixingParser(*, parser, retry_chain)[source] Bases: langchain.schema.BaseOutputParser[langchain.output_parsers.fix.T] Wraps a parser and tries to fix parsing errors. Parameters parser (langchain.schema.BaseOutputParser[langchain.output_parsers.fix.T]) – retry_chain (langchain.chains.llm.LLMChain) – Return type None attribute parser: langchain.schema.BaseOutputParser[langchain.output_parsers.fix.T] [Required] attribute retry_chain: langchain.chains.llm.LLMChain [Required] classmethod from_llm(llm, parser, prompt=PromptTemplate(input_variables=['completion', 'error', 'instructions'], output_parser=None, partial_variables={}, template='Instructions:\n--------------\n{instructions}\n--------------\nCompletion:\n--------------\n{completion}\n--------------\n\nAbove, the Completion did not satisfy the constraints given in the Instructions.\nError:\n--------------\n{error}\n--------------\n\nPlease try again. Please only respond with an answer that satisfies the constraints laid out in the Instructions:', template_format='f-string', validate_template=True))[source] Parameters llm (langchain.base_language.BaseLanguageModel) – parser (langchain.schema.BaseOutputParser[langchain.output_parsers.fix.T]) – prompt (langchain.prompts.base.BasePromptTemplate) – Return type langchain.output_parsers.fix.OutputFixingParser[langchain.output_parsers.fix.T] get_format_instructions()[source] Instructions on how the LLM output should be formatted. Return type str parse(completion)[source] Parse the output of an LLM call. A method which takes in a string (assumed output of a language model ) and parses it into some structure. Parameters text – output of language model completion (str) – Returns structured output Return type langchain.output_parsers.fix.T class langchain.output_parsers.PydanticOutputParser(*, pydantic_object)[source] Bases: langchain.schema.BaseOutputParser[langchain.output_parsers.pydantic.T] Parameters pydantic_object (Type[langchain.output_parsers.pydantic.T]) – Return type None attribute pydantic_object: Type[langchain.output_parsers.pydantic.T] [Required] get_format_instructions()[source] Instructions on how the LLM output should be formatted. Return type str parse(text)[source] Parse the output of an LLM call. A method which takes in a string (assumed output of a language model ) and parses it into some structure. Parameters text (str) – output of language model Returns structured output Return type langchain.output_parsers.pydantic.T class langchain.output_parsers.RegexDictParser(*, regex_pattern="{}:\\s?([^.'\\n']*)\\.?", output_key_to_format, no_update_value=None)[source] Bases: langchain.schema.BaseOutputParser Class to parse the output into a dictionary. Parameters regex_pattern (str) – output_key_to_format (Dict[str, str]) – no_update_value (Optional[str]) – Return type None attribute no_update_value: Optional[str] = None attribute output_key_to_format: Dict[str, str] [Required] attribute regex_pattern: str = "{}:\\s?([^.'\\n']*)\\.?" parse(text)[source] Parse the output of an LLM call. Parameters text (str) – Return type Dict[str, str] class langchain.output_parsers.RegexParser(*, regex, output_keys, default_output_key=None)[source] Bases: langchain.schema.BaseOutputParser Class to parse the output into a dictionary. Parameters regex (str) – output_keys (List[str]) – default_output_key (Optional[str]) – Return type None attribute default_output_key: Optional[str] = None attribute output_keys: List[str] [Required] attribute regex: str [Required] parse(text)[source] Parse the output of an LLM call. Parameters text (str) – Return type Dict[str, str] class langchain.output_parsers.ResponseSchema(*, name, description, type='string')[source] Bases: pydantic.main.BaseModel Parameters name (str) – description (str) – type (str) – Return type None attribute description: str [Required] attribute name: str [Required] attribute type: str = 'string' class langchain.output_parsers.RetryOutputParser(*, parser, retry_chain)[source] Bases: langchain.schema.BaseOutputParser[langchain.output_parsers.retry.T] Wraps a parser and tries to fix parsing errors. Does this by passing the original prompt and the completion to another LLM, and telling it the completion did not satisfy criteria in the prompt. Parameters parser (langchain.schema.BaseOutputParser[langchain.output_parsers.retry.T]) – retry_chain (langchain.chains.llm.LLMChain) – Return type None attribute parser: langchain.schema.BaseOutputParser[langchain.output_parsers.retry.T] [Required] attribute retry_chain: langchain.chains.llm.LLMChain [Required] classmethod from_llm(llm, parser, prompt=PromptTemplate(input_variables=['completion', 'prompt'], output_parser=None, partial_variables={}, template='Prompt:\n{prompt}\nCompletion:\n{completion}\n\nAbove, the Completion did not satisfy the constraints given in the Prompt.\nPlease try again:', template_format='f-string', validate_template=True))[source] Parameters llm (langchain.base_language.BaseLanguageModel) – parser (langchain.schema.BaseOutputParser[langchain.output_parsers.retry.T]) – prompt (langchain.prompts.base.BasePromptTemplate) – Return type langchain.output_parsers.retry.RetryOutputParser[langchain.output_parsers.retry.T] get_format_instructions()[source] Instructions on how the LLM output should be formatted. Return type str parse(completion)[source] Parse the output of an LLM call. A method which takes in a string (assumed output of a language model ) and parses it into some structure. Parameters text – output of language model completion (str) – Returns structured output Return type langchain.output_parsers.retry.T parse_with_prompt(completion, prompt_value)[source] Optional method to parse the output of an LLM call with a prompt. The prompt is largely provided in the event the OutputParser wants to retry or fix the output in some way, and needs information from the prompt to do so. Parameters completion (str) – output of language model prompt – prompt value prompt_value (langchain.schema.PromptValue) – Returns structured output Return type langchain.output_parsers.retry.T class langchain.output_parsers.RetryWithErrorOutputParser(*, parser, retry_chain)[source] Bases: langchain.schema.BaseOutputParser[langchain.output_parsers.retry.T] Wraps a parser and tries to fix parsing errors. Does this by passing the original prompt, the completion, AND the error that was raised to another language model and telling it that the completion did not work, and raised the given error. Differs from RetryOutputParser in that this implementation provides the error that was raised back to the LLM, which in theory should give it more information on how to fix it. Parameters parser (langchain.schema.BaseOutputParser[langchain.output_parsers.retry.T]) – retry_chain (langchain.chains.llm.LLMChain) – Return type None attribute parser: langchain.schema.BaseOutputParser[langchain.output_parsers.retry.T] [Required] attribute retry_chain: langchain.chains.llm.LLMChain [Required] classmethod from_llm(llm, parser, prompt=PromptTemplate(input_variables=['completion', 'error', 'prompt'], output_parser=None, partial_variables={}, template='Prompt:\n{prompt}\nCompletion:\n{completion}\n\nAbove, the Completion did not satisfy the constraints given in the Prompt.\nDetails: {error}\nPlease try again:', template_format='f-string', validate_template=True))[source] Parameters llm (langchain.base_language.BaseLanguageModel) – parser (langchain.schema.BaseOutputParser[langchain.output_parsers.retry.T]) – prompt (langchain.prompts.base.BasePromptTemplate) – Return type langchain.output_parsers.retry.RetryWithErrorOutputParser[langchain.output_parsers.retry.T] get_format_instructions()[source] Instructions on how the LLM output should be formatted. Return type str parse(completion)[source] Parse the output of an LLM call. A method which takes in a string (assumed output of a language model ) and parses it into some structure. Parameters text – output of language model completion (str) – Returns structured output Return type langchain.output_parsers.retry.T parse_with_prompt(completion, prompt_value)[source] Optional method to parse the output of an LLM call with a prompt. The prompt is largely provided in the event the OutputParser wants to retry or fix the output in some way, and needs information from the prompt to do so. Parameters completion (str) – output of language model prompt – prompt value prompt_value (langchain.schema.PromptValue) – Returns structured output Return type langchain.output_parsers.retry.T class langchain.output_parsers.StructuredOutputParser(*, response_schemas)[source] Bases: langchain.schema.BaseOutputParser Parameters response_schemas (List[langchain.output_parsers.structured.ResponseSchema]) – Return type None attribute response_schemas: List[langchain.output_parsers.structured.ResponseSchema] [Required] classmethod from_response_schemas(response_schemas)[source] Parameters response_schemas (List[langchain.output_parsers.structured.ResponseSchema]) – Return type langchain.output_parsers.structured.StructuredOutputParser get_format_instructions()[source] Instructions on how the LLM output should be formatted. Return type str parse(text)[source] Parse the output of an LLM call. A method which takes in a string (assumed output of a language model ) and parses it into some structure. Parameters text (str) – output of language model Returns structured output Return type Any
https://api.python.langchain.com/en/latest/modules/output_parsers.html
392b7fe6-4662-40df-96c9-d94e21dcb8f8
Tools Core toolkit implementations. class langchain.tools.AIPluginTool(*, name, description, args_schema=<class 'langchain.tools.plugin.AIPluginToolSchema'>, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, plugin, api_spec)[source] Bases: langchain.tools.base.BaseTool Parameters name (str) – description (str) – args_schema (Type[langchain.tools.plugin.AIPluginToolSchema]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – plugin (langchain.tools.plugin.AIPlugin) – api_spec (str) – Return type None attribute api_spec: str [Required] attribute args_schema: Type[AIPluginToolSchema] = <class 'langchain.tools.plugin.AIPluginToolSchema'> Pydantic model class to validate and parse the tool’s input arguments. attribute plugin: AIPlugin [Required] classmethod from_plugin_url(url)[source] Parameters url (str) – Return type langchain.tools.plugin.AIPluginTool class langchain.tools.APIOperation(*, operation_id, description=None, base_url, path, method, properties, request_body=None)[source] Bases: pydantic.main.BaseModel A model for a single API operation. Parameters operation_id (str) – description (Optional[str]) – base_url (str) – path (str) – method (langchain.utilities.openapi.HTTPVerb) – properties (Sequence[langchain.tools.openapi.utils.api_models.APIProperty]) – request_body (Optional[langchain.tools.openapi.utils.api_models.APIRequestBody]) – Return type None attribute base_url: str [Required] The base URL of the operation. attribute description: Optional[str] = None The description of the operation. attribute method: langchain.utilities.openapi.HTTPVerb [Required] The HTTP method of the operation. attribute operation_id: str [Required] The unique identifier of the operation. attribute path: str [Required] The path of the operation. attribute properties: Sequence[langchain.tools.openapi.utils.api_models.APIProperty] [Required] attribute request_body: Optional[langchain.tools.openapi.utils.api_models.APIRequestBody] = None The request body of the operation. classmethod from_openapi_spec(spec, path, method)[source] Create an APIOperation from an OpenAPI spec. Parameters spec (langchain.utilities.openapi.OpenAPISpec) – path (str) – method (str) – Return type langchain.tools.openapi.utils.api_models.APIOperation classmethod from_openapi_url(spec_url, path, method)[source] Create an APIOperation from an OpenAPI URL. Parameters spec_url (str) – path (str) – method (str) – Return type langchain.tools.openapi.utils.api_models.APIOperation to_typescript()[source] Get typescript string representation of the operation. Return type str static ts_type_from_python(type_)[source] Parameters type_ (Union[str, Type, tuple, None, enum.Enum]) – Return type str property body_params: List[str] property path_params: List[str] property query_params: List[str] class langchain.tools.ArxivQueryRun(*, name='arxiv', description='A wrapper around Arxiv.org Useful for when you need to answer questions about Physics, Mathematics, Computer Science, Quantitative Biology, Quantitative Finance, Statistics, Electrical Engineering, and Economics from scientific articles on arxiv.org. Input should be a search query.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, api_wrapper=None)[source] Bases: langchain.tools.base.BaseTool Tool that adds the capability to search using the Arxiv API. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – api_wrapper (langchain.utilities.arxiv.ArxivAPIWrapper) – Return type None attribute api_wrapper: langchain.utilities.arxiv.ArxivAPIWrapper [Optional] class langchain.tools.AzureCogsFormRecognizerTool(*, name='azure_cognitive_services_form_recognizer', description='A wrapper around Azure Cognitive Services Form Recognizer. Useful for when you need to extract text, tables, and key-value pairs from documents. Input should be a url to a document.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, azure_cogs_key='', azure_cogs_endpoint='', doc_analysis_client=None)[source] Bases: langchain.tools.base.BaseTool Tool that queries the Azure Cognitive Services Form Recognizer API. In order to set this up, follow instructions at: https://learn.microsoft.com/en-us/azure/applied-ai-services/form-recognizer/quickstarts/get-started-sdks-rest-api?view=form-recog-3.0.0&pivots=programming-language-python Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – azure_cogs_key (str) – azure_cogs_endpoint (str) – doc_analysis_client (Any) – Return type None class langchain.tools.AzureCogsImageAnalysisTool(*, name='azure_cognitive_services_image_analysis', description='A wrapper around Azure Cognitive Services Image Analysis. Useful for when you need to analyze images. Input should be a url to an image.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, azure_cogs_key='', azure_cogs_endpoint='', vision_service=None, analysis_options=None)[source] Bases: langchain.tools.base.BaseTool Tool that queries the Azure Cognitive Services Image Analysis API. In order to set this up, follow instructions at: https://learn.microsoft.com/en-us/azure/cognitive-services/computer-vision/quickstarts-sdk/image-analysis-client-library-40 Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – azure_cogs_key (str) – azure_cogs_endpoint (str) – vision_service (Any) – analysis_options (Any) – Return type None class langchain.tools.AzureCogsSpeech2TextTool(*, name='azure_cognitive_services_speech2text', description='A wrapper around Azure Cognitive Services Speech2Text. Useful for when you need to transcribe audio to text. Input should be a url to an audio file.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, azure_cogs_key='', azure_cogs_region='', speech_language='en-US', speech_config=None)[source] Bases: langchain.tools.base.BaseTool Tool that queries the Azure Cognitive Services Speech2Text API. In order to set this up, follow instructions at: https://learn.microsoft.com/en-us/azure/cognitive-services/speech-service/get-started-speech-to-text?pivots=programming-language-python Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – azure_cogs_key (str) – azure_cogs_region (str) – speech_language (str) – speech_config (Any) – Return type None class langchain.tools.AzureCogsText2SpeechTool(*, name='azure_cognitive_services_text2speech', description='A wrapper around Azure Cognitive Services Text2Speech. Useful for when you need to convert text to speech. ', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, azure_cogs_key='', azure_cogs_region='', speech_language='en-US', speech_config=None)[source] Bases: langchain.tools.base.BaseTool Tool that queries the Azure Cognitive Services Text2Speech API. In order to set this up, follow instructions at: https://learn.microsoft.com/en-us/azure/cognitive-services/speech-service/get-started-text-to-speech?pivots=programming-language-python Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – azure_cogs_key (str) – azure_cogs_region (str) – speech_language (str) – speech_config (Any) – Return type None class langchain.tools.BaseGraphQLTool(*, name='query_graphql', description="Β Β Β  Input to this tool is a detailed and correct GraphQL query, output is a result from the API.\nΒ Β Β  If the query is not correct, an error message will be returned.\nΒ Β Β  If an error is returned with 'Bad request' in it, rewrite the query and try again.\nΒ Β Β  If an error is returned with 'Unauthorized' in it, do not try again, but tell the user to change their authentication.\n\nΒ Β Β  Example Input: query {{ allUsers {{ id, name, email }} }}Β Β Β  ", args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, graphql_wrapper)[source] Bases: langchain.tools.base.BaseTool Base tool for querying a GraphQL API. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – graphql_wrapper (langchain.utilities.graphql.GraphQLAPIWrapper) – Return type None attribute graphql_wrapper: langchain.utilities.graphql.GraphQLAPIWrapper [Required] class langchain.tools.BaseRequestsTool(*, requests_wrapper)[source] Bases: pydantic.main.BaseModel Base class for requests tools. Parameters requests_wrapper (langchain.requests.TextRequestsWrapper) – Return type None attribute requests_wrapper: langchain.requests.TextRequestsWrapper [Required] class langchain.tools.BaseSQLDatabaseTool(*, db)[source] Bases: pydantic.main.BaseModel Base tool for interacting with a SQL database. Parameters db (langchain.sql_database.SQLDatabase) – Return type None attribute db: langchain.sql_database.SQLDatabase [Required] class langchain.tools.BaseSparkSQLTool(*, db)[source] Bases: pydantic.main.BaseModel Base tool for interacting with Spark SQL. Parameters db (langchain.utilities.spark_sql.SparkSQL) – Return type None attribute db: langchain.utilities.spark_sql.SparkSQL [Required] class langchain.tools.BaseTool(*, name, description, args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False)[source] Bases: abc.ABC, pydantic.main.BaseModel Interface LangChain tools must implement. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – Return type None attribute args_schema: Optional[Type[pydantic.main.BaseModel]] = None Pydantic model class to validate and parse the tool’s input arguments. attribute callback_manager: Optional[langchain.callbacks.base.BaseCallbackManager] = None Deprecated. Please use callbacks instead. attribute callbacks: Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]] = None Callbacks to be called during tool execution. attribute description: str [Required] Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute handle_tool_error: Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]] = False Handle the content of the ToolException thrown. attribute name: str [Required] The unique name of the tool that clearly communicates its purpose. attribute return_direct: bool = False Whether to return the tool’s output directly. Setting this to True means that after the tool is called, the AgentExecutor will stop looping. attribute verbose: bool = False Whether to log the tool’s progress. async arun(tool_input, verbose=None, start_color='green', color='green', callbacks=None, **kwargs)[source] Run the tool asynchronously. Parameters tool_input (Union[str, Dict]) – verbose (Optional[bool]) – start_color (Optional[str]) – color (Optional[str]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type Any run(tool_input, verbose=None, start_color='green', color='green', callbacks=None, **kwargs)[source] Run the tool. Parameters tool_input (Union[str, Dict]) – verbose (Optional[bool]) – start_color (Optional[str]) – color (Optional[str]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type Any property args: dict property is_single_input: bool Whether the tool only accepts a single input. class langchain.tools.BingSearchResults(*, name='Bing Search Results JSON', description='A wrapper around Bing Search. Useful for when you need to answer questions about current events. Input should be a search query. Output is a JSON array of the query results', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, num_results=4, api_wrapper)[source] Bases: langchain.tools.base.BaseTool Tool that has capability to query the Bing Search API and get back json. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – num_results (int) – api_wrapper (langchain.utilities.bing_search.BingSearchAPIWrapper) – Return type None attribute api_wrapper: langchain.utilities.bing_search.BingSearchAPIWrapper [Required] attribute num_results: int = 4 class langchain.tools.BingSearchRun(*, name='bing_search', description='A wrapper around Bing Search. Useful for when you need to answer questions about current events. Input should be a search query.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, api_wrapper)[source] Bases: langchain.tools.base.BaseTool Tool that adds the capability to query the Bing search API. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – api_wrapper (langchain.utilities.bing_search.BingSearchAPIWrapper) – Return type None attribute api_wrapper: langchain.utilities.bing_search.BingSearchAPIWrapper [Required] class langchain.tools.BraveSearch(*, name='brave_search', description='a search engine. useful for when you need to answer questions about current events. input should be a search query.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, search_wrapper)[source] Bases: langchain.tools.base.BaseTool Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – search_wrapper (langchain.utilities.brave_search.BraveSearchWrapper) – Return type None attribute search_wrapper: BraveSearchWrapper [Required] classmethod from_api_key(api_key, search_kwargs=None, **kwargs)[source] Parameters api_key (str) – search_kwargs (Optional[dict]) – kwargs (Any) – Return type langchain.tools.brave_search.tool.BraveSearch class langchain.tools.ClickTool(*, name='click_element', description='Click on an element with the given CSS selector', args_schema=<class 'langchain.tools.playwright.click.ClickToolInput'>, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, sync_browser=None, async_browser=None, visible_only=True, playwright_strict=False, playwright_timeout=1000)[source] Bases: langchain.tools.playwright.base.BaseBrowserTool Parameters name (str) – description (str) – args_schema (Type[pydantic.main.BaseModel]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – sync_browser (Optional['SyncBrowser']) – async_browser (Optional['AsyncBrowser']) – visible_only (bool) – playwright_strict (bool) – playwright_timeout (float) – Return type None attribute args_schema: Type[BaseModel] = <class 'langchain.tools.playwright.click.ClickToolInput'> Pydantic model class to validate and parse the tool’s input arguments. attribute description: str = 'Click on an element with the given CSS selector' Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute name: str = 'click_element' The unique name of the tool that clearly communicates its purpose. attribute playwright_strict: bool = False Whether to employ Playwright’s strict mode when clicking on elements. attribute playwright_timeout: float = 1000 Timeout (in ms) for Playwright to wait for element to be ready. attribute visible_only: bool = True Whether to consider only visible elements. class langchain.tools.CopyFileTool(*, name='copy_file', description='Create a copy of a file in a specified location', args_schema=<class 'langchain.tools.file_management.copy.FileCopyInput'>, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, root_dir=None)[source] Bases: langchain.tools.file_management.utils.BaseFileToolMixin, langchain.tools.base.BaseTool Parameters name (str) – description (str) – args_schema (Type[pydantic.main.BaseModel]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – root_dir (Optional[str]) – Return type None attribute args_schema: Type[pydantic.main.BaseModel] = <class 'langchain.tools.file_management.copy.FileCopyInput'> Pydantic model class to validate and parse the tool’s input arguments. attribute description: str = 'Create a copy of a file in a specified location' Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute name: str = 'copy_file' The unique name of the tool that clearly communicates its purpose. class langchain.tools.CurrentWebPageTool(*, name='current_webpage', description='Returns the URL of the current page', args_schema=<class 'pydantic.main.BaseModel'>, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, sync_browser=None, async_browser=None)[source] Bases: langchain.tools.playwright.base.BaseBrowserTool Parameters name (str) – description (str) – args_schema (Type[pydantic.main.BaseModel]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – sync_browser (Optional['SyncBrowser']) – async_browser (Optional['AsyncBrowser']) – Return type None attribute args_schema: Type[BaseModel] = <class 'pydantic.main.BaseModel'> Pydantic model class to validate and parse the tool’s input arguments. attribute description: str = 'Returns the URL of the current page' Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute name: str = 'current_webpage' The unique name of the tool that clearly communicates its purpose. class langchain.tools.DeleteFileTool(*, name='file_delete', description='Delete a file', args_schema=<class 'langchain.tools.file_management.delete.FileDeleteInput'>, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, root_dir=None)[source] Bases: langchain.tools.file_management.utils.BaseFileToolMixin, langchain.tools.base.BaseTool Parameters name (str) – description (str) – args_schema (Type[pydantic.main.BaseModel]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – root_dir (Optional[str]) – Return type None attribute args_schema: Type[pydantic.main.BaseModel] = <class 'langchain.tools.file_management.delete.FileDeleteInput'> Pydantic model class to validate and parse the tool’s input arguments. attribute description: str = 'Delete a file' Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute name: str = 'file_delete' The unique name of the tool that clearly communicates its purpose. class langchain.tools.DuckDuckGoSearchResults(*, name='DuckDuckGo Results JSON', description='A wrapper around Duck Duck Go Search. Useful for when you need to answer questions about current events. Input should be a search query. Output is a JSON array of the query results', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, num_results=4, api_wrapper=None)[source] Bases: langchain.tools.base.BaseTool Tool that queries the Duck Duck Go Search API and get back json. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – num_results (int) – api_wrapper (langchain.utilities.duckduckgo_search.DuckDuckGoSearchAPIWrapper) – Return type None attribute api_wrapper: langchain.utilities.duckduckgo_search.DuckDuckGoSearchAPIWrapper [Optional] attribute num_results: int = 4 class langchain.tools.DuckDuckGoSearchRun(*, name='duckduckgo_search', description='A wrapper around DuckDuckGo Search. Useful for when you need to answer questions about current events. Input should be a search query.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, api_wrapper=None)[source] Bases: langchain.tools.base.BaseTool Tool that adds the capability to query the DuckDuckGo search API. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – api_wrapper (langchain.utilities.duckduckgo_search.DuckDuckGoSearchAPIWrapper) – Return type None attribute api_wrapper: langchain.utilities.duckduckgo_search.DuckDuckGoSearchAPIWrapper [Optional] class langchain.tools.ExtractHyperlinksTool(*, name='extract_hyperlinks', description='Extract all hyperlinks on the current webpage', args_schema=<class 'langchain.tools.playwright.extract_hyperlinks.ExtractHyperlinksToolInput'>, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, sync_browser=None, async_browser=None)[source] Bases: langchain.tools.playwright.base.BaseBrowserTool Extract all hyperlinks on the page. Parameters name (str) – description (str) – args_schema (Type[pydantic.main.BaseModel]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – sync_browser (Optional['SyncBrowser']) – async_browser (Optional['AsyncBrowser']) – Return type None attribute args_schema: Type[BaseModel] = <class 'langchain.tools.playwright.extract_hyperlinks.ExtractHyperlinksToolInput'> Pydantic model class to validate and parse the tool’s input arguments. attribute description: str = 'Extract all hyperlinks on the current webpage' Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute name: str = 'extract_hyperlinks' The unique name of the tool that clearly communicates its purpose. static scrape_page(page, html_content, absolute_urls)[source] Parameters page (Any) – html_content (str) – absolute_urls (bool) – Return type str class langchain.tools.ExtractTextTool(*, name='extract_text', description='Extract all the text on the current webpage', args_schema=<class 'pydantic.main.BaseModel'>, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, sync_browser=None, async_browser=None)[source] Bases: langchain.tools.playwright.base.BaseBrowserTool Parameters name (str) – description (str) – args_schema (Type[pydantic.main.BaseModel]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – sync_browser (Optional['SyncBrowser']) – async_browser (Optional['AsyncBrowser']) – Return type None attribute args_schema: Type[BaseModel] = <class 'pydantic.main.BaseModel'> Pydantic model class to validate and parse the tool’s input arguments. attribute description: str = 'Extract all the text on the current webpage' Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute name: str = 'extract_text' The unique name of the tool that clearly communicates its purpose. class langchain.tools.FileSearchTool(*, name='file_search', description='Recursively search for files in a subdirectory that match the regex pattern', args_schema=<class 'langchain.tools.file_management.file_search.FileSearchInput'>, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, root_dir=None)[source] Bases: langchain.tools.file_management.utils.BaseFileToolMixin, langchain.tools.base.BaseTool Parameters name (str) – description (str) – args_schema (Type[pydantic.main.BaseModel]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – root_dir (Optional[str]) – Return type None attribute args_schema: Type[pydantic.main.BaseModel] = <class 'langchain.tools.file_management.file_search.FileSearchInput'> Pydantic model class to validate and parse the tool’s input arguments. attribute description: str = 'Recursively search for files in a subdirectory that match the regex pattern' Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute name: str = 'file_search' The unique name of the tool that clearly communicates its purpose. class langchain.tools.GetElementsTool(*, name='get_elements', description='Retrieve elements in the current web page matching the given CSS selector', args_schema=<class 'langchain.tools.playwright.get_elements.GetElementsToolInput'>, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, sync_browser=None, async_browser=None)[source] Bases: langchain.tools.playwright.base.BaseBrowserTool Parameters name (str) – description (str) – args_schema (Type[pydantic.main.BaseModel]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – sync_browser (Optional['SyncBrowser']) – async_browser (Optional['AsyncBrowser']) – Return type None attribute args_schema: Type[BaseModel] = <class 'langchain.tools.playwright.get_elements.GetElementsToolInput'> Pydantic model class to validate and parse the tool’s input arguments. attribute description: str = 'Retrieve elements in the current web page matching the given CSS selector' Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute name: str = 'get_elements' The unique name of the tool that clearly communicates its purpose. class langchain.tools.GmailCreateDraft(*, name='create_gmail_draft', description='Use this tool to create a draft email with the provided message fields.', args_schema=<class 'langchain.tools.gmail.create_draft.CreateDraftSchema'>, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, api_resource=None)[source] Bases: langchain.tools.gmail.base.GmailBaseTool Parameters name (str) – description (str) – args_schema (Type[langchain.tools.gmail.create_draft.CreateDraftSchema]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – api_resource (Resource) – Return type None attribute args_schema: Type[langchain.tools.gmail.create_draft.CreateDraftSchema] = <class 'langchain.tools.gmail.create_draft.CreateDraftSchema'> Pydantic model class to validate and parse the tool’s input arguments. attribute description: str = 'Use this tool to create a draft email with the provided message fields.' Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute name: str = 'create_gmail_draft' The unique name of the tool that clearly communicates its purpose. class langchain.tools.GmailGetMessage(*, name='get_gmail_message', description='Use this tool to fetch an email by message ID. Returns the thread ID, snipet, body, subject, and sender.', args_schema=<class 'langchain.tools.gmail.get_message.SearchArgsSchema'>, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, api_resource=None)[source] Bases: langchain.tools.gmail.base.GmailBaseTool Parameters name (str) – description (str) – args_schema (Type[langchain.tools.gmail.get_message.SearchArgsSchema]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – api_resource (Resource) – Return type None attribute args_schema: Type[langchain.tools.gmail.get_message.SearchArgsSchema] = <class 'langchain.tools.gmail.get_message.SearchArgsSchema'> Pydantic model class to validate and parse the tool’s input arguments. attribute description: str = 'Use this tool to fetch an email by message ID. Returns the thread ID, snipet, body, subject, and sender.' Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute name: str = 'get_gmail_message' The unique name of the tool that clearly communicates its purpose. class langchain.tools.GmailGetThread(*, name='get_gmail_thread', description='Use this tool to search for email messages. The input must be a valid Gmail query. The output is a JSON list of messages.', args_schema=<class 'langchain.tools.gmail.get_thread.GetThreadSchema'>, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, api_resource=None)[source] Bases: langchain.tools.gmail.base.GmailBaseTool Parameters name (str) – description (str) – args_schema (Type[langchain.tools.gmail.get_thread.GetThreadSchema]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – api_resource (Resource) – Return type None attribute args_schema: Type[langchain.tools.gmail.get_thread.GetThreadSchema] = <class 'langchain.tools.gmail.get_thread.GetThreadSchema'> Pydantic model class to validate and parse the tool’s input arguments. attribute description: str = 'Use this tool to search for email messages. The input must be a valid Gmail query. The output is a JSON list of messages.' Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute name: str = 'get_gmail_thread' The unique name of the tool that clearly communicates its purpose. class langchain.tools.GmailSearch(*, name='search_gmail', description='Use this tool to search for email messages or threads. The input must be a valid Gmail query. The output is a JSON list of the requested resource.', args_schema=<class 'langchain.tools.gmail.search.SearchArgsSchema'>, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, api_resource=None)[source] Bases: langchain.tools.gmail.base.GmailBaseTool Parameters name (str) – description (str) – args_schema (Type[langchain.tools.gmail.search.SearchArgsSchema]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – api_resource (Resource) – Return type None attribute args_schema: Type[langchain.tools.gmail.search.SearchArgsSchema] = <class 'langchain.tools.gmail.search.SearchArgsSchema'> Pydantic model class to validate and parse the tool’s input arguments. attribute description: str = 'Use this tool to search for email messages or threads. The input must be a valid Gmail query. The output is a JSON list of the requested resource.' Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute name: str = 'search_gmail' The unique name of the tool that clearly communicates its purpose. class langchain.tools.GmailSendMessage(*, name='send_gmail_message', description='Use this tool to send email messages. The input is the message, recipents', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, api_resource=None)[source] Bases: langchain.tools.gmail.base.GmailBaseTool Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – api_resource (Resource) – Return type None attribute description: str = 'Use this tool to send email messages. The input is the message, recipents' Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute name: str = 'send_gmail_message' The unique name of the tool that clearly communicates its purpose. class langchain.tools.GooglePlacesTool(*, name='google_places', description='A wrapper around Google Places. Useful for when you need to validate or discover addressed from ambiguous text. Input should be a search query.', args_schema=<class 'langchain.tools.google_places.tool.GooglePlacesSchema'>, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, api_wrapper=None)[source] Bases: langchain.tools.base.BaseTool Tool that adds the capability to query the Google places API. Parameters name (str) – description (str) – args_schema (Type[pydantic.main.BaseModel]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – api_wrapper (langchain.utilities.google_places_api.GooglePlacesAPIWrapper) – Return type None attribute api_wrapper: langchain.utilities.google_places_api.GooglePlacesAPIWrapper [Optional] attribute args_schema: Type[pydantic.main.BaseModel] = <class 'langchain.tools.google_places.tool.GooglePlacesSchema'> Pydantic model class to validate and parse the tool’s input arguments. class langchain.tools.GoogleSearchResults(*, name='Google Search Results JSON', description='A wrapper around Google Search. Useful for when you need to answer questions about current events. Input should be a search query. Output is a JSON array of the query results', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, num_results=4, api_wrapper)[source] Bases: langchain.tools.base.BaseTool Tool that has capability to query the Google Search API and get back json. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – num_results (int) – api_wrapper (langchain.utilities.google_search.GoogleSearchAPIWrapper) – Return type None attribute api_wrapper: langchain.utilities.google_search.GoogleSearchAPIWrapper [Required] attribute num_results: int = 4 class langchain.tools.GoogleSearchRun(*, name='google_search', description='A wrapper around Google Search. Useful for when you need to answer questions about current events. Input should be a search query.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, api_wrapper)[source] Bases: langchain.tools.base.BaseTool Tool that adds the capability to query the Google search API. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – api_wrapper (langchain.utilities.google_search.GoogleSearchAPIWrapper) – Return type None attribute api_wrapper: langchain.utilities.google_search.GoogleSearchAPIWrapper [Required] class langchain.tools.GoogleSerperResults(*, name='Google Serrper Results JSON', description='A low-cost Google Search API.Useful for when you need to answer questions about current events.Input should be a search query. Output is a JSON object of the query results', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, api_wrapper=None)[source] Bases: langchain.tools.base.BaseTool Tool that has capability to query the Serper.dev Google Search API and get back json. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – api_wrapper (langchain.utilities.google_serper.GoogleSerperAPIWrapper) – Return type None attribute api_wrapper: langchain.utilities.google_serper.GoogleSerperAPIWrapper [Optional] class langchain.tools.GoogleSerperRun(*, name='google_serper', description='A low-cost Google Search API.Useful for when you need to answer questions about current events.Input should be a search query.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, api_wrapper)[source] Bases: langchain.tools.base.BaseTool Tool that adds the capability to query the Serper.dev Google search API. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – api_wrapper (langchain.utilities.google_serper.GoogleSerperAPIWrapper) – Return type None attribute api_wrapper: langchain.utilities.google_serper.GoogleSerperAPIWrapper [Required] class langchain.tools.HumanInputRun(*, name='human', description='You can ask a human for guidance when you think you got stuck or you are not sure what to do next. The input should be a question for the human.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, prompt_func=None, input_func=None)[source] Bases: langchain.tools.base.BaseTool Tool that adds the capability to ask user for input. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – prompt_func (Callable[[str], None]) – input_func (Callable) – Return type None attribute input_func: Callable [Optional] attribute prompt_func: Callable[[str], None] [Optional] class langchain.tools.IFTTTWebhook(*, name, description, args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, url)[source] Bases: langchain.tools.base.BaseTool IFTTT Webhook. Parameters name (str) – name of the tool description (str) – description of the tool url (str) – url to hit with the json event. args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – Return type None attribute url: str [Required] class langchain.tools.InfoPowerBITool(*, name='schema_powerbi', description='\nΒ Β Β  Input to this tool is a comma-separated list of tables, output is the schema and sample rows for those tables.\nΒ Β Β  Be sure that the tables actually exist by calling list_tables_powerbi first!\n\nΒ Β Β  Example Input: "table1, table2, table3"\nΒ Β Β  ', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, powerbi)[source] Bases: langchain.tools.base.BaseTool Tool for getting metadata about a PowerBI Dataset. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – powerbi (langchain.utilities.powerbi.PowerBIDataset) – Return type None attribute powerbi: langchain.utilities.powerbi.PowerBIDataset [Required] class langchain.tools.InfoSQLDatabaseTool(*, name='sql_db_schema', description='\nΒ Β Β  Input to this tool is a comma-separated list of tables, output is the schema and sample rows for those tables.Β Β Β  \n\nΒ Β Β  Example Input: "table1, table2, table3"\nΒ Β Β  ', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, db)[source] Bases: langchain.tools.sql_database.tool.BaseSQLDatabaseTool, langchain.tools.base.BaseTool Tool for getting metadata about a SQL database. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – db (langchain.sql_database.SQLDatabase) – Return type None class langchain.tools.InfoSparkSQLTool(*, name='schema_sql_db', description='\nΒ Β Β  Input to this tool is a comma-separated list of tables, output is the schema and sample rows for those tables.\nΒ Β Β  Be sure that the tables actually exist by calling list_tables_sql_db first!\n\nΒ Β Β  Example Input: "table1, table2, table3"\nΒ Β Β  ', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, db)[source] Bases: langchain.tools.spark_sql.tool.BaseSparkSQLTool, langchain.tools.base.BaseTool Tool for getting metadata about a Spark SQL. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – db (langchain.utilities.spark_sql.SparkSQL) – Return type None class langchain.tools.JiraAction(*, name='', description='', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, api_wrapper=None, mode)[source] Bases: langchain.tools.base.BaseTool Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – api_wrapper (langchain.utilities.jira.JiraAPIWrapper) – mode (str) – Return type None attribute api_wrapper: langchain.utilities.jira.JiraAPIWrapper [Optional] attribute mode: str [Required] class langchain.tools.JsonGetValueTool(*, name='json_spec_get_value', description='\nΒ Β Β  Can be used to see value in string format at a given path.\nΒ Β Β  Before calling this you should be SURE that the path to this exists.\nΒ Β Β  The input is a text representation of the path to the dict in Python syntax (e.g. data["key1"][0]["key2"]).\nΒ Β Β  ', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, spec)[source] Bases: langchain.tools.base.BaseTool Tool for getting a value in a JSON spec. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – spec (langchain.tools.json.tool.JsonSpec) – Return type None attribute spec: JsonSpec [Required] class langchain.tools.JsonListKeysTool(*, name='json_spec_list_keys', description='\nΒ Β Β  Can be used to list all keys at a given path. \nΒ Β Β  Before calling this you should be SURE that the path to this exists.\nΒ Β Β  The input is a text representation of the path to the dict in Python syntax (e.g. data["key1"][0]["key2"]).\nΒ Β Β  ', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, spec)[source] Bases: langchain.tools.base.BaseTool Tool for listing keys in a JSON spec. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – spec (langchain.tools.json.tool.JsonSpec) – Return type None attribute spec: JsonSpec [Required] class langchain.tools.ListDirectoryTool(*, name='list_directory', description='List files and directories in a specified folder', args_schema=<class 'langchain.tools.file_management.list_dir.DirectoryListingInput'>, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, root_dir=None)[source] Bases: langchain.tools.file_management.utils.BaseFileToolMixin, langchain.tools.base.BaseTool Parameters name (str) – description (str) – args_schema (Type[pydantic.main.BaseModel]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – root_dir (Optional[str]) – Return type None attribute args_schema: Type[pydantic.main.BaseModel] = <class 'langchain.tools.file_management.list_dir.DirectoryListingInput'> Pydantic model class to validate and parse the tool’s input arguments. attribute description: str = 'List files and directories in a specified folder' Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute name: str = 'list_directory' The unique name of the tool that clearly communicates its purpose. class langchain.tools.ListPowerBITool(*, name='list_tables_powerbi', description='Input is an empty string, output is a comma separated list of tables in the database.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, powerbi)[source] Bases: langchain.tools.base.BaseTool Tool for getting tables names. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – powerbi (langchain.utilities.powerbi.PowerBIDataset) – Return type None attribute powerbi: langchain.utilities.powerbi.PowerBIDataset [Required] class langchain.tools.ListSQLDatabaseTool(*, name='sql_db_list_tables', description='Input is an empty string, output is a comma separated list of tables in the database.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, db)[source] Bases: langchain.tools.sql_database.tool.BaseSQLDatabaseTool, langchain.tools.base.BaseTool Tool for getting tables names. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – db (langchain.sql_database.SQLDatabase) – Return type None class langchain.tools.ListSparkSQLTool(*, name='list_tables_sql_db', description='Input is an empty string, output is a comma separated list of tables in the Spark SQL.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, db)[source] Bases: langchain.tools.spark_sql.tool.BaseSparkSQLTool, langchain.tools.base.BaseTool Tool for getting tables names. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – db (langchain.utilities.spark_sql.SparkSQL) – Return type None class langchain.tools.MetaphorSearchResults(*, name='metaphor_search_results_json', description='A wrapper around Metaphor Search. Input should be a Metaphor-optimized query. Output is a JSON array of the query results', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, api_wrapper)[source] Bases: langchain.tools.base.BaseTool Tool that has capability to query the Metaphor Search API and get back json. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – api_wrapper (langchain.utilities.metaphor_search.MetaphorSearchAPIWrapper) – Return type None attribute api_wrapper: langchain.utilities.metaphor_search.MetaphorSearchAPIWrapper [Required] class langchain.tools.MoveFileTool(*, name='move_file', description='Move or rename a file from one location to another', args_schema=<class 'langchain.tools.file_management.move.FileMoveInput'>, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, root_dir=None)[source] Bases: langchain.tools.file_management.utils.BaseFileToolMixin, langchain.tools.base.BaseTool Parameters name (str) – description (str) – args_schema (Type[pydantic.main.BaseModel]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – root_dir (Optional[str]) – Return type None attribute args_schema: Type[pydantic.main.BaseModel] = <class 'langchain.tools.file_management.move.FileMoveInput'> Pydantic model class to validate and parse the tool’s input arguments. attribute description: str = 'Move or rename a file from one location to another' Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute name: str = 'move_file' The unique name of the tool that clearly communicates its purpose. class langchain.tools.NavigateBackTool(*, name='previous_webpage', description='Navigate back to the previous page in the browser history', args_schema=<class 'pydantic.main.BaseModel'>, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, sync_browser=None, async_browser=None)[source] Bases: langchain.tools.playwright.base.BaseBrowserTool Navigate back to the previous page in the browser history. Parameters name (str) – description (str) – args_schema (Type[pydantic.main.BaseModel]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – sync_browser (Optional['SyncBrowser']) – async_browser (Optional['AsyncBrowser']) – Return type None attribute args_schema: Type[BaseModel] = <class 'pydantic.main.BaseModel'> Pydantic model class to validate and parse the tool’s input arguments. attribute description: str = 'Navigate back to the previous page in the browser history' Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute name: str = 'previous_webpage' The unique name of the tool that clearly communicates its purpose. class langchain.tools.NavigateTool(*, name='navigate_browser', description='Navigate a browser to the specified URL', args_schema=<class 'langchain.tools.playwright.navigate.NavigateToolInput'>, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, sync_browser=None, async_browser=None)[source] Bases: langchain.tools.playwright.base.BaseBrowserTool Parameters name (str) – description (str) – args_schema (Type[pydantic.main.BaseModel]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – sync_browser (Optional['SyncBrowser']) – async_browser (Optional['AsyncBrowser']) – Return type None attribute args_schema: Type[BaseModel] = <class 'langchain.tools.playwright.navigate.NavigateToolInput'> Pydantic model class to validate and parse the tool’s input arguments. attribute description: str = 'Navigate a browser to the specified URL' Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute name: str = 'navigate_browser' The unique name of the tool that clearly communicates its purpose. class langchain.tools.OpenAPISpec(*, openapi='3.1.0', info, jsonSchemaDialect=None, servers=[Server(url='/', description=None, variables=None)], paths=None, webhooks=None, components=None, security=None, tags=None, externalDocs=None)[source] Bases: openapi_schema_pydantic.v3.v3_1_0.open_api.OpenAPI OpenAPI Model that removes misformatted parts of the spec. Parameters openapi (str) – info (openapi_schema_pydantic.v3.v3_1_0.info.Info) – jsonSchemaDialect (Optional[str]) – servers (List[openapi_schema_pydantic.v3.v3_1_0.server.Server]) – paths (Optional[Dict[str, openapi_schema_pydantic.v3.v3_1_0.path_item.PathItem]]) – webhooks (Optional[Dict[str, Union[openapi_schema_pydantic.v3.v3_1_0.path_item.PathItem, openapi_schema_pydantic.v3.v3_1_0.reference.Reference]]]) – components (Optional[openapi_schema_pydantic.v3.v3_1_0.components.Components]) – security (Optional[List[Dict[str, List[str]]]]) – tags (Optional[List[openapi_schema_pydantic.v3.v3_1_0.tag.Tag]]) – externalDocs (Optional[openapi_schema_pydantic.v3.v3_1_0.external_documentation.ExternalDocumentation]) – Return type None classmethod from_file(path)[source] Get an OpenAPI spec from a file path. Parameters path (Union[str, pathlib.Path]) – Return type langchain.utilities.openapi.OpenAPISpec classmethod from_spec_dict(spec_dict)[source] Get an OpenAPI spec from a dict. Parameters spec_dict (dict) – Return type langchain.utilities.openapi.OpenAPISpec classmethod from_text(text)[source] Get an OpenAPI spec from a text. Parameters text (str) – Return type langchain.utilities.openapi.OpenAPISpec classmethod from_url(url)[source] Get an OpenAPI spec from a URL. Parameters url (str) – Return type langchain.utilities.openapi.OpenAPISpec static get_cleaned_operation_id(operation, path, method)[source] Get a cleaned operation id from an operation id. Parameters operation (openapi_schema_pydantic.v3.v3_1_0.operation.Operation) – path (str) – method (str) – Return type str get_methods_for_path(path)[source] Return a list of valid methods for the specified path. Parameters path (str) – Return type List[str] get_operation(path, method)[source] Get the operation object for a given path and HTTP method. Parameters path (str) – method (str) – Return type openapi_schema_pydantic.v3.v3_1_0.operation.Operation get_parameters_for_operation(operation)[source] Get the components for a given operation. Parameters operation (openapi_schema_pydantic.v3.v3_1_0.operation.Operation) – Return type List[openapi_schema_pydantic.v3.v3_1_0.parameter.Parameter] get_parameters_for_path(path)[source] Parameters path (str) – Return type List[openapi_schema_pydantic.v3.v3_1_0.parameter.Parameter] get_referenced_schema(ref)[source] Get a schema (or nested reference) or err. Parameters ref (openapi_schema_pydantic.v3.v3_1_0.reference.Reference) – Return type openapi_schema_pydantic.v3.v3_1_0.schema.Schema get_request_body_for_operation(operation)[source] Get the request body for a given operation. Parameters operation (openapi_schema_pydantic.v3.v3_1_0.operation.Operation) – Return type Optional[openapi_schema_pydantic.v3.v3_1_0.request_body.RequestBody] get_schema(schema)[source] Parameters schema (Union[openapi_schema_pydantic.v3.v3_1_0.reference.Reference, openapi_schema_pydantic.v3.v3_1_0.schema.Schema]) – Return type openapi_schema_pydantic.v3.v3_1_0.schema.Schema classmethod parse_obj(obj)[source] Parameters obj (dict) – Return type langchain.utilities.openapi.OpenAPISpec property base_url: str Get the base url. class langchain.tools.OpenWeatherMapQueryRun(*, name='OpenWeatherMap', description='A wrapper around OpenWeatherMap API. Useful for fetching current weather information for a specified location. Input should be a location string (e.g. London,GB).', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, api_wrapper=None)[source] Bases: langchain.tools.base.BaseTool Tool that adds the capability to query using the OpenWeatherMap API. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – api_wrapper (langchain.utilities.openweathermap.OpenWeatherMapAPIWrapper) – Return type None attribute api_wrapper: langchain.utilities.openweathermap.OpenWeatherMapAPIWrapper [Optional] class langchain.tools.PubmedQueryRun(*, name='PubMed', description='A wrapper around PubMed.org Useful for when you need to answer questions about Physics, Mathematics, Computer Science, Quantitative Biology, Quantitative Finance, Statistics, Electrical Engineering, and Economics from scientific articles on PubMed.org. Input should be a search query.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, api_wrapper=None)[source] Bases: langchain.tools.base.BaseTool Tool that adds the capability to search using the PubMed API. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – api_wrapper (langchain.utilities.pupmed.PubMedAPIWrapper) – Return type None attribute api_wrapper: langchain.utilities.pupmed.PubMedAPIWrapper [Optional] class langchain.tools.PythonAstREPLTool(*, name='python_repl_ast', description='A Python shell. Use this to execute python commands. Input should be a valid python command. When using this tool, sometimes output is abbreviated - make sure it does not look abbreviated before using it in your answer.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, globals=None, locals=None, sanitize_input=True)[source] Bases: langchain.tools.base.BaseTool A tool for running python code in a REPL. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – globals (Optional[Dict]) – locals (Optional[Dict]) – sanitize_input (bool) – Return type None attribute globals: Optional[Dict] [Optional] attribute locals: Optional[Dict] [Optional] attribute sanitize_input: bool = True class langchain.tools.PythonREPLTool(*, name='Python_REPL', description='A Python shell. Use this to execute python commands. Input should be a valid python command. If you want to see the output of a value, you should print it out with `print(...)`.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, python_repl=None, sanitize_input=True)[source] Bases: langchain.tools.base.BaseTool A tool for running python code in a REPL. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – python_repl (langchain.utilities.python.PythonREPL) – sanitize_input (bool) – Return type None attribute python_repl: langchain.utilities.python.PythonREPL [Optional] attribute sanitize_input: bool = True class langchain.tools.QueryCheckerTool(*, name='query_checker_sql_db', description='\nΒ Β Β  Use this tool to double check if your query is correct before executing it.\nΒ Β Β  Always use this tool before executing a query with query_sql_db!\nΒ Β Β  ', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, db, template='\n{query}\nDouble check the Spark SQL query above for common mistakes, including:\n- Using NOT IN with NULL values\n- Using UNION when UNION ALL should have been used\n- Using BETWEEN for exclusive ranges\n- Data type mismatch in predicates\n- Properly quoting identifiers\n- Using the correct number of arguments for functions\n- Casting to the correct data type\n- Using the proper columns for joins\n\nIf there are any of the above mistakes, rewrite the query. If there are no mistakes, just reproduce the original query.', llm, llm_chain)[source] Bases: langchain.tools.spark_sql.tool.BaseSparkSQLTool, langchain.tools.base.BaseTool Use an LLM to check if a query is correct. Adapted from https://www.patterns.app/blog/2023/01/18/crunchbot-sql-analyst-gpt/ Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – db (langchain.utilities.spark_sql.SparkSQL) – template (str) – llm (langchain.base_language.BaseLanguageModel) – llm_chain (langchain.chains.llm.LLMChain) – Return type None attribute llm: langchain.base_language.BaseLanguageModel [Required] attribute llm_chain: langchain.chains.llm.LLMChain [Required] attribute template: str = '\n{query}\nDouble check the Spark SQL query above for common mistakes, including:\n- Using NOT IN with NULL values\n- Using UNION when UNION ALL should have been used\n- Using BETWEEN for exclusive ranges\n- Data type mismatch in predicates\n- Properly quoting identifiers\n- Using the correct number of arguments for functions\n- Casting to the correct data type\n- Using the proper columns for joins\n\nIf there are any of the above mistakes, rewrite the query. If there are no mistakes, just reproduce the original query.' class langchain.tools.QueryPowerBITool(*, name='query_powerbi', description='\nΒ Β Β  Input to this tool is a detailed question about the dataset, output is a result from the dataset. It will try to answer the question using the dataset, and if it cannot, it will ask for clarification.\n\nΒ Β Β  Example Input: "How many rows are in table1?"\nΒ Β Β  ', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, llm_chain, powerbi, template='\nAnswer the question below with a DAX query that can be sent to Power BI. DAX queries have a simple syntax comprised of just one required keyword, EVALUATE, and several optional keywords: ORDER BY, START AT, DEFINE, MEASURE, VAR, TABLE, and COLUMN. Each keyword defines a statement used for the duration of the query. Any time < or > are used in the text below it means that those values need to be replaced by table, columns or other things. If the question is not something you can answer with a DAX query, reply with "I cannot answer this" and the question will be escalated to a human.\n\nSome DAX functions return a table instead of a scalar, and must be wrapped in a function that evaluates the table and returns a scalar; unless the table is a single column, single row table, then it is treated as a scalar value. Most DAX functions require one or more arguments, which can include tables, columns, expressions, and values. However, some functions, such as PI, do not require any arguments, but always require parentheses to indicate the null argument. For example, you must always type PI(), not PI. You can also nest functions within other functions. \n\nSome commonly used functions are:\nEVALUATE <table> - At the most basic level, a DAX query is an EVALUATE statement containing a table expression. At least one EVALUATE statement is required, however, a query can contain any number of EVALUATE statements.\nEVALUATE <table> ORDER BY <expression> ASC or DESC - The optional ORDER BY keyword defines one or more expressions used to sort query results. Any expression that can be evaluated for each row of the result is valid.\nEVALUATE <table> ORDER BY <expression> ASC or DESC START AT <value> or <parameter> - The optional START AT keyword is used inside an ORDER BY clause. It defines the value at which the query results begin.\nDEFINE MEASURE | VAR; EVALUATE <table> - The optional DEFINE keyword introduces one or more calculated entity definitions that exist only for the duration of the query. Definitions precede the EVALUATE statement and are valid for all EVALUATE statements in the query. Definitions can be variables, measures, tables1, and columns1. Definitions can reference other definitions that appear before or after the current definition. At least one definition is required if the DEFINE keyword is included in a query.\nMEASURE <table name>[<measure name>] = <scalar expression> - Introduces a measure definition in a DEFINE statement of a DAX query.\nVAR <name> = <expression> - Stores the result of an expression as a named variable, which can then be passed as an argument to other measure expressions. Once resultant values have been calculated for a variable expression, those values do not change, even if the variable is referenced in another expression.\n\nFILTER(<table>,<filter>) - Returns a table that represents a subset of another table or expression, where <filter> is a Boolean expression that is to be evaluated for each row of the table. For example, [Amount] > 0 or [Region] = "France"\nROW(<name>, <expression>) - Returns a table with a single row containing values that result from the expressions given to each column.\nDISTINCT(<column>) - Returns a one-column table that contains the distinct values from the specified column. In other words, duplicate values are removed and only unique values are returned. This function cannot be used to Return values into a cell or column on a worksheet; rather, you nest the DISTINCT function within a formula, to get a list of distinct values that can be passed to another function and then counted, summed, or used for other operations.\nDISTINCT(<table>) - Returns a table by removing duplicate rows from another table or expression.\n\nAggregation functions, names with a A in it, handle booleans and empty strings in appropriate ways, while the same function without A only uses the numeric values in a column. Functions names with an X in it can include a expression as an argument, this will be evaluated for each row in the table and the result will be used in the regular function calculation, these are the functions:\nCOUNT(<column>), COUNTA(<column>), COUNTX(<table>,<expression>), COUNTAX(<table>,<expression>), COUNTROWS([<table>]), COUNTBLANK(<column>), DISTINCTCOUNT(<column>), DISTINCTCOUNTNOBLANK (<column>) - these are all variantions of count functions.\nAVERAGE(<column>), AVERAGEA(<column>), AVERAGEX(<table>,<expression>) - these are all variantions of average functions.\nMAX(<column>), MAXA(<column>), MAXX(<table>,<expression>) - these are all variantions of max functions.\nMIN(<column>), MINA(<column>), MINX(<table>,<expression>) - these are all variantions of min functions.\nPRODUCT(<column>), PRODUCTX(<table>,<expression>) - these are all variantions of product functions.\nSUM(<column>), SUMX(<table>,<expression>) - these are all variantions of sum functions.\n\nDate and time functions:\nDATE(year, month, day) - Returns a date value that represents the specified year, month, and day.\nDATEDIFF(date1, date2, <interval>) - Returns the difference between two date values, in the specified interval, that can be SECOND, MINUTE, HOUR, DAY, WEEK, MONTH, QUARTER, YEAR.\nDATEVALUE(<date_text>) - Returns a date value that represents the specified date.\nYEAR(<date>), QUARTER(<date>), MONTH(<date>), DAY(<date>), HOUR(<date>), MINUTE(<date>), SECOND(<date>) - Returns the part of the date for the specified date.\n\nFinally, make sure to escape double quotes with a single backslash, and make sure that only table names have single quotes around them, while names of measures or the values of columns that you want to compare against are in escaped double quotes. Newlines are not necessary and can be skipped. The queries are serialized as json and so will have to fit be compliant with json syntax. Sometimes you will get a question, a DAX query and a error, in that case you need to rewrite the DAX query to get the correct answer.\n\nThe following tables exist: {tables}\n\nand the schema\'s for some are given here:\n{schemas}\n\nExamples:\n{examples}\n\nQuestion: {tool_input}\nDAX: \n', examples='\nQuestion: How many rows are in the table <table>?\nDAX: EVALUATE ROW("Number of rows", COUNTROWS(<table>))\n----\nQuestion: How many rows are in the table <table> where <column> is not empty?\nDAX: EVALUATE ROW("Number of rows", COUNTROWS(FILTER(<table>, <table>[<column>] <> "")))\n----\nQuestion: What was the average of <column> in <table>?\nDAX: EVALUATE ROW("Average", AVERAGE(<table>[<column>]))\n----\n', session_cache=None, max_iterations=5)[source] Bases: langchain.tools.base.BaseTool Tool for querying a Power BI Dataset. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – llm_chain (langchain.chains.llm.LLMChain) – powerbi (langchain.utilities.powerbi.PowerBIDataset) – template (Optional[str]) – examples (Optional[str]) – session_cache (Dict[str, Any]) – max_iterations (int) – Return type None attribute examples: Optional[str] = '\nQuestion: How many rows are in the table <table>?\nDAX: EVALUATE ROW("Number of rows", COUNTROWS(<table>))\n----\nQuestion: How many rows are in the table <table> where <column> is not empty?\nDAX: EVALUATE ROW("Number of rows", COUNTROWS(FILTER(<table>, <table>[<column>] <> "")))\n----\nQuestion: What was the average of <column> in <table>?\nDAX: EVALUATE ROW("Average", AVERAGE(<table>[<column>]))\n----\n' attribute llm_chain: langchain.chains.llm.LLMChain [Required] attribute max_iterations: int = 5 attribute powerbi: langchain.utilities.powerbi.PowerBIDataset [Required] attribute session_cache: Dict[str, Any] [Optional] attribute template: Optional[str] = '\nAnswer the question below with a DAX query that can be sent to Power BI. DAX queries have a simple syntax comprised of just one required keyword, EVALUATE, and several optional keywords: ORDER BY, START AT, DEFINE, MEASURE, VAR, TABLE, and COLUMN. Each keyword defines a statement used for the duration of the query. Any time < or > are used in the text below it means that those values need to be replaced by table, columns or other things. If the question is not something you can answer with a DAX query, reply with "I cannot answer this" and the question will be escalated to a human.\n\nSome DAX functions return a table instead of a scalar, and must be wrapped in a function that evaluates the table and returns a scalar; unless the table is a single column, single row table, then it is treated as a scalar value. Most DAX functions require one or more arguments, which can include tables, columns, expressions, and values. However, some functions, such as PI, do not require any arguments, but always require parentheses to indicate the null argument. For example, you must always type PI(), not PI. You can also nest functions within other functions. \n\nSome commonly used functions are:\nEVALUATE <table> - At the most basic level, a DAX query is an EVALUATE statement containing a table expression. At least one EVALUATE statement is required, however, a query can contain any number of EVALUATE statements.\nEVALUATE <table> ORDER BY <expression> ASC or DESC - The optional ORDER BY keyword defines one or more expressions used to sort query results. Any expression that can be evaluated for each row of the result is valid.\nEVALUATE <table> ORDER BY <expression> ASC or DESC START AT <value> or <parameter> - The optional START AT keyword is used inside an ORDER BY clause. It defines the value at which the query results begin.\nDEFINE MEASURE | VAR; EVALUATE <table> - The optional DEFINE keyword introduces one or more calculated entity definitions that exist only for the duration of the query. Definitions precede the EVALUATE statement and are valid for all EVALUATE statements in the query. Definitions can be variables, measures, tables1, and columns1. Definitions can reference other definitions that appear before or after the current definition. At least one definition is required if the DEFINE keyword is included in a query.\nMEASURE <table name>[<measure name>] = <scalar expression> - Introduces a measure definition in a DEFINE statement of a DAX query.\nVAR <name> = <expression> - Stores the result of an expression as a named variable, which can then be passed as an argument to other measure expressions. Once resultant values have been calculated for a variable expression, those values do not change, even if the variable is referenced in another expression.\n\nFILTER(<table>,<filter>) - Returns a table that represents a subset of another table or expression, where <filter> is a Boolean expression that is to be evaluated for each row of the table. For example, [Amount] > 0 or [Region] = "France"\nROW(<name>, <expression>) - Returns a table with a single row containing values that result from the expressions given to each column.\nDISTINCT(<column>) - Returns a one-column table that contains the distinct values from the specified column. In other words, duplicate values are removed and only unique values are returned. This function cannot be used to Return values into a cell or column on a worksheet; rather, you nest the DISTINCT function within a formula, to get a list of distinct values that can be passed to another function and then counted, summed, or used for other operations.\nDISTINCT(<table>) - Returns a table by removing duplicate rows from another table or expression.\n\nAggregation functions, names with a A in it, handle booleans and empty strings in appropriate ways, while the same function without A only uses the numeric values in a column. Functions names with an X in it can include a expression as an argument, this will be evaluated for each row in the table and the result will be used in the regular function calculation, these are the functions:\nCOUNT(<column>), COUNTA(<column>), COUNTX(<table>,<expression>), COUNTAX(<table>,<expression>), COUNTROWS([<table>]), COUNTBLANK(<column>), DISTINCTCOUNT(<column>), DISTINCTCOUNTNOBLANK (<column>) - these are all variantions of count functions.\nAVERAGE(<column>), AVERAGEA(<column>), AVERAGEX(<table>,<expression>) - these are all variantions of average functions.\nMAX(<column>), MAXA(<column>), MAXX(<table>,<expression>) - these are all variantions of max functions.\nMIN(<column>), MINA(<column>), MINX(<table>,<expression>) - these are all variantions of min functions.\nPRODUCT(<column>), PRODUCTX(<table>,<expression>) - these are all variantions of product functions.\nSUM(<column>), SUMX(<table>,<expression>) - these are all variantions of sum functions.\n\nDate and time functions:\nDATE(year, month, day) - Returns a date value that represents the specified year, month, and day.\nDATEDIFF(date1, date2, <interval>) - Returns the difference between two date values, in the specified interval, that can be SECOND, MINUTE, HOUR, DAY, WEEK, MONTH, QUARTER, YEAR.\nDATEVALUE(<date_text>) - Returns a date value that represents the specified date.\nYEAR(<date>), QUARTER(<date>), MONTH(<date>), DAY(<date>), HOUR(<date>), MINUTE(<date>), SECOND(<date>) - Returns the part of the date for the specified date.\n\nFinally, make sure to escape double quotes with a single backslash, and make sure that only table names have single quotes around them, while names of measures or the values of columns that you want to compare against are in escaped double quotes. Newlines are not necessary and can be skipped. The queries are serialized as json and so will have to fit be compliant with json syntax. Sometimes you will get a question, a DAX query and a error, in that case you need to rewrite the DAX query to get the correct answer.\n\nThe following tables exist: {tables}\n\nand the schema\'s for some are given here:\n{schemas}\n\nExamples:\n{examples}\n\nQuestion: {tool_input}\nDAX: \n' class langchain.tools.QuerySQLCheckerTool(*, name='sql_db_query_checker', description='\nΒ Β Β  Use this tool to double check if your query is correct before executing it.\nΒ Β Β  Always use this tool before executing a query with query_sql_db!\nΒ Β Β  ', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, db, template='\n{query}\nDouble check the {dialect} query above for common mistakes, including:\n- Using NOT IN with NULL values\n- Using UNION when UNION ALL should have been used\n- Using BETWEEN for exclusive ranges\n- Data type mismatch in predicates\n- Properly quoting identifiers\n- Using the correct number of arguments for functions\n- Casting to the correct data type\n- Using the proper columns for joins\n\nIf there are any of the above mistakes, rewrite the query. If there are no mistakes, just reproduce the original query.', llm, llm_chain)[source] Bases: langchain.tools.sql_database.tool.BaseSQLDatabaseTool, langchain.tools.base.BaseTool Use an LLM to check if a query is correct. Adapted from https://www.patterns.app/blog/2023/01/18/crunchbot-sql-analyst-gpt/ Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – db (langchain.sql_database.SQLDatabase) – template (str) – llm (langchain.base_language.BaseLanguageModel) – llm_chain (langchain.chains.llm.LLMChain) – Return type None attribute llm: langchain.base_language.BaseLanguageModel [Required] attribute llm_chain: langchain.chains.llm.LLMChain [Required] attribute template: str = '\n{query}\nDouble check the {dialect} query above for common mistakes, including:\n- Using NOT IN with NULL values\n- Using UNION when UNION ALL should have been used\n- Using BETWEEN for exclusive ranges\n- Data type mismatch in predicates\n- Properly quoting identifiers\n- Using the correct number of arguments for functions\n- Casting to the correct data type\n- Using the proper columns for joins\n\nIf there are any of the above mistakes, rewrite the query. If there are no mistakes, just reproduce the original query.' class langchain.tools.QuerySQLDataBaseTool(*, name='sql_db_query', description='\nΒ Β Β  Input to this tool is a detailed and correct SQL query, output is a result from the database.\nΒ Β Β  If the query is not correct, an error message will be returned.\nΒ Β Β  If an error is returned, rewrite the query, check the query, and try again.\nΒ Β Β  ', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, db)[source] Bases: langchain.tools.sql_database.tool.BaseSQLDatabaseTool, langchain.tools.base.BaseTool Tool for querying a SQL database. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – db (langchain.sql_database.SQLDatabase) – Return type None class langchain.tools.QuerySparkSQLTool(*, name='query_sql_db', description='\nΒ Β Β  Input to this tool is a detailed and correct SQL query, output is a result from the Spark SQL.\nΒ Β Β  If the query is not correct, an error message will be returned.\nΒ Β Β  If an error is returned, rewrite the query, check the query, and try again.\nΒ Β Β  ', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, db)[source] Bases: langchain.tools.spark_sql.tool.BaseSparkSQLTool, langchain.tools.base.BaseTool Tool for querying a Spark SQL. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – db (langchain.utilities.spark_sql.SparkSQL) – Return type None class langchain.tools.ReadFileTool(*, name='read_file', description='Read file from disk', args_schema=<class 'langchain.tools.file_management.read.ReadFileInput'>, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, root_dir=None)[source] Bases: langchain.tools.file_management.utils.BaseFileToolMixin, langchain.tools.base.BaseTool Parameters name (str) – description (str) – args_schema (Type[pydantic.main.BaseModel]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – root_dir (Optional[str]) – Return type None attribute args_schema: Type[pydantic.main.BaseModel] = <class 'langchain.tools.file_management.read.ReadFileInput'> Pydantic model class to validate and parse the tool’s input arguments. attribute description: str = 'Read file from disk' Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute name: str = 'read_file' The unique name of the tool that clearly communicates its purpose. class langchain.tools.RequestsDeleteTool(*, name='requests_delete', description='A portal to the internet. Use this when you need to make a DELETE request to a URL. Input should be a specific url, and the output will be the text response of the DELETE request.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, requests_wrapper)[source] Bases: langchain.tools.requests.tool.BaseRequestsTool, langchain.tools.base.BaseTool Tool for making a DELETE request to an API endpoint. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – requests_wrapper (langchain.requests.TextRequestsWrapper) – Return type None class langchain.tools.RequestsGetTool(*, name='requests_get', description='A portal to the internet. Use this when you need to get specific content from a website. Input should be aΒ  url (i.e. https://www.google.com). The output will be the text response of the GET request.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, requests_wrapper)[source] Bases: langchain.tools.requests.tool.BaseRequestsTool, langchain.tools.base.BaseTool Tool for making a GET request to an API endpoint. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – requests_wrapper (langchain.requests.TextRequestsWrapper) – Return type None class langchain.tools.RequestsPatchTool(*, name='requests_patch', description='Use this when you want to PATCH to a website.\nΒ Β Β  Input should be a json string with two keys: "url" and "data".\nΒ Β Β  The value of "url" should be a string, and the value of "data" should be a dictionary of \nΒ Β Β  key-value pairs you want to PATCH to the url.\nΒ Β Β  Be careful to always use double quotes for strings in the json string\nΒ Β Β  The output will be the text response of the PATCH request.\nΒ Β Β  ', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, requests_wrapper)[source] Bases: langchain.tools.requests.tool.BaseRequestsTool, langchain.tools.base.BaseTool Tool for making a PATCH request to an API endpoint. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – requests_wrapper (langchain.requests.TextRequestsWrapper) – Return type None class langchain.tools.RequestsPostTool(*, name='requests_post', description='Use this when you want to POST to a website.\nΒ Β Β  Input should be a json string with two keys: "url" and "data".\nΒ Β Β  The value of "url" should be a string, and the value of "data" should be a dictionary of \nΒ Β Β  key-value pairs you want to POST to the url.\nΒ Β Β  Be careful to always use double quotes for strings in the json string\nΒ Β Β  The output will be the text response of the POST request.\nΒ Β Β  ', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, requests_wrapper)[source] Bases: langchain.tools.requests.tool.BaseRequestsTool, langchain.tools.base.BaseTool Tool for making a POST request to an API endpoint. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – requests_wrapper (langchain.requests.TextRequestsWrapper) – Return type None class langchain.tools.RequestsPutTool(*, name='requests_put', description='Use this when you want to PUT to a website.\nΒ Β Β  Input should be a json string with two keys: "url" and "data".\nΒ Β Β  The value of "url" should be a string, and the value of "data" should be a dictionary of \nΒ Β Β  key-value pairs you want to PUT to the url.\nΒ Β Β  Be careful to always use double quotes for strings in the json string.\nΒ Β Β  The output will be the text response of the PUT request.\nΒ Β Β  ', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, requests_wrapper)[source] Bases: langchain.tools.requests.tool.BaseRequestsTool, langchain.tools.base.BaseTool Tool for making a PUT request to an API endpoint. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – requests_wrapper (langchain.requests.TextRequestsWrapper) – Return type None class langchain.tools.SceneXplainTool(*, name='image_explainer', description='An Image Captioning Tool: Use this tool to generate a detailed caption for an image. The input can be an image file of any format, and the output will be a text description that covers every detail of the image.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, api_wrapper=None)[source] Bases: langchain.tools.base.BaseTool Tool that adds the capability to explain images. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – api_wrapper (langchain.utilities.scenexplain.SceneXplainAPIWrapper) – Return type None attribute api_wrapper: langchain.utilities.scenexplain.SceneXplainAPIWrapper [Optional] class langchain.tools.SearxSearchResults(*, name='Searx Search Results', description='A meta search engine.Useful for when you need to answer questions about current events.Input should be a search query. Output is a JSON array of the query results', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, wrapper, num_results=4, kwargs=None, **extra_data)[source] Bases: langchain.tools.base.BaseTool Tool that has the capability to query a Searx instance and get back json. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – wrapper (langchain.utilities.searx_search.SearxSearchWrapper) – num_results (int) – kwargs (dict) – extra_data (Any) – Return type None attribute kwargs: dict [Optional] attribute num_results: int = 4 attribute wrapper: langchain.utilities.searx_search.SearxSearchWrapper [Required] class langchain.tools.SearxSearchRun(*, name='searx_search', description='A meta search engine.Useful for when you need to answer questions about current events.Input should be a search query.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, wrapper, kwargs=None)[source] Bases: langchain.tools.base.BaseTool Tool that adds the capability to query a Searx instance. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – wrapper (langchain.utilities.searx_search.SearxSearchWrapper) – kwargs (dict) – Return type None attribute kwargs: dict [Optional] attribute wrapper: langchain.utilities.searx_search.SearxSearchWrapper [Required] class langchain.tools.ShellTool(*, name='terminal', description='Run shell commands on this Linux machine.', args_schema=<class 'langchain.tools.shell.tool.ShellInput'>, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, process=None)[source] Bases: langchain.tools.base.BaseTool Tool to run shell commands. Parameters name (str) – description (str) – args_schema (Type[pydantic.main.BaseModel]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – process (langchain.utilities.bash.BashProcess) – Return type None attribute args_schema: Type[pydantic.main.BaseModel] = <class 'langchain.tools.shell.tool.ShellInput'> Schema for input arguments. attribute description: str = 'Run shell commands on this Linux machine.' Description of tool. attribute name: str = 'terminal' Name of tool. attribute process: langchain.utilities.bash.BashProcess [Optional] Bash process to run commands. class langchain.tools.SleepTool(*, name='sleep', description='Make agent sleep for a specified number of seconds.', args_schema=<class 'langchain.tools.sleep.tool.SleepInput'>, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False)[source] Bases: langchain.tools.base.BaseTool Tool that adds the capability to sleep. Parameters name (str) – description (str) – args_schema (Type[pydantic.main.BaseModel]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – Return type None attribute args_schema: Type[pydantic.main.BaseModel] = <class 'langchain.tools.sleep.tool.SleepInput'> Pydantic model class to validate and parse the tool’s input arguments. langchain.tools.StdInInquireTool(*args, **kwargs)[source] Tool for asking the user for input. Parameters args (Any) – kwargs (Any) – Return type langchain.tools.human.tool.HumanInputRun class langchain.tools.SteamshipImageGenerationTool(*, name='GenerateImage', description='Useful for when you need to generate an image.Input: A detailed text-2-image prompt describing an imageOutput: the UUID of a generated image', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, model_name, size='512x512', steamship, return_urls=False)[source] Bases: langchain.tools.base.BaseTool Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – model_name (langchain.tools.steamship_image_generation.tool.ModelName) – size (Optional[str]) – steamship (Steamship) – return_urls (Optional[bool]) – Return type None attribute model_name: ModelName [Required] attribute return_urls: Optional[bool] = False attribute size: Optional[str] = '512x512' attribute steamship: Steamship [Required] class langchain.tools.StructuredTool(*, name, description='', args_schema, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, func, coroutine=None)[source] Bases: langchain.tools.base.BaseTool Tool that can operate on any number of inputs. Parameters name (str) – description (str) – args_schema (Type[pydantic.main.BaseModel]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – func (Callable[[...], Any]) – coroutine (Optional[Callable[[...], Awaitable[Any]]]) – Return type None attribute args_schema: Type[pydantic.main.BaseModel] [Required] The input arguments’ schema. The tool schema. attribute coroutine: Optional[Callable[[...], Awaitable[Any]]] = None The asynchronous version of the function. attribute description: str = '' Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute func: Callable[[...], Any] [Required] The function to run when the tool is called. classmethod from_function(func, name=None, description=None, return_direct=False, args_schema=None, infer_schema=True, **kwargs)[source] Create tool from a given function. A classmethod that helps to create a tool from a function. Parameters func (Callable) – The function from which to create a tool name (Optional[str]) – The name of the tool. Defaults to the function name description (Optional[str]) – The description of the tool. Defaults to the function docstring return_direct (bool) – Whether to return the result directly or as a callback args_schema (Optional[Type[pydantic.main.BaseModel]]) – The schema of the tool’s input arguments infer_schema (bool) – Whether to infer the schema from the function’s signature **kwargs – Additional arguments to pass to the tool kwargs (Any) – Returns The tool Return type langchain.tools.base.StructuredTool Examples … code-block:: python def add(a: int, b: int) -> int:β€œβ€β€Add two numbers””” return a + b tool = StructuredTool.from_function(add) tool.run(1, 2) # 3 property args: dict The tool’s input arguments. class langchain.tools.Tool(name, func, description, *, args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, coroutine=None)[source] Bases: langchain.tools.base.BaseTool Tool that takes in function or coroutine directly. Parameters name (str) – func (Callable[[...], str]) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – coroutine (Optional[Callable[[...], Awaitable[str]]]) – Return type None attribute args_schema: Optional[Type[pydantic.main.BaseModel]] = None Pydantic model class to validate and parse the tool’s input arguments. attribute callback_manager: Optional[langchain.callbacks.base.BaseCallbackManager] = None Deprecated. Please use callbacks instead. attribute callbacks: Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]] = None Callbacks to be called during tool execution. attribute coroutine: Optional[Callable[[...], Awaitable[str]]] = None The asynchronous version of the function. attribute description: str = '' Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute func: Callable[[...], str] [Required] The function to run when the tool is called. attribute handle_tool_error: Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]] = False Handle the content of the ToolException thrown. attribute name: str [Required] The unique name of the tool that clearly communicates its purpose. attribute return_direct: bool = False Whether to return the tool’s output directly. Setting this to True means that after the tool is called, the AgentExecutor will stop looping. attribute verbose: bool = False Whether to log the tool’s progress. classmethod from_function(func, name, description, return_direct=False, args_schema=None, **kwargs)[source] Initialize tool from a function. Parameters func (Callable) – name (str) – description (str) – return_direct (bool) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – kwargs (Any) – Return type langchain.tools.base.Tool property args: dict The tool’s input arguments. class langchain.tools.VectorStoreQATool(*, name, description, args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, vectorstore, llm=None)[source] Bases: langchain.tools.vectorstore.tool.BaseVectorStoreTool, langchain.tools.base.BaseTool Tool for the VectorDBQA chain. To be initialized with name and chain. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – vectorstore (langchain.vectorstores.base.VectorStore) – llm (langchain.base_language.BaseLanguageModel) – Return type None static get_description(name, description)[source] Parameters name (str) – description (str) – Return type str class langchain.tools.VectorStoreQAWithSourcesTool(*, name, description, args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, vectorstore, llm=None)[source] Bases: langchain.tools.vectorstore.tool.BaseVectorStoreTool, langchain.tools.base.BaseTool Tool for the VectorDBQAWithSources chain. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – vectorstore (langchain.vectorstores.base.VectorStore) – llm (langchain.base_language.BaseLanguageModel) – Return type None static get_description(name, description)[source] Parameters name (str) – description (str) – Return type str class langchain.tools.WikipediaQueryRun(*, name='Wikipedia', description='A wrapper around Wikipedia. Useful for when you need to answer general questions about people, places, companies, facts, historical events, or other subjects. Input should be a search query.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, api_wrapper)[source] Bases: langchain.tools.base.BaseTool Tool that adds the capability to search using the Wikipedia API. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – api_wrapper (langchain.utilities.wikipedia.WikipediaAPIWrapper) – Return type None attribute api_wrapper: langchain.utilities.wikipedia.WikipediaAPIWrapper [Required] class langchain.tools.WolframAlphaQueryRun(*, name='wolfram_alpha', description='A wrapper around Wolfram Alpha. Useful for when you need to answer questions about Math, Science, Technology, Culture, Society and Everyday Life. Input should be a search query.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, api_wrapper)[source] Bases: langchain.tools.base.BaseTool Tool that adds the capability to query using the Wolfram Alpha SDK. Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – api_wrapper (langchain.utilities.wolfram_alpha.WolframAlphaAPIWrapper) – Return type None attribute api_wrapper: langchain.utilities.wolfram_alpha.WolframAlphaAPIWrapper [Required] class langchain.tools.WriteFileTool(*, name='write_file', description='Write file to disk', args_schema=<class 'langchain.tools.file_management.write.WriteFileInput'>, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, root_dir=None)[source] Bases: langchain.tools.file_management.utils.BaseFileToolMixin, langchain.tools.base.BaseTool Parameters name (str) – description (str) – args_schema (Type[pydantic.main.BaseModel]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – root_dir (Optional[str]) – Return type None attribute args_schema: Type[pydantic.main.BaseModel] = <class 'langchain.tools.file_management.write.WriteFileInput'> Pydantic model class to validate and parse the tool’s input arguments. attribute description: str = 'Write file to disk' Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute name: str = 'write_file' The unique name of the tool that clearly communicates its purpose. class langchain.tools.YouTubeSearchTool(*, name='youtube_search', description='search for youtube videos associated with a person. the input to this tool should be a comma separated list, the first part contains a person name and the second a number that is the maximum number of video results to return aka num_results. the second part is optional', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False)[source] Bases: langchain.tools.base.BaseTool Parameters name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – Return type None class langchain.tools.ZapierNLAListActions(*, name='ZapierNLA_list_actions', description='A wrapper around Zapier NLA actions. The input to this tool is a natural language instruction, for example "get the latest email from my bank" or "send a slack message to the #general channel". Each tool will have params associated with it that are specified as a list. You MUST take into account the params when creating the instruction. For example, if the params are [\'Message_Text\', \'Channel\'], your instruction should be something like \'send a slack message to the #general channel with the text hello world\'. Another example: if the params are [\'Calendar\', \'Search_Term\'], your instruction should be something like \'find the meeting in my personal calendar at 3pm\'. Do not make up params, they will be explicitly specified in the tool description. If you do not have enough information to fill in the params, just say \'not enough information provided in the instruction, missing <param>\'. If you get a none or null response, STOP EXECUTION, do not try to another tool!This tool specifically used for: {zapier_description}, and has params: {params}This tool returns a list of the user\'s exposed actions.', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, api_wrapper=None)[source] Bases: langchain.tools.base.BaseTool Returns a list of all exposed (enabled) actions associated withcurrent user (associated with the set api_key). Change your exposed actions here: https://nla.zapier.com/demo/start/ The return list can be empty if no actions exposed. Else will contain a list of action objects: [{β€œid”: str, β€œdescription”: str, β€œparams”: Dict[str, str] }] params will always contain an instructions key, the only required param. All others optional and if provided will override any AI guesses (see β€œunderstanding the AI guessing flow” here: https://nla.zapier.com/api/v1/docs) Parameters None – name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – api_wrapper (langchain.utilities.zapier.ZapierNLAWrapper) – Return type None attribute api_wrapper: langchain.utilities.zapier.ZapierNLAWrapper [Optional] class langchain.tools.ZapierNLARunAction(*, name='', description='', args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, api_wrapper=None, action_id, params=None, base_prompt='A wrapper around Zapier NLA actions. The input to this tool is a natural language instruction, for example "get the latest email from my bank" or "send a slack message to the #general channel". Each tool will have params associated with it that are specified as a list. You MUST take into account the params when creating the instruction. For example, if the params are [\'Message_Text\', \'Channel\'], your instruction should be something like \'send a slack message to the #general channel with the text hello world\'. Another example: if the params are [\'Calendar\', \'Search_Term\'], your instruction should be something like \'find the meeting in my personal calendar at 3pm\'. Do not make up params, they will be explicitly specified in the tool description. If you do not have enough information to fill in the params, just say \'not enough information provided in the instruction, missing <param>\'. If you get a none or null response, STOP EXECUTION, do not try to another tool!This tool specifically used for: {zapier_description}, and has params: {params}', zapier_description, params_schema=None)[source] Bases: langchain.tools.base.BaseTool Executes an action that is identified by action_id, must be exposed(enabled) by the current user (associated with the set api_key). Change your exposed actions here: https://nla.zapier.com/demo/start/ The return JSON is guaranteed to be less than ~500 words (350 tokens) making it safe to inject into the prompt of another LLM call. Parameters action_id (str) – a specific action ID (from list actions) of the action to execute (the set api_key must be associated with the action owner) instructions – a natural language instruction string for using the action (eg. β€œget the latest email from Mike Knoop” for β€œGmail: find email” action) params (Optional[dict]) – a dict, optional. Any params provided will override AI guesses from instructions (see β€œunderstanding the AI guessing flow” here: https://nla.zapier.com/api/v1/docs) name (str) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – api_wrapper (langchain.utilities.zapier.ZapierNLAWrapper) – base_prompt (str) – zapier_description (str) – params_schema (Dict[str, str]) – Return type None attribute action_id: str [Required] attribute api_wrapper: langchain.utilities.zapier.ZapierNLAWrapper [Optional] attribute base_prompt: str = 'A wrapper around Zapier NLA actions. The input to this tool is a natural language instruction, for example "get the latest email from my bank" or "send a slack message to the #general channel". Each tool will have params associated with it that are specified as a list. You MUST take into account the params when creating the instruction. For example, if the params are [\'Message_Text\', \'Channel\'], your instruction should be something like \'send a slack message to the #general channel with the text hello world\'. Another example: if the params are [\'Calendar\', \'Search_Term\'], your instruction should be something like \'find the meeting in my personal calendar at 3pm\'. Do not make up params, they will be explicitly specified in the tool description. If you do not have enough information to fill in the params, just say \'not enough information provided in the instruction, missing <param>\'. If you get a none or null response, STOP EXECUTION, do not try to another tool!This tool specifically used for: {zapier_description}, and has params: {params}' attribute params: Optional[dict] = None attribute params_schema: Dict[str, str] [Optional] attribute zapier_description: str [Required] langchain.tools.format_tool_to_openai_function(tool)[source] Format tool into the OpenAI function API. Parameters tool (langchain.tools.base.BaseTool) – Return type langchain.tools.convert_to_openai.FunctionDescription langchain.tools.tool(*args, return_direct=False, args_schema=None, infer_schema=True)[source] Make tools out of functions, can be used with or without arguments. Parameters *args – The arguments to the tool. return_direct (bool) – Whether to return directly from the tool rather than continuing the agent loop. args_schema (Optional[Type[pydantic.main.BaseModel]]) – optional argument schema for user to specify infer_schema (bool) – Whether to infer the schema of the arguments from the function’s signature. This also makes the resultant tool accept a dictionary input to its run() function. args (Union[str, Callable]) – Return type Callable Requires: Function must be of type (str) -> str Function must have a docstring Examples @tool def search_api(query: str) -> str: # Searches the API for the query. return @tool("search", return_direct=True) def search_api(query: str) -> str: # Searches the API for the query. return
https://api.python.langchain.com/en/latest/modules/tools.html
18521a36-3fee-40f9-86fe-e77cfb18d8be
Callbacks Callback handlers that allow listening to events in LangChain. class langchain.callbacks.AimCallbackHandler(repo=None, experiment_name=None, system_tracking_interval=10, log_system_params=True)[source] Bases: langchain.callbacks.aim_callback.BaseMetadataCallbackHandler, langchain.callbacks.base.BaseCallbackHandler Callback Handler that logs to Aim. Parameters repo (str, optional) – Aim repository path or Repo object to which Run object is bound. If skipped, default Repo is used. experiment_name (str, optional) – Sets Run’s experiment property. β€˜default’ if not specified. Can be used later to query runs/sequences. system_tracking_interval (int, optional) – Sets the tracking interval in seconds for system usage metrics (CPU, Memory, etc.). Set to None to disable system metrics tracking. log_system_params (bool, optional) – Enable/Disable logging of system params such as installed packages, git info, environment variables, etc. Return type None This handler will utilize the associated callback method called and formats the input of each callback function with metadata regarding the state of LLM run and then logs the response to Aim. setup(**kwargs)[source] Parameters kwargs (Any) – Return type None on_llm_start(serialized, prompts, **kwargs)[source] Run when LLM starts. Parameters serialized (Dict[str, Any]) – prompts (List[str]) – kwargs (Any) – Return type None on_llm_end(response, **kwargs)[source] Run when LLM ends running. Parameters response (langchain.schema.LLMResult) – kwargs (Any) – Return type None on_llm_new_token(token, **kwargs)[source] Run when LLM generates a new token. Parameters token (str) – kwargs (Any) – Return type None on_llm_error(error, **kwargs)[source] Run when LLM errors. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_chain_start(serialized, inputs, **kwargs)[source] Run when chain starts running. Parameters serialized (Dict[str, Any]) – inputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_end(outputs, **kwargs)[source] Run when chain ends running. Parameters outputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_error(error, **kwargs)[source] Run when chain errors. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_tool_start(serialized, input_str, **kwargs)[source] Run when tool starts running. Parameters serialized (Dict[str, Any]) – input_str (str) – kwargs (Any) – Return type None on_tool_end(output, **kwargs)[source] Run when tool ends running. Parameters output (str) – kwargs (Any) – Return type None on_tool_error(error, **kwargs)[source] Run when tool errors. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_text(text, **kwargs)[source] Run when agent is ending. Parameters text (str) – kwargs (Any) – Return type None on_agent_finish(finish, **kwargs)[source] Run when agent ends running. Parameters finish (langchain.schema.AgentFinish) – kwargs (Any) – Return type None on_agent_action(action, **kwargs)[source] Run on agent action. Parameters action (langchain.schema.AgentAction) – kwargs (Any) – Return type Any flush_tracker(repo=None, experiment_name=None, system_tracking_interval=10, log_system_params=True, langchain_asset=None, reset=True, finish=False)[source] Flush the tracker and reset the session. Parameters repo (str, optional) – Aim repository path or Repo object to which Run object is bound. If skipped, default Repo is used. experiment_name (str, optional) – Sets Run’s experiment property. β€˜default’ if not specified. Can be used later to query runs/sequences. system_tracking_interval (int, optional) – Sets the tracking interval in seconds for system usage metrics (CPU, Memory, etc.). Set to None to disable system metrics tracking. log_system_params (bool, optional) – Enable/Disable logging of system params such as installed packages, git info, environment variables, etc. langchain_asset (Any) – The langchain asset to save. reset (bool) – Whether to reset the session. finish (bool) – Whether to finish the run. Returns – None Return type None class langchain.callbacks.ArgillaCallbackHandler(dataset_name, workspace_name=None, api_url=None, api_key=None)[source] Bases: langchain.callbacks.base.BaseCallbackHandler Callback Handler that logs into Argilla. Parameters dataset_name (str) – name of the FeedbackDataset in Argilla. Note that it must exist in advance. If you need help on how to create a FeedbackDataset in Argilla, please visit https://docs.argilla.io/en/latest/guides/llms/practical_guides/use_argilla_callback_in_langchain.html. workspace_name (Optional[str]) – name of the workspace in Argilla where the specified FeedbackDataset lives in. Defaults to None, which means that the default workspace will be used. api_url (Optional[str]) – URL of the Argilla Server that we want to use, and where the FeedbackDataset lives in. Defaults to None, which means that either ARGILLA_API_URL environment variable or the default http://localhost:6900 will be used. api_key (Optional[str]) – API Key to connect to the Argilla Server. Defaults to None, which means that either ARGILLA_API_KEY environment variable or the default argilla.apikey will be used. Raises ImportError – if the argilla package is not installed. ConnectionError – if the connection to Argilla fails. FileNotFoundError – if the FeedbackDataset retrieval from Argilla fails. Return type None Examples >>> from langchain.llms import OpenAI >>> from langchain.callbacks import ArgillaCallbackHandler >>> argilla_callback = ArgillaCallbackHandler( ... dataset_name="my-dataset", ... workspace_name="my-workspace", ... api_url="http://localhost:6900", ... api_key="argilla.apikey", ... ) >>> llm = OpenAI( ... temperature=0, ... callbacks=[argilla_callback], ... verbose=True, ... openai_api_key="API_KEY_HERE", ... ) >>> llm.generate([ ... "What is the best NLP-annotation tool out there? (no bias at all)", ... ]) "Argilla, no doubt about it." on_llm_start(serialized, prompts, **kwargs)[source] Save the prompts in memory when an LLM starts. Parameters serialized (Dict[str, Any]) – prompts (List[str]) – kwargs (Any) – Return type None on_llm_new_token(token, **kwargs)[source] Do nothing when a new token is generated. Parameters token (str) – kwargs (Any) – Return type None on_llm_end(response, **kwargs)[source] Log records to Argilla when an LLM ends. Parameters response (langchain.schema.LLMResult) – kwargs (Any) – Return type None on_llm_error(error, **kwargs)[source] Do nothing when LLM outputs an error. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_chain_start(serialized, inputs, **kwargs)[source] If the key input is in inputs, then save it in self.prompts using either the parent_run_id or the run_id as the key. This is done so that we don’t log the same input prompt twice, once when the LLM starts and once when the chain starts. Parameters serialized (Dict[str, Any]) – inputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_end(outputs, **kwargs)[source] If either the parent_run_id or the run_id is in self.prompts, then log the outputs to Argilla, and pop the run from self.prompts. The behavior differs if the output is a list or not. Parameters outputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_error(error, **kwargs)[source] Do nothing when LLM chain outputs an error. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_tool_start(serialized, input_str, **kwargs)[source] Do nothing when tool starts. Parameters serialized (Dict[str, Any]) – input_str (str) – kwargs (Any) – Return type None on_agent_action(action, **kwargs)[source] Do nothing when agent takes a specific action. Parameters action (langchain.schema.AgentAction) – kwargs (Any) – Return type Any on_tool_end(output, observation_prefix=None, llm_prefix=None, **kwargs)[source] Do nothing when tool ends. Parameters output (str) – observation_prefix (Optional[str]) – llm_prefix (Optional[str]) – kwargs (Any) – Return type None on_tool_error(error, **kwargs)[source] Do nothing when tool outputs an error. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_text(text, **kwargs)[source] Do nothing Parameters text (str) – kwargs (Any) – Return type None on_agent_finish(finish, **kwargs)[source] Do nothing Parameters finish (langchain.schema.AgentFinish) – kwargs (Any) – Return type None class langchain.callbacks.ArizeCallbackHandler(model_id=None, model_version=None, SPACE_KEY=None, API_KEY=None)[source] Bases: langchain.callbacks.base.BaseCallbackHandler Callback Handler that logs to Arize. Parameters model_id (Optional[str]) – model_version (Optional[str]) – SPACE_KEY (Optional[str]) – API_KEY (Optional[str]) – Return type None on_llm_start(serialized, prompts, **kwargs)[source] Run when LLM starts running. Parameters serialized (Dict[str, Any]) – prompts (List[str]) – kwargs (Any) – Return type None on_llm_new_token(token, **kwargs)[source] Do nothing. Parameters token (str) – kwargs (Any) – Return type None on_llm_end(response, **kwargs)[source] Run when LLM ends running. Parameters response (langchain.schema.LLMResult) – kwargs (Any) – Return type None on_llm_error(error, **kwargs)[source] Do nothing. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_chain_start(serialized, inputs, **kwargs)[source] Run when chain starts running. Parameters serialized (Dict[str, Any]) – inputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_end(outputs, **kwargs)[source] Do nothing. Parameters outputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_error(error, **kwargs)[source] Do nothing. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_tool_start(serialized, input_str, **kwargs)[source] Run when tool starts running. Parameters serialized (Dict[str, Any]) – input_str (str) – kwargs (Any) – Return type None on_agent_action(action, **kwargs)[source] Do nothing. Parameters action (langchain.schema.AgentAction) – kwargs (Any) – Return type Any on_tool_end(output, observation_prefix=None, llm_prefix=None, **kwargs)[source] Run when tool ends running. Parameters output (str) – observation_prefix (Optional[str]) – llm_prefix (Optional[str]) – kwargs (Any) – Return type None on_tool_error(error, **kwargs)[source] Run when tool errors. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_text(text, **kwargs)[source] Run on arbitrary text. Parameters text (str) – kwargs (Any) – Return type None on_agent_finish(finish, **kwargs)[source] Run on agent end. Parameters finish (langchain.schema.AgentFinish) – kwargs (Any) – Return type None class langchain.callbacks.AsyncIteratorCallbackHandler[source] Bases: langchain.callbacks.base.AsyncCallbackHandler Callback handler that returns an async iterator. Return type None property always_verbose: bool queue: asyncio.queues.Queue[str] done: asyncio.locks.Event async on_llm_start(serialized, prompts, **kwargs)[source] Run when LLM starts running. Parameters serialized (Dict[str, Any]) – prompts (List[str]) – kwargs (Any) – Return type None async on_llm_new_token(token, **kwargs)[source] Run on new LLM token. Only available when streaming is enabled. Parameters token (str) – kwargs (Any) – Return type None async on_llm_end(response, **kwargs)[source] Run when LLM ends running. Parameters response (langchain.schema.LLMResult) – kwargs (Any) – Return type None async on_llm_error(error, **kwargs)[source] Run when LLM errors. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None async aiter()[source] Return type AsyncIterator[str] class langchain.callbacks.ClearMLCallbackHandler(task_type='inference', project_name='langchain_callback_demo', tags=None, task_name=None, visualize=False, complexity_metrics=False, stream_logs=False)[source] Bases: langchain.callbacks.utils.BaseMetadataCallbackHandler, langchain.callbacks.base.BaseCallbackHandler Callback Handler that logs to ClearML. Parameters job_type (str) – The type of clearml task such as β€œinference”, β€œtesting” or β€œqc” project_name (str) – The clearml project name tags (list) – Tags to add to the task task_name (str) – Name of the clearml task visualize (bool) – Whether to visualize the run. complexity_metrics (bool) – Whether to log complexity metrics stream_logs (bool) – Whether to stream callback actions to ClearML task_type (Optional[str]) – Return type None This handler will utilize the associated callback method and formats the input of each callback function with metadata regarding the state of LLM run, and adds the response to the list of records for both the {method}_records and action. It then logs the response to the ClearML console. on_llm_start(serialized, prompts, **kwargs)[source] Run when LLM starts. Parameters serialized (Dict[str, Any]) – prompts (List[str]) – kwargs (Any) – Return type None on_llm_new_token(token, **kwargs)[source] Run when LLM generates a new token. Parameters token (str) – kwargs (Any) – Return type None on_llm_end(response, **kwargs)[source] Run when LLM ends running. Parameters response (langchain.schema.LLMResult) – kwargs (Any) – Return type None on_llm_error(error, **kwargs)[source] Run when LLM errors. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_chain_start(serialized, inputs, **kwargs)[source] Run when chain starts running. Parameters serialized (Dict[str, Any]) – inputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_end(outputs, **kwargs)[source] Run when chain ends running. Parameters outputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_error(error, **kwargs)[source] Run when chain errors. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_tool_start(serialized, input_str, **kwargs)[source] Run when tool starts running. Parameters serialized (Dict[str, Any]) – input_str (str) – kwargs (Any) – Return type None on_tool_end(output, **kwargs)[source] Run when tool ends running. Parameters output (str) – kwargs (Any) – Return type None on_tool_error(error, **kwargs)[source] Run when tool errors. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_text(text, **kwargs)[source] Run when agent is ending. Parameters text (str) – kwargs (Any) – Return type None on_agent_finish(finish, **kwargs)[source] Run when agent ends running. Parameters finish (langchain.schema.AgentFinish) – kwargs (Any) – Return type None on_agent_action(action, **kwargs)[source] Run on agent action. Parameters action (langchain.schema.AgentAction) – kwargs (Any) – Return type Any analyze_text(text)[source] Analyze text using textstat and spacy. Parameters text (str) – The text to analyze. Returns A dictionary containing the complexity metrics. Return type (dict) flush_tracker(name=None, langchain_asset=None, finish=False)[source] Flush the tracker and setup the session. Everything after this will be a new table. Parameters name (Optional[str]) – Name of the preformed session so far so it is identifyable langchain_asset (Any) – The langchain asset to save. finish (bool) – Whether to finish the run. Returns – None Return type None class langchain.callbacks.CometCallbackHandler(task_type='inference', workspace=None, project_name=None, tags=None, name=None, visualizations=None, complexity_metrics=False, custom_metrics=None, stream_logs=True)[source] Bases: langchain.callbacks.utils.BaseMetadataCallbackHandler, langchain.callbacks.base.BaseCallbackHandler Callback Handler that logs to Comet. Parameters job_type (str) – The type of comet_ml task such as β€œinference”, β€œtesting” or β€œqc” project_name (str) – The comet_ml project name tags (list) – Tags to add to the task task_name (str) – Name of the comet_ml task visualize (bool) – Whether to visualize the run. complexity_metrics (bool) – Whether to log complexity metrics stream_logs (bool) – Whether to stream callback actions to Comet task_type (Optional[str]) – workspace (Optional[str]) – name (Optional[str]) – visualizations (Optional[List[str]]) – custom_metrics (Optional[Callable]) – Return type None This handler will utilize the associated callback method and formats the input of each callback function with metadata regarding the state of LLM run, and adds the response to the list of records for both the {method}_records and action. It then logs the response to Comet. on_llm_start(serialized, prompts, **kwargs)[source] Run when LLM starts. Parameters serialized (Dict[str, Any]) – prompts (List[str]) – kwargs (Any) – Return type None on_llm_new_token(token, **kwargs)[source] Run when LLM generates a new token. Parameters token (str) – kwargs (Any) – Return type None on_llm_end(response, **kwargs)[source] Run when LLM ends running. Parameters response (langchain.schema.LLMResult) – kwargs (Any) – Return type None on_llm_error(error, **kwargs)[source] Run when LLM errors. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_chain_start(serialized, inputs, **kwargs)[source] Run when chain starts running. Parameters serialized (Dict[str, Any]) – inputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_end(outputs, **kwargs)[source] Run when chain ends running. Parameters outputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_error(error, **kwargs)[source] Run when chain errors. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_tool_start(serialized, input_str, **kwargs)[source] Run when tool starts running. Parameters serialized (Dict[str, Any]) – input_str (str) – kwargs (Any) – Return type None on_tool_end(output, **kwargs)[source] Run when tool ends running. Parameters output (str) – kwargs (Any) – Return type None on_tool_error(error, **kwargs)[source] Run when tool errors. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_text(text, **kwargs)[source] Run when agent is ending. Parameters text (str) – kwargs (Any) – Return type None on_agent_finish(finish, **kwargs)[source] Run when agent ends running. Parameters finish (langchain.schema.AgentFinish) – kwargs (Any) – Return type None on_agent_action(action, **kwargs)[source] Run on agent action. Parameters action (langchain.schema.AgentAction) – kwargs (Any) – Return type Any flush_tracker(langchain_asset=None, task_type='inference', workspace=None, project_name='comet-langchain-demo', tags=None, name=None, visualizations=None, complexity_metrics=False, custom_metrics=None, finish=False, reset=False)[source] Flush the tracker and setup the session. Everything after this will be a new table. Parameters name (Optional[str]) – Name of the preformed session so far so it is identifyable langchain_asset (Any) – The langchain asset to save. finish (bool) – Whether to finish the run. Returns – None task_type (Optional[str]) – workspace (Optional[str]) – project_name (Optional[str]) – tags (Optional[Sequence]) – visualizations (Optional[List[str]]) – complexity_metrics (bool) – custom_metrics (Optional[Callable]) – reset (bool) – Return type None class langchain.callbacks.FileCallbackHandler(filename, mode='a', color=None)[source] Bases: langchain.callbacks.base.BaseCallbackHandler Callback Handler that writes to a file. Parameters filename (str) – mode (str) – color (Optional[str]) – Return type None on_chain_start(serialized, inputs, **kwargs)[source] Print out that we are entering a chain. Parameters serialized (Dict[str, Any]) – inputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_end(outputs, **kwargs)[source] Print out that we finished a chain. Parameters outputs (Dict[str, Any]) – kwargs (Any) – Return type None on_agent_action(action, color=None, **kwargs)[source] Run on agent action. Parameters action (langchain.schema.AgentAction) – color (Optional[str]) – kwargs (Any) – Return type Any on_tool_end(output, color=None, observation_prefix=None, llm_prefix=None, **kwargs)[source] If not the final action, print out observation. Parameters output (str) – color (Optional[str]) – observation_prefix (Optional[str]) – llm_prefix (Optional[str]) – kwargs (Any) – Return type None on_text(text, color=None, end='', **kwargs)[source] Run when agent ends. Parameters text (str) – color (Optional[str]) – end (str) – kwargs (Any) – Return type None on_agent_finish(finish, color=None, **kwargs)[source] Run on agent end. Parameters finish (langchain.schema.AgentFinish) – color (Optional[str]) – kwargs (Any) – Return type None class langchain.callbacks.FinalStreamingStdOutCallbackHandler(*, answer_prefix_tokens=None, strip_tokens=True, stream_prefix=False)[source] Bases: langchain.callbacks.streaming_stdout.StreamingStdOutCallbackHandler Callback handler for streaming in agents. Only works with agents using LLMs that support streaming. Only the final output of the agent will be streamed. Parameters answer_prefix_tokens (Optional[List[str]]) – strip_tokens (bool) – stream_prefix (bool) – Return type None append_to_last_tokens(token)[source] Parameters token (str) – Return type None check_if_answer_reached()[source] Return type bool on_llm_start(serialized, prompts, **kwargs)[source] Run when LLM starts running. Parameters serialized (Dict[str, Any]) – prompts (List[str]) – kwargs (Any) – Return type None on_llm_new_token(token, **kwargs)[source] Run on new LLM token. Only available when streaming is enabled. Parameters token (str) – kwargs (Any) – Return type None class langchain.callbacks.HumanApprovalCallbackHandler(approve=<function _default_approve>, should_check=<function _default_true>)[source] Bases: langchain.callbacks.base.BaseCallbackHandler Callback for manually validating values. Parameters approve (Callable[[Any], bool]) – should_check (Callable[[Dict[str, Any]], bool]) – raise_error: bool = True on_tool_start(serialized, input_str, *, run_id, parent_run_id=None, **kwargs)[source] Run when tool starts running. Parameters serialized (Dict[str, Any]) – input_str (str) – run_id (uuid.UUID) – parent_run_id (Optional[uuid.UUID]) – kwargs (Any) – Return type Any class langchain.callbacks.InfinoCallbackHandler(model_id=None, model_version=None, verbose=False)[source] Bases: langchain.callbacks.base.BaseCallbackHandler Callback Handler that logs to Infino. Parameters model_id (Optional[str]) – model_version (Optional[str]) – verbose (bool) – Return type None on_llm_start(serialized, prompts, **kwargs)[source] Log the prompts to Infino, and set start time and error flag. Parameters serialized (Dict[str, Any]) – prompts (List[str]) – kwargs (Any) – Return type None on_llm_new_token(token, **kwargs)[source] Do nothing when a new token is generated. Parameters token (str) – kwargs (Any) – Return type None on_llm_end(response, **kwargs)[source] Log the latency, error, token usage, and response to Infino. Parameters response (langchain.schema.LLMResult) – kwargs (Any) – Return type None on_llm_error(error, **kwargs)[source] Set the error flag. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_chain_start(serialized, inputs, **kwargs)[source] Do nothing when LLM chain starts. Parameters serialized (Dict[str, Any]) – inputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_end(outputs, **kwargs)[source] Do nothing when LLM chain ends. Parameters outputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_error(error, **kwargs)[source] Need to log the error. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_tool_start(serialized, input_str, **kwargs)[source] Do nothing when tool starts. Parameters serialized (Dict[str, Any]) – input_str (str) – kwargs (Any) – Return type None on_agent_action(action, **kwargs)[source] Do nothing when agent takes a specific action. Parameters action (langchain.schema.AgentAction) – kwargs (Any) – Return type Any on_tool_end(output, observation_prefix=None, llm_prefix=None, **kwargs)[source] Do nothing when tool ends. Parameters output (str) – observation_prefix (Optional[str]) – llm_prefix (Optional[str]) – kwargs (Any) – Return type None on_tool_error(error, **kwargs)[source] Do nothing when tool outputs an error. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_text(text, **kwargs)[source] Do nothing. Parameters text (str) – kwargs (Any) – Return type None on_agent_finish(finish, **kwargs)[source] Do nothing. Parameters finish (langchain.schema.AgentFinish) – kwargs (Any) – Return type None class langchain.callbacks.MlflowCallbackHandler(name='langchainrun-%', experiment='langchain', tags={}, tracking_uri=None)[source] Bases: langchain.callbacks.utils.BaseMetadataCallbackHandler, langchain.callbacks.base.BaseCallbackHandler Callback Handler that logs metrics and artifacts to mlflow server. Parameters name (str) – Name of the run. experiment (str) – Name of the experiment. tags (dict) – Tags to be attached for the run. tracking_uri (str) – MLflow tracking server uri. Return type None This handler will utilize the associated callback method called and formats the input of each callback function with metadata regarding the state of LLM run, and adds the response to the list of records for both the {method}_records and action. It then logs the response to mlflow server. on_llm_start(serialized, prompts, **kwargs)[source] Run when LLM starts. Parameters serialized (Dict[str, Any]) – prompts (List[str]) – kwargs (Any) – Return type None on_llm_new_token(token, **kwargs)[source] Run when LLM generates a new token. Parameters token (str) – kwargs (Any) – Return type None on_llm_end(response, **kwargs)[source] Run when LLM ends running. Parameters response (langchain.schema.LLMResult) – kwargs (Any) – Return type None on_llm_error(error, **kwargs)[source] Run when LLM errors. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_chain_start(serialized, inputs, **kwargs)[source] Run when chain starts running. Parameters serialized (Dict[str, Any]) – inputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_end(outputs, **kwargs)[source] Run when chain ends running. Parameters outputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_error(error, **kwargs)[source] Run when chain errors. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_tool_start(serialized, input_str, **kwargs)[source] Run when tool starts running. Parameters serialized (Dict[str, Any]) – input_str (str) – kwargs (Any) – Return type None on_tool_end(output, **kwargs)[source] Run when tool ends running. Parameters output (str) – kwargs (Any) – Return type None on_tool_error(error, **kwargs)[source] Run when tool errors. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_text(text, **kwargs)[source] Run when agent is ending. Parameters text (str) – kwargs (Any) – Return type None on_agent_finish(finish, **kwargs)[source] Run when agent ends running. Parameters finish (langchain.schema.AgentFinish) – kwargs (Any) – Return type None on_agent_action(action, **kwargs)[source] Run on agent action. Parameters action (langchain.schema.AgentAction) – kwargs (Any) – Return type Any flush_tracker(langchain_asset=None, finish=False)[source] Parameters langchain_asset (Any) – finish (bool) – Return type None class langchain.callbacks.OpenAICallbackHandler[source] Bases: langchain.callbacks.base.BaseCallbackHandler Callback Handler that tracks OpenAI info. total_tokens: int = 0 prompt_tokens: int = 0 completion_tokens: int = 0 successful_requests: int = 0 total_cost: float = 0.0 property always_verbose: bool Whether to call verbose callbacks even if verbose is False. on_llm_start(serialized, prompts, **kwargs)[source] Print out the prompts. Parameters serialized (Dict[str, Any]) – prompts (List[str]) – kwargs (Any) – Return type None on_llm_new_token(token, **kwargs)[source] Print out the token. Parameters token (str) – kwargs (Any) – Return type None on_llm_end(response, **kwargs)[source] Collect token usage. Parameters response (langchain.schema.LLMResult) – kwargs (Any) – Return type None class langchain.callbacks.StdOutCallbackHandler(color=None)[source] Bases: langchain.callbacks.base.BaseCallbackHandler Callback Handler that prints to std out. Parameters color (Optional[str]) – Return type None on_llm_start(serialized, prompts, **kwargs)[source] Print out the prompts. Parameters serialized (Dict[str, Any]) – prompts (List[str]) – kwargs (Any) – Return type None on_llm_end(response, **kwargs)[source] Do nothing. Parameters response (langchain.schema.LLMResult) – kwargs (Any) – Return type None on_llm_new_token(token, **kwargs)[source] Do nothing. Parameters token (str) – kwargs (Any) – Return type None on_llm_error(error, **kwargs)[source] Do nothing. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_chain_start(serialized, inputs, **kwargs)[source] Print out that we are entering a chain. Parameters serialized (Dict[str, Any]) – inputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_end(outputs, **kwargs)[source] Print out that we finished a chain. Parameters outputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_error(error, **kwargs)[source] Do nothing. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_tool_start(serialized, input_str, **kwargs)[source] Do nothing. Parameters serialized (Dict[str, Any]) – input_str (str) – kwargs (Any) – Return type None on_agent_action(action, color=None, **kwargs)[source] Run on agent action. Parameters action (langchain.schema.AgentAction) – color (Optional[str]) – kwargs (Any) – Return type Any on_tool_end(output, color=None, observation_prefix=None, llm_prefix=None, **kwargs)[source] If not the final action, print out observation. Parameters output (str) – color (Optional[str]) – observation_prefix (Optional[str]) – llm_prefix (Optional[str]) – kwargs (Any) – Return type None on_tool_error(error, **kwargs)[source] Do nothing. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_text(text, color=None, end='', **kwargs)[source] Run when agent ends. Parameters text (str) – color (Optional[str]) – end (str) – kwargs (Any) – Return type None on_agent_finish(finish, color=None, **kwargs)[source] Run on agent end. Parameters finish (langchain.schema.AgentFinish) – color (Optional[str]) – kwargs (Any) – Return type None class langchain.callbacks.StreamingStdOutCallbackHandler[source] Bases: langchain.callbacks.base.BaseCallbackHandler Callback handler for streaming. Only works with LLMs that support streaming. on_llm_start(serialized, prompts, **kwargs)[source] Run when LLM starts running. Parameters serialized (Dict[str, Any]) – prompts (List[str]) – kwargs (Any) – Return type None on_llm_new_token(token, **kwargs)[source] Run on new LLM token. Only available when streaming is enabled. Parameters token (str) – kwargs (Any) – Return type None on_llm_end(response, **kwargs)[source] Run when LLM ends running. Parameters response (langchain.schema.LLMResult) – kwargs (Any) – Return type None on_llm_error(error, **kwargs)[source] Run when LLM errors. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_chain_start(serialized, inputs, **kwargs)[source] Run when chain starts running. Parameters serialized (Dict[str, Any]) – inputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_end(outputs, **kwargs)[source] Run when chain ends running. Parameters outputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_error(error, **kwargs)[source] Run when chain errors. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_tool_start(serialized, input_str, **kwargs)[source] Run when tool starts running. Parameters serialized (Dict[str, Any]) – input_str (str) – kwargs (Any) – Return type None on_agent_action(action, **kwargs)[source] Run on agent action. Parameters action (langchain.schema.AgentAction) – kwargs (Any) – Return type Any on_tool_end(output, **kwargs)[source] Run when tool ends running. Parameters output (str) – kwargs (Any) – Return type None on_tool_error(error, **kwargs)[source] Run when tool errors. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_text(text, **kwargs)[source] Run on arbitrary text. Parameters text (str) – kwargs (Any) – Return type None on_agent_finish(finish, **kwargs)[source] Run on agent end. Parameters finish (langchain.schema.AgentFinish) – kwargs (Any) – Return type None langchain.callbacks.StreamlitCallbackHandler(parent_container, *, max_thought_containers=4, expand_new_thoughts=True, collapse_completed_thoughts=True, thought_labeler=None)[source] Construct a new StreamlitCallbackHandler. This CallbackHandler is geared towards use with a LangChain Agent; it displays the Agent’s LLM and tool-usage β€œthoughts” inside a series of Streamlit expanders. Parameters parent_container (DeltaGenerator) – The st.container that will contain all the Streamlit elements that the Handler creates. max_thought_containers (int) – The max number of completed LLM thought containers to show at once. When this threshold is reached, a new thought will cause the oldest thoughts to be collapsed into a β€œHistory” expander. Defaults to 4. expand_new_thoughts (bool) – Each LLM β€œthought” gets its own st.expander. This param controls whether that expander is expanded by default. Defaults to True. collapse_completed_thoughts (bool) – If True, LLM thought expanders will be collapsed when completed. Defaults to True. thought_labeler (Optional[LLMThoughtLabeler]) – An optional custom LLMThoughtLabeler instance. If unspecified, the handler will use the default thought labeling logic. Defaults to None. Returns A new StreamlitCallbackHandler instance. Note that this is an β€œauto-updating” API (if the installed version of Streamlit) has a more recent StreamlitCallbackHandler implementation, an instance of that class will be used. Return type BaseCallbackHandler class langchain.callbacks.LLMThoughtLabeler[source] Bases: object Generates markdown labels for LLMThought containers. Pass a custom subclass of this to StreamlitCallbackHandler to override its default labeling logic. get_initial_label()[source] Return the markdown label for a new LLMThought that doesn’t have an associated tool yet. Return type str get_tool_label(tool, is_complete)[source] Return the label for an LLMThought that has an associated tool. Parameters tool (langchain.callbacks.streamlit.streamlit_callback_handler.ToolRecord) – The tool’s ToolRecord is_complete (bool) – True if the thought is complete; False if the thought is still receiving input. Return type The markdown label for the thought’s container. get_history_label()[source] Return a markdown label for the special β€˜history’ container that contains overflow thoughts. Return type str get_final_agent_thought_label()[source] Return the markdown label for the agent’s final thought - the β€œNow I have the answer” thought, that doesn’t involve a tool. Return type str class langchain.callbacks.WandbCallbackHandler(job_type=None, project='langchain_callback_demo', entity=None, tags=None, group=None, name=None, notes=None, visualize=False, complexity_metrics=False, stream_logs=False)[source] Bases: langchain.callbacks.utils.BaseMetadataCallbackHandler, langchain.callbacks.base.BaseCallbackHandler Callback Handler that logs to Weights and Biases. Parameters job_type (str) – The type of job. project (str) – The project to log to. entity (str) – The entity to log to. tags (list) – The tags to log. group (str) – The group to log to. name (str) – The name of the run. notes (str) – The notes to log. visualize (bool) – Whether to visualize the run. complexity_metrics (bool) – Whether to log complexity metrics. stream_logs (bool) – Whether to stream callback actions to W&B Return type None This handler will utilize the associated callback method called and formats the input of each callback function with metadata regarding the state of LLM run, and adds the response to the list of records for both the {method}_records and action. It then logs the response using the run.log() method to Weights and Biases. on_llm_start(serialized, prompts, **kwargs)[source] Run when LLM starts. Parameters serialized (Dict[str, Any]) – prompts (List[str]) – kwargs (Any) – Return type None on_llm_new_token(token, **kwargs)[source] Run when LLM generates a new token. Parameters token (str) – kwargs (Any) – Return type None on_llm_end(response, **kwargs)[source] Run when LLM ends running. Parameters response (langchain.schema.LLMResult) – kwargs (Any) – Return type None on_llm_error(error, **kwargs)[source] Run when LLM errors. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_chain_start(serialized, inputs, **kwargs)[source] Run when chain starts running. Parameters serialized (Dict[str, Any]) – inputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_end(outputs, **kwargs)[source] Run when chain ends running. Parameters outputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_error(error, **kwargs)[source] Run when chain errors. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_tool_start(serialized, input_str, **kwargs)[source] Run when tool starts running. Parameters serialized (Dict[str, Any]) – input_str (str) – kwargs (Any) – Return type None on_tool_end(output, **kwargs)[source] Run when tool ends running. Parameters output (str) – kwargs (Any) – Return type None on_tool_error(error, **kwargs)[source] Run when tool errors. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_text(text, **kwargs)[source] Run when agent is ending. Parameters text (str) – kwargs (Any) – Return type None on_agent_finish(finish, **kwargs)[source] Run when agent ends running. Parameters finish (langchain.schema.AgentFinish) – kwargs (Any) – Return type None on_agent_action(action, **kwargs)[source] Run on agent action. Parameters action (langchain.schema.AgentAction) – kwargs (Any) – Return type Any flush_tracker(langchain_asset=None, reset=True, finish=False, job_type=None, project=None, entity=None, tags=None, group=None, name=None, notes=None, visualize=None, complexity_metrics=None)[source] Flush the tracker and reset the session. Parameters langchain_asset (Any) – The langchain asset to save. reset (bool) – Whether to reset the session. finish (bool) – Whether to finish the run. job_type (Optional[str]) – The job type. project (Optional[str]) – The project. entity (Optional[str]) – The entity. tags (Optional[Sequence]) – The tags. group (Optional[str]) – The group. name (Optional[str]) – The name. notes (Optional[str]) – The notes. visualize (Optional[bool]) – Whether to visualize. complexity_metrics (Optional[bool]) – Whether to compute complexity metrics. Returns – None Return type None class langchain.callbacks.WhyLabsCallbackHandler(logger)[source] Bases: langchain.callbacks.base.BaseCallbackHandler WhyLabs CallbackHandler. Parameters logger (Logger) – on_llm_start(serialized, prompts, **kwargs)[source] Pass the input prompts to the logger Parameters serialized (Dict[str, Any]) – prompts (List[str]) – kwargs (Any) – Return type None on_llm_end(response, **kwargs)[source] Pass the generated response to the logger. Parameters response (langchain.schema.LLMResult) – kwargs (Any) – Return type None on_llm_new_token(token, **kwargs)[source] Do nothing. Parameters token (str) – kwargs (Any) – Return type None on_llm_error(error, **kwargs)[source] Do nothing. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_chain_start(serialized, inputs, **kwargs)[source] Do nothing. Parameters serialized (Dict[str, Any]) – inputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_end(outputs, **kwargs)[source] Do nothing. Parameters outputs (Dict[str, Any]) – kwargs (Any) – Return type None on_chain_error(error, **kwargs)[source] Do nothing. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_tool_start(serialized, input_str, **kwargs)[source] Do nothing. Parameters serialized (Dict[str, Any]) – input_str (str) – kwargs (Any) – Return type None on_agent_action(action, color=None, **kwargs)[source] Do nothing. Parameters action (langchain.schema.AgentAction) – color (Optional[str]) – kwargs (Any) – Return type Any on_tool_end(output, color=None, observation_prefix=None, llm_prefix=None, **kwargs)[source] Do nothing. Parameters output (str) – color (Optional[str]) – observation_prefix (Optional[str]) – llm_prefix (Optional[str]) – kwargs (Any) – Return type None on_tool_error(error, **kwargs)[source] Do nothing. Parameters error (Union[Exception, KeyboardInterrupt]) – kwargs (Any) – Return type None on_text(text, **kwargs)[source] Do nothing. Parameters text (str) – kwargs (Any) – Return type None on_agent_finish(finish, color=None, **kwargs)[source] Run on agent end. Parameters finish (langchain.schema.AgentFinish) – color (Optional[str]) – kwargs (Any) – Return type None flush()[source] Return type None close()[source] Return type None classmethod from_params(*, api_key=None, org_id=None, dataset_id=None, sentiment=False, toxicity=False, themes=False)[source] Instantiate whylogs Logger from params. Parameters api_key (Optional[str]) – WhyLabs API key. Optional because the preferred way to specify the API key is with environment variable WHYLABS_API_KEY. org_id (Optional[str]) – WhyLabs organization id to write profiles to. If not set must be specified in environment variable WHYLABS_DEFAULT_ORG_ID. dataset_id (Optional[str]) – The model or dataset this callback is gathering telemetry for. If not set must be specified in environment variable WHYLABS_DEFAULT_DATASET_ID. sentiment (bool) – If True will initialize a model to perform sentiment analysis compound score. Defaults to False and will not gather this metric. toxicity (bool) – If True will initialize a model to score toxicity. Defaults to False and will not gather this metric. themes (bool) – If True will initialize a model to calculate distance to configured themes. Defaults to None and will not gather this metric. Return type Logger langchain.callbacks.get_openai_callback()[source] Get the OpenAI callback handler in a context manager. which conveniently exposes token and cost information. Returns The OpenAI callback handler. Return type OpenAICallbackHandler Example >>> with get_openai_callback() as cb: ... # Use the OpenAI callback handler langchain.callbacks.tracing_enabled(session_name='default')[source] Get the Deprecated LangChainTracer in a context manager. Parameters session_name (str, optional) – The name of the session. Defaults to β€œdefault”. Returns The LangChainTracer session. Return type TracerSessionV1 Example >>> with tracing_enabled() as session: ... # Use the LangChainTracer session langchain.callbacks.wandb_tracing_enabled(session_name='default')[source] Get the WandbTracer in a context manager. Parameters session_name (str, optional) – The name of the session. Defaults to β€œdefault”. Returns None Return type Generator[None, None, None] Example >>> with wandb_tracing_enabled() as session: ... # Use the WandbTracer session
https://api.python.langchain.com/en/latest/modules/callbacks.html
fb470470-dfc6-4e38-8a18-8c5a864d5999
Document Loaders All different types of document loaders. class langchain.document_loaders.AcreomLoader(path, encoding='UTF-8', collect_metadata=True)[source] Bases: langchain.document_loaders.base.BaseLoader Parameters path (str) – encoding (str) – collect_metadata (bool) – FRONT_MATTER_REGEX = re.compile('^---\\n(.*?)\\n---\\n', re.MULTILINE|re.DOTALL) lazy_load()[source] A lazy loader for document content. Return type Iterator[langchain.schema.Document] load()[source] Load data into document objects. Return type List[langchain.schema.Document] class langchain.document_loaders.AZLyricsLoader(web_path, header_template=None, verify=True)[source] Bases: langchain.document_loaders.web_base.WebBaseLoader Loader that loads AZLyrics webpages. Parameters web_path (Union[str, List[str]]) – header_template (Optional[dict]) – verify (Optional[bool]) – load()[source] Load webpage. Return type List[langchain.schema.Document] class langchain.document_loaders.AirbyteJSONLoader(file_path)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that loads local airbyte json files. Parameters file_path (str) – load()[source] Load file. Return type List[langchain.schema.Document] class langchain.document_loaders.AirtableLoader(api_token, table_id, base_id)[source] Bases: langchain.document_loaders.base.BaseLoader Loader for Airtable tables. Parameters api_token (str) – table_id (str) – base_id (str) – lazy_load()[source] Lazy load records from table. Return type Iterator[langchain.schema.Document] load()[source] Load Table. Return type List[langchain.schema.Document] class langchain.document_loaders.ApifyDatasetLoader(dataset_id, dataset_mapping_function)[source] Bases: langchain.document_loaders.base.BaseLoader, pydantic.main.BaseModel Logic for loading documents from Apify datasets. Parameters dataset_id (str) – dataset_mapping_function (Callable[[Dict], langchain.schema.Document]) – Return type None attribute apify_client: Any = None attribute dataset_id: str [Required] The ID of the dataset on the Apify platform. attribute dataset_mapping_function: Callable[[Dict], langchain.schema.Document] [Required] A custom function that takes a single dictionary (an Apify dataset item) and converts it to an instance of the Document class. load()[source] Load documents. Return type List[langchain.schema.Document] class langchain.document_loaders.ArxivLoader(query, load_max_docs=100, load_all_available_meta=False)[source] Bases: langchain.document_loaders.base.BaseLoader Loads a query result from arxiv.org into a list of Documents. Each document represents one Document. The loader converts the original PDF format into the text. Parameters query (str) – load_max_docs (Optional[int]) – load_all_available_meta (Optional[bool]) – load()[source] Load data into document objects. Return type List[langchain.schema.Document] class langchain.document_loaders.AzureBlobStorageContainerLoader(conn_str, container, prefix='')[source] Bases: langchain.document_loaders.base.BaseLoader Loading logic for loading documents from Azure Blob Storage. Parameters conn_str (str) – container (str) – prefix (str) – load()[source] Load documents. Return type List[langchain.schema.Document] class langchain.document_loaders.AzureBlobStorageFileLoader(conn_str, container, blob_name)[source] Bases: langchain.document_loaders.base.BaseLoader Loading logic for loading documents from Azure Blob Storage. Parameters conn_str (str) – container (str) – blob_name (str) – load()[source] Load documents. Return type List[langchain.schema.Document] class langchain.document_loaders.BSHTMLLoader(file_path, open_encoding=None, bs_kwargs=None, get_text_separator='')[source] Bases: langchain.document_loaders.base.BaseLoader Loader that uses beautiful soup to parse HTML files. Parameters file_path (str) – open_encoding (Optional[str]) – bs_kwargs (Optional[dict]) – get_text_separator (str) – Return type None load()[source] Load data into document objects. Return type List[langchain.schema.Document] class langchain.document_loaders.BibtexLoader(file_path, *, parser=None, max_docs=None, max_content_chars=4000, load_extra_metadata=False, file_pattern='[^:]+\\.pdf')[source] Bases: langchain.document_loaders.base.BaseLoader Loads a bibtex file into a list of Documents. Each document represents one entry from the bibtex file. If a PDF file is present in the file bibtex field, the original PDF is loaded into the document text. If no such file entry is present, the abstract field is used instead. Parameters file_path (str) – parser (Optional[langchain.utilities.bibtex.BibtexparserWrapper]) – max_docs (Optional[int]) – max_content_chars (Optional[int]) – load_extra_metadata (bool) – file_pattern (str) – lazy_load()[source] Load bibtex file using bibtexparser and get the article texts plus the article metadata. See https://bibtexparser.readthedocs.io/en/master/ Returns a list of documents with the document.page_content in text format Return type Iterator[langchain.schema.Document] load()[source] Load bibtex file documents from the given bibtex file path. See https://bibtexparser.readthedocs.io/en/master/ Parameters file_path – the path to the bibtex file Returns a list of documents with the document.page_content in text format Return type List[langchain.schema.Document] class langchain.document_loaders.BigQueryLoader(query, project=None, page_content_columns=None, metadata_columns=None, credentials=None)[source] Bases: langchain.document_loaders.base.BaseLoader Loads a query result from BigQuery into a list of documents. Each document represents one row of the result. The page_content_columns are written into the page_content of the document. The metadata_columns are written into the metadata of the document. By default, all columns are written into the page_content and none into the metadata. Parameters query (str) – project (Optional[str]) – page_content_columns (Optional[List[str]]) – metadata_columns (Optional[List[str]]) – credentials (Optional[Credentials]) – load()[source] Load data into document objects. Return type List[langchain.schema.Document] class langchain.document_loaders.BiliBiliLoader(video_urls)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that loads bilibili transcripts. Parameters video_urls (List[str]) – load()[source] Load from bilibili url. Return type List[langchain.schema.Document] class langchain.document_loaders.BlackboardLoader(blackboard_course_url, bbrouter, load_all_recursively=True, basic_auth=None, cookies=None)[source] Bases: langchain.document_loaders.web_base.WebBaseLoader Loader that loads all documents from a Blackboard course. This loader is not compatible with all Blackboard courses. It is only compatible with courses that use the new Blackboard interface. To use this loader, you must have the BbRouter cookie. You can get this cookie by logging into the course and then copying the value of the BbRouter cookie from the browser’s developer tools. Example from langchain.document_loaders import BlackboardLoader loader = BlackboardLoader( blackboard_course_url="https://blackboard.example.com/webapps/blackboard/execute/announcement?method=search&context=course_entry&course_id=_123456_1", bbrouter="expires:12345...", ) documents = loader.load() Parameters blackboard_course_url (str) – bbrouter (str) – load_all_recursively (bool) – basic_auth (Optional[Tuple[str, str]]) – cookies (Optional[dict]) – folder_path: str base_url: str load_all_recursively: bool check_bs4()[source] Check if BeautifulSoup4 is installed. Raises ImportError – If BeautifulSoup4 is not installed. Return type None load()[source] Load data into document objects. Returns List of documents. Return type List[langchain.schema.Document] download(path)[source] Download a file from a url. Parameters path (str) – Path to the file. Return type None parse_filename(url)[source] Parse the filename from a url. Parameters url (str) – Url to parse the filename from. Returns The filename. Return type str class langchain.document_loaders.Blob(*, data=None, mimetype=None, encoding='utf-8', path=None)[source] Bases: pydantic.main.BaseModel A blob is used to represent raw data by either reference or value. Provides an interface to materialize the blob in different representations, and help to decouple the development of data loaders from the downstream parsing of the raw data. Inspired by: https://developer.mozilla.org/en-US/docs/Web/API/Blob Parameters data (Optional[Union[bytes, str]]) – mimetype (Optional[str]) – encoding (str) – path (Optional[Union[str, pathlib.PurePath]]) – Return type None attribute data: Optional[Union[bytes, str]] = None attribute encoding: str = 'utf-8' attribute mimetype: Optional[str] = None attribute path: Optional[Union[str, pathlib.PurePath]] = None as_bytes()[source] Read data as bytes. Return type bytes as_bytes_io()[source] Read data as a byte stream. Return type Generator[Union[_io.BytesIO, _io.BufferedReader], None, None] as_string()[source] Read data as a string. Return type str classmethod from_data(data, *, encoding='utf-8', mime_type=None, path=None)[source] Initialize the blob from in-memory data. Parameters data (Union[str, bytes]) – the in-memory data associated with the blob encoding (str) – Encoding to use if decoding the bytes into a string mime_type (Optional[str]) – if provided, will be set as the mime-type of the data path (Optional[str]) – if provided, will be set as the source from which the data came Returns Blob instance Return type langchain.document_loaders.blob_loaders.schema.Blob classmethod from_path(path, *, encoding='utf-8', mime_type=None, guess_type=True)[source] Load the blob from a path like object. Parameters path (Union[str, pathlib.PurePath]) – path like object to file to be read encoding (str) – Encoding to use if decoding the bytes into a string mime_type (Optional[str]) – if provided, will be set as the mime-type of the data guess_type (bool) – If True, the mimetype will be guessed from the file extension, if a mime-type was not provided Returns Blob instance Return type langchain.document_loaders.blob_loaders.schema.Blob property source: Optional[str] The source location of the blob as string if known otherwise none. class langchain.document_loaders.BlobLoader[source] Bases: abc.ABC Abstract interface for blob loaders implementation. Implementer should be able to load raw content from a storage system according to some criteria and return the raw content lazily as a stream of blobs. abstract yield_blobs()[source] A lazy loader for raw data represented by LangChain’s Blob object. Returns A generator over blobs Return type Iterable[langchain.document_loaders.blob_loaders.schema.Blob] class langchain.document_loaders.BlockchainDocumentLoader(contract_address, blockchainType=BlockchainType.ETH_MAINNET, api_key='docs-demo', startToken='', get_all_tokens=False, max_execution_time=None)[source] Bases: langchain.document_loaders.base.BaseLoader Loads elements from a blockchain smart contract into Langchain documents. The supported blockchains are: Ethereum mainnet, Ethereum Goerli testnet, Polygon mainnet, and Polygon Mumbai testnet. If no BlockchainType is specified, the default is Ethereum mainnet. The Loader uses the Alchemy API to interact with the blockchain. ALCHEMY_API_KEY environment variable must be set to use this loader. The API returns 100 NFTs per request and can be paginated using the startToken parameter. If get_all_tokens is set to True, the loader will get all tokens on the contract. Note that for contracts with a large number of tokens, this may take a long time (e.g. 10k tokens is 100 requests). Default value is false for this reason. The max_execution_time (sec) can be set to limit the execution time of the loader. Future versions of this loader can: Support additional Alchemy APIs (e.g. getTransactions, etc.) Support additional blockain APIs (e.g. Infura, Opensea, etc.) Parameters contract_address (str) – blockchainType (langchain.document_loaders.blockchain.BlockchainType) – api_key (str) – startToken (str) – get_all_tokens (bool) – max_execution_time (Optional[int]) – load()[source] Load data into document objects. Return type List[langchain.schema.Document] class langchain.document_loaders.CSVLoader(file_path, source_column=None, csv_args=None, encoding=None)[source] Bases: langchain.document_loaders.base.BaseLoader Loads a CSV file into a list of documents. Each document represents one row of the CSV file. Every row is converted into a key/value pair and outputted to a new line in the document’s page_content. The source for each document loaded from csv is set to the value of the file_path argument for all doucments by default. You can override this by setting the source_column argument to the name of a column in the CSV file. The source of each document will then be set to the value of the column with the name specified in source_column. Output Example:column1: value1 column2: value2 column3: value3 Parameters file_path (str) – source_column (Optional[str]) – csv_args (Optional[Dict]) – encoding (Optional[str]) – load()[source] Load data into document objects. Return type List[langchain.schema.Document] class langchain.document_loaders.ChatGPTLoader(log_file, num_logs=- 1)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that loads conversations from exported ChatGPT data. Parameters log_file (str) – num_logs (int) – load()[source] Load data into document objects. Return type List[langchain.schema.Document] class langchain.document_loaders.CoNLLULoader(file_path)[source] Bases: langchain.document_loaders.base.BaseLoader Load CoNLL-U files. Parameters file_path (str) – load()[source] Load from file path. Return type List[langchain.schema.Document] class langchain.document_loaders.CollegeConfidentialLoader(web_path, header_template=None, verify=True)[source] Bases: langchain.document_loaders.web_base.WebBaseLoader Loader that loads College Confidential webpages. Parameters web_path (Union[str, List[str]]) – header_template (Optional[dict]) – verify (Optional[bool]) – load()[source] Load webpage. Return type List[langchain.schema.Document] class langchain.document_loaders.ConfluenceLoader(url, api_key=None, username=None, oauth2=None, token=None, cloud=True, number_of_retries=3, min_retry_seconds=2, max_retry_seconds=10, confluence_kwargs=None)[source] Bases: langchain.document_loaders.base.BaseLoader Load Confluence pages. Port of https://llamahub.ai/l/confluence This currently supports username/api_key, Oauth2 login or personal access token authentication. Specify a list page_ids and/or space_key to load in the corresponding pages into Document objects, if both are specified the union of both sets will be returned. You can also specify a boolean include_attachments to include attachments, this is set to False by default, if set to True all attachments will be downloaded and ConfluenceReader will extract the text from the attachments and add it to the Document object. Currently supported attachment types are: PDF, PNG, JPEG/JPG, SVG, Word and Excel. Confluence API supports difference format of page content. The storage format is the raw XML representation for storage. The view format is the HTML representation for viewing with macros are rendered as though it is viewed by users. You can pass a enum content_format argument to load() to specify the content format, this is set to ContentFormat.STORAGE by default. Hint: space_key and page_id can both be found in the URL of a page in Confluence - https://yoursite.atlassian.com/wiki/spaces/<space_key>/pages/<page_id> Example from langchain.document_loaders import ConfluenceLoader loader = ConfluenceLoader( url="https://yoursite.atlassian.com/wiki", username="me", api_key="12345" ) documents = loader.load(space_key="SPACE",limit=50) Parameters url (str) – _description_ api_key (str, optional) – _description_, defaults to None username (str, optional) – _description_, defaults to None oauth2 (dict, optional) – _description_, defaults to {} token (str, optional) – _description_, defaults to None cloud (bool, optional) – _description_, defaults to True number_of_retries (Optional[int], optional) – How many times to retry, defaults to 3 min_retry_seconds (Optional[int], optional) – defaults to 2 max_retry_seconds (Optional[int], optional) – defaults to 10 confluence_kwargs (dict, optional) – additional kwargs to initialize confluence with Raises ValueError – Errors while validating input ImportError – Required dependencies not installed. static validate_init_args(url=None, api_key=None, username=None, oauth2=None, token=None)[source] Validates proper combinations of init arguments Parameters url (Optional[str]) – api_key (Optional[str]) – username (Optional[str]) – oauth2 (Optional[dict]) – token (Optional[str]) – Return type Optional[List] load(space_key=None, page_ids=None, label=None, cql=None, include_restricted_content=False, include_archived_content=False, include_attachments=False, include_comments=False, content_format=ContentFormat.STORAGE, limit=50, max_pages=1000, ocr_languages=None)[source] Parameters space_key (Optional[str], optional) – Space key retrieved from a confluence URL, defaults to None page_ids (Optional[List[str]], optional) – List of specific page IDs to load, defaults to None label (Optional[str], optional) – Get all pages with this label, defaults to None cql (Optional[str], optional) – CQL Expression, defaults to None include_restricted_content (bool, optional) – defaults to False include_archived_content (bool, optional) – Whether to include archived content, defaults to False include_attachments (bool, optional) – defaults to False include_comments (bool, optional) – defaults to False content_format (ContentFormat) – Specify content format, defaults to ContentFormat.STORAGE limit (int, optional) – Maximum number of pages to retrieve per request, defaults to 50 max_pages (int, optional) – Maximum number of pages to retrieve in total, defaults 1000 ocr_languages (str, optional) – The languages to use for the Tesseract agent. To use a language, you’ll first need to install the appropriate Tesseract language pack. Raises ValueError – _description_ ImportError – _description_ Returns _description_ Return type List[Document] paginate_request(retrieval_method, **kwargs)[source] Paginate the various methods to retrieve groups of pages. Unfortunately, due to page size, sometimes the Confluence API doesn’t match the limit value. If limit is >100 confluence seems to cap the response to 100. Also, due to the Atlassian Python package, we don’t get the β€œnext” values from the β€œ_links” key because they only return the value from the results key. So here, the pagination starts from 0 and goes until the max_pages, getting the limit number of pages with each request. We have to manually check if there are more docs based on the length of the returned list of pages, rather than just checking for the presence of a next key in the response like this page would have you do: https://developer.atlassian.com/server/confluence/pagination-in-the-rest-api/ Parameters retrieval_method (callable) – Function used to retrieve docs kwargs (Any) – Returns List of documents Return type List is_public_page(page)[source] Check if a page is publicly accessible. Parameters page (dict) – Return type bool process_pages(pages, include_restricted_content, include_attachments, include_comments, content_format, ocr_languages=None)[source] Process a list of pages into a list of documents. Parameters pages (List[dict]) – include_restricted_content (bool) – include_attachments (bool) – include_comments (bool) – content_format (langchain.document_loaders.confluence.ContentFormat) – ocr_languages (Optional[str]) – Return type List[langchain.schema.Document] process_page(page, include_attachments, include_comments, content_format, ocr_languages=None)[source] Parameters page (dict) – include_attachments (bool) – include_comments (bool) – content_format (langchain.document_loaders.confluence.ContentFormat) – ocr_languages (Optional[str]) – Return type langchain.schema.Document process_attachment(page_id, ocr_languages=None)[source] Parameters page_id (str) – ocr_languages (Optional[str]) – Return type List[str] process_pdf(link, ocr_languages=None)[source] Parameters link (str) – ocr_languages (Optional[str]) – Return type str process_image(link, ocr_languages=None)[source] Parameters link (str) – ocr_languages (Optional[str]) – Return type str process_doc(link)[source] Parameters link (str) – Return type str process_xls(link)[source] Parameters link (str) – Return type str process_svg(link, ocr_languages=None)[source] Parameters link (str) – ocr_languages (Optional[str]) – Return type str class langchain.document_loaders.DataFrameLoader(data_frame, page_content_column='text')[source] Bases: langchain.document_loaders.base.BaseLoader Load Pandas DataFrames. Parameters data_frame (Any) – page_content_column (str) – lazy_load()[source] Lazy load records from dataframe. Return type Iterator[langchain.schema.Document] load()[source] Load full dataframe. Return type List[langchain.schema.Document] class langchain.document_loaders.DiffbotLoader(api_token, urls, continue_on_failure=True)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that loads Diffbot file json. Parameters api_token (str) – urls (List[str]) – continue_on_failure (bool) – load()[source] Extract text from Diffbot on all the URLs and return Document instances Return type List[langchain.schema.Document] class langchain.document_loaders.DirectoryLoader(path, glob='**/[!.]*', silent_errors=False, load_hidden=False, loader_cls=<class 'langchain.document_loaders.unstructured.UnstructuredFileLoader'>, loader_kwargs=None, recursive=False, show_progress=False, use_multithreading=False, max_concurrency=4)[source] Bases: langchain.document_loaders.base.BaseLoader Loading logic for loading documents from a directory. Parameters path (str) – glob (str) – silent_errors (bool) – load_hidden (bool) – loader_cls (Union[Type[langchain.document_loaders.unstructured.UnstructuredFileLoader], Type[langchain.document_loaders.text.TextLoader], Type[langchain.document_loaders.html_bs.BSHTMLLoader]]) – loader_kwargs (Optional[dict]) – recursive (bool) – show_progress (bool) – use_multithreading (bool) – max_concurrency (int) – load_file(item, path, docs, pbar)[source] Parameters item (pathlib.Path) – path (pathlib.Path) – docs (List[langchain.schema.Document]) – pbar (Optional[Any]) – Return type None load()[source] Load documents. Return type List[langchain.schema.Document] class langchain.document_loaders.DiscordChatLoader(chat_log, user_id_col='ID')[source] Bases: langchain.document_loaders.base.BaseLoader Load Discord chat logs. Parameters chat_log (pd.DataFrame) – user_id_col (str) – load()[source] Load all chat messages. Return type List[langchain.schema.Document] class langchain.document_loaders.DocugamiLoader(*, api='https://api.docugami.com/v1preview1', access_token=None, docset_id=None, document_ids=None, file_paths=None, min_chunk_size=32)[source] Bases: langchain.document_loaders.base.BaseLoader, pydantic.main.BaseModel Loader that loads processed docs from Docugami. To use, you should have the lxml python package installed. Parameters api (str) – access_token (Optional[str]) – docset_id (Optional[str]) – document_ids (Optional[Sequence[str]]) – file_paths (Optional[Sequence[Union[pathlib.Path, str]]]) – min_chunk_size (int) – Return type None attribute access_token: Optional[str] = None attribute api: str = 'https://api.docugami.com/v1preview1' attribute docset_id: Optional[str] = None attribute document_ids: Optional[Sequence[str]] = None attribute file_paths: Optional[Sequence[Union[pathlib.Path, str]]] = None attribute min_chunk_size: int = 32 load()[source] Load documents. Return type List[langchain.schema.Document] class langchain.document_loaders.Docx2txtLoader(file_path)[source] Bases: langchain.document_loaders.base.BaseLoader, abc.ABC Loads a DOCX with docx2txt and chunks at character level. Defaults to check for local file, but if the file is a web path, it will download it to a temporary file, and use that, then clean up the temporary file after completion Parameters file_path (str) – load()[source] Load given path as single page. Return type List[langchain.schema.Document] class langchain.document_loaders.DuckDBLoader(query, database=':memory:', read_only=False, config=None, page_content_columns=None, metadata_columns=None)[source] Bases: langchain.document_loaders.base.BaseLoader Loads a query result from DuckDB into a list of documents. Each document represents one row of the result. The page_content_columns are written into the page_content of the document. The metadata_columns are written into the metadata of the document. By default, all columns are written into the page_content and none into the metadata. Parameters query (str) – database (str) – read_only (bool) – config (Optional[Dict[str, str]]) – page_content_columns (Optional[List[str]]) – metadata_columns (Optional[List[str]]) – load()[source] Load data into document objects. Return type List[langchain.schema.Document] class langchain.document_loaders.EmbaasBlobLoader(*, embaas_api_key=None, api_url='https://api.embaas.io/v1/document/extract-text/bytes/', params={})[source] Bases: langchain.document_loaders.embaas.BaseEmbaasLoader, langchain.document_loaders.base.BaseBlobParser Wrapper around embaas’s document byte loader service. To use, you should have the environment variable EMBAAS_API_KEY set with your API key, or pass it as a named parameter to the constructor. Example # Default parsing from langchain.document_loaders.embaas import EmbaasBlobLoader loader = EmbaasBlobLoader() blob = Blob.from_path(path="example.mp3") documents = loader.parse(blob=blob) # Custom api parameters (create embeddings automatically) from langchain.document_loaders.embaas import EmbaasBlobLoader loader = EmbaasBlobLoader( params={ "should_embed": True, "model": "e5-large-v2", "chunk_size": 256, "chunk_splitter": "CharacterTextSplitter" } ) blob = Blob.from_path(path="example.pdf") documents = loader.parse(blob=blob) Parameters embaas_api_key (Optional[str]) – api_url (str) – params (langchain.document_loaders.embaas.EmbaasDocumentExtractionParameters) – Return type None lazy_parse(blob)[source] Lazy parsing interface. Subclasses are required to implement this method. Parameters blob (langchain.document_loaders.blob_loaders.schema.Blob) – Blob instance Returns Generator of documents Return type Iterator[langchain.schema.Document] class langchain.document_loaders.EmbaasLoader(*, embaas_api_key=None, api_url='https://api.embaas.io/v1/document/extract-text/bytes/', params={}, file_path, blob_loader=None)[source] Bases: langchain.document_loaders.embaas.BaseEmbaasLoader, langchain.document_loaders.base.BaseLoader Wrapper around embaas’s document loader service. To use, you should have the environment variable EMBAAS_API_KEY set with your API key, or pass it as a named parameter to the constructor. Example # Default parsing from langchain.document_loaders.embaas import EmbaasLoader loader = EmbaasLoader(file_path="example.mp3") documents = loader.load() # Custom api parameters (create embeddings automatically) from langchain.document_loaders.embaas import EmbaasBlobLoader loader = EmbaasBlobLoader( file_path="example.pdf", params={ "should_embed": True, "model": "e5-large-v2", "chunk_size": 256, "chunk_splitter": "CharacterTextSplitter" } ) documents = loader.load() Parameters embaas_api_key (Optional[str]) – api_url (str) – params (langchain.document_loaders.embaas.EmbaasDocumentExtractionParameters) – file_path (str) – blob_loader (Optional[langchain.document_loaders.embaas.EmbaasBlobLoader]) – Return type None attribute blob_loader: Optional[langchain.document_loaders.embaas.EmbaasBlobLoader] = None The blob loader to use. If not provided, a default one will be created. attribute file_path: str [Required] The path to the file to load. lazy_load()[source] Load the documents from the file path lazily. Return type Iterator[langchain.schema.Document] load()[source] Load data into document objects. Return type List[langchain.schema.Document] load_and_split(text_splitter=None)[source] Load documents and split into chunks. Parameters text_splitter (Optional[langchain.text_splitter.TextSplitter]) – Return type List[langchain.schema.Document] class langchain.document_loaders.EverNoteLoader(file_path, load_single_document=True)[source] Bases: langchain.document_loaders.base.BaseLoader EverNote Loader. Loads an EverNote notebook export file e.g. my_notebook.enex into Documents. Instructions on producing this file can be found at https://help.evernote.com/hc/en-us/articles/209005557-Export-notes-and-notebooks-as-ENEX-or-HTML Currently only the plain text in the note is extracted and stored as the contents of the Document, any non content metadata (e.g. β€˜author’, β€˜created’, β€˜updated’ etc. but not β€˜content-raw’ or β€˜resource’) tags on the note will be extracted and stored as metadata on the Document. Parameters file_path (str) – The path to the notebook export with a .enex extension load_single_document (bool) – Whether or not to concatenate the content of all notes into a single long Document. True (If this is set to) – the β€˜source’ which contains the file name of the export. load()[source] Load documents from EverNote export file. Return type List[langchain.schema.Document] class langchain.document_loaders.FacebookChatLoader(path)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that loads Facebook messages json directory dump. Parameters path (str) – load()[source] Load documents. Return type List[langchain.schema.Document] class langchain.document_loaders.FaunaLoader(query, page_content_field, secret, metadata_fields=None)[source] Bases: langchain.document_loaders.base.BaseLoader FaunaDB Loader. Parameters query (str) – page_content_field (str) – secret (str) – metadata_fields (Optional[Sequence[str]]) – query The FQL query string to execute. Type str page_content_field The field that contains the content of each page. Type str secret The secret key for authenticating to FaunaDB. Type str metadata_fields Optional list of field names to include in metadata. Type Optional[Sequence[str]] load()[source] Load data into document objects. Return type List[langchain.schema.Document] lazy_load()[source] A lazy loader for document content. Return type Iterator[langchain.schema.Document] class langchain.document_loaders.FigmaFileLoader(access_token, ids, key)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that loads Figma file json. Parameters access_token (str) – ids (str) – key (str) – load()[source] Load file Return type List[langchain.schema.Document] class langchain.document_loaders.FileSystemBlobLoader(path, *, glob='**/[!.]*', suffixes=None, show_progress=False)[source] Bases: langchain.document_loaders.blob_loaders.schema.BlobLoader Blob loader for the local file system. Example: from langchain.document_loaders.blob_loaders import FileSystemBlobLoader loader = FileSystemBlobLoader("/path/to/directory") for blob in loader.yield_blobs(): print(blob) Parameters path (Union[str, pathlib.Path]) – glob (str) – suffixes (Optional[Sequence[str]]) – show_progress (bool) – Return type None yield_blobs()[source] Yield blobs that match the requested pattern. Return type Iterable[langchain.document_loaders.blob_loaders.schema.Blob] count_matching_files()[source] Count files that match the pattern without loading them. Return type int class langchain.document_loaders.GCSDirectoryLoader(project_name, bucket, prefix='')[source] Bases: langchain.document_loaders.base.BaseLoader Loading logic for loading documents from GCS. Parameters project_name (str) – bucket (str) – prefix (str) – load()[source] Load documents. Return type List[langchain.schema.Document] class langchain.document_loaders.GCSFileLoader(project_name, bucket, blob)[source] Bases: langchain.document_loaders.base.BaseLoader Loading logic for loading documents from GCS. Parameters project_name (str) – bucket (str) – blob (str) – load()[source] Load documents. Return type List[langchain.schema.Document] class langchain.document_loaders.GitHubIssuesLoader(*, repo, access_token, include_prs=True, milestone=None, state=None, assignee=None, creator=None, mentioned=None, labels=None, sort=None, direction=None, since=None)[source] Bases: langchain.document_loaders.github.BaseGitHubLoader Parameters repo (str) – access_token (str) – include_prs (bool) – milestone (Optional[Union[int, Literal['*', 'none']]]) – state (Optional[Literal['open', 'closed', 'all']]) – assignee (Optional[str]) – creator (Optional[str]) – mentioned (Optional[str]) – labels (Optional[List[str]]) – sort (Optional[Literal['created', 'updated', 'comments']]) – direction (Optional[Literal['asc', 'desc']]) – since (Optional[str]) – Return type None attribute assignee: Optional[str] = None Filter on assigned user. Pass β€˜none’ for no user and β€˜*’ for any user. attribute creator: Optional[str] = None Filter on the user that created the issue. attribute direction: Optional[Literal['asc', 'desc']] = None The direction to sort the results by. Can be one of: β€˜asc’, β€˜desc’. attribute include_prs: bool = True If True include Pull Requests in results, otherwise ignore them. attribute labels: Optional[List[str]] = None Label names to filter one. Example: bug,ui,@high. attribute mentioned: Optional[str] = None Filter on a user that’s mentioned in the issue. attribute milestone: Optional[Union[int, Literal['*', 'none']]] = None If integer is passed, it should be a milestone’s number field. If the string β€˜*’ is passed, issues with any milestone are accepted. If the string β€˜none’ is passed, issues without milestones are returned. attribute since: Optional[str] = None Only show notifications updated after the given time. This is a timestamp in ISO 8601 format: YYYY-MM-DDTHH:MM:SSZ. attribute sort: Optional[Literal['created', 'updated', 'comments']] = None What to sort results by. Can be one of: β€˜created’, β€˜updated’, β€˜comments’. Default is β€˜created’. attribute state: Optional[Literal['open', 'closed', 'all']] = None Filter on issue state. Can be one of: β€˜open’, β€˜closed’, β€˜all’. lazy_load()[source] Get issues of a GitHub repository. Returns page_content metadata url title creator created_at last_update_time closed_time number of comments state labels assignee assignees milestone locked number is_pull_request Return type A list of Documents with attributes load()[source] Get issues of a GitHub repository. Returns page_content metadata url title creator created_at last_update_time closed_time number of comments state labels assignee assignees milestone locked number is_pull_request Return type A list of Documents with attributes parse_issue(issue)[source] Create Document objects from a list of GitHub issues. Parameters issue (dict) – Return type langchain.schema.Document property query_params: str property url: str class langchain.document_loaders.GitLoader(repo_path, clone_url=None, branch='main', file_filter=None)[source] Bases: langchain.document_loaders.base.BaseLoader Loads files from a Git repository into a list of documents. Repository can be local on disk available at repo_path, or remote at clone_url that will be cloned to repo_path. Currently supports only text files. Each document represents one file in the repository. The path points to the local Git repository, and the branch specifies the branch to load files from. By default, it loads from the main branch. Parameters repo_path (str) – clone_url (Optional[str]) – branch (Optional[str]) – file_filter (Optional[Callable[[str], bool]]) – load()[source] Load data into document objects. Return type List[langchain.schema.Document] class langchain.document_loaders.GitbookLoader(web_page, load_all_paths=False, base_url=None, content_selector='main')[source] Bases: langchain.document_loaders.web_base.WebBaseLoader Load GitBook data. load from either a single page, or load all (relative) paths in the navbar. Parameters web_page (str) – load_all_paths (bool) – base_url (Optional[str]) – content_selector (str) – load()[source] Fetch text from one single GitBook page. Return type List[langchain.schema.Document] class langchain.document_loaders.GoogleApiClient(credentials_path=PosixPath('/home/docs/.credentials/credentials.json'), service_account_path=PosixPath('/home/docs/.credentials/credentials.json'), token_path=PosixPath('/home/docs/.credentials/token.json'))[source] Bases: object A Generic Google Api Client. To use, you should have the google_auth_oauthlib,youtube_transcript_api,google python package installed. As the google api expects credentials you need to set up a google account and register your Service. β€œhttps://developers.google.com/docs/api/quickstart/python” Example from langchain.document_loaders import GoogleApiClient google_api_client = GoogleApiClient( service_account_path=Path("path_to_your_sec_file.json") ) Parameters credentials_path (pathlib.Path) – service_account_path (pathlib.Path) – token_path (pathlib.Path) – Return type None credentials_path: pathlib.Path = PosixPath('/home/docs/.credentials/credentials.json') service_account_path: pathlib.Path = PosixPath('/home/docs/.credentials/credentials.json') token_path: pathlib.Path = PosixPath('/home/docs/.credentials/token.json') classmethod validate_channel_or_videoIds_is_set(values)[source] Validate that either folder_id or document_ids is set, but not both. Parameters values (Dict[str, Any]) – Return type Dict[str, Any] class langchain.document_loaders.GoogleApiYoutubeLoader(google_api_client, channel_name=None, video_ids=None, add_video_info=True, captions_language='en', continue_on_failure=False)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that loads all Videos from a Channel To use, you should have the googleapiclient,youtube_transcript_api python package installed. As the service needs a google_api_client, you first have to initialize the GoogleApiClient. Additionally you have to either provide a channel name or a list of videoids β€œhttps://developers.google.com/docs/api/quickstart/python” Example from langchain.document_loaders import GoogleApiClient from langchain.document_loaders import GoogleApiYoutubeLoader google_api_client = GoogleApiClient( service_account_path=Path("path_to_your_sec_file.json") ) loader = GoogleApiYoutubeLoader( google_api_client=google_api_client, channel_name = "CodeAesthetic" ) load.load() Parameters google_api_client (langchain.document_loaders.youtube.GoogleApiClient) – channel_name (Optional[str]) – video_ids (Optional[List[str]]) – add_video_info (bool) – captions_language (str) – continue_on_failure (bool) – Return type None google_api_client: langchain.document_loaders.youtube.GoogleApiClient channel_name: Optional[str] = None video_ids: Optional[List[str]] = None add_video_info: bool = True captions_language: str = 'en' continue_on_failure: bool = False classmethod validate_channel_or_videoIds_is_set(values)[source] Validate that either folder_id or document_ids is set, but not both. Parameters values (Dict[str, Any]) – Return type Dict[str, Any] load()[source] Load documents. Return type List[langchain.schema.Document] class langchain.document_loaders.GoogleDriveLoader(*, service_account_key=PosixPath('/home/docs/.credentials/keys.json'), credentials_path=PosixPath('/home/docs/.credentials/credentials.json'), token_path=PosixPath('/home/docs/.credentials/token.json'), folder_id=None, document_ids=None, file_ids=None, recursive=False, file_types=None, load_trashed_files=False, file_loader_cls=None, file_loader_kwargs={})[source] Bases: langchain.document_loaders.base.BaseLoader, pydantic.main.BaseModel Loader that loads Google Docs from Google Drive. Parameters service_account_key (pathlib.Path) – credentials_path (pathlib.Path) – token_path (pathlib.Path) – folder_id (Optional[str]) – document_ids (Optional[List[str]]) – file_ids (Optional[List[str]]) – recursive (bool) – file_types (Optional[Sequence[str]]) – load_trashed_files (bool) – file_loader_cls (Any) – file_loader_kwargs (Dict[str, Any]) – Return type None attribute credentials_path: pathlib.Path = PosixPath('/home/docs/.credentials/credentials.json') attribute document_ids: Optional[List[str]] = None attribute file_ids: Optional[List[str]] = None attribute file_loader_cls: Any = None attribute file_loader_kwargs: Dict[str, Any] = {} attribute file_types: Optional[Sequence[str]] = None attribute folder_id: Optional[str] = None attribute load_trashed_files: bool = False attribute recursive: bool = False attribute service_account_key: pathlib.Path = PosixPath('/home/docs/.credentials/keys.json') attribute token_path: pathlib.Path = PosixPath('/home/docs/.credentials/token.json') load()[source] Load documents. Return type List[langchain.schema.Document] class langchain.document_loaders.GutenbergLoader(file_path)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that uses urllib to load .txt web files. Parameters file_path (str) – load()[source] Load file. Return type List[langchain.schema.Document] class langchain.document_loaders.HNLoader(web_path, header_template=None, verify=True)[source] Bases: langchain.document_loaders.web_base.WebBaseLoader Load Hacker News data from either main page results or the comments page. Parameters web_path (Union[str, List[str]]) – header_template (Optional[dict]) – verify (Optional[bool]) – load()[source] Get important HN webpage information. Components are: title content source url, time of post author of the post number of comments rank of the post Return type List[langchain.schema.Document] load_comments(soup_info)[source] Load comments from a HN post. Parameters soup_info (Any) – Return type List[langchain.schema.Document] load_results(soup)[source] Load items from an HN page. Parameters soup (Any) – Return type List[langchain.schema.Document] class langchain.document_loaders.HuggingFaceDatasetLoader(path, page_content_column='text', name=None, data_dir=None, data_files=None, cache_dir=None, keep_in_memory=None, save_infos=False, use_auth_token=None, num_proc=None)[source] Bases: langchain.document_loaders.base.BaseLoader Loading logic for loading documents from the Hugging Face Hub. Parameters path (str) – page_content_column (str) – name (Optional[str]) – data_dir (Optional[str]) – data_files (Optional[Union[str, Sequence[str], Mapping[str, Union[str, Sequence[str]]]]]) – cache_dir (Optional[str]) – keep_in_memory (Optional[bool]) – save_infos (bool) – use_auth_token (Optional[Union[bool, str]]) – num_proc (Optional[int]) – lazy_load()[source] Load documents lazily. Return type Iterator[langchain.schema.Document] load()[source] Load documents. Return type List[langchain.schema.Document] class langchain.document_loaders.IFixitLoader(web_path)[source] Bases: langchain.document_loaders.base.BaseLoader Load iFixit repair guides, device wikis and answers. iFixit is the largest, open repair community on the web. The site contains nearly 100k repair manuals, 200k Questions & Answers on 42k devices, and all the data is licensed under CC-BY. This loader will allow you to download the text of a repair guide, text of Q&A’s and wikis from devices on iFixit using their open APIs and web scraping. Parameters web_path (str) – load()[source] Load data into document objects. Return type List[langchain.schema.Document] static load_suggestions(query='', doc_type='all')[source] Parameters query (str) – doc_type (str) – Return type List[langchain.schema.Document] load_questions_and_answers(url_override=None)[source] Parameters url_override (Optional[str]) – Return type List[langchain.schema.Document] load_device(url_override=None, include_guides=True)[source] Parameters url_override (Optional[str]) – include_guides (bool) – Return type List[langchain.schema.Document] load_guide(url_override=None)[source] Parameters url_override (Optional[str]) – Return type List[langchain.schema.Document] class langchain.document_loaders.IMSDbLoader(web_path, header_template=None, verify=True)[source] Bases: langchain.document_loaders.web_base.WebBaseLoader Loader that loads IMSDb webpages. Parameters web_path (Union[str, List[str]]) – header_template (Optional[dict]) – verify (Optional[bool]) – load()[source] Load webpage. Return type List[langchain.schema.Document] class langchain.document_loaders.ImageCaptionLoader(path_images, blip_processor='Salesforce/blip-image-captioning-base', blip_model='Salesforce/blip-image-captioning-base')[source] Bases: langchain.document_loaders.base.BaseLoader Loader that loads the captions of an image Parameters path_images (Union[str, List[str]]) – blip_processor (str) – blip_model (str) – load()[source] Load from a list of image files Return type List[langchain.schema.Document] class langchain.document_loaders.IuguLoader(resource, api_token=None)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that fetches data from IUGU. Parameters resource (str) – api_token (Optional[str]) – Return type None load()[source] Load data into document objects. Return type List[langchain.schema.Document] class langchain.document_loaders.JSONLoader(file_path, jq_schema, content_key=None, metadata_func=None, text_content=True)[source] Bases: langchain.document_loaders.base.BaseLoader Loads a JSON file and references a jq schema provided to load the text into documents. Example [{β€œtext”: …}, {β€œtext”: …}, {β€œtext”: …}] -> schema = .[].text {β€œkey”: [{β€œtext”: …}, {β€œtext”: …}, {β€œtext”: …}]} -> schema = .key[].text [β€œβ€, β€œβ€, β€œβ€] -> schema = .[] Parameters file_path (Union[str, pathlib.Path]) – jq_schema (str) – content_key (Optional[str]) – metadata_func (Optional[Callable[[Dict, Dict], Dict]]) – text_content (bool) – load()[source] Load and return documents from the JSON file. Return type List[langchain.schema.Document] class langchain.document_loaders.JoplinLoader(access_token=None, port=41184, host='localhost')[source] Bases: langchain.document_loaders.base.BaseLoader Loader that fetches notes from Joplin. In order to use this loader, you need to have Joplin running with the Web Clipper enabled (look for β€œWeb Clipper” in the app settings). To get the access token, you need to go to the Web Clipper options and under β€œAdvanced Options” you will find the access token. You can find more information about the Web Clipper service here: https://joplinapp.org/clipper/ Parameters access_token (Optional[str]) – port (int) – host (str) – Return type None lazy_load()[source] A lazy loader for document content. Return type Iterator[langchain.schema.Document] load()[source] Load data into document objects. Return type List[langchain.schema.Document] class langchain.document_loaders.MWDumpLoader(file_path, encoding='utf8')[source] Bases: langchain.document_loaders.base.BaseLoader Load MediaWiki dump from XML file .. rubric:: Example from langchain.document_loaders import MWDumpLoader loader = MWDumpLoader( file_path="myWiki.xml", encoding="utf8" ) docs = loader.load() from langchain.text_splitter import RecursiveCharacterTextSplitter text_splitter = RecursiveCharacterTextSplitter( chunk_size=1000, chunk_overlap=0 ) texts = text_splitter.split_documents(docs) Parameters file_path (str) – XML local file path encoding (str, optional) – Charset encoding, defaults to β€œutf8” load()[source] Load from file path. Return type List[langchain.schema.Document] class langchain.document_loaders.MastodonTootsLoader(mastodon_accounts, number_toots=100, exclude_replies=False, access_token=None, api_base_url='https://mastodon.social')[source] Bases: langchain.document_loaders.base.BaseLoader Mastodon toots loader. Parameters mastodon_accounts (Sequence[str]) – number_toots (Optional[int]) – exclude_replies (bool) – access_token (Optional[str]) – api_base_url (str) – load()[source] Load toots into documents. Return type List[langchain.schema.Document] class langchain.document_loaders.MathpixPDFLoader(file_path, processed_file_format='mmd', max_wait_time_seconds=500, should_clean_pdf=False, **kwargs)[source] Bases: langchain.document_loaders.pdf.BasePDFLoader Parameters file_path (str) – processed_file_format (str) – max_wait_time_seconds (int) – should_clean_pdf (bool) – kwargs (Any) – Return type None property headers: dict property url: str property data: dict send_pdf()[source] Return type str wait_for_processing(pdf_id)[source] Parameters pdf_id (str) – Return type None get_processed_pdf(pdf_id)[source] Parameters pdf_id (str) – Return type str clean_pdf(contents)[source] Parameters contents (str) – Return type str load()[source] Load data into document objects. Return type List[langchain.schema.Document] class langchain.document_loaders.MaxComputeLoader(query, api_wrapper, *, page_content_columns=None, metadata_columns=None)[source] Bases: langchain.document_loaders.base.BaseLoader Loads a query result from Alibaba Cloud MaxCompute table into documents. Parameters query (str) – api_wrapper (MaxComputeAPIWrapper) – page_content_columns (Optional[Sequence[str]]) – metadata_columns (Optional[Sequence[str]]) – classmethod from_params(query, endpoint, project, *, access_id=None, secret_access_key=None, **kwargs)[source] Convenience constructor that builds the MaxCompute API wrapper fromgiven parameters. Parameters query (str) – SQL query to execute. endpoint (str) – MaxCompute endpoint. project (str) – A project is a basic organizational unit of MaxCompute, which is similar to a database. access_id (Optional[str]) – MaxCompute access ID. Should be passed in directly or set as the environment variable MAX_COMPUTE_ACCESS_ID. secret_access_key (Optional[str]) – MaxCompute secret access key. Should be passed in directly or set as the environment variable MAX_COMPUTE_SECRET_ACCESS_KEY. kwargs (Any) – Return type langchain.document_loaders.max_compute.MaxComputeLoader lazy_load()[source] A lazy loader for document content. Return type Iterator[langchain.schema.Document] load()[source] Load data into document objects. Return type List[langchain.schema.Document] class langchain.document_loaders.MergedDataLoader(loaders)[source] Bases: langchain.document_loaders.base.BaseLoader Merge documents from a list of loaders Parameters loaders (List) – lazy_load()[source] Lazy load docs from each individual loader. Return type Iterator[langchain.schema.Document] load()[source] Load docs. Return type List[langchain.schema.Document] class langchain.document_loaders.MHTMLLoader(file_path, open_encoding=None, bs_kwargs=None, get_text_separator='')[source] Bases: langchain.document_loaders.base.BaseLoader Loader that uses beautiful soup to parse HTML files. Parameters file_path (str) – open_encoding (Optional[str]) – bs_kwargs (Optional[dict]) – get_text_separator (str) – Return type None load()[source] Load data into document objects. Return type List[langchain.schema.Document] class langchain.document_loaders.ModernTreasuryLoader(resource, organization_id=None, api_key=None)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that fetches data from Modern Treasury. Parameters resource (str) – organization_id (Optional[str]) – api_key (Optional[str]) – Return type None load()[source] Load data into document objects. Return type List[langchain.schema.Document] class langchain.document_loaders.NotebookLoader(path, include_outputs=False, max_output_length=10, remove_newline=False, traceback=False)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that loads .ipynb notebook files. Parameters path (str) – include_outputs (bool) – max_output_length (int) – remove_newline (bool) – traceback (bool) – load()[source] Load documents. Return type List[langchain.schema.Document] class langchain.document_loaders.NotionDBLoader(integration_token, database_id, request_timeout_sec=10)[source] Bases: langchain.document_loaders.base.BaseLoader Notion DB Loader. Reads content from pages within a Noton Database. :param integration_token: Notion integration token. :type integration_token: str :param database_id: Notion database id. :type database_id: str :param request_timeout_sec: Timeout for Notion requests in seconds. :type request_timeout_sec: int Parameters integration_token (str) – database_id (str) – request_timeout_sec (Optional[int]) – Return type None load()[source] Load documents from the Notion database. :returns: List of documents. :rtype: List[Document] Return type List[langchain.schema.Document] load_page(page_summary)[source] Read a page. Parameters page_summary (Dict[str, Any]) – Return type langchain.schema.Document class langchain.document_loaders.NotionDirectoryLoader(path)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that loads Notion directory dump. Parameters path (str) – load()[source] Load documents. Return type List[langchain.schema.Document] class langchain.document_loaders.ObsidianLoader(path, encoding='UTF-8', collect_metadata=True)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that loads Obsidian files from disk. Parameters path (str) – encoding (str) – collect_metadata (bool) – FRONT_MATTER_REGEX = re.compile('^---\\n(.*?)\\n---\\n', re.MULTILINE|re.DOTALL) load()[source] Load documents. Return type List[langchain.schema.Document] class langchain.document_loaders.OneDriveFileLoader(*, file)[source] Bases: langchain.document_loaders.base.BaseLoader, pydantic.main.BaseModel Parameters file (File) – Return type None attribute file: File [Required] load()[source] Load Documents Return type List[langchain.schema.Document] class langchain.document_loaders.OneDriveLoader(*, settings=None, drive_id, folder_path=None, object_ids=None, auth_with_token=False)[source] Bases: langchain.document_loaders.base.BaseLoader, pydantic.main.BaseModel Parameters settings (langchain.document_loaders.onedrive._OneDriveSettings) – drive_id (str) – folder_path (Optional[str]) – object_ids (Optional[List[str]]) – auth_with_token (bool) – Return type None attribute auth_with_token: bool = False attribute drive_id: str [Required] attribute folder_path: Optional[str] = None attribute object_ids: Optional[List[str]] = None attribute settings: langchain.document_loaders.onedrive._OneDriveSettings [Optional] load()[source] Loads all supported document files from the specified OneDrive drive a nd returns a list of Document objects. Returns A list of Document objects representing the loaded documents. Return type List[Document] Raises ValueError – If the specified drive ID does not correspond to a drive in the OneDrive storage. – class langchain.document_loaders.OnlinePDFLoader(file_path)[source] Bases: langchain.document_loaders.pdf.BasePDFLoader Loader that loads online PDFs. Parameters file_path (str) – load()[source] Load documents. Return type List[langchain.schema.Document] class langchain.document_loaders.OutlookMessageLoader(file_path)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that loads Outlook Message files using extract_msg. https://github.com/TeamMsgExtractor/msg-extractor Parameters file_path (str) – load()[source] Load data into document objects. Return type List[langchain.schema.Document] class langchain.document_loaders.OpenCityDataLoader(city_id, dataset_id, limit)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that loads Open city data. Parameters city_id (str) – dataset_id (str) – limit (int) – lazy_load()[source] Lazy load records. Return type Iterator[langchain.schema.Document] load()[source] Load records. Return type List[langchain.schema.Document] class langchain.document_loaders.PDFMinerLoader(file_path)[source] Bases: langchain.document_loaders.pdf.BasePDFLoader Loader that uses PDFMiner to load PDF files. Parameters file_path (str) – Return type None load()[source] Eagerly load the content. Return type List[langchain.schema.Document] lazy_load()[source] Lazily lod documents. Return type Iterator[langchain.schema.Document] class langchain.document_loaders.PDFMinerPDFasHTMLLoader(file_path)[source] Bases: langchain.document_loaders.pdf.BasePDFLoader Loader that uses PDFMiner to load PDF files as HTML content. Parameters file_path (str) – load()[source] Load file. Return type List[langchain.schema.Document] class langchain.document_loaders.PDFPlumberLoader(file_path, text_kwargs=None)[source] Bases: langchain.document_loaders.pdf.BasePDFLoader Loader that uses pdfplumber to load PDF files. Parameters file_path (str) – text_kwargs (Optional[Mapping[str, Any]]) – Return type None load()[source] Load file. Return type List[langchain.schema.Document] langchain.document_loaders.PagedPDFSplitter alias of langchain.document_loaders.pdf.PyPDFLoader class langchain.document_loaders.PlaywrightURLLoader(urls, continue_on_failure=True, headless=True, remove_selectors=None)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that uses Playwright and to load a page and unstructured to load the html. This is useful for loading pages that require javascript to render. Parameters urls (List[str]) – continue_on_failure (bool) – headless (bool) – remove_selectors (Optional[List[str]]) – urls List of URLs to load. Type List[str] continue_on_failure If True, continue loading other URLs on failure. Type bool headless If True, the browser will run in headless mode. Type bool load()[source] Load the specified URLs using Playwright and create Document instances. Returns A list of Document instances with loaded content. Return type List[Document] class langchain.document_loaders.PsychicLoader(api_key, connector_id, connection_id)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that loads documents from Psychic.dev. Parameters api_key (str) – connector_id (str) – connection_id (str) – load()[source] Load documents. Return type List[langchain.schema.Document] class langchain.document_loaders.PyMuPDFLoader(file_path)[source] Bases: langchain.document_loaders.pdf.BasePDFLoader Loader that uses PyMuPDF to load PDF files. Parameters file_path (str) – Return type None load(**kwargs)[source] Load file. Parameters kwargs (Optional[Any]) – Return type List[langchain.schema.Document] class langchain.document_loaders.PyPDFDirectoryLoader(path, glob='**/[!.]*.pdf', silent_errors=False, load_hidden=False, recursive=False)[source] Bases: langchain.document_loaders.base.BaseLoader Loads a directory with PDF files with pypdf and chunks at character level. Loader also stores page numbers in metadatas. Parameters path (str) – glob (str) – silent_errors (bool) – load_hidden (bool) – recursive (bool) – load()[source] Load data into document objects. Return type List[langchain.schema.Document] class langchain.document_loaders.PyPDFLoader(file_path)[source] Bases: langchain.document_loaders.pdf.BasePDFLoader Loads a PDF with pypdf and chunks at character level. Loader also stores page numbers in metadatas. Parameters file_path (str) – Return type None load()[source] Load given path as pages. Return type List[langchain.schema.Document] lazy_load()[source] Lazy load given path as pages. Return type Iterator[langchain.schema.Document] class langchain.document_loaders.PyPDFium2Loader(file_path)[source] Bases: langchain.document_loaders.pdf.BasePDFLoader Loads a PDF with pypdfium2 and chunks at character level. Parameters file_path (str) – load()[source] Load given path as pages. Return type List[langchain.schema.Document] lazy_load()[source] Lazy load given path as pages. Return type Iterator[langchain.schema.Document] class langchain.document_loaders.PySparkDataFrameLoader(spark_session=None, df=None, page_content_column='text', fraction_of_memory=0.1)[source] Bases: langchain.document_loaders.base.BaseLoader Load PySpark DataFrames Parameters spark_session (Optional[SparkSession]) – df (Optional[Any]) – page_content_column (str) – fraction_of_memory (float) – get_num_rows()[source] Gets the amount of β€œfeasible” rows for the DataFrame Return type Tuple[int, int] lazy_load()[source] A lazy loader for document content. Return type Iterator[langchain.schema.Document] load()[source] Load from the dataframe. Return type List[langchain.schema.Document] class langchain.document_loaders.PythonLoader(file_path)[source] Bases: langchain.document_loaders.text.TextLoader Load Python files, respecting any non-default encoding if specified. Parameters file_path (str) – class langchain.document_loaders.ReadTheDocsLoader(path, encoding=None, errors=None, custom_html_tag=None, **kwargs)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that loads ReadTheDocs documentation directory dump. Parameters path (Union[str, pathlib.Path]) – encoding (Optional[str]) – errors (Optional[str]) – custom_html_tag (Optional[Tuple[str, dict]]) – kwargs (Optional[Any]) – load()[source] Load documents. Return type List[langchain.schema.Document] class langchain.document_loaders.RecursiveUrlLoader(url, exclude_dirs=None)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that loads all child links from a given url. Parameters url (str) – exclude_dirs (Optional[str]) – Return type None get_child_links_recursive(url, visited=None)[source] Recursively get all child links starting with the path of the input URL. Parameters url (str) – visited (Optional[Set[str]]) – Return type Set[str] lazy_load()[source] A lazy loader for document content. Return type Iterator[langchain.schema.Document] load()[source] Load web pages. Return type List[langchain.schema.Document] class langchain.document_loaders.RedditPostsLoader(client_id, client_secret, user_agent, search_queries, mode, categories=['new'], number_posts=10)[source] Bases: langchain.document_loaders.base.BaseLoader Reddit posts loader. Read posts on a subreddit. First you need to go to https://www.reddit.com/prefs/apps/ and create your application Parameters client_id (str) – client_secret (str) – user_agent (str) – search_queries (Sequence[str]) – mode (str) – categories (Sequence[str]) – number_posts (Optional[int]) – load()[source] Load reddits. Return type List[langchain.schema.Document] class langchain.document_loaders.RoamLoader(path)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that loads Roam files from disk. Parameters path (str) – load()[source] Load documents. Return type List[langchain.schema.Document] class langchain.document_loaders.S3DirectoryLoader(bucket, prefix='')[source] Bases: langchain.document_loaders.base.BaseLoader Loading logic for loading documents from s3. Parameters bucket (str) – prefix (str) – load()[source] Load documents. Return type List[langchain.schema.Document] class langchain.document_loaders.S3FileLoader(bucket, key)[source] Bases: langchain.document_loaders.base.BaseLoader Loading logic for loading documents from s3. Parameters bucket (str) – key (str) – load()[source] Load documents. Return type List[langchain.schema.Document] class langchain.document_loaders.SRTLoader(file_path)[source] Bases: langchain.document_loaders.base.BaseLoader Loader for .srt (subtitle) files. Parameters file_path (str) – load()[source] Load using pysrt file. Return type List[langchain.schema.Document] class langchain.document_loaders.SeleniumURLLoader(urls, continue_on_failure=True, browser='chrome', binary_location=None, executable_path=None, headless=True, arguments=[])[source] Bases: langchain.document_loaders.base.BaseLoader Loader that uses Selenium and to load a page and unstructured to load the html. This is useful for loading pages that require javascript to render. Parameters urls (List[str]) – continue_on_failure (bool) – browser (Literal['chrome', 'firefox']) – binary_location (Optional[str]) – executable_path (Optional[str]) – headless (bool) – arguments (List[str]) – urls List of URLs to load. Type List[str] continue_on_failure If True, continue loading other URLs on failure. Type bool browser The browser to use, either β€˜chrome’ or β€˜firefox’. Type str binary_location The location of the browser binary. Type Optional[str] executable_path The path to the browser executable. Type Optional[str] headless If True, the browser will run in headless mode. Type bool arguments [List[str]] List of arguments to pass to the browser. load()[source] Load the specified URLs using Selenium and create Document instances. Returns A list of Document instances with loaded content. Return type List[Document] class langchain.document_loaders.SitemapLoader(web_path, filter_urls=None, parsing_function=None, blocksize=None, blocknum=0, meta_function=None, is_local=False)[source] Bases: langchain.document_loaders.web_base.WebBaseLoader Loader that fetches a sitemap and loads those URLs. Parameters web_path (str) – filter_urls (Optional[List[str]]) – parsing_function (Optional[Callable]) – blocksize (Optional[int]) – blocknum (int) – meta_function (Optional[Callable]) – is_local (bool) – parse_sitemap(soup)[source] Parse sitemap xml and load into a list of dicts. Parameters soup (Any) – Return type List[dict] load()[source] Load sitemap. Return type List[langchain.schema.Document] class langchain.document_loaders.SlackDirectoryLoader(zip_path, workspace_url=None)[source] Bases: langchain.document_loaders.base.BaseLoader Loader for loading documents from a Slack directory dump. Parameters zip_path (str) – workspace_url (Optional[str]) – load()[source] Load and return documents from the Slack directory dump. Return type List[langchain.schema.Document] class langchain.document_loaders.SnowflakeLoader(query, user, password, account, warehouse, role, database, schema, parameters=None, page_content_columns=None, metadata_columns=None)[source] Bases: langchain.document_loaders.base.BaseLoader Loads a query result from Snowflake into a list of documents. Each document represents one row of the result. The page_content_columns are written into the page_content of the document. The metadata_columns are written into the metadata of the document. By default, all columns are written into the page_content and none into the metadata. Parameters query (str) – user (str) – password (str) – account (str) – warehouse (str) – role (str) – database (str) – schema (str) – parameters (Optional[Dict[str, Any]]) – page_content_columns (Optional[List[str]]) – metadata_columns (Optional[List[str]]) – lazy_load()[source] A lazy loader for document content. Return type Iterator[langchain.schema.Document] load()[source] Load data into document objects. Return type List[langchain.schema.Document] class langchain.document_loaders.SpreedlyLoader(access_token, resource)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that fetches data from Spreedly API. Parameters access_token (str) – resource (str) – Return type None load()[source] Load data into document objects. Return type List[langchain.schema.Document] class langchain.document_loaders.StripeLoader(resource, access_token=None)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that fetches data from Stripe. Parameters resource (str) – access_token (Optional[str]) – Return type None load()[source] Load data into document objects. Return type List[langchain.schema.Document] class langchain.document_loaders.TelegramChatApiLoader(chat_entity=None, api_id=None, api_hash=None, username=None, file_path='telegram_data.json')[source] Bases: langchain.document_loaders.base.BaseLoader Loader that loads Telegram chat json directory dump. Parameters chat_entity (Optional[EntityLike]) – api_id (Optional[int]) – api_hash (Optional[str]) – username (Optional[str]) – file_path (str) – async fetch_data_from_telegram()[source] Fetch data from Telegram API and save it as a JSON file. Return type None load()[source] Load documents. Return type List[langchain.schema.Document] class langchain.document_loaders.TelegramChatFileLoader(path)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that loads Telegram chat json directory dump. Parameters path (str) – load()[source] Load documents. Return type List[langchain.schema.Document] langchain.document_loaders.TelegramChatLoader alias of langchain.document_loaders.telegram.TelegramChatFileLoader class langchain.document_loaders.TextLoader(file_path, encoding=None, autodetect_encoding=False)[source] Bases: langchain.document_loaders.base.BaseLoader Load text files. Parameters file_path (str) – Path to the file to load. encoding (Optional[str]) – File encoding to use. If None, the file will be loaded encoding. (with the default system) – autodetect_encoding (bool) – Whether to try to autodetect the file encoding if the specified encoding fails. load()[source] Load from file path. Return type List[langchain.schema.Document] class langchain.document_loaders.ToMarkdownLoader(url, api_key)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that loads HTML to markdown using 2markdown. Parameters url (str) – api_key (str) – lazy_load()[source] Lazily load the file. Return type Iterator[langchain.schema.Document] load()[source] Load file. Return type List[langchain.schema.Document] class langchain.document_loaders.TomlLoader(source)[source] Bases: langchain.document_loaders.base.BaseLoader A TOML document loader that inherits from the BaseLoader class. This class can be initialized with either a single source file or a source directory containing TOML files. Parameters source (Union[str, pathlib.Path]) – load()[source] Load and return all documents. Return type List[langchain.schema.Document] lazy_load()[source] Lazily load the TOML documents from the source file or directory. Return type Iterator[langchain.schema.Document] class langchain.document_loaders.TrelloLoader(client, board_name, *, include_card_name=True, include_comments=True, include_checklist=True, card_filter='all', extra_metadata=('due_date', 'labels', 'list', 'closed'))[source] Bases: langchain.document_loaders.base.BaseLoader Trello loader. Reads all cards from a Trello board. Parameters client (TrelloClient) – board_name (str) – include_card_name (bool) – include_comments (bool) – include_checklist (bool) – card_filter (Literal['closed', 'open', 'all']) – extra_metadata (Tuple[str, ...]) – classmethod from_credentials(board_name, *, api_key=None, token=None, **kwargs)[source] Convenience constructor that builds TrelloClient init param for you. Parameters board_name (str) – The name of the Trello board. api_key (Optional[str]) – Trello API key. Can also be specified as environment variable TRELLO_API_KEY. token (Optional[str]) – Trello token. Can also be specified as environment variable TRELLO_TOKEN. include_card_name – Whether to include the name of the card in the document. include_comments – Whether to include the comments on the card in the document. include_checklist – Whether to include the checklist on the card in the document. card_filter – Filter on card status. Valid values are β€œclosed”, β€œopen”, β€œall”. extra_metadata – List of additional metadata fields to include as document metadata.Valid values are β€œdue_date”, β€œlabels”, β€œlist”, β€œclosed”. kwargs (Any) – Return type langchain.document_loaders.trello.TrelloLoader load()[source] Loads all cards from the specified Trello board. You can filter the cards, metadata and text included by using the optional parameters. Returns:A list of documents, one for each card in the board. Return type List[langchain.schema.Document] class langchain.document_loaders.TwitterTweetLoader(auth_handler, twitter_users, number_tweets=100)[source] Bases: langchain.document_loaders.base.BaseLoader Twitter tweets loader. Read tweets of user twitter handle. First you need to go to https://developer.twitter.com/en/docs/twitter-api /getting-started/getting-access-to-the-twitter-api to get your token. And create a v2 version of the app. Parameters auth_handler (Union[OAuthHandler, OAuth2BearerHandler]) – twitter_users (Sequence[str]) – number_tweets (Optional[int]) – load()[source] Load tweets. Return type List[langchain.schema.Document] classmethod from_bearer_token(oauth2_bearer_token, twitter_users, number_tweets=100)[source] Create a TwitterTweetLoader from OAuth2 bearer token. Parameters oauth2_bearer_token (str) – twitter_users (Sequence[str]) – number_tweets (Optional[int]) – Return type langchain.document_loaders.twitter.TwitterTweetLoader classmethod from_secrets(access_token, access_token_secret, consumer_key, consumer_secret, twitter_users, number_tweets=100)[source] Create a TwitterTweetLoader from access tokens and secrets. Parameters access_token (str) – access_token_secret (str) – consumer_key (str) – consumer_secret (str) – twitter_users (Sequence[str]) – number_tweets (Optional[int]) – Return type langchain.document_loaders.twitter.TwitterTweetLoader class langchain.document_loaders.UnstructuredAPIFileIOLoader(file, mode='single', url='https://api.unstructured.io/general/v0/general', api_key='', **unstructured_kwargs)[source] Bases: langchain.document_loaders.unstructured.UnstructuredFileIOLoader Loader that uses the unstructured web API to load file IO objects. Parameters file (Union[IO, Sequence[IO]]) – mode (str) – url (str) – api_key (str) – unstructured_kwargs (Any) – class langchain.document_loaders.UnstructuredAPIFileLoader(file_path='', mode='single', url='https://api.unstructured.io/general/v0/general', api_key='', **unstructured_kwargs)[source] Bases: langchain.document_loaders.unstructured.UnstructuredFileLoader Loader that uses the unstructured web API to load files. Parameters file_path (Union[str, List[str]]) – mode (str) – url (str) – api_key (str) – unstructured_kwargs (Any) – class langchain.document_loaders.UnstructuredCSVLoader(file_path, mode='single', **unstructured_kwargs)[source] Bases: langchain.document_loaders.unstructured.UnstructuredFileLoader Loader that uses unstructured to load CSV files. Parameters file_path (str) – mode (str) – unstructured_kwargs (Any) – class langchain.document_loaders.UnstructuredEPubLoader(file_path, mode='single', **unstructured_kwargs)[source] Bases: langchain.document_loaders.unstructured.UnstructuredFileLoader Loader that uses unstructured to load epub files. Parameters file_path (Union[str, List[str]]) – mode (str) – unstructured_kwargs (Any) – class langchain.document_loaders.UnstructuredEmailLoader(file_path, mode='single', **unstructured_kwargs)[source] Bases: langchain.document_loaders.unstructured.UnstructuredFileLoader Loader that uses unstructured to load email files. Parameters file_path (Union[str, List[str]]) – mode (str) – unstructured_kwargs (Any) – class langchain.document_loaders.UnstructuredExcelLoader(file_path, mode='single', **unstructured_kwargs)[source] Bases: langchain.document_loaders.unstructured.UnstructuredFileLoader Loader that uses unstructured to load Microsoft Excel files. Parameters file_path (str) – mode (str) – unstructured_kwargs (Any) – class langchain.document_loaders.UnstructuredFileIOLoader(file, mode='single', **unstructured_kwargs)[source] Bases: langchain.document_loaders.unstructured.UnstructuredBaseLoader Loader that uses unstructured to load file IO objects. Parameters file (Union[IO, Sequence[IO]]) – mode (str) – unstructured_kwargs (Any) – class langchain.document_loaders.UnstructuredFileLoader(file_path, mode='single', **unstructured_kwargs)[source] Bases: langchain.document_loaders.unstructured.UnstructuredBaseLoader Loader that uses unstructured to load files. Parameters file_path (Union[str, List[str]]) – mode (str) – unstructured_kwargs (Any) – class langchain.document_loaders.UnstructuredHTMLLoader(file_path, mode='single', **unstructured_kwargs)[source] Bases: langchain.document_loaders.unstructured.UnstructuredFileLoader Loader that uses unstructured to load HTML files. Parameters file_path (Union[str, List[str]]) – mode (str) – unstructured_kwargs (Any) – class langchain.document_loaders.UnstructuredImageLoader(file_path, mode='single', **unstructured_kwargs)[source] Bases: langchain.document_loaders.unstructured.UnstructuredFileLoader Loader that uses unstructured to load image files, such as PNGs and JPGs. Parameters file_path (Union[str, List[str]]) – mode (str) – unstructured_kwargs (Any) – class langchain.document_loaders.UnstructuredMarkdownLoader(file_path, mode='single', **unstructured_kwargs)[source] Bases: langchain.document_loaders.unstructured.UnstructuredFileLoader Loader that uses unstructured to load markdown files. Parameters file_path (Union[str, List[str]]) – mode (str) – unstructured_kwargs (Any) – class langchain.document_loaders.UnstructuredODTLoader(file_path, mode='single', **unstructured_kwargs)[source] Bases: langchain.document_loaders.unstructured.UnstructuredFileLoader Loader that uses unstructured to load open office ODT files. Parameters file_path (str) – mode (str) – unstructured_kwargs (Any) – class langchain.document_loaders.UnstructuredPDFLoader(file_path, mode='single', **unstructured_kwargs)[source] Bases: langchain.document_loaders.unstructured.UnstructuredFileLoader Loader that uses unstructured to load PDF files. Parameters file_path (Union[str, List[str]]) – mode (str) – unstructured_kwargs (Any) – class langchain.document_loaders.UnstructuredPowerPointLoader(file_path, mode='single', **unstructured_kwargs)[source] Bases: langchain.document_loaders.unstructured.UnstructuredFileLoader Loader that uses unstructured to load powerpoint files. Parameters file_path (Union[str, List[str]]) – mode (str) – unstructured_kwargs (Any) – class langchain.document_loaders.UnstructuredRSTLoader(file_path, mode='single', **unstructured_kwargs)[source] Bases: langchain.document_loaders.unstructured.UnstructuredFileLoader Loader that uses unstructured to load RST files. Parameters file_path (str) – mode (str) – unstructured_kwargs (Any) – class langchain.document_loaders.UnstructuredRTFLoader(file_path, mode='single', **unstructured_kwargs)[source] Bases: langchain.document_loaders.unstructured.UnstructuredFileLoader Loader that uses unstructured to load rtf files. Parameters file_path (str) – mode (str) – unstructured_kwargs (Any) – class langchain.document_loaders.UnstructuredURLLoader(urls, continue_on_failure=True, mode='single', show_progress_bar=False, **unstructured_kwargs)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that uses unstructured to load HTML files. Parameters urls (List[str]) – continue_on_failure (bool) – mode (str) – show_progress_bar (bool) – unstructured_kwargs (Any) – load()[source] Load file. Return type List[langchain.schema.Document] class langchain.document_loaders.UnstructuredWordDocumentLoader(file_path, mode='single', **unstructured_kwargs)[source] Bases: langchain.document_loaders.unstructured.UnstructuredFileLoader Loader that uses unstructured to load word documents. Parameters file_path (Union[str, List[str]]) – mode (str) – unstructured_kwargs (Any) – class langchain.document_loaders.UnstructuredXMLLoader(file_path, mode='single', **unstructured_kwargs)[source] Bases: langchain.document_loaders.unstructured.UnstructuredFileLoader Loader that uses unstructured to load XML files. Parameters file_path (str) – mode (str) – unstructured_kwargs (Any) – class langchain.document_loaders.WeatherDataLoader(client, places)[source] Bases: langchain.document_loaders.base.BaseLoader Weather Reader. Reads the forecast & current weather of any location using OpenWeatherMap’s free API. Checkout β€˜https://openweathermap.org/appid’ for more on how to generate a free OpenWeatherMap API. Parameters client (OpenWeatherMapAPIWrapper) – places (Sequence[str]) – Return type None classmethod from_params(places, *, openweathermap_api_key=None)[source] Parameters places (Sequence[str]) – openweathermap_api_key (Optional[str]) – Return type langchain.document_loaders.weather.WeatherDataLoader lazy_load()[source] Lazily load weather data for the given locations. Return type Iterator[langchain.schema.Document] load()[source] Load weather data for the given locations. Return type List[langchain.schema.Document] class langchain.document_loaders.WebBaseLoader(web_path, header_template=None, verify=True)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that uses urllib and beautiful soup to load webpages. Parameters web_path (Union[str, List[str]]) – header_template (Optional[dict]) – verify (Optional[bool]) – requests_per_second: int = 2 Max number of concurrent requests to make. default_parser: str = 'html.parser' Default parser to use for BeautifulSoup. requests_kwargs: Dict[str, Any] = {} kwargs for requests bs_get_text_kwargs: Dict[str, Any] = {} kwargs for beatifulsoup4 get_text web_paths: List[str] property web_path: str async fetch_all(urls)[source] Fetch all urls concurrently with rate limiting. Parameters urls (List[str]) – Return type Any scrape_all(urls, parser=None)[source] Fetch all urls, then return soups for all results. Parameters urls (List[str]) – parser (Optional[str]) – Return type List[Any] scrape(parser=None)[source] Scrape data from webpage and return it in BeautifulSoup format. Parameters parser (Optional[str]) – Return type Any lazy_load()[source] Lazy load text from the url(s) in web_path. Return type Iterator[langchain.schema.Document] load()[source] Load text from the url(s) in web_path. Return type List[langchain.schema.Document] aload()[source] Load text from the urls in web_path async into Documents. Return type List[langchain.schema.Document] class langchain.document_loaders.WhatsAppChatLoader(path)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that loads WhatsApp messages text file. Parameters path (str) – load()[source] Load documents. Return type List[langchain.schema.Document] class langchain.document_loaders.WikipediaLoader(query, lang='en', load_max_docs=100, load_all_available_meta=False, doc_content_chars_max=4000)[source] Bases: langchain.document_loaders.base.BaseLoader Loads a query result from www.wikipedia.org into a list of Documents. The hard limit on the number of downloaded Documents is 300 for now. Each wiki page represents one Document. Parameters query (str) – lang (str) – load_max_docs (Optional[int]) – load_all_available_meta (Optional[bool]) – doc_content_chars_max (Optional[int]) – load()[source] Loads the query result from Wikipedia into a list of Documents. Returns A list of Document objects representing the loadedWikipedia pages. Return type List[Document] class langchain.document_loaders.YoutubeAudioLoader(urls, save_dir)[source] Bases: langchain.document_loaders.blob_loaders.schema.BlobLoader Load YouTube urls as audio file(s). Parameters urls (List[str]) – save_dir (str) – yield_blobs()[source] Yield audio blobs for each url. Return type Iterable[langchain.document_loaders.blob_loaders.schema.Blob] class langchain.document_loaders.YoutubeLoader(video_id, add_video_info=False, language='en', translation='en', continue_on_failure=False)[source] Bases: langchain.document_loaders.base.BaseLoader Loader that loads Youtube transcripts. Parameters video_id (str) – add_video_info (bool) – language (Union[str, Sequence[str]]) – translation (str) – continue_on_failure (bool) – static extract_video_id(youtube_url)[source] Extract video id from common YT urls. Parameters youtube_url (str) – Return type str classmethod from_youtube_url(youtube_url, **kwargs)[source] Given youtube URL, load video. Parameters youtube_url (str) – kwargs (Any) – Return type langchain.document_loaders.youtube.YoutubeLoader load()[source] Load documents. Return type List[langchain.schema.Document]
https://api.python.langchain.com/en/latest/modules/document_loaders.html
30182965-8330-44d6-a813-300a47ec8ed1
Experimental This module contains experimental modules and reproductions of existing work using LangChain primitives. Autonomous agents Here, we document the BabyAGI and AutoGPT classes from the langchain.experimental module. class langchain.experimental.BabyAGI(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, task_list=None, task_creation_chain, task_prioritization_chain, execution_chain, task_id_counter=1, vectorstore, max_iterations=None)[source] Bases: langchain.chains.base.Chain, pydantic.main.BaseModel Controller model for the BabyAGI agent. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – task_list (collections.deque) – task_creation_chain (langchain.chains.base.Chain) – task_prioritization_chain (langchain.chains.base.Chain) – execution_chain (langchain.chains.base.Chain) – task_id_counter (int) – vectorstore (langchain.vectorstores.base.VectorStore) – max_iterations (Optional[int]) – Return type None model Config[source] Bases: object Configuration for this pydantic object. arbitrary_types_allowed = True property input_keys: List[str] Input keys this chain expects. property output_keys: List[str] Output keys this chain expects. get_next_task(result, task_description, objective)[source] Get the next task. Parameters result (str) – task_description (str) – objective (str) – Return type List[Dict] prioritize_tasks(this_task_id, objective)[source] Prioritize tasks. Parameters this_task_id (int) – objective (str) – Return type List[Dict] execute_task(objective, task, k=5)[source] Execute a task. Parameters objective (str) – task (str) – k (int) – Return type str classmethod from_llm(llm, vectorstore, verbose=False, task_execution_chain=None, **kwargs)[source] Initialize the BabyAGI Controller. Parameters llm (langchain.base_language.BaseLanguageModel) – vectorstore (langchain.vectorstores.base.VectorStore) – verbose (bool) – task_execution_chain (Optional[langchain.chains.base.Chain]) – kwargs (Dict[str, Any]) – Return type langchain.experimental.autonomous_agents.baby_agi.baby_agi.BabyAGI class langchain.experimental.AutoGPT(ai_name, memory, chain, output_parser, tools, feedback_tool=None, chat_history_memory=None)[source] Bases: object Agent class for interacting with Auto-GPT. Parameters ai_name (str) – memory (VectorStoreRetriever) – chain (LLMChain) – output_parser (BaseAutoGPTOutputParser) – tools (List[BaseTool]) – feedback_tool (Optional[HumanInputRun]) – chat_history_memory (Optional[BaseChatMessageHistory]) – Generative agents Here, we document the GenerativeAgent and GenerativeAgentMemory classes from the langchain.experimental module. class langchain.experimental.GenerativeAgent(*, name, age=None, traits='N/A', status, memory, llm, verbose=False, summary='', summary_refresh_seconds=3600, last_refreshed=None, daily_summaries=None)[source] Bases: pydantic.main.BaseModel A character with memory and innate characteristics. Parameters name (str) – age (Optional[int]) – traits (str) – status (str) – memory (langchain.experimental.generative_agents.memory.GenerativeAgentMemory) – llm (langchain.base_language.BaseLanguageModel) – verbose (bool) – summary (str) – summary_refresh_seconds (int) – last_refreshed (datetime.datetime) – daily_summaries (List[str]) – Return type None attribute name: str [Required] The character’s name. attribute age: Optional[int] = None The optional age of the character. attribute traits: str = 'N/A' Permanent traits to ascribe to the character. attribute status: str [Required] The traits of the character you wish not to change. attribute memory: langchain.experimental.generative_agents.memory.GenerativeAgentMemory [Required] The memory object that combines relevance, recency, and β€˜importance’. attribute llm: langchain.base_language.BaseLanguageModel [Required] The underlying language model. attribute summary: str = '' Stateful self-summary generated via reflection on the character’s memory. attribute summary_refresh_seconds: int = 3600 How frequently to re-generate the summary. attribute last_refreshed: datetime.datetime [Optional] The last time the character’s summary was regenerated. attribute daily_summaries: List[str] [Optional] Summary of the events in the plan that the agent took. model Config[source] Bases: object Configuration for this pydantic object. arbitrary_types_allowed = True summarize_related_memories(observation)[source] Summarize memories that are most relevant to an observation. Parameters observation (str) – Return type str generate_reaction(observation, now=None)[source] React to a given observation. Parameters observation (str) – now (Optional[datetime.datetime]) – Return type Tuple[bool, str] generate_dialogue_response(observation, now=None)[source] React to a given observation. Parameters observation (str) – now (Optional[datetime.datetime]) – Return type Tuple[bool, str] get_summary(force_refresh=False, now=None)[source] Return a descriptive summary of the agent. Parameters force_refresh (bool) – now (Optional[datetime.datetime]) – Return type str get_full_header(force_refresh=False, now=None)[source] Return a full header of the agent’s status, summary, and current time. Parameters force_refresh (bool) – now (Optional[datetime.datetime]) – Return type str class langchain.experimental.GenerativeAgentMemory(*, llm, memory_retriever, verbose=False, reflection_threshold=None, current_plan=[], importance_weight=0.15, aggregate_importance=0.0, max_tokens_limit=1200, queries_key='queries', most_recent_memories_token_key='recent_memories_token', add_memory_key='add_memory', relevant_memories_key='relevant_memories', relevant_memories_simple_key='relevant_memories_simple', most_recent_memories_key='most_recent_memories', now_key='now', reflecting=False)[source] Bases: langchain.schema.BaseMemory Parameters llm (langchain.base_language.BaseLanguageModel) – memory_retriever (langchain.retrievers.time_weighted_retriever.TimeWeightedVectorStoreRetriever) – verbose (bool) – reflection_threshold (Optional[float]) – current_plan (List[str]) – importance_weight (float) – aggregate_importance (float) – max_tokens_limit (int) – queries_key (str) – most_recent_memories_token_key (str) – add_memory_key (str) – relevant_memories_key (str) – relevant_memories_simple_key (str) – most_recent_memories_key (str) – now_key (str) – reflecting (bool) – Return type None attribute llm: langchain.base_language.BaseLanguageModel [Required] The core language model. attribute memory_retriever: langchain.retrievers.time_weighted_retriever.TimeWeightedVectorStoreRetriever [Required] The retriever to fetch related memories. attribute reflection_threshold: Optional[float] = None When aggregate_importance exceeds reflection_threshold, stop to reflect. attribute current_plan: List[str] = [] The current plan of the agent. attribute importance_weight: float = 0.15 How much weight to assign the memory importance. attribute aggregate_importance: float = 0.0 Track the sum of the β€˜importance’ of recent memories. Triggers reflection when it reaches reflection_threshold. pause_to_reflect(now=None)[source] Reflect on recent observations and generate β€˜insights’. Parameters now (Optional[datetime.datetime]) – Return type List[str] add_memories(memory_content, now=None)[source] Add an observations or memories to the agent’s memory. Parameters memory_content (str) – now (Optional[datetime.datetime]) – Return type List[str] add_memory(memory_content, now=None)[source] Add an observation or memory to the agent’s memory. Parameters memory_content (str) – now (Optional[datetime.datetime]) – Return type List[str] fetch_memories(observation, now=None)[source] Fetch related memories. Parameters observation (str) – now (Optional[datetime.datetime]) – Return type List[langchain.schema.Document] property memory_variables: List[str] Input keys this memory class will load dynamically. load_memory_variables(inputs)[source] Return key-value pairs given the text input to the chain. Parameters inputs (Dict[str, Any]) – Return type Dict[str, str] save_context(inputs, outputs)[source] Save the context of this model run to memory. Parameters inputs (Dict[str, Any]) – outputs (Dict[str, Any]) – Return type None clear()[source] Clear memory contents. Return type None
https://api.python.langchain.com/en/latest/modules/experimental.html
93f10def-2cda-45f9-a6d1-df7aedd7480d
Utilities General utilities. class langchain.utilities.ApifyWrapper(*, apify_client=None, apify_client_async=None)[source] Bases: pydantic.main.BaseModel Wrapper around Apify. To use, you should have the apify-client python package installed, and the environment variable APIFY_API_TOKEN set with your API key, or pass apify_api_token as a named parameter to the constructor. Parameters apify_client (Any) – apify_client_async (Any) – Return type None attribute apify_client: Any = None attribute apify_client_async: Any = None async acall_actor(actor_id, run_input, dataset_mapping_function, *, build=None, memory_mbytes=None, timeout_secs=None)[source] Run an Actor on the Apify platform and wait for results to be ready. Parameters actor_id (str) – The ID or name of the Actor on the Apify platform. run_input (Dict) – The input object of the Actor that you’re trying to run. dataset_mapping_function (Callable) – A function that takes a single dictionary (an Apify dataset item) and converts it to an instance of the Document class. build (str, optional) – Optionally specifies the actor build to run. It can be either a build tag or build number. memory_mbytes (int, optional) – Optional memory limit for the run, in megabytes. timeout_secs (int, optional) – Optional timeout for the run, in seconds. Returns A loader that will fetch the records from theActor run’s default dataset. Return type ApifyDatasetLoader call_actor(actor_id, run_input, dataset_mapping_function, *, build=None, memory_mbytes=None, timeout_secs=None)[source] Run an Actor on the Apify platform and wait for results to be ready. Parameters actor_id (str) – The ID or name of the Actor on the Apify platform. run_input (Dict) – The input object of the Actor that you’re trying to run. dataset_mapping_function (Callable) – A function that takes a single dictionary (an Apify dataset item) and converts it to an instance of the Document class. build (str, optional) – Optionally specifies the actor build to run. It can be either a build tag or build number. memory_mbytes (int, optional) – Optional memory limit for the run, in megabytes. timeout_secs (int, optional) – Optional timeout for the run, in seconds. Returns A loader that will fetch the records from theActor run’s default dataset. Return type ApifyDatasetLoader class langchain.utilities.ArxivAPIWrapper(*, arxiv_search=None, arxiv_exceptions=None, top_k_results=3, load_max_docs=100, load_all_available_meta=False, doc_content_chars_max=4000, ARXIV_MAX_QUERY_LENGTH=300)[source] Bases: pydantic.main.BaseModel Wrapper around ArxivAPI. To use, you should have the arxiv python package installed. https://lukasschwab.me/arxiv.py/index.html This wrapper will use the Arxiv API to conduct searches and fetch document summaries. By default, it will return the document summaries of the top-k results. It limits the Document content by doc_content_chars_max. Set doc_content_chars_max=None if you don’t want to limit the content size. Parameters top_k_results (int) – number of the top-scored document used for the arxiv tool ARXIV_MAX_QUERY_LENGTH (int) – the cut limit on the query used for the arxiv tool. load_max_docs (int) – a limit to the number of loaded documents load_all_available_meta (bool) – if True: the metadata of the loaded Documents gets all available meta info(see https://lukasschwab.me/arxiv.py/index.html#Result), if False: the metadata gets only the most informative fields. arxiv_search (Any) – arxiv_exceptions (Any) – doc_content_chars_max (Optional[int]) – Return type None attribute arxiv_exceptions: Any = None attribute doc_content_chars_max: Optional[int] = 4000 attribute load_all_available_meta: bool = False attribute load_max_docs: int = 100 attribute top_k_results: int = 3 load(query)[source] Run Arxiv search and get the article texts plus the article meta information. See https://lukasschwab.me/arxiv.py/index.html#Search Returns: a list of documents with the document.page_content in text format Parameters query (str) – Return type List[langchain.schema.Document] run(query)[source] Run Arxiv search and get the article meta information. See https://lukasschwab.me/arxiv.py/index.html#Search See https://lukasschwab.me/arxiv.py/index.html#Result It uses only the most informative fields of article meta information. Parameters query (str) – Return type str class langchain.utilities.BashProcess(strip_newlines=False, return_err_output=False, persistent=False)[source] Bases: object Executes bash commands and returns the output. Parameters strip_newlines (bool) – return_err_output (bool) – persistent (bool) – run(commands)[source] Run commands and return final output. Parameters commands (Union[str, List[str]]) – Return type str process_output(output, command)[source] Parameters output (str) – command (str) – Return type str class langchain.utilities.BibtexparserWrapper[source] Bases: pydantic.main.BaseModel Wrapper around bibtexparser. To use, you should have the bibtexparser python package installed. https://bibtexparser.readthedocs.io/en/master/ This wrapper will use bibtexparser to load a collection of references from a bibtex file and fetch document summaries. Return type None get_metadata(entry, load_extra=False)[source] Get metadata for the given entry. Parameters entry (Mapping[str, Any]) – load_extra (bool) – Return type Dict[str, Any] load_bibtex_entries(path)[source] Load bibtex entries from the bibtex file at the given path. Parameters path (str) – Return type List[Dict[str, Any]] class langchain.utilities.BingSearchAPIWrapper(*, bing_subscription_key, bing_search_url, k=10)[source] Bases: pydantic.main.BaseModel Wrapper for Bing Search API. In order to set this up, follow instructions at: https://levelup.gitconnected.com/api-tutorial-how-to-use-bing-web-search-api-in-python-4165d5592a7e Parameters bing_subscription_key (str) – bing_search_url (str) – k (int) – Return type None attribute bing_search_url: str [Required] attribute bing_subscription_key: str [Required] attribute k: int = 10 results(query, num_results)[source] Run query through BingSearch and return metadata. Parameters query (str) – The query to search for. num_results (int) – The number of results to return. Returns snippet - The description of the result. title - The title of the result. link - The link to the result. Return type A list of dictionaries with the following keys run(query)[source] Run query through BingSearch and parse result. Parameters query (str) – Return type str class langchain.utilities.BraveSearchWrapper(*, api_key, search_kwargs=None)[source] Bases: pydantic.main.BaseModel Parameters api_key (str) – search_kwargs (dict) – Return type None attribute api_key: str [Required] attribute search_kwargs: dict [Optional] run(query)[source] Parameters query (str) – Return type str class langchain.utilities.DuckDuckGoSearchAPIWrapper(*, k=10, region='wt-wt', safesearch='moderate', time='y', max_results=5)[source] Bases: pydantic.main.BaseModel Wrapper for DuckDuckGo Search API. Free and does not require any setup Parameters k (int) – region (Optional[str]) – safesearch (str) – time (Optional[str]) – max_results (int) – Return type None attribute k: int = 10 attribute max_results: int = 5 attribute region: Optional[str] = 'wt-wt' attribute safesearch: str = 'moderate' attribute time: Optional[str] = 'y' get_snippets(query)[source] Run query through DuckDuckGo and return concatenated results. Parameters query (str) – Return type List[str] results(query, num_results)[source] Run query through DuckDuckGo and return metadata. Parameters query (str) – The query to search for. num_results (int) – The number of results to return. Returns snippet - The description of the result. title - The title of the result. link - The link to the result. Return type A list of dictionaries with the following keys run(query)[source] Parameters query (str) – Return type str class langchain.utilities.GooglePlacesAPIWrapper(*, gplaces_api_key=None, google_map_client=None, top_k_results=None)[source] Bases: pydantic.main.BaseModel Wrapper around Google Places API. To use, you should have the googlemaps python package installed,an API key for the google maps platform, and the enviroment variable β€˜β€™GPLACES_API_KEY’’ set with your API key , or pass β€˜gplaces_api_key’ as a named parameter to the constructor. By default, this will return the all the results on the input query.You can use the top_k_results argument to limit the number of results. Example from langchain import GooglePlacesAPIWrapper gplaceapi = GooglePlacesAPIWrapper() Parameters gplaces_api_key (Optional[str]) – google_map_client (Any) – top_k_results (Optional[int]) – Return type None attribute gplaces_api_key: Optional[str] = None attribute top_k_results: Optional[int] = None fetch_place_details(place_id)[source] Parameters place_id (str) – Return type Optional[str] format_place_details(place_details)[source] Parameters place_details (Dict[str, Any]) – Return type Optional[str] run(query)[source] Run Places search and get k number of places that exists that match. Parameters query (str) – Return type str class langchain.utilities.GoogleSearchAPIWrapper(*, search_engine=None, google_api_key=None, google_cse_id=None, k=10, siterestrict=False)[source] Bases: pydantic.main.BaseModel Wrapper for Google Search API. Adapted from: Instructions adapted from https://stackoverflow.com/questions/ 37083058/ programmatically-searching-google-in-python-using-custom-search TODO: DOCS for using it 1. Install google-api-python-client - If you don’t already have a Google account, sign up. - If you have never created a Google APIs Console project, read the Managing Projects page and create a project in the Google API Console. - Install the library using pip install google-api-python-client The current version of the library is 2.70.0 at this time 2. To create an API key: - Navigate to the APIs & Servicesβ†’Credentials panel in Cloud Console. - Select Create credentials, then select API key from the drop-down menu. - The API key created dialog box displays your newly created key. - You now have an API_KEY 3. Setup Custom Search Engine so you can search the entire web - Create a custom search engine in this link. - In Sites to search, add any valid URL (i.e. www.stackoverflow.com). - That’s all you have to fill up, the rest doesn’t matter. In the left-side menu, click Edit search engine β†’ {your search engine name} β†’ Setup Set Search the entire web to ON. Remove the URL you added from the list of Sites to search. - Under Search engine ID you’ll find the search-engine-ID. 4. Enable the Custom Search API - Navigate to the APIs & Servicesβ†’Dashboard panel in Cloud Console. - Click Enable APIs and Services. - Search for Custom Search API and click on it. - Click Enable. URL for it: https://console.cloud.google.com/apis/library/customsearch.googleapis .com Parameters search_engine (Any) – google_api_key (Optional[str]) – google_cse_id (Optional[str]) – k (int) – siterestrict (bool) – Return type None attribute google_api_key: Optional[str] = None attribute google_cse_id: Optional[str] = None attribute k: int = 10 attribute siterestrict: bool = False results(query, num_results)[source] Run query through GoogleSearch and return metadata. Parameters query (str) – The query to search for. num_results (int) – The number of results to return. Returns snippet - The description of the result. title - The title of the result. link - The link to the result. Return type A list of dictionaries with the following keys run(query)[source] Run query through GoogleSearch and parse result. Parameters query (str) – Return type str class langchain.utilities.GoogleSerperAPIWrapper(*, k=10, gl='us', hl='en', type='search', tbs=None, serper_api_key=None, aiosession=None, result_key_for_type={'images': 'images', 'news': 'news', 'places': 'places', 'search': 'organic'})[source] Bases: pydantic.main.BaseModel Wrapper around the Serper.dev Google Search API. You can create a free API key at https://serper.dev. To use, you should have the environment variable SERPER_API_KEY set with your API key, or pass serper_api_key as a named parameter to the constructor. Example from langchain import GoogleSerperAPIWrapper google_serper = GoogleSerperAPIWrapper() Parameters k (int) – gl (str) – hl (str) – type (Literal['news', 'search', 'places', 'images']) – tbs (Optional[str]) – serper_api_key (Optional[str]) – aiosession (Optional[aiohttp.client.ClientSession]) – result_key_for_type (dict) – Return type None attribute aiosession: Optional[aiohttp.client.ClientSession] = None attribute gl: str = 'us' attribute hl: str = 'en' attribute k: int = 10 attribute serper_api_key: Optional[str] = None attribute tbs: Optional[str] = None attribute type: Literal['news', 'search', 'places', 'images'] = 'search' async aresults(query, **kwargs)[source] Run query through GoogleSearch. Parameters query (str) – kwargs (Any) – Return type Dict async arun(query, **kwargs)[source] Run query through GoogleSearch and parse result async. Parameters query (str) – kwargs (Any) – Return type str results(query, **kwargs)[source] Run query through GoogleSearch. Parameters query (str) – kwargs (Any) – Return type Dict run(query, **kwargs)[source] Run query through GoogleSearch and parse result. Parameters query (str) – kwargs (Any) – Return type str class langchain.utilities.GraphQLAPIWrapper(*, custom_headers=None, graphql_endpoint, gql_client=None, gql_function)[source] Bases: pydantic.main.BaseModel Wrapper around GraphQL API. To use, you should have the gql python package installed. This wrapper will use the GraphQL API to conduct queries. Parameters custom_headers (Optional[Dict[str, str]]) – graphql_endpoint (str) – gql_client (Any) – gql_function (Callable[[str], Any]) – Return type None attribute custom_headers: Optional[Dict[str, str]] = None attribute graphql_endpoint: str [Required] run(query)[source] Run a GraphQL query and get the results. Parameters query (str) – Return type str class langchain.utilities.JiraAPIWrapper(*, jira=None, confluence=None, jira_username=None, jira_api_token=None, jira_instance_url=None, operations=[{'mode': 'jql', 'name': 'JQL Query', 'description': '\nΒ Β Β  This tool is a wrapper around atlassian-python-api\'s Jira jql API, useful when you need to search for Jira issues.\nΒ Β Β  The input to this tool is a JQL query string, and will be passed into atlassian-python-api\'s Jira `jql` function,\nΒ Β Β  For example, to find all the issues in project "Test" assigned to the me, you would pass in the following string:\nΒ Β Β  project = Test AND assignee = currentUser()\nΒ Β Β  or to find issues with summaries that contain the word "test", you would pass in the following string:\nΒ Β Β  summary ~ \'test\'\nΒ Β Β  '}, {'mode': 'get_projects', 'name': 'Get Projects', 'description': "\nΒ Β Β  This tool is a wrapper around atlassian-python-api's Jira project API, \nΒ Β Β  useful when you need to fetch all the projects the user has access to, find out how many projects there are, or as an intermediary step that involv searching by projects. \nΒ Β Β  there is no input to this tool.\nΒ Β Β  "}, {'mode': 'create_issue', 'name': 'Create Issue', 'description': '\nΒ Β Β  This tool is a wrapper around atlassian-python-api\'s Jira issue_create API, useful when you need to create a Jira issue. \nΒ Β Β  The input to this tool is a dictionary specifying the fields of the Jira issue, and will be passed into atlassian-python-api\'s Jira `issue_create` function.\nΒ Β Β  For example, to create a low priority task called "test issue" with description "test description", you would pass in the following dictionary: \nΒ Β Β  {{"summary": "test issue", "description": "test description", "issuetype": {{"name": "Task"}}, "priority": {{"name": "Low"}}}}\nΒ Β Β  '}, {'mode': 'other', 'name': 'Catch all Jira API call', 'description': '\nΒ Β Β  This tool is a wrapper around atlassian-python-api\'s Jira API.\nΒ Β Β  There are other dedicated tools for fetching all projects, and creating and searching for issues, \nΒ Β Β  use this tool if you need to perform any other actions allowed by the atlassian-python-api Jira API.\nΒ Β Β  The input to this tool is line of python code that calls a function from atlassian-python-api\'s Jira API\nΒ Β Β  For example, to update the summary field of an issue, you would pass in the following string:\nΒ Β Β  self.jira.update_issue_field(key, {{"summary": "New summary"}})\nΒ Β Β  or to find out how many projects are in the Jira instance, you would pass in the following string:\nΒ Β Β  self.jira.projects()\nΒ Β Β  For more information on the Jira API, refer to https://atlassian-python-api.readthedocs.io/jira.html\nΒ Β Β  '}, {'mode': 'create_page', 'name': 'Create confluence page', 'description': 'This tool is a wrapper around atlassian-python-api\'s Confluence \natlassian-python-api API, useful when you need to create a Confluence page. The input to this tool is a dictionary \nspecifying the fields of the Confluence page, and will be passed into atlassian-python-api\'s Confluence `create_page` \nfunction. For example, to create a page in the DEMO space titled "This is the title" with body "This is the body. You can use \n<strong>HTML tags</strong>!", you would pass in the following dictionary: {{"space": "DEMO", "title":"This is the \ntitle","body":"This is the body. You can use <strong>HTML tags</strong>!"}} '}])[source] Bases: pydantic.main.BaseModel Wrapper for Jira API. Parameters jira (Any) – confluence (Any) – jira_username (Optional[str]) – jira_api_token (Optional[str]) – jira_instance_url (Optional[str]) – operations (List[Dict]) – Return type None attribute confluence: Any = None attribute jira_api_token: Optional[str] = None attribute jira_instance_url: Optional[str] = None attribute jira_username: Optional[str] = None attribute operations: List[Dict] = [{'mode': 'jql', 'name': 'JQL Query', 'description': '\nΒ Β Β  This tool is a wrapper around atlassian-python-api\'s Jira jql API, useful when you need to search for Jira issues.\nΒ Β Β  The input to this tool is a JQL query string, and will be passed into atlassian-python-api\'s Jira `jql` function,\nΒ Β Β  For example, to find all the issues in project "Test" assigned to the me, you would pass in the following string:\nΒ Β Β  project = Test AND assignee = currentUser()\nΒ Β Β  or to find issues with summaries that contain the word "test", you would pass in the following string:\nΒ Β Β  summary ~ \'test\'\nΒ Β Β  '}, {'mode': 'get_projects', 'name': 'Get Projects', 'description': "\nΒ Β Β  This tool is a wrapper around atlassian-python-api's Jira project API, \nΒ Β Β  useful when you need to fetch all the projects the user has access to, find out how many projects there are, or as an intermediary step that involv searching by projects. \nΒ Β Β  there is no input to this tool.\nΒ Β Β  "}, {'mode': 'create_issue', 'name': 'Create Issue', 'description': '\nΒ Β Β  This tool is a wrapper around atlassian-python-api\'s Jira issue_create API, useful when you need to create a Jira issue. \nΒ Β Β  The input to this tool is a dictionary specifying the fields of the Jira issue, and will be passed into atlassian-python-api\'s Jira `issue_create` function.\nΒ Β Β  For example, to create a low priority task called "test issue" with description "test description", you would pass in the following dictionary: \nΒ Β Β  {{"summary": "test issue", "description": "test description", "issuetype": {{"name": "Task"}}, "priority": {{"name": "Low"}}}}\nΒ Β Β  '}, {'mode': 'other', 'name': 'Catch all Jira API call', 'description': '\nΒ Β Β  This tool is a wrapper around atlassian-python-api\'s Jira API.\nΒ Β Β  There are other dedicated tools for fetching all projects, and creating and searching for issues, \nΒ Β Β  use this tool if you need to perform any other actions allowed by the atlassian-python-api Jira API.\nΒ Β Β  The input to this tool is line of python code that calls a function from atlassian-python-api\'s Jira API\nΒ Β Β  For example, to update the summary field of an issue, you would pass in the following string:\nΒ Β Β  self.jira.update_issue_field(key, {{"summary": "New summary"}})\nΒ Β Β  or to find out how many projects are in the Jira instance, you would pass in the following string:\nΒ Β Β  self.jira.projects()\nΒ Β Β  For more information on the Jira API, refer to https://atlassian-python-api.readthedocs.io/jira.html\nΒ Β Β  '}, {'mode': 'create_page', 'name': 'Create confluence page', 'description': 'This tool is a wrapper around atlassian-python-api\'s Confluence \natlassian-python-api API, useful when you need to create a Confluence page. The input to this tool is a dictionary \nspecifying the fields of the Confluence page, and will be passed into atlassian-python-api\'s Confluence `create_page` \nfunction. For example, to create a page in the DEMO space titled "This is the title" with body "This is the body. You can use \n<strong>HTML tags</strong>!", you would pass in the following dictionary: {{"space": "DEMO", "title":"This is the \ntitle","body":"This is the body. You can use <strong>HTML tags</strong>!"}} '}] issue_create(query)[source] Parameters query (str) – Return type str list()[source] Return type List[Dict] other(query)[source] Parameters query (str) – Return type str page_create(query)[source] Parameters query (str) – Return type str parse_issues(issues)[source] Parameters issues (Dict) – Return type List[dict] parse_projects(projects)[source] Parameters projects (List[dict]) – Return type List[dict] project()[source] Return type str run(mode, query)[source] Parameters mode (str) – query (str) – Return type str search(query)[source] Parameters query (str) – Return type str class langchain.utilities.LambdaWrapper(*, lambda_client=None, function_name=None, awslambda_tool_name=None, awslambda_tool_description=None)[source] Bases: pydantic.main.BaseModel Wrapper for AWS Lambda SDK. Docs for using: pip install boto3 Create a lambda function using the AWS Console or CLI Run aws configure and enter your AWS credentials Parameters lambda_client (Any) – function_name (Optional[str]) – awslambda_tool_name (Optional[str]) – awslambda_tool_description (Optional[str]) – Return type None attribute awslambda_tool_description: Optional[str] = None attribute awslambda_tool_name: Optional[str] = None attribute function_name: Optional[str] = None run(query)[source] Invoke Lambda function and parse result. Parameters query (str) – Return type str class langchain.utilities.MaxComputeAPIWrapper(client)[source] Bases: object Interface for querying Alibaba Cloud MaxCompute tables. Parameters client (ODPS) – classmethod from_params(endpoint, project, *, access_id=None, secret_access_key=None)[source] Convenience constructor that builds the odsp.ODPS MaxCompute client fromgiven parameters. Parameters endpoint (str) – MaxCompute endpoint. project (str) – A project is a basic organizational unit of MaxCompute, which is similar to a database. access_id (Optional[str]) – MaxCompute access ID. Should be passed in directly or set as the environment variable MAX_COMPUTE_ACCESS_ID. secret_access_key (Optional[str]) – MaxCompute secret access key. Should be passed in directly or set as the environment variable MAX_COMPUTE_SECRET_ACCESS_KEY. Return type langchain.utilities.max_compute.MaxComputeAPIWrapper lazy_query(query)[source] Parameters query (str) – Return type Iterator[dict] query(query)[source] Parameters query (str) – Return type List[dict] class langchain.utilities.MetaphorSearchAPIWrapper(*, metaphor_api_key, k=10)[source] Bases: pydantic.main.BaseModel Wrapper for Metaphor Search API. Parameters metaphor_api_key (str) – k (int) – Return type None attribute k: int = 10 attribute metaphor_api_key: str [Required] results(query, num_results, include_domains=None, exclude_domains=None, start_crawl_date=None, end_crawl_date=None, start_published_date=None, end_published_date=None)[source] Run query through Metaphor Search and return metadata. Parameters query (str) – The query to search for. num_results (int) – The number of results to return. include_domains (Optional[List[str]]) – exclude_domains (Optional[List[str]]) – start_crawl_date (Optional[str]) – end_crawl_date (Optional[str]) – start_published_date (Optional[str]) – end_published_date (Optional[str]) – Returns title - The title of the url - The url author - Author of the content, if applicable. Otherwise, None. published_date - Estimated date published in YYYY-MM-DD format. Otherwise, None. Return type A list of dictionaries with the following keys async results_async(query, num_results, include_domains=None, exclude_domains=None, start_crawl_date=None, end_crawl_date=None, start_published_date=None, end_published_date=None)[source] Get results from the Metaphor Search API asynchronously. Parameters query (str) – num_results (int) – include_domains (Optional[List[str]]) – exclude_domains (Optional[List[str]]) – start_crawl_date (Optional[str]) – end_crawl_date (Optional[str]) – start_published_date (Optional[str]) – end_published_date (Optional[str]) – Return type List[Dict] class langchain.utilities.OpenWeatherMapAPIWrapper(*, owm=None, openweathermap_api_key=None)[source] Bases: pydantic.main.BaseModel Wrapper for OpenWeatherMap API using PyOWM. Docs for using: Go to OpenWeatherMap and sign up for an API key Save your API KEY into OPENWEATHERMAP_API_KEY env variable pip install pyowm Parameters owm (Any) – openweathermap_api_key (Optional[str]) – Return type None attribute openweathermap_api_key: Optional[str] = None attribute owm: Any = None run(location)[source] Get the current weather information for a specified location. Parameters location (str) – Return type str class langchain.utilities.PowerBIDataset(*, dataset_id, table_names, group_id=None, credential=None, token=None, impersonated_user_name=None, sample_rows_in_table_info=1, schemas=None, aiosession=None)[source] Bases: pydantic.main.BaseModel Create PowerBI engine from dataset ID and credential or token. Use either the credential or a supplied token to authenticate. If both are supplied the credential is used to generate a token. The impersonated_user_name is the UPN of a user to be impersonated. If the model is not RLS enabled, this will be ignored. Parameters dataset_id (str) – table_names (List[str]) – group_id (Optional[str]) – credential (Optional[TokenCredential]) – token (Optional[str]) – impersonated_user_name (Optional[str]) – sample_rows_in_table_info (langchain.utilities.powerbi.ConstrainedIntValue) – schemas (Dict[str, str]) – aiosession (Optional[aiohttp.client.ClientSession]) – Return type None attribute aiosession: Optional[aiohttp.ClientSession] = None attribute credential: Optional[TokenCredential] = None attribute dataset_id: str [Required] attribute group_id: Optional[str] = None attribute impersonated_user_name: Optional[str] = None attribute sample_rows_in_table_info: int = 1 Constraints exclusiveMinimum = 0 maximum = 10 attribute schemas: Dict[str, str] [Optional] attribute table_names: List[str] [Required] attribute token: Optional[str] = None async aget_table_info(table_names=None)[source] Get information about specified tables. Parameters table_names (Optional[Union[List[str], str]]) – Return type str async arun(command)[source] Execute a DAX command and return the result asynchronously. Parameters command (str) – Return type Any get_schemas()[source] Get the available schema’s. Return type str get_table_info(table_names=None)[source] Get information about specified tables. Parameters table_names (Optional[Union[List[str], str]]) – Return type str get_table_names()[source] Get names of tables available. Return type Iterable[str] run(command)[source] Execute a DAX command and return a json representing the results. Parameters command (str) – Return type Any property headers: Dict[str, str] Get the token. property request_url: str Get the request url. property table_info: str Information about all tables in the database. class langchain.utilities.PubMedAPIWrapper(*, top_k_results=3, load_max_docs=25, doc_content_chars_max=2000, load_all_available_meta=False, email='your_email@example.com', base_url_esearch='https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?', base_url_efetch='https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?', max_retry=5, sleep_time=0.2, ARXIV_MAX_QUERY_LENGTH=300)[source] Bases: pydantic.main.BaseModel Wrapper around PubMed API. This wrapper will use the PubMed API to conduct searches and fetch document summaries. By default, it will return the document summaries of the top-k results of an input search. Parameters top_k_results (int) – number of the top-scored document used for the PubMed tool load_max_docs (int) – a limit to the number of loaded documents load_all_available_meta (bool) – if True: the metadata of the loaded Documents gets all available meta info(see https://www.ncbi.nlm.nih.gov/books/NBK25499/#chapter4.ESearch) if False: the metadata gets only the most informative fields. doc_content_chars_max (int) – email (str) – base_url_esearch (str) – base_url_efetch (str) – max_retry (int) – sleep_time (float) – ARXIV_MAX_QUERY_LENGTH (int) – Return type None attribute doc_content_chars_max: int = 2000 attribute email: str = 'your_email@example.com' attribute load_all_available_meta: bool = False attribute load_max_docs: int = 25 attribute top_k_results: int = 3 load(query)[source] Search PubMed for documents matching the query. Return a list of dictionaries containing the document metadata. Parameters query (str) – Return type List[dict] load_docs(query)[source] Parameters query (str) – Return type List[langchain.schema.Document] retrieve_article(uid, webenv)[source] Parameters uid (str) – webenv (str) – Return type dict run(query)[source] Run PubMed search and get the article meta information. See https://www.ncbi.nlm.nih.gov/books/NBK25499/#chapter4.ESearch It uses only the most informative fields of article meta information. Parameters query (str) – Return type str class langchain.utilities.PythonREPL(*, _globals=None, _locals=None)[source] Bases: pydantic.main.BaseModel Simulates a standalone Python REPL. Parameters _globals (Optional[Dict]) – _locals (Optional[Dict]) – Return type None attribute globals: Optional[Dict] [Optional] (alias '_globals') attribute locals: Optional[Dict] [Optional] (alias '_locals') run(command)[source] Run command with own globals/locals and returns anything printed. Parameters command (str) – Return type str pydantic settings langchain.utilities.SceneXplainAPIWrapper[source] Bases: pydantic.env_settings.BaseSettings, pydantic.main.BaseModel Wrapper for SceneXplain API. In order to set this up, you need API key for the SceneXplain API. You can obtain a key by following the steps below. - Sign up for a free account at https://scenex.jina.ai/. - Navigate to the API Access page (https://scenex.jina.ai/api) and create a new API key. Show JSON schema{ "title": "SceneXplainAPIWrapper", "description": "Wrapper for SceneXplain API.\n\nIn order to set this up, you need API key for the SceneXplain API.\nYou can obtain a key by following the steps below.\n- Sign up for a free account at https://scenex.jina.ai/.\n- Navigate to the API Access page (https://scenex.jina.ai/api)\n and create a new API key.", "type": "object", "properties": { "scenex_api_key": { "title": "Scenex Api Key", "env": "SCENEX_API_KEY", "env_names": "{'scenex_api_key'}", "type": "string" }, "scenex_api_url": { "title": "Scenex Api Url", "default": "https://us-central1-causal-diffusion.cloudfunctions.net/describe", "env_names": "{'scenex_api_url'}", "type": "string" } }, "required": [ "scenex_api_key" ], "additionalProperties": false } Fields scenex_api_key (str) scenex_api_url (str) attribute scenex_api_key: str [Required] attribute scenex_api_url: str = 'https://us-central1-causal-diffusion.cloudfunctions.net/describe' run(image)[source] Run SceneXplain image explainer. Parameters image (str) – Return type str validator validate_environmentΒ  »  all fields[source] Validate that api key exists in environment. Parameters values (Dict) – Return type Dict class langchain.utilities.SearxSearchWrapper(*, searx_host='', unsecure=False, params=None, headers=None, engines=[], categories=[], query_suffix='', k=10, aiosession=None)[source] Bases: pydantic.main.BaseModel Wrapper for Searx API. To use you need to provide the searx host by passing the named parameter searx_host or exporting the environment variable SEARX_HOST. In some situations you might want to disable SSL verification, for example if you are running searx locally. You can do this by passing the named parameter unsecure. You can also pass the host url scheme as http to disable SSL. Example from langchain.utilities import SearxSearchWrapper searx = SearxSearchWrapper(searx_host="http://localhost:8888") Example with SSL disabled:from langchain.utilities import SearxSearchWrapper # note the unsecure parameter is not needed if you pass the url scheme as # http searx = SearxSearchWrapper(searx_host="http://localhost:8888", unsecure=True) Parameters searx_host (str) – unsecure (bool) – params (dict) – headers (Optional[dict]) – engines (Optional[List[str]]) – categories (Optional[List[str]]) – query_suffix (Optional[str]) – k (int) – aiosession (Optional[Any]) – Return type None attribute aiosession: Optional[Any] = None attribute categories: Optional[List[str]] = [] attribute engines: Optional[List[str]] = [] attribute headers: Optional[dict] = None attribute k: int = 10 attribute params: dict [Optional] attribute query_suffix: Optional[str] = '' attribute searx_host: str = '' attribute unsecure: bool = False async aresults(query, num_results, engines=None, query_suffix='', **kwargs)[source] Asynchronously query with json results. Uses aiohttp. See results for more info. Parameters query (str) – num_results (int) – engines (Optional[List[str]]) – query_suffix (Optional[str]) – kwargs (Any) – Return type List[Dict] async arun(query, engines=None, query_suffix='', **kwargs)[source] Asynchronously version of run. Parameters query (str) – engines (Optional[List[str]]) – query_suffix (Optional[str]) – kwargs (Any) – Return type str results(query, num_results, engines=None, categories=None, query_suffix='', **kwargs)[source] Run query through Searx API and returns the results with metadata. Parameters query (str) – The query to search for. query_suffix (Optional[str]) – Extra suffix appended to the query. num_results (int) – Limit the number of results to return. engines (Optional[List[str]]) – List of engines to use for the query. categories (Optional[List[str]]) – List of categories to use for the query. **kwargs – extra parameters to pass to the searx API. kwargs (Any) – Returns {snippet: The description of the result. title: The title of the result. link: The link to the result. engines: The engines used for the result. category: Searx category of the result. } Return type Dict with the following keys run(query, engines=None, categories=None, query_suffix='', **kwargs)[source] Run query through Searx API and parse results. You can pass any other params to the searx query API. Parameters query (str) – The query to search for. query_suffix (Optional[str]) – Extra suffix appended to the query. engines (Optional[List[str]]) – List of engines to use for the query. categories (Optional[List[str]]) – List of categories to use for the query. **kwargs – extra parameters to pass to the searx API. kwargs (Any) – Returns The result of the query. Return type str Raises ValueError – If an error occured with the query. Example This will make a query to the qwant engine: from langchain.utilities import SearxSearchWrapper searx = SearxSearchWrapper(searx_host="http://my.searx.host") searx.run("what is the weather in France ?", engine="qwant") # the same result can be achieved using the `!` syntax of searx # to select the engine using `query_suffix` searx.run("what is the weather in France ?", query_suffix="!qwant") class langchain.utilities.SerpAPIWrapper(*, search_engine=None, params={'engine': 'google', 'gl': 'us', 'google_domain': 'google.com', 'hl': 'en'}, serpapi_api_key=None, aiosession=None)[source] Bases: pydantic.main.BaseModel Wrapper around SerpAPI. To use, you should have the google-search-results python package installed, and the environment variable SERPAPI_API_KEY set with your API key, or pass serpapi_api_key as a named parameter to the constructor. Example from langchain import SerpAPIWrapper serpapi = SerpAPIWrapper() Parameters search_engine (Any) – params (dict) – serpapi_api_key (Optional[str]) – aiosession (Optional[aiohttp.client.ClientSession]) – Return type None attribute aiosession: Optional[aiohttp.client.ClientSession] = None attribute params: dict = {'engine': 'google', 'gl': 'us', 'google_domain': 'google.com', 'hl': 'en'} attribute serpapi_api_key: Optional[str] = None async aresults(query)[source] Use aiohttp to run query through SerpAPI and return the results async. Parameters query (str) – Return type dict async arun(query, **kwargs)[source] Run query through SerpAPI and parse result async. Parameters query (str) – kwargs (Any) – Return type str get_params(query)[source] Get parameters for SerpAPI. Parameters query (str) – Return type Dict[str, str] results(query)[source] Run query through SerpAPI and return the raw result. Parameters query (str) – Return type dict run(query, **kwargs)[source] Run query through SerpAPI and parse result. Parameters query (str) – kwargs (Any) – Return type str class langchain.utilities.SparkSQL(spark_session=None, catalog=None, schema=None, ignore_tables=None, include_tables=None, sample_rows_in_table_info=3)[source] Bases: object Parameters spark_session (Optional[SparkSession]) – catalog (Optional[str]) – schema (Optional[str]) – ignore_tables (Optional[List[str]]) – include_tables (Optional[List[str]]) – sample_rows_in_table_info (int) – classmethod from_uri(database_uri, engine_args=None, **kwargs)[source] Creating a remote Spark Session via Spark connect. For example: SparkSQL.from_uri(β€œsc://localhost:15002”) Parameters database_uri (str) – engine_args (Optional[dict]) – kwargs (Any) – Return type langchain.utilities.spark_sql.SparkSQL get_usable_table_names()[source] Get names of tables available. Return type Iterable[str] get_table_info(table_names=None)[source] Parameters table_names (Optional[List[str]]) – Return type str run(command, fetch='all')[source] Parameters command (str) – fetch (str) – Return type str get_table_info_no_throw(table_names=None)[source] Get information about specified tables. Follows best practices as specified in: Rajkumar et al, 2022 (https://arxiv.org/abs/2204.00498) If sample_rows_in_table_info, the specified number of sample rows will be appended to each table description. This can increase performance as demonstrated in the paper. Parameters table_names (Optional[List[str]]) – Return type str run_no_throw(command, fetch='all')[source] Execute a SQL command and return a string representing the results. If the statement returns rows, a string of the results is returned. If the statement returns no rows, an empty string is returned. If the statement throws an error, the error message is returned. Parameters command (str) – fetch (str) – Return type str class langchain.utilities.TextRequestsWrapper(*, headers=None, aiosession=None)[source] Bases: pydantic.main.BaseModel Lightweight wrapper around requests library. The main purpose of this wrapper is to always return a text output. Parameters headers (Optional[Dict[str, str]]) – aiosession (Optional[aiohttp.client.ClientSession]) – Return type None attribute aiosession: Optional[aiohttp.client.ClientSession] = None attribute headers: Optional[Dict[str, str]] = None async adelete(url, **kwargs)[source] DELETE the URL and return the text asynchronously. Parameters url (str) – kwargs (Any) – Return type str async aget(url, **kwargs)[source] GET the URL and return the text asynchronously. Parameters url (str) – kwargs (Any) – Return type str async apatch(url, data, **kwargs)[source] PATCH the URL and return the text asynchronously. Parameters url (str) – data (Dict[str, Any]) – kwargs (Any) – Return type str async apost(url, data, **kwargs)[source] POST to the URL and return the text asynchronously. Parameters url (str) – data (Dict[str, Any]) – kwargs (Any) – Return type str async aput(url, data, **kwargs)[source] PUT the URL and return the text asynchronously. Parameters url (str) – data (Dict[str, Any]) – kwargs (Any) – Return type str delete(url, **kwargs)[source] DELETE the URL and return the text. Parameters url (str) – kwargs (Any) – Return type str get(url, **kwargs)[source] GET the URL and return the text. Parameters url (str) – kwargs (Any) – Return type str patch(url, data, **kwargs)[source] PATCH the URL and return the text. Parameters url (str) – data (Dict[str, Any]) – kwargs (Any) – Return type str post(url, data, **kwargs)[source] POST to the URL and return the text. Parameters url (str) – data (Dict[str, Any]) – kwargs (Any) – Return type str put(url, data, **kwargs)[source] PUT the URL and return the text. Parameters url (str) – data (Dict[str, Any]) – kwargs (Any) – Return type str property requests: langchain.requests.Requests class langchain.utilities.TwilioAPIWrapper(*, client=None, account_sid=None, auth_token=None, from_number=None)[source] Bases: pydantic.main.BaseModel Messaging Client using Twilio. To use, you should have the twilio python package installed, and the environment variables TWILIO_ACCOUNT_SID, TWILIO_AUTH_TOKEN, and TWILIO_FROM_NUMBER, or pass account_sid, auth_token, and from_number as named parameters to the constructor. Example from langchain.utilities.twilio import TwilioAPIWrapper twilio = TwilioAPIWrapper( account_sid="ACxxx", auth_token="xxx", from_number="+10123456789" ) twilio.run('test', '+12484345508') Parameters client (Any) – account_sid (Optional[str]) – auth_token (Optional[str]) – from_number (Optional[str]) – Return type None attribute account_sid: Optional[str] = None Twilio account string identifier. attribute auth_token: Optional[str] = None Twilio auth token. attribute from_number: Optional[str] = None A Twilio phone number in [E.164](https://www.twilio.com/docs/glossary/what-e164) format, an [alphanumeric sender ID](https://www.twilio.com/docs/sms/send-messages#use-an-alphanumeric-sender-id), or a [Channel Endpoint address](https://www.twilio.com/docs/sms/channels#channel-addresses) that is enabled for the type of message you want to send. Phone numbers or [short codes](https://www.twilio.com/docs/sms/api/short-code) purchased from Twilio also work here. You cannot, for example, spoof messages from a private cell phone number. If you are using messaging_service_sid, this parameter must be empty. run(body, to)[source] Run body through Twilio and respond with message sid. Parameters body (str) – The text of the message you want to send. Can be up to 1,600 characters in length. to (str) – The destination phone number in [E.164](https://www.twilio.com/docs/glossary/what-e164) format for SMS/MMS or [Channel user address](https://www.twilio.com/docs/sms/channels#channel-addresses) for other 3rd-party channels. Return type str class langchain.utilities.WikipediaAPIWrapper(*, wiki_client=None, top_k_results=3, lang='en', load_all_available_meta=False, doc_content_chars_max=4000)[source] Bases: pydantic.main.BaseModel Wrapper around WikipediaAPI. To use, you should have the wikipedia python package installed. This wrapper will use the Wikipedia API to conduct searches and fetch page summaries. By default, it will return the page summaries of the top-k results. It limits the Document content by doc_content_chars_max. Parameters wiki_client (Any) – top_k_results (int) – lang (str) – load_all_available_meta (bool) – doc_content_chars_max (int) – Return type None attribute doc_content_chars_max: int = 4000 attribute lang: str = 'en' attribute load_all_available_meta: bool = False attribute top_k_results: int = 3 load(query)[source] Run Wikipedia search and get the article text plus the meta information. See Returns: a list of documents. Parameters query (str) – Return type List[langchain.schema.Document] run(query)[source] Run Wikipedia search and get page summaries. Parameters query (str) – Return type str class langchain.utilities.WolframAlphaAPIWrapper(*, wolfram_client=None, wolfram_alpha_appid=None)[source] Bases: pydantic.main.BaseModel Wrapper for Wolfram Alpha. Docs for using: Go to wolfram alpha and sign up for a developer account Create an app and get your APP ID Save your APP ID into WOLFRAM_ALPHA_APPID env variable pip install wolframalpha Parameters wolfram_client (Any) – wolfram_alpha_appid (Optional[str]) – Return type None attribute wolfram_alpha_appid: Optional[str] = None run(query)[source] Run query through WolframAlpha and parse result. Parameters query (str) – Return type str class langchain.utilities.ZapierNLAWrapper(*, zapier_nla_api_key, zapier_nla_oauth_access_token, zapier_nla_api_base='https://nla.zapier.com/api/v1/')[source] Bases: pydantic.main.BaseModel Wrapper for Zapier NLA. Full docs here: https://nla.zapier.com/api/v1/docs Note: this wrapper currently only implemented the api_key auth method for testingand server-side production use cases (using the developer’s connected accounts on Zapier.com) For use-cases where LangChain + Zapier NLA is powering a user-facing application, and LangChain needs access to the end-user’s connected accounts on Zapier.com, you’ll need to use oauth. Review the full docs above and reach out to nla@zapier.com for developer support. Parameters zapier_nla_api_key (str) – zapier_nla_oauth_access_token (str) – zapier_nla_api_base (str) – Return type None attribute zapier_nla_api_base: str = 'https://nla.zapier.com/api/v1/' attribute zapier_nla_api_key: str [Required] attribute zapier_nla_oauth_access_token: str [Required] list()[source] Returns a list of all exposed (enabled) actions associated with current user (associated with the set api_key). Change your exposed actions here: https://nla.zapier.com/demo/start/ The return list can be empty if no actions exposed. Else will contain a list of action objects: [{β€œid”: str, β€œdescription”: str, β€œparams”: Dict[str, str] }] params will always contain an instructions key, the only required param. All others optional and if provided will override any AI guesses (see β€œunderstanding the AI guessing flow” here: https://nla.zapier.com/api/v1/docs) Return type List[Dict] list_as_str()[source] Same as list, but returns a stringified version of the JSON for insertting back into an LLM. Return type str preview(action_id, instructions, params=None)[source] Same as run, but instead of actually executing the action, will instead return a preview of params that have been guessed by the AI in case you need to explicitly review before executing. Parameters action_id (str) – instructions (str) – params (Optional[Dict]) – Return type Dict preview_as_str(*args, **kwargs)[source] Same as preview, but returns a stringified version of the JSON for insertting back into an LLM. Return type str run(action_id, instructions, params=None)[source] Executes an action that is identified by action_id, must be exposed (enabled) by the current user (associated with the set api_key). Change your exposed actions here: https://nla.zapier.com/demo/start/ The return JSON is guaranteed to be less than ~500 words (350 tokens) making it safe to inject into the prompt of another LLM call. Parameters action_id (str) – instructions (str) – params (Optional[Dict]) – Return type Dict run_as_str(*args, **kwargs)[source] Same as run, but returns a stringified version of the JSON for insertting back into an LLM. Return type str
https://api.python.langchain.com/en/latest/modules/utilities.html
90376b9d-3a12-4a0f-b352-1793ecf6a7af
Vector Stores Wrappers on top of vector stores. class langchain.vectorstores.AlibabaCloudOpenSearch(embedding, config, **kwargs)[source] Bases: langchain.vectorstores.base.VectorStore Alibaba Cloud OpenSearch Vector Store Parameters embedding (langchain.embeddings.base.Embeddings) – config (langchain.vectorstores.alibabacloud_opensearch.AlibabaCloudOpenSearchSettings) – kwargs (Any) – Return type None add_texts(texts, metadatas=None, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Parameters texts (Iterable[str]) – Iterable of strings to add to the vectorstore. metadatas (Optional[List[dict]]) – Optional list of metadatas associated with the texts. kwargs (Any) – vectorstore specific parameters Returns List of ids from adding the texts into the vectorstore. Return type List[str] similarity_search(query, k=4, search_filter=None, **kwargs)[source] Return docs most similar to query. Parameters query (str) – k (int) – search_filter (Optional[Dict[str, Any]]) – kwargs (Any) – Return type List[langchain.schema.Document] similarity_search_with_relevance_scores(query, k=4, search_filter=None, **kwargs)[source] Return docs and relevance scores in the range [0, 1]. 0 is dissimilar, 1 is most similar. Parameters query (str) – input text k (int) – Number of Documents to return. Defaults to 4. **kwargs – kwargs to be passed to similarity search. Should include: score_threshold: Optional, a floating point value between 0 to 1 to filter the resulting set of retrieved docs search_filter (Optional[dict]) – kwargs (Any) – Returns List of Tuples of (doc, similarity_score) Return type List[Tuple[langchain.schema.Document, float]] similarity_search_by_vector(embedding, k=4, search_filter=None, **kwargs)[source] Return docs most similar to embedding vector. Parameters embedding (List[float]) – Embedding to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. search_filter (Optional[dict]) – kwargs (Any) – Returns List of Documents most similar to the query vector. Return type List[langchain.schema.Document] inner_embedding_query(embedding, search_filter=None, k=4)[source] Parameters embedding (List[float]) – search_filter (Optional[Dict[str, Any]]) – k (int) – Return type Dict[str, Any] create_results(json_result)[source] Parameters json_result (Dict[str, Any]) – Return type List[langchain.schema.Document] create_results_with_score(json_result)[source] Parameters json_result (Dict[str, Any]) – Return type List[Tuple[langchain.schema.Document, float]] classmethod from_texts(texts, embedding, metadatas=None, config=None, **kwargs)[source] Return VectorStore initialized from texts and embeddings. Parameters texts (List[str]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – config (Optional[langchain.vectorstores.alibabacloud_opensearch.AlibabaCloudOpenSearchSettings]) – kwargs (Any) – Return type langchain.vectorstores.alibabacloud_opensearch.AlibabaCloudOpenSearch classmethod from_documents(documents, embedding, ids=None, config=None, **kwargs)[source] Return VectorStore initialized from documents and embeddings. Parameters documents (List[langchain.schema.Document]) – embedding (langchain.embeddings.base.Embeddings) – ids (Optional[List[str]]) – config (Optional[langchain.vectorstores.alibabacloud_opensearch.AlibabaCloudOpenSearchSettings]) – kwargs (Any) – Return type langchain.vectorstores.alibabacloud_opensearch.AlibabaCloudOpenSearch class langchain.vectorstores.AlibabaCloudOpenSearchSettings(endpoint, instance_id, username, password, datasource_name, embedding_index_name, field_name_mapping)[source] Bases: object Opensearch Client Configuration Attribute: endpoint (str) : The endpoint of opensearch instance, You can find it from the console of Alibaba Cloud OpenSearch. instance_id (str) : The identify of opensearch instance, You can find it from the console of Alibaba Cloud OpenSearch. datasource_name (str): The name of the data source specified when creating it. username (str) : The username specified when purchasing the instance. password (str) : The password specified when purchasing the instance. embedding_index_name (str) : The name of the vector attribute specified when configuring the instance attributes. field_name_mapping (Dict) : Using field name mapping between opensearch vector store and opensearch instance configuration table field names: { β€˜id’: β€˜The id field name map of index document.’, β€˜document’: β€˜The text field name map of index document.’, β€˜embedding’: β€˜In the embedding field of the opensearch instance, the values must be in float16 multivalue type and separated by commas.’, β€˜metadata_field_x’: β€˜Metadata field mapping includes the mapped field name and operator in the mapping value, separated by a comma between the mapped field name and the operator.’, } Parameters endpoint (str) – instance_id (str) – username (str) – password (str) – datasource_name (str) – embedding_index_name (str) – field_name_mapping (Dict[str, str]) – Return type None endpoint: str instance_id: str username: str password: str datasource_name: str embedding_index_name: str field_name_mapping: Dict[str, str] = {'document': 'document', 'embedding': 'embedding', 'id': 'id', 'metadata_field_x': 'metadata_field_x,operator'} class langchain.vectorstores.AnalyticDB(connection_string, embedding_function, embedding_dimension=1536, collection_name='langchain_document', pre_delete_collection=False, logger=None)[source] Bases: langchain.vectorstores.base.VectorStore VectorStore implementation using AnalyticDB. AnalyticDB is a distributed full PostgresSQL syntax cloud-native database. - connection_string is a postgres connection string. - embedding_function any embedding function implementing langchain.embeddings.base.Embeddings interface. collection_name is the name of the collection to use. (default: langchain) NOTE: This is not the name of the table, but the name of the collection.The tables will be created when initializing the store (if not exists) So, make sure the user has the right permissions to create tables. pre_delete_collection if True, will delete the collection if it exists.(default: False) - Useful for testing. Parameters connection_string (str) – embedding_function (Embeddings) – embedding_dimension (int) – collection_name (str) – pre_delete_collection (bool) – logger (Optional[logging.Logger]) – Return type None create_table_if_not_exists()[source] Return type None create_collection()[source] Return type None delete_collection()[source] Return type None add_texts(texts, metadatas=None, ids=None, batch_size=500, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Parameters texts (Iterable[str]) – Iterable of strings to add to the vectorstore. metadatas (Optional[List[dict]]) – Optional list of metadatas associated with the texts. kwargs (Any) – vectorstore specific parameters ids (Optional[List[str]]) – batch_size (int) – Returns List of ids from adding the texts into the vectorstore. Return type List[str] similarity_search(query, k=4, filter=None, **kwargs)[source] Run similarity search with AnalyticDB with distance. Parameters query (str) – Query text to search for. k (int) – Number of results to return. Defaults to 4. filter (Optional[Dict[str, str]]) – Filter by metadata. Defaults to None. kwargs (Any) – Returns List of Documents most similar to the query. Return type List[langchain.schema.Document] similarity_search_with_score(query, k=4, filter=None)[source] Return docs most similar to query. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. filter (Optional[Dict[str, str]]) – Filter by metadata. Defaults to None. Returns List of Documents most similar to the query and score for each Return type List[Tuple[langchain.schema.Document, float]] similarity_search_with_score_by_vector(embedding, k=4, filter=None)[source] Parameters embedding (List[float]) – k (int) – filter (Optional[dict]) – Return type List[Tuple[langchain.schema.Document, float]] similarity_search_by_vector(embedding, k=4, filter=None, **kwargs)[source] Return docs most similar to embedding vector. Parameters embedding (List[float]) – Embedding to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. filter (Optional[Dict[str, str]]) – Filter by metadata. Defaults to None. kwargs (Any) – Returns List of Documents most similar to the query vector. Return type List[langchain.schema.Document] classmethod from_texts(texts, embedding, metadatas=None, embedding_dimension=1536, collection_name='langchain_document', ids=None, pre_delete_collection=False, **kwargs)[source] Return VectorStore initialized from texts and embeddings. Postgres Connection string is required Either pass it as a parameter or set the PG_CONNECTION_STRING environment variable. Parameters texts (List[str]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – embedding_dimension (int) – collection_name (str) – ids (Optional[List[str]]) – pre_delete_collection (bool) – kwargs (Any) – Return type langchain.vectorstores.analyticdb.AnalyticDB classmethod get_connection_string(kwargs)[source] Parameters kwargs (Dict[str, Any]) – Return type str classmethod from_documents(documents, embedding, embedding_dimension=1536, collection_name='langchain_document', ids=None, pre_delete_collection=False, **kwargs)[source] Return VectorStore initialized from documents and embeddings. Postgres Connection string is required Either pass it as a parameter or set the PG_CONNECTION_STRING environment variable. Parameters documents (List[langchain.schema.Document]) – embedding (langchain.embeddings.base.Embeddings) – embedding_dimension (int) – collection_name (str) – ids (Optional[List[str]]) – pre_delete_collection (bool) – kwargs (Any) – Return type langchain.vectorstores.analyticdb.AnalyticDB classmethod connection_string_from_db_params(driver, host, port, database, user, password)[source] Return connection string from database parameters. Parameters driver (str) – host (str) – port (int) – database (str) – user (str) – password (str) – Return type str class langchain.vectorstores.Annoy(embedding_function, index, metric, docstore, index_to_docstore_id)[source] Bases: langchain.vectorstores.base.VectorStore Wrapper around Annoy vector database. To use, you should have the annoy python package installed. Example from langchain import Annoy db = Annoy(embedding_function, index, docstore, index_to_docstore_id) Parameters embedding_function (Callable) – index (Any) – metric (str) – docstore (Docstore) – index_to_docstore_id (Dict[int, str]) – add_texts(texts, metadatas=None, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Parameters texts (Iterable[str]) – Iterable of strings to add to the vectorstore. metadatas (Optional[List[dict]]) – Optional list of metadatas associated with the texts. kwargs (Any) – vectorstore specific parameters Returns List of ids from adding the texts into the vectorstore. Return type List[str] process_index_results(idxs, dists)[source] Turns annoy results into a list of documents and scores. Parameters idxs (List[int]) – List of indices of the documents in the index. dists (List[float]) – List of distances of the documents in the index. Returns List of Documents and scores. Return type List[Tuple[langchain.schema.Document, float]] similarity_search_with_score_by_vector(embedding, k=4, search_k=- 1)[source] Return docs most similar to query. Parameters query – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. search_k (int) – inspect up to search_k nodes which defaults to n_trees * n if not provided embedding (List[float]) – Returns List of Documents most similar to the query and score for each Return type List[Tuple[langchain.schema.Document, float]] similarity_search_with_score_by_index(docstore_index, k=4, search_k=- 1)[source] Return docs most similar to query. Parameters query – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. search_k (int) – inspect up to search_k nodes which defaults to n_trees * n if not provided docstore_index (int) – Returns List of Documents most similar to the query and score for each Return type List[Tuple[langchain.schema.Document, float]] similarity_search_with_score(query, k=4, search_k=- 1)[source] Return docs most similar to query. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. search_k (int) – inspect up to search_k nodes which defaults to n_trees * n if not provided Returns List of Documents most similar to the query and score for each Return type List[Tuple[langchain.schema.Document, float]] similarity_search_by_vector(embedding, k=4, search_k=- 1, **kwargs)[source] Return docs most similar to embedding vector. Parameters embedding (List[float]) – Embedding to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. search_k (int) – inspect up to search_k nodes which defaults to n_trees * n if not provided kwargs (Any) – Returns List of Documents most similar to the embedding. Return type List[langchain.schema.Document] similarity_search_by_index(docstore_index, k=4, search_k=- 1, **kwargs)[source] Return docs most similar to docstore_index. Parameters docstore_index (int) – Index of document in docstore k (int) – Number of Documents to return. Defaults to 4. search_k (int) – inspect up to search_k nodes which defaults to n_trees * n if not provided kwargs (Any) – Returns List of Documents most similar to the embedding. Return type List[langchain.schema.Document] similarity_search(query, k=4, search_k=- 1, **kwargs)[source] Return docs most similar to query. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. search_k (int) – inspect up to search_k nodes which defaults to n_trees * n if not provided kwargs (Any) – Returns List of Documents most similar to the query. Return type List[langchain.schema.Document] max_marginal_relevance_search_by_vector(embedding, k=4, fetch_k=20, lambda_mult=0.5, **kwargs)[source] Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Parameters embedding (List[float]) – Embedding to look up documents similar to. fetch_k (int) – Number of Documents to fetch to pass to MMR algorithm. k (int) – Number of Documents to return. Defaults to 4. lambda_mult (float) – Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. kwargs (Any) – Returns List of Documents selected by maximal marginal relevance. Return type List[langchain.schema.Document] max_marginal_relevance_search(query, k=4, fetch_k=20, lambda_mult=0.5, **kwargs)[source] Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. fetch_k (int) – Number of Documents to fetch to pass to MMR algorithm. lambda_mult (float) – Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. kwargs (Any) – Returns List of Documents selected by maximal marginal relevance. Return type List[langchain.schema.Document] classmethod from_texts(texts, embedding, metadatas=None, metric='angular', trees=100, n_jobs=- 1, **kwargs)[source] Construct Annoy wrapper from raw documents. Parameters texts (List[str]) – List of documents to index. embedding (langchain.embeddings.base.Embeddings) – Embedding function to use. metadatas (Optional[List[dict]]) – List of metadata dictionaries to associate with documents. metric (str) – Metric to use for indexing. Defaults to β€œangular”. trees (int) – Number of trees to use for indexing. Defaults to 100. n_jobs (int) – Number of jobs to use for indexing. Defaults to -1. kwargs (Any) – Return type langchain.vectorstores.annoy.Annoy This is a user friendly interface that: Embeds documents. Creates an in memory docstore Initializes the Annoy database This is intended to be a quick way to get started. Example from langchain import Annoy from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() index = Annoy.from_texts(texts, embeddings) classmethod from_embeddings(text_embeddings, embedding, metadatas=None, metric='angular', trees=100, n_jobs=- 1, **kwargs)[source] Construct Annoy wrapper from embeddings. Parameters text_embeddings (List[Tuple[str, List[float]]]) – List of tuples of (text, embedding) embedding (langchain.embeddings.base.Embeddings) – Embedding function to use. metadatas (Optional[List[dict]]) – List of metadata dictionaries to associate with documents. metric (str) – Metric to use for indexing. Defaults to β€œangular”. trees (int) – Number of trees to use for indexing. Defaults to 100. n_jobs (int) – Number of jobs to use for indexing. Defaults to -1 kwargs (Any) – Return type langchain.vectorstores.annoy.Annoy This is a user friendly interface that: Creates an in memory docstore with provided embeddings Initializes the Annoy database This is intended to be a quick way to get started. Example from langchain import Annoy from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() text_embeddings = embeddings.embed_documents(texts) text_embedding_pairs = list(zip(texts, text_embeddings)) db = Annoy.from_embeddings(text_embedding_pairs, embeddings) save_local(folder_path, prefault=False)[source] Save Annoy index, docstore, and index_to_docstore_id to disk. Parameters folder_path (str) – folder path to save index, docstore, and index_to_docstore_id to. prefault (bool) – Whether to pre-load the index into memory. Return type None classmethod load_local(folder_path, embeddings)[source] Load Annoy index, docstore, and index_to_docstore_id to disk. Parameters folder_path (str) – folder path to load index, docstore, and index_to_docstore_id from. embeddings (langchain.embeddings.base.Embeddings) – Embeddings to use when generating queries. Return type langchain.vectorstores.annoy.Annoy class langchain.vectorstores.AtlasDB(name, embedding_function=None, api_key=None, description='A description for your project', is_public=True, reset_project_if_exists=False)[source] Bases: langchain.vectorstores.base.VectorStore Wrapper around Atlas: Nomic’s neural database and rhizomatic instrument. To use, you should have the nomic python package installed. Example from langchain.vectorstores import AtlasDB from langchain.embeddings.openai import OpenAIEmbeddings embeddings = OpenAIEmbeddings() vectorstore = AtlasDB("my_project", embeddings.embed_query) Parameters name (str) – embedding_function (Optional[Embeddings]) – api_key (Optional[str]) – description (str) – is_public (bool) – reset_project_if_exists (bool) – Return type None add_texts(texts, metadatas=None, ids=None, refresh=True, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Parameters texts (Iterable[str]) – Texts to add to the vectorstore. metadatas (Optional[List[dict]], optional) – Optional list of metadatas. ids (Optional[List[str]]) – An optional list of ids. refresh (bool) – Whether or not to refresh indices with the updated data. Default True. kwargs (Any) – Returns List of IDs of the added texts. Return type List[str] create_index(**kwargs)[source] Creates an index in your project. See https://docs.nomic.ai/atlas_api.html#nomic.project.AtlasProject.create_index for full detail. Parameters kwargs (Any) – Return type Any similarity_search(query, k=4, **kwargs)[source] Run similarity search with AtlasDB Parameters query (str) – Query text to search for. k (int) – Number of results to return. Defaults to 4. kwargs (Any) – Returns List of documents most similar to the query text. Return type List[Document] classmethod from_texts(texts, embedding=None, metadatas=None, ids=None, name=None, api_key=None, description='A description for your project', is_public=True, reset_project_if_exists=False, index_kwargs=None, **kwargs)[source] Create an AtlasDB vectorstore from a raw documents. Parameters texts (List[str]) – The list of texts to ingest. name (str) – Name of the project to create. api_key (str) – Your nomic API key, embedding (Optional[Embeddings]) – Embedding function. Defaults to None. metadatas (Optional[List[dict]]) – List of metadatas. Defaults to None. ids (Optional[List[str]]) – Optional list of document IDs. If None, ids will be auto created description (str) – A description for your project. is_public (bool) – Whether your project is publicly accessible. True by default. reset_project_if_exists (bool) – Whether to reset this project if it already exists. Default False. Generally userful during development and testing. index_kwargs (Optional[dict]) – Dict of kwargs for index creation. See https://docs.nomic.ai/atlas_api.html kwargs (Any) – Returns Nomic’s neural database and finest rhizomatic instrument Return type AtlasDB classmethod from_documents(documents, embedding=None, ids=None, name=None, api_key=None, persist_directory=None, description='A description for your project', is_public=True, reset_project_if_exists=False, index_kwargs=None, **kwargs)[source] Create an AtlasDB vectorstore from a list of documents. Parameters name (str) – Name of the collection to create. api_key (str) – Your nomic API key, documents (List[Document]) – List of documents to add to the vectorstore. embedding (Optional[Embeddings]) – Embedding function. Defaults to None. ids (Optional[List[str]]) – Optional list of document IDs. If None, ids will be auto created description (str) – A description for your project. is_public (bool) – Whether your project is publicly accessible. True by default. reset_project_if_exists (bool) – Whether to reset this project if it already exists. Default False. Generally userful during development and testing. index_kwargs (Optional[dict]) – Dict of kwargs for index creation. See https://docs.nomic.ai/atlas_api.html persist_directory (Optional[str]) – kwargs (Any) – Returns Nomic’s neural database and finest rhizomatic instrument Return type AtlasDB class langchain.vectorstores.AwaDB(table_name='langchain_awadb', embedding_model=None, log_and_data_dir=None, client=None)[source] Bases: langchain.vectorstores.base.VectorStore Interface implemented by AwaDB vector stores. Parameters table_name (str) – embedding_model (Optional[Embeddings]) – log_and_data_dir (Optional[str]) – client (Optional[awadb.Client]) – Return type None add_texts(texts, metadatas=None, is_duplicate_texts=None, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. :param texts: Iterable of strings to add to the vectorstore. :param metadatas: Optional list of metadatas associated with the texts. :param is_duplicate_texts: Optional whether to duplicate texts. :param kwargs: vectorstore specific parameters. Returns List of ids from adding the texts into the vectorstore. Parameters texts (Iterable[str]) – metadatas (Optional[List[dict]]) – is_duplicate_texts (Optional[bool]) – kwargs (Any) – Return type List[str] load_local(table_name, **kwargs)[source] Parameters table_name (str) – kwargs (Any) – Return type bool similarity_search(query, k=4, **kwargs)[source] Return docs most similar to query. Parameters query (str) – k (int) – kwargs (Any) – Return type List[langchain.schema.Document] similarity_search_with_score(query, k=4, **kwargs)[source] Return docs and relevance scores, normalized on a scale from 0 to 1. 0 is dissimilar, 1 is most similar. Parameters query (str) – k (int) – kwargs (Any) – Return type List[Tuple[langchain.schema.Document, float]] similarity_search_with_relevance_scores(query, k=4, **kwargs)[source] Return docs and relevance scores, normalized on a scale from 0 to 1. 0 is dissimilar, 1 is most similar. Parameters query (str) – k (int) – kwargs (Any) – Return type List[Tuple[langchain.schema.Document, float]] similarity_search_by_vector(embedding=None, k=4, scores=None, **kwargs)[source] Return docs most similar to embedding vector. Parameters embedding (Optional[List[float]]) – Embedding to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. scores (Optional[list]) – kwargs (Any) – Returns List of Documents most similar to the query vector. Return type List[langchain.schema.Document] create_table(table_name, **kwargs)[source] Create a new table. Parameters table_name (str) – kwargs (Any) – Return type bool use(table_name, **kwargs)[source] Use the specified table. Don’t know the tables, please invoke list_tables. Parameters table_name (str) – kwargs (Any) – Return type bool list_tables(**kwargs)[source] List all the tables created by the client. Parameters kwargs (Any) – Return type List[str] get_current_table(**kwargs)[source] Get the current table. Parameters kwargs (Any) – Return type str classmethod from_texts(texts, embedding=None, metadatas=None, table_name='langchain_awadb', logging_and_data_dir=None, client=None, **kwargs)[source] Create an AwaDB vectorstore from a raw documents. Parameters texts (List[str]) – List of texts to add to the table. embedding (Optional[Embeddings]) – Embedding function. Defaults to None. metadatas (Optional[List[dict]]) – List of metadatas. Defaults to None. table_name (str) – Name of the table to create. logging_and_data_dir (Optional[str]) – Directory of logging and persistence. client (Optional[awadb.Client]) – AwaDB client kwargs (Any) – Returns AwaDB vectorstore. Return type AwaDB classmethod from_documents(documents, embedding=None, table_name='langchain_awadb', logging_and_data_dir=None, client=None, **kwargs)[source] Create an AwaDB vectorstore from a list of documents. If a logging_and_data_dir specified, the table will be persisted there. Parameters documents (List[Document]) – List of documents to add to the vectorstore. embedding (Optional[Embeddings]) – Embedding function. Defaults to None. table_name (str) – Name of the table to create. logging_and_data_dir (Optional[str]) – Directory to persist the table. client (Optional[awadb.Client]) – AwaDB client kwargs (Any) – Returns AwaDB vectorstore. Return type AwaDB class langchain.vectorstores.AzureSearch(azure_search_endpoint, azure_search_key, index_name, embedding_function, search_type='hybrid', semantic_configuration_name=None, semantic_query_language='en-us', **kwargs)[source] Bases: langchain.vectorstores.base.VectorStore Parameters azure_search_endpoint (str) – azure_search_key (str) – index_name (str) – embedding_function (Callable) – search_type (str) – semantic_configuration_name (Optional[str]) – semantic_query_language (str) – kwargs (Any) – add_texts(texts, metadatas=None, **kwargs)[source] Add texts data to an existing index. Parameters texts (Iterable[str]) – metadatas (Optional[List[dict]]) – kwargs (Any) – Return type List[str] similarity_search(query, k=4, **kwargs)[source] Return docs most similar to query. Parameters query (str) – k (int) – kwargs (Any) – Return type List[langchain.schema.Document] vector_search(query, k=4, **kwargs)[source] Returns the most similar indexed documents to the query text. Parameters query (str) – The query text for which to find similar documents. k (int) – The number of documents to return. Default is 4. kwargs (Any) – Returns A list of documents that are most similar to the query text. Return type List[Document] vector_search_with_score(query, k=4, filters=None)[source] Return docs most similar to query. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. filters (Optional[str]) – Returns List of Documents most similar to the query and score for each Return type List[Tuple[langchain.schema.Document, float]] hybrid_search(query, k=4, **kwargs)[source] Returns the most similar indexed documents to the query text. Parameters query (str) – The query text for which to find similar documents. k (int) – The number of documents to return. Default is 4. kwargs (Any) – Returns A list of documents that are most similar to the query text. Return type List[Document] hybrid_search_with_score(query, k=4, filters=None)[source] Return docs most similar to query with an hybrid query. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. filters (Optional[str]) – Returns List of Documents most similar to the query and score for each Return type List[Tuple[langchain.schema.Document, float]] semantic_hybrid_search(query, k=4, **kwargs)[source] Returns the most similar indexed documents to the query text. Parameters query (str) – The query text for which to find similar documents. k (int) – The number of documents to return. Default is 4. kwargs (Any) – Returns A list of documents that are most similar to the query text. Return type List[Document] semantic_hybrid_search_with_score(query, k=4, filters=None)[source] Return docs most similar to query with an hybrid query. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. filters (Optional[str]) – Returns List of Documents most similar to the query and score for each Return type List[Tuple[langchain.schema.Document, float]] classmethod from_texts(texts, embedding, metadatas=None, azure_search_endpoint='', azure_search_key='', index_name='langchain-index', **kwargs)[source] Return VectorStore initialized from texts and embeddings. Parameters texts (List[str]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – azure_search_endpoint (str) – azure_search_key (str) – index_name (str) – kwargs (Any) – Return type langchain.vectorstores.azuresearch.AzureSearch class langchain.vectorstores.Cassandra(embedding, session, keyspace, table_name, ttl_seconds=None)[source] Bases: langchain.vectorstores.base.VectorStore Wrapper around Cassandra embeddings platform. There is no notion of a default table name, since each embedding function implies its own vector dimension, which is part of the schema. Example from langchain.vectorstores import Cassandra from langchain.embeddings.openai import OpenAIEmbeddings embeddings = OpenAIEmbeddings() session = ... keyspace = 'my_keyspace' vectorstore = Cassandra(embeddings, session, keyspace, 'my_doc_archive') Parameters embedding (Embeddings) – session (Session) – keyspace (str) – table_name (str) – ttl_seconds (int | None) – Return type None delete_collection()[source] Just an alias for clear (to better align with other VectorStore implementations). Return type None clear()[source] Empty the collection. Return type None delete_by_document_id(document_id)[source] Parameters document_id (str) – Return type None add_texts(texts, metadatas=None, ids=None, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Parameters texts (Iterable[str]) – Texts to add to the vectorstore. metadatas (Optional[List[dict]], optional) – Optional list of metadatas. ids (Optional[List[str]], optional) – Optional list of IDs. kwargs (Any) – Returns List of IDs of the added texts. Return type List[str] similarity_search_with_score_id_by_vector(embedding, k=4)[source] Return docs most similar to embedding vector. No support for filter query (on metadata) along with vector search. Parameters embedding (str) – Embedding to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. Returns List of (Document, score, id), the most similar to the query vector. Return type List[Tuple[langchain.schema.Document, float, str]] similarity_search_with_score_id(query, k=4, **kwargs)[source] Parameters query (str) – k (int) – kwargs (Any) – Return type List[Tuple[langchain.schema.Document, float, str]] similarity_search_with_score_by_vector(embedding, k=4)[source] Return docs most similar to embedding vector. No support for filter query (on metadata) along with vector search. Parameters embedding (str) – Embedding to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. Returns List of (Document, score), the most similar to the query vector. Return type List[Tuple[langchain.schema.Document, float]] similarity_search(query, k=4, **kwargs)[source] Return docs most similar to query. Parameters query (str) – k (int) – kwargs (Any) – Return type List[langchain.schema.Document] similarity_search_by_vector(embedding, k=4, **kwargs)[source] Return docs most similar to embedding vector. Parameters embedding (List[float]) – Embedding to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. kwargs (Any) – Returns List of Documents most similar to the query vector. Return type List[langchain.schema.Document] similarity_search_with_score(query, k=4, **kwargs)[source] Parameters query (str) – k (int) – kwargs (Any) – Return type List[Tuple[langchain.schema.Document, float]] max_marginal_relevance_search_by_vector(embedding, k=4, fetch_k=20, lambda_mult=0.5, **kwargs)[source] Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. :param embedding: Embedding to look up documents similar to. :param k: Number of Documents to return. :param fetch_k: Number of Documents to fetch to pass to MMR algorithm. :param lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Returns List of Documents selected by maximal marginal relevance. Parameters embedding (List[float]) – k (int) – fetch_k (int) – lambda_mult (float) – kwargs (Any) – Return type List[langchain.schema.Document] max_marginal_relevance_search(query, k=4, fetch_k=20, lambda_mult=0.5, **kwargs)[source] Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. :param query: Text to look up documents similar to. :param k: Number of Documents to return. :param fetch_k: Number of Documents to fetch to pass to MMR algorithm. :param lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Optional. Returns List of Documents selected by maximal marginal relevance. Parameters query (str) – k (int) – fetch_k (int) – lambda_mult (float) – kwargs (Any) – Return type List[langchain.schema.Document] classmethod from_texts(texts, embedding, metadatas=None, **kwargs)[source] Create a Cassandra vectorstore from raw texts. No support for specifying text IDs Returns a Cassandra vectorstore. Parameters texts (List[str]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – kwargs (Any) – Return type langchain.vectorstores.cassandra.CVST classmethod from_documents(documents, embedding, **kwargs)[source] Create a Cassandra vectorstore from a document list. No support for specifying text IDs Returns a Cassandra vectorstore. Parameters documents (List[langchain.schema.Document]) – embedding (langchain.embeddings.base.Embeddings) – kwargs (Any) – Return type langchain.vectorstores.cassandra.CVST class langchain.vectorstores.Chroma(collection_name='langchain', embedding_function=None, persist_directory=None, client_settings=None, collection_metadata=None, client=None)[source] Bases: langchain.vectorstores.base.VectorStore Wrapper around ChromaDB embeddings platform. To use, you should have the chromadb python package installed. Example from langchain.vectorstores import Chroma from langchain.embeddings.openai import OpenAIEmbeddings embeddings = OpenAIEmbeddings() vectorstore = Chroma("langchain_store", embeddings) Parameters collection_name (str) – embedding_function (Optional[Embeddings]) – persist_directory (Optional[str]) – client_settings (Optional[chromadb.config.Settings]) – collection_metadata (Optional[Dict]) – client (Optional[chromadb.Client]) – Return type None add_texts(texts, metadatas=None, ids=None, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Parameters texts (Iterable[str]) – Texts to add to the vectorstore. metadatas (Optional[List[dict]], optional) – Optional list of metadatas. ids (Optional[List[str]], optional) – Optional list of IDs. kwargs (Any) – Returns List of IDs of the added texts. Return type List[str] similarity_search(query, k=4, filter=None, **kwargs)[source] Run similarity search with Chroma. Parameters query (str) – Query text to search for. k (int) – Number of results to return. Defaults to 4. filter (Optional[Dict[str, str]]) – Filter by metadata. Defaults to None. kwargs (Any) – Returns List of documents most similar to the query text. Return type List[Document] similarity_search_by_vector(embedding, k=4, filter=None, **kwargs)[source] Return docs most similar to embedding vector. :param embedding: Embedding to look up documents similar to. :type embedding: str :param k: Number of Documents to return. Defaults to 4. :type k: int :param filter: Filter by metadata. Defaults to None. :type filter: Optional[Dict[str, str]] Returns List of Documents most similar to the query vector. Parameters embedding (List[float]) – k (int) – filter (Optional[Dict[str, str]]) – kwargs (Any) – Return type List[langchain.schema.Document] similarity_search_with_score(query, k=4, filter=None, **kwargs)[source] Run similarity search with Chroma with distance. Parameters query (str) – Query text to search for. k (int) – Number of results to return. Defaults to 4. filter (Optional[Dict[str, str]]) – Filter by metadata. Defaults to None. kwargs (Any) – Returns List of documents most similar to the query text and cosine distance in float for each. Lower score represents more similarity. Return type List[Tuple[Document, float]] max_marginal_relevance_search_by_vector(embedding, k=4, fetch_k=20, lambda_mult=0.5, filter=None, **kwargs)[source] Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Parameters embedding (List[float]) – Embedding to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. fetch_k (int) – Number of Documents to fetch to pass to MMR algorithm. lambda_mult (float) – Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. filter (Optional[Dict[str, str]]) – Filter by metadata. Defaults to None. kwargs (Any) – Returns List of Documents selected by maximal marginal relevance. Return type List[langchain.schema.Document] max_marginal_relevance_search(query, k=4, fetch_k=20, lambda_mult=0.5, filter=None, **kwargs)[source] Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. fetch_k (int) – Number of Documents to fetch to pass to MMR algorithm. lambda_mult (float) – Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. filter (Optional[Dict[str, str]]) – Filter by metadata. Defaults to None. kwargs (Any) – Returns List of Documents selected by maximal marginal relevance. Return type List[langchain.schema.Document] delete_collection()[source] Delete the collection. Return type None get(ids=None, where=None, limit=None, offset=None, where_document=None, include=None)[source] Gets the collection. Parameters ids (Optional[OneOrMany[ID]]) – The ids of the embeddings to get. Optional. where (Optional[Where]) – A Where type dict used to filter results by. E.g. {β€œcolor” : β€œred”, β€œprice”: 4.20}. Optional. limit (Optional[int]) – The number of documents to return. Optional. offset (Optional[int]) – The offset to start returning results from. Useful for paging results with limit. Optional. where_document (Optional[WhereDocument]) – A WhereDocument type dict used to filter by the documents. E.g. {$contains: {β€œtext”: β€œhello”}}. Optional. include (Optional[List[str]]) – A list of what to include in the results. Can contain β€œembeddings”, β€œmetadatas”, β€œdocuments”. Ids are always included. Defaults to [β€œmetadatas”, β€œdocuments”]. Optional. Return type Dict[str, Any] persist()[source] Persist the collection. This can be used to explicitly persist the data to disk. It will also be called automatically when the object is destroyed. Return type None update_document(document_id, document)[source] Update a document in the collection. Parameters document_id (str) – ID of the document to update. document (Document) – Document to update. Return type None classmethod from_texts(texts, embedding=None, metadatas=None, ids=None, collection_name='langchain', persist_directory=None, client_settings=None, client=None, **kwargs)[source] Create a Chroma vectorstore from a raw documents. If a persist_directory is specified, the collection will be persisted there. Otherwise, the data will be ephemeral in-memory. Parameters texts (List[str]) – List of texts to add to the collection. collection_name (str) – Name of the collection to create. persist_directory (Optional[str]) – Directory to persist the collection. embedding (Optional[Embeddings]) – Embedding function. Defaults to None. metadatas (Optional[List[dict]]) – List of metadatas. Defaults to None. ids (Optional[List[str]]) – List of document IDs. Defaults to None. client_settings (Optional[chromadb.config.Settings]) – Chroma client settings client (Optional[chromadb.Client]) – kwargs (Any) – Returns Chroma vectorstore. Return type Chroma classmethod from_documents(documents, embedding=None, ids=None, collection_name='langchain', persist_directory=None, client_settings=None, client=None, **kwargs)[source] Create a Chroma vectorstore from a list of documents. If a persist_directory is specified, the collection will be persisted there. Otherwise, the data will be ephemeral in-memory. Parameters collection_name (str) – Name of the collection to create. persist_directory (Optional[str]) – Directory to persist the collection. ids (Optional[List[str]]) – List of document IDs. Defaults to None. documents (List[Document]) – List of documents to add to the vectorstore. embedding (Optional[Embeddings]) – Embedding function. Defaults to None. client_settings (Optional[chromadb.config.Settings]) – Chroma client settings client (Optional[chromadb.Client]) – kwargs (Any) – Returns Chroma vectorstore. Return type Chroma delete(ids)[source] Delete by vector IDs. Parameters ids (List[str]) – List of ids to delete. Return type None class langchain.vectorstores.Clickhouse(embedding, config=None, **kwargs)[source] Bases: langchain.vectorstores.base.VectorStore Wrapper around ClickHouse vector database You need a clickhouse-connect python package, and a valid account to connect to ClickHouse. ClickHouse can not only search with simple vector indexes, it also supports complex query with multiple conditions, constraints and even sub-queries. For more information, please visit[ClickHouse official site](https://clickhouse.com/clickhouse) Parameters embedding (Embeddings) – config (Optional[ClickhouseSettings]) – kwargs (Any) – Return type None escape_str(value)[source] Parameters value (str) – Return type str add_texts(texts, metadatas=None, batch_size=32, ids=None, **kwargs)[source] Insert more texts through the embeddings and add to the VectorStore. Parameters texts (Iterable[str]) – Iterable of strings to add to the VectorStore. ids (Optional[Iterable[str]]) – Optional list of ids to associate with the texts. batch_size (int) – Batch size of insertion metadata – Optional column data to be inserted metadatas (Optional[List[dict]]) – kwargs (Any) – Returns List of ids from adding the texts into the VectorStore. Return type List[str] classmethod from_texts(texts, embedding, metadatas=None, config=None, text_ids=None, batch_size=32, **kwargs)[source] Create ClickHouse wrapper with existing texts Parameters embedding_function (Embeddings) – Function to extract text embedding texts (Iterable[str]) – List or tuple of strings to be added config (ClickHouseSettings, Optional) – ClickHouse configuration text_ids (Optional[Iterable], optional) – IDs for the texts. Defaults to None. batch_size (int, optional) – Batchsize when transmitting data to ClickHouse. Defaults to 32. metadata (List[dict], optional) – metadata to texts. Defaults to None. into (Other keyword arguments will pass) – [clickhouse-connect](https://clickhouse.com/docs/en/integrations/python#clickhouse-connect-driver-api) embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[Dict[Any, Any]]]) – kwargs (Any) – Returns ClickHouse Index Return type langchain.vectorstores.clickhouse.Clickhouse similarity_search(query, k=4, where_str=None, **kwargs)[source] Perform a similarity search with ClickHouse Parameters query (str) – query string k (int, optional) – Top K neighbors to retrieve. Defaults to 4. where_str (Optional[str], optional) – where condition string. Defaults to None. NOTE – Please do not let end-user to fill this and always be aware of SQL injection. When dealing with metadatas, remember to use {self.metadata_column}.attribute instead of attribute alone. The default name for it is metadata. kwargs (Any) – Returns List of Documents Return type List[Document] similarity_search_by_vector(embedding, k=4, where_str=None, **kwargs)[source] Perform a similarity search with ClickHouse by vectors Parameters query (str) – query string k (int, optional) – Top K neighbors to retrieve. Defaults to 4. where_str (Optional[str], optional) – where condition string. Defaults to None. NOTE – Please do not let end-user to fill this and always be aware of SQL injection. When dealing with metadatas, remember to use {self.metadata_column}.attribute instead of attribute alone. The default name for it is metadata. embedding (List[float]) – kwargs (Any) – Returns List of (Document, similarity) Return type List[Document] similarity_search_with_relevance_scores(query, k=4, where_str=None, **kwargs)[source] Perform a similarity search with ClickHouse Parameters query (str) – query string k (int, optional) – Top K neighbors to retrieve. Defaults to 4. where_str (Optional[str], optional) – where condition string. Defaults to None. NOTE – Please do not let end-user to fill this and always be aware of SQL injection. When dealing with metadatas, remember to use {self.metadata_column}.attribute instead of attribute alone. The default name for it is metadata. kwargs (Any) – Returns List of documents Return type List[Document] drop()[source] Helper function: Drop data Return type None property metadata_column: str pydantic settings langchain.vectorstores.ClickhouseSettings[source] Bases: pydantic.env_settings.BaseSettings ClickHouse Client Configuration Attribute: clickhouse_host (str)An URL to connect to MyScale backend.Defaults to β€˜localhost’. clickhouse_port (int) : URL port to connect with HTTP. Defaults to 8443. username (str) : Username to login. Defaults to None. password (str) : Password to login. Defaults to None. index_type (str): index type string. index_param (list): index build parameter. index_query_params(dict): index query parameters. database (str) : Database name to find the table. Defaults to β€˜default’. table (str) : Table name to operate on. Defaults to β€˜vector_table’. metric (str)Metric to compute distance,supported are (β€˜angular’, β€˜euclidean’, β€˜manhattan’, β€˜hamming’, β€˜dot’). Defaults to β€˜angular’. https://github.com/spotify/annoy/blob/main/src/annoymodule.cc#L149-L169 column_map (Dict)Column type map to project column name onto langchainsemantics. Must have keys: text, id, vector, must be same size to number of columns. For example: .. code-block:: python {β€˜id’: β€˜text_id’, β€˜uuid’: β€˜global_unique_id’ β€˜embedding’: β€˜text_embedding’, β€˜document’: β€˜text_plain’, β€˜metadata’: β€˜metadata_dictionary_in_json’, } Defaults to identity map. Show JSON schema{ "title": "ClickhouseSettings", "description": "ClickHouse Client Configuration\n\nAttribute:\n clickhouse_host (str) : An URL to connect to MyScale backend.\n Defaults to 'localhost'.\n clickhouse_port (int) : URL port to connect with HTTP. Defaults to 8443.\n username (str) : Username to login. Defaults to None.\n password (str) : Password to login. Defaults to None.\n index_type (str): index type string.\n index_param (list): index build parameter.\n index_query_params(dict): index query parameters.\n database (str) : Database name to find the table. Defaults to 'default'.\n table (str) : Table name to operate on.\n Defaults to 'vector_table'.\n metric (str) : Metric to compute distance,\n supported are ('angular', 'euclidean', 'manhattan', 'hamming',\n 'dot'). Defaults to 'angular'.\n https://github.com/spotify/annoy/blob/main/src/annoymodule.cc#L149-L169\n\n column_map (Dict) : Column type map to project column name onto langchain\n semantics. Must have keys: `text`, `id`, `vector`,\n must be same size to number of columns. For example:\n .. code-block:: python\n\n {\n 'id': 'text_id',\n 'uuid': 'global_unique_id'\n 'embedding': 'text_embedding',\n 'document': 'text_plain',\n 'metadata': 'metadata_dictionary_in_json',\n }\n\n Defaults to identity map.", "type": "object", "properties": { "host": { "title": "Host", "default": "localhost", "env_names": "{'clickhouse_host'}", "type": "string" }, "port": { "title": "Port", "default": 8123, "env_names": "{'clickhouse_port'}", "type": "integer" }, "username": { "title": "Username", "env_names": "{'clickhouse_username'}", "type": "string" }, "password": { "title": "Password", "env_names": "{'clickhouse_password'}", "type": "string" }, "index_type": { "title": "Index Type", "default": "annoy", "env_names": "{'clickhouse_index_type'}", "type": "string" }, "index_param": { "title": "Index Param", "default": [ "'L2Distance'", 100 ], "env_names": "{'clickhouse_index_param'}", "anyOf": [ { "type": "array", "items": {} }, { "type": "object" } ] }, "index_query_params": { "title": "Index Query Params", "default": {}, "env_names": "{'clickhouse_index_query_params'}", "type": "object", "additionalProperties": { "type": "string" } }, "column_map": { "title": "Column Map", "default": { "id": "id", "uuid": "uuid", "document": "document", "embedding": "embedding", "metadata": "metadata" }, "env_names": "{'clickhouse_column_map'}", "type": "object", "additionalProperties": { "type": "string" } }, "database": { "title": "Database", "default": "default", "env_names": "{'clickhouse_database'}", "type": "string" }, "table": { "title": "Table", "default": "langchain", "env_names": "{'clickhouse_table'}", "type": "string" }, "metric": { "title": "Metric", "default": "angular", "env_names": "{'clickhouse_metric'}", "type": "string" } }, "additionalProperties": false } Config env_file: str = .env env_file_encoding: str = utf-8 env_prefix: str = clickhouse_ Fields column_map (Dict[str, str]) database (str) host (str) index_param (Optional[Union[List, Dict]]) index_query_params (Dict[str, str]) index_type (str) metric (str) password (Optional[str]) port (int) table (str) username (Optional[str]) attribute column_map: Dict[str, str] = {'document': 'document', 'embedding': 'embedding', 'id': 'id', 'metadata': 'metadata', 'uuid': 'uuid'} attribute database: str = 'default' attribute host: str = 'localhost' attribute index_param: Optional[Union[List, Dict]] = ["'L2Distance'", 100] attribute index_query_params: Dict[str, str] = {} attribute index_type: str = 'annoy' attribute metric: str = 'angular' attribute password: Optional[str] = None attribute port: int = 8123 attribute table: str = 'langchain' attribute username: Optional[str] = None class langchain.vectorstores.DeepLake(dataset_path='./deeplake/', token=None, embedding_function=None, read_only=False, ingestion_batch_size=1000, num_workers=0, verbose=True, exec_option='python', **kwargs)[source] Bases: langchain.vectorstores.base.VectorStore Wrapper around Deep Lake, a data lake for deep learning applications. We integrated deeplake’s similarity search and filtering for fast prototyping, Now, it supports Tensor Query Language (TQL) for production use cases over billion rows. Why Deep Lake? Not only stores embeddings, but also the original data with version control. Serverless, doesn’t require another service and can be used with majorcloud providers (S3, GCS, etc.) More than just a multi-modal vector store. You can use the datasetto fine-tune your own LLM models. To use, you should have the deeplake python package installed. Example from langchain.vectorstores import DeepLake from langchain.embeddings.openai import OpenAIEmbeddings embeddings = OpenAIEmbeddings() vectorstore = DeepLake("langchain_store", embeddings.embed_query) Parameters dataset_path (str) – token (Optional[str]) – embedding_function (Optional[Embeddings]) – read_only (bool) – ingestion_batch_size (int) – num_workers (int) – verbose (bool) – exec_option (str) – kwargs (Any) – Return type None add_texts(texts, metadatas=None, ids=None, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Examples >>> ids = deeplake_vectorstore.add_texts( ... texts = <list_of_texts>, ... metadatas = <list_of_metadata_jsons>, ... ids = <list_of_ids>, ... ) Parameters texts (Iterable[str]) – Texts to add to the vectorstore. metadatas (Optional[List[dict]], optional) – Optional list of metadatas. ids (Optional[List[str]], optional) – Optional list of IDs. **kwargs – other optional keyword arguments. kwargs (Any) – Returns List of IDs of the added texts. Return type List[str] similarity_search(query, k=4, **kwargs)[source] Return docs most similar to query. Examples >>> # Search using an embedding >>> data = vector_store.similarity_search( ... query=<your_query>, ... k=<num_items>, ... exec_option=<preferred_exec_option>, ... ) >>> # Run tql search: >>> data = vector_store.tql_search( ... tql_query="SELECT * WHERE id == <id>", ... exec_option="compute_engine", ... ) Parameters k (int) – Number of Documents to return. Defaults to 4. query (str) – Text to look up similar documents. **kwargs – Additional keyword arguments include: embedding (Callable): Embedding function to use. Defaults to None. distance_metric (str): β€˜L2’ for Euclidean, β€˜L1’ for Nuclear, β€˜max’ for L-infinity, β€˜cos’ for cosine, β€˜dot’ for dot product. Defaults to β€˜L2’. filter (Union[Dict, Callable], optional): Additional filterbefore embedding search. - Dict: Key-value search on tensors of htype json, (sample must satisfy all key-value filters) Dict = {β€œtensor_1”: {β€œkey”: value}, β€œtensor_2”: {β€œkey”: value}} Function: Compatible with deeplake.filter. Defaults to None. exec_option (str): Supports 3 ways to perform searching.’python’, β€˜compute_engine’, or β€˜tensor_db’. Defaults to β€˜python’. - β€˜python’: Pure-python implementation for the client. WARNING: not recommended for big datasets. ’compute_engine’: C++ implementation of the Compute Engine forthe client. Not for in-memory or local datasets. ’tensor_db’: Managed Tensor Database for storage and query.Only for data in Deep Lake Managed Database. Use runtime = {β€œdb_engine”: True} during dataset creation. kwargs (Any) – Returns List of Documents most similar to the query vector. Return type List[Document] similarity_search_by_vector(embedding, k=4, **kwargs)[source] Return docs most similar to embedding vector. Examples >>> # Search using an embedding >>> data = vector_store.similarity_search_by_vector( ... embedding=<your_embedding>, ... k=<num_items_to_return>, ... exec_option=<preferred_exec_option>, ... ) Parameters embedding (Union[List[float], np.ndarray]) – Embedding to find similar docs. k (int) – Number of Documents to return. Defaults to 4. **kwargs – Additional keyword arguments including: filter (Union[Dict, Callable], optional): Additional filter before embedding search. - Dict - Key-value search on tensors of htype json. True if all key-value filters are satisfied. Dict = {β€œtensor_name_1”: {β€œkey”: value}, ”tensor_name_2”: {β€œkey”: value}} Function - Any function compatible withdeeplake.filter. Defaults to None. exec_option (str): Options for search execution include”python”, β€œcompute_engine”, or β€œtensor_db”. Defaults to β€œpython”. - β€œpython” - Pure-python implementation running on the client. Can be used for data stored anywhere. WARNING: using this option with big datasets is discouraged due to potential memory issues. ”compute_engine” - Performant C++ implementation of the DeepLake Compute Engine. Runs on the client and can be used for any data stored in or connected to Deep Lake. It cannot be used with in-memory or local datasets. ”tensor_db” - Performant, fully-hosted Managed Tensor Database.Responsible for storage and query execution. Only available for data stored in the Deep Lake Managed Database. To store datasets in this database, specify runtime = {β€œdb_engine”: True} during dataset creation. distance_metric (str): L2 for Euclidean, L1 for Nuclear,max for L-infinity distance, cos for cosine similarity, β€˜dot’ for dot product. Defaults to L2. kwargs (Any) – Returns List of Documents most similar to the query vector. Return type List[Document] similarity_search_with_score(query, k=4, **kwargs)[source] Run similarity search with Deep Lake with distance returned. Examples: >>> data = vector_store.similarity_search_with_score( … query=<your_query>, … embedding=<your_embedding_function> … k=<number_of_items_to_return>, … exec_option=<preferred_exec_option>, … ) Parameters query (str) – Query text to search for. k (int) – Number of results to return. Defaults to 4. **kwargs – Additional keyword arguments. Some of these arguments are: distance_metric: L2 for Euclidean, L1 for Nuclear, max L-infinity distance, cos for cosine similarity, β€˜dot’ for dot product. Defaults to L2. filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None.embedding_function (Callable): Embedding function to use. Defaults to None. exec_option (str): DeepLakeVectorStore supports 3 ways to performsearching. It could be either β€œpython”, β€œcompute_engine” or β€œtensor_db”. Defaults to β€œpython”. - β€œpython” - Pure-python implementation running on the client. Can be used for data stored anywhere. WARNING: using this option with big datasets is discouraged due to potential memory issues. ”compute_engine” - Performant C++ implementation of the DeepLake Compute Engine. Runs on the client and can be used for any data stored in or connected to Deep Lake. It cannot be used with in-memory or local datasets. ”tensor_db” - Performant, fully-hosted Managed Tensor Database.Responsible for storage and query execution. Only available for data stored in the Deep Lake Managed Database. To store datasets in this database, specify runtime = {β€œdb_engine”: True} during dataset creation. kwargs (Any) – Returns List of documents most similar to the querytext with distance in float. Return type List[Tuple[Document, float]] max_marginal_relevance_search_by_vector(embedding, k=4, fetch_k=20, lambda_mult=0.5, exec_option=None, **kwargs)[source] Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected docs. Examples: >>> data = vector_store.max_marginal_relevance_search_by_vector( … embedding=<your_embedding>, … fetch_k=<elements_to_fetch_before_mmr_search>, … k=<number_of_items_to_return>, … exec_option=<preferred_exec_option>, … ) Parameters embedding (List[float]) – Embedding to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. fetch_k (int) – Number of Documents to fetch for MMR algorithm. lambda_mult (float) – Number between 0 and 1 determining the degree of diversity. 0 corresponds to max diversity and 1 to min diversity. Defaults to 0.5. exec_option (str) – DeepLakeVectorStore supports 3 ways for searching. Could be β€œpython”, β€œcompute_engine” or β€œtensor_db”. Defaults to β€œpython”. - β€œpython” - Pure-python implementation running on the client. Can be used for data stored anywhere. WARNING: using this option with big datasets is discouraged due to potential memory issues. ”compute_engine” - Performant C++ implementation of the DeepLake Compute Engine. Runs on the client and can be used for any data stored in or connected to Deep Lake. It cannot be used with in-memory or local datasets. ”tensor_db” - Performant, fully-hosted Managed Tensor Database.Responsible for storage and query execution. Only available for data stored in the Deep Lake Managed Database. To store datasets in this database, specify runtime = {β€œdb_engine”: True} during dataset creation. **kwargs – Additional keyword arguments. kwargs (Any) – Returns List[Documents] - A list of documents. Return type List[langchain.schema.Document] max_marginal_relevance_search(query, k=4, fetch_k=20, lambda_mult=0.5, exec_option=None, **kwargs)[source] Return docs selected using maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Examples: >>> # Search using an embedding >>> data = vector_store.max_marginal_relevance_search( … query = <query_to_search>, … embedding_function = <embedding_function_for_query>, … k = <number_of_items_to_return>, … exec_option = <preferred_exec_option>, … ) Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. fetch_k (int) – Number of Documents for MMR algorithm. lambda_mult (float) – Value between 0 and 1. 0 corresponds to maximum diversity and 1 to minimum. Defaults to 0.5. exec_option (str) – Supports 3 ways to perform searching. - β€œpython” - Pure-python implementation running on the client. Can be used for data stored anywhere. WARNING: using this option with big datasets is discouraged due to potential memory issues. ”compute_engine” - Performant C++ implementation of the DeepLake Compute Engine. Runs on the client and can be used for any data stored in or connected to Deep Lake. It cannot be used with in-memory or local datasets. ”tensor_db” - Performant, fully-hosted Managed Tensor Database.Responsible for storage and query execution. Only available for data stored in the Deep Lake Managed Database. To store datasets in this database, specify runtime = {β€œdb_engine”: True} during dataset creation. **kwargs – Additional keyword arguments kwargs (Any) – Returns List of Documents selected by maximal marginal relevance. Raises ValueError – when MRR search is on but embedding function is not specified. Return type List[langchain.schema.Document] classmethod from_texts(texts, embedding=None, metadatas=None, ids=None, dataset_path='./deeplake/', **kwargs)[source] Create a Deep Lake dataset from a raw documents. If a dataset_path is specified, the dataset will be persisted in that location, otherwise by default at ./deeplake Examples: >>> # Search using an embedding >>> vector_store = DeepLake.from_texts( … texts = <the_texts_that_you_want_to_embed>, … embedding_function = <embedding_function_for_query>, … k = <number_of_items_to_return>, … exec_option = <preferred_exec_option>, … ) Parameters dataset_path (str) – The full path to the dataset. Can be: Deep Lake cloud path of the form hub://username/dataset_name.To write to Deep Lake cloud datasets, ensure that you are logged in to Deep Lake (use β€˜activeloop login’ from command line) AWS S3 path of the form s3://bucketname/path/to/dataset.Credentials are required in either the environment Google Cloud Storage path of the formgcs://bucketname/path/to/dataset Credentials are required in either the environment Local file system path of the form ./path/to/dataset or~/path/to/dataset or path/to/dataset. In-memory path of the form mem://path/to/dataset which doesn’tsave the dataset, but keeps it in memory instead. Should be used only for testing as it does not persist. texts (List[Document]) – List of documents to add. embedding (Optional[Embeddings]) – Embedding function. Defaults to None. Note, in other places, it is called embedding_function. metadatas (Optional[List[dict]]) – List of metadatas. Defaults to None. ids (Optional[List[str]]) – List of document IDs. Defaults to None. **kwargs – Additional keyword arguments. kwargs (Any) – Returns Deep Lake dataset. Return type DeepLake Raises ValueError – If β€˜embedding’ is provided in kwargs. This is deprecated, please use embedding_function instead. delete(ids=None, filter=None, delete_all=None)[source] Delete the entities in the dataset. Parameters ids (Optional[List[str]], optional) – The document_ids to delete. Defaults to None. filter (Optional[Dict[str, str]], optional) – The filter to delete by. Defaults to None. delete_all (Optional[bool], optional) – Whether to drop the dataset. Defaults to None. Returns Whether the delete operation was successful. Return type bool classmethod force_delete_by_path(path)[source] Force delete dataset by path. Parameters path (str) – path of the dataset to delete. Raises ValueError – if deeplake is not installed. Return type None delete_dataset()[source] Delete the collection. Return type None class langchain.vectorstores.DocArrayHnswSearch(doc_index, embedding)[source] Bases: langchain.vectorstores.docarray.base.DocArrayIndex Wrapper around HnswLib storage. To use it, you should have the docarray package with version >=0.32.0 installed. You can install it with pip install β€œlangchain[docarray]”. Parameters doc_index (BaseDocIndex) – embedding (langchain.embeddings.base.Embeddings) – classmethod from_params(embedding, work_dir, n_dim, dist_metric='cosine', max_elements=1024, index=True, ef_construction=200, ef=10, M=16, allow_replace_deleted=True, num_threads=1, **kwargs)[source] Initialize DocArrayHnswSearch store. Parameters embedding (Embeddings) – Embedding function. work_dir (str) – path to the location where all the data will be stored. n_dim (int) – dimension of an embedding. dist_metric (str) – Distance metric for DocArrayHnswSearch can be one of: β€œcosine”, β€œip”, and β€œl2”. Defaults to β€œcosine”. max_elements (int) – Maximum number of vectors that can be stored. Defaults to 1024. index (bool) – Whether an index should be built for this field. Defaults to True. ef_construction (int) – defines a construction time/accuracy trade-off. Defaults to 200. ef (int) – parameter controlling query time/accuracy trade-off. Defaults to 10. M (int) – parameter that defines the maximum number of outgoing connections in the graph. Defaults to 16. allow_replace_deleted (bool) – Enables replacing of deleted elements with new added ones. Defaults to True. num_threads (int) – Sets the number of cpu threads to use. Defaults to 1. **kwargs – Other keyword arguments to be passed to the get_doc_cls method. kwargs (Any) – Return type langchain.vectorstores.docarray.hnsw.DocArrayHnswSearch classmethod from_texts(texts, embedding, metadatas=None, work_dir=None, n_dim=None, **kwargs)[source] Create an DocArrayHnswSearch store and insert data. Parameters texts (List[str]) – Text data. embedding (Embeddings) – Embedding function. metadatas (Optional[List[dict]]) – Metadata for each text if it exists. Defaults to None. work_dir (str) – path to the location where all the data will be stored. n_dim (int) – dimension of an embedding. **kwargs – Other keyword arguments to be passed to the __init__ method. kwargs (Any) – Returns DocArrayHnswSearch Vector Store Return type langchain.vectorstores.docarray.hnsw.DocArrayHnswSearch class langchain.vectorstores.DocArrayInMemorySearch(doc_index, embedding)[source] Bases: langchain.vectorstores.docarray.base.DocArrayIndex Wrapper around in-memory storage for exact search. To use it, you should have the docarray package with version >=0.32.0 installed. You can install it with pip install β€œlangchain[docarray]”. Parameters doc_index (BaseDocIndex) – embedding (langchain.embeddings.base.Embeddings) – classmethod from_params(embedding, metric='cosine_sim', **kwargs)[source] Initialize DocArrayInMemorySearch store. Parameters embedding (Embeddings) – Embedding function. metric (str) – metric for exact nearest-neighbor search. Can be one of: β€œcosine_sim”, β€œeuclidean_dist” and β€œsqeuclidean_dist”. Defaults to β€œcosine_sim”. **kwargs – Other keyword arguments to be passed to the get_doc_cls method. kwargs (Any) – Return type langchain.vectorstores.docarray.in_memory.DocArrayInMemorySearch classmethod from_texts(texts, embedding, metadatas=None, **kwargs)[source] Create an DocArrayInMemorySearch store and insert data. Parameters texts (List[str]) – Text data. embedding (Embeddings) – Embedding function. metadatas (Optional[List[Dict[Any, Any]]]) – Metadata for each text if it exists. Defaults to None. metric (str) – metric for exact nearest-neighbor search. Can be one of: β€œcosine_sim”, β€œeuclidean_dist” and β€œsqeuclidean_dist”. Defaults to β€œcosine_sim”. kwargs (Any) – Returns DocArrayInMemorySearch Vector Store Return type langchain.vectorstores.docarray.in_memory.DocArrayInMemorySearch class langchain.vectorstores.ElasticVectorSearch(elasticsearch_url, index_name, embedding, *, ssl_verify=None)[source] Bases: langchain.vectorstores.base.VectorStore, abc.ABC Wrapper around Elasticsearch as a vector database. To connect to an Elasticsearch instance that does not require login credentials, pass the Elasticsearch URL and index name along with the embedding object to the constructor. Example from langchain import ElasticVectorSearch from langchain.embeddings import OpenAIEmbeddings embedding = OpenAIEmbeddings() elastic_vector_search = ElasticVectorSearch( elasticsearch_url="http://localhost:9200", index_name="test_index", embedding=embedding ) To connect to an Elasticsearch instance that requires login credentials, including Elastic Cloud, use the Elasticsearch URL format https://username:password@es_host:9243. For example, to connect to Elastic Cloud, create the Elasticsearch URL with the required authentication details and pass it to the ElasticVectorSearch constructor as the named parameter elasticsearch_url. You can obtain your Elastic Cloud URL and login credentials by logging in to the Elastic Cloud console at https://cloud.elastic.co, selecting your deployment, and navigating to the β€œDeployments” page. To obtain your Elastic Cloud password for the default β€œelastic” user: Log in to the Elastic Cloud console at https://cloud.elastic.co Go to β€œSecurity” > β€œUsers” Locate the β€œelastic” user and click β€œEdit” Click β€œReset password” Follow the prompts to reset the password The format for Elastic Cloud URLs is https://username:password@cluster_id.region_id.gcp.cloud.es.io:9243. Example from langchain import ElasticVectorSearch from langchain.embeddings import OpenAIEmbeddings embedding = OpenAIEmbeddings() elastic_host = "cluster_id.region_id.gcp.cloud.es.io" elasticsearch_url = f"https://username:password@{elastic_host}:9243" elastic_vector_search = ElasticVectorSearch( elasticsearch_url=elasticsearch_url, index_name="test_index", embedding=embedding ) Parameters elasticsearch_url (str) – The URL for the Elasticsearch instance. index_name (str) – The name of the Elasticsearch index for the embeddings. embedding (Embeddings) – An object that provides the ability to embed text. It should be an instance of a class that subclasses the Embeddings abstract base class, such as OpenAIEmbeddings() ssl_verify (Optional[Dict[str, Any]]) – Raises ValueError – If the elasticsearch python package is not installed. add_texts(texts, metadatas=None, refresh_indices=True, ids=None, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Parameters texts (Iterable[str]) – Iterable of strings to add to the vectorstore. metadatas (Optional[List[dict]]) – Optional list of metadatas associated with the texts. refresh_indices (bool) – bool to refresh ElasticSearch indices ids (Optional[List[str]]) – kwargs (Any) – Returns List of ids from adding the texts into the vectorstore. Return type List[str] similarity_search(query, k=4, filter=None, **kwargs)[source] Return docs most similar to query. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. filter (Optional[dict]) – kwargs (Any) – Returns List of Documents most similar to the query. Return type List[langchain.schema.Document] similarity_search_with_score(query, k=4, filter=None, **kwargs)[source] Return docs most similar to query. :param query: Text to look up documents similar to. :param k: Number of Documents to return. Defaults to 4. Returns List of Documents most similar to the query. Parameters query (str) – k (int) – filter (Optional[dict]) – kwargs (Any) – Return type List[Tuple[langchain.schema.Document, float]] classmethod from_texts(texts, embedding, metadatas=None, elasticsearch_url=None, index_name=None, refresh_indices=True, **kwargs)[source] Construct ElasticVectorSearch wrapper from raw documents. This is a user-friendly interface that: Embeds documents. Creates a new index for the embeddings in the Elasticsearch instance. Adds the documents to the newly created Elasticsearch index. This is intended to be a quick way to get started. Example from langchain import ElasticVectorSearch from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() elastic_vector_search = ElasticVectorSearch.from_texts( texts, embeddings, elasticsearch_url="http://localhost:9200" ) Parameters texts (List[str]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – elasticsearch_url (Optional[str]) – index_name (Optional[str]) – refresh_indices (bool) – kwargs (Any) – Return type langchain.vectorstores.elastic_vector_search.ElasticVectorSearch create_index(client, index_name, mapping)[source] Parameters client (Any) – index_name (str) – mapping (Dict) – Return type None client_search(client, index_name, script_query, size)[source] Parameters client (Any) – index_name (str) – script_query (Dict) – size (int) – Return type Any delete(ids)[source] Delete by vector IDs. Parameters ids (List[str]) – List of ids to delete. Return type None class langchain.vectorstores.FAISS(embedding_function, index, docstore, index_to_docstore_id, relevance_score_fn=<function _default_relevance_score_fn>, normalize_L2=False)[source] Bases: langchain.vectorstores.base.VectorStore Wrapper around FAISS vector database. To use, you should have the faiss python package installed. Example from langchain import FAISS faiss = FAISS(embedding_function, index, docstore, index_to_docstore_id) Parameters embedding_function (Callable) – index (Any) – docstore (Docstore) – index_to_docstore_id (Dict[int, str]) – relevance_score_fn (Optional[Callable[[float], float]]) – normalize_L2 (bool) – add_texts(texts, metadatas=None, ids=None, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Parameters texts (Iterable[str]) – Iterable of strings to add to the vectorstore. metadatas (Optional[List[dict]]) – Optional list of metadatas associated with the texts. ids (Optional[List[str]]) – Optional list of unique IDs. kwargs (Any) – Returns List of ids from adding the texts into the vectorstore. Return type List[str] add_embeddings(text_embeddings, metadatas=None, ids=None, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Parameters text_embeddings (Iterable[Tuple[str, List[float]]]) – Iterable pairs of string and embedding to add to the vectorstore. metadatas (Optional[List[dict]]) – Optional list of metadatas associated with the texts. ids (Optional[List[str]]) – Optional list of unique IDs. kwargs (Any) – Returns List of ids from adding the texts into the vectorstore. Return type List[str] similarity_search_with_score_by_vector(embedding, k=4, filter=None, fetch_k=20, **kwargs)[source] Return docs most similar to query. Parameters embedding (List[float]) – Embedding vector to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. filter (Optional[Dict[str, Any]]) – Filter by metadata. Defaults to None. fetch_k (int) – (Optional[int]) Number of Documents to fetch before filtering. Defaults to 20. **kwargs – kwargs to be passed to similarity search. Can include: score_threshold: Optional, a floating point value between 0 to 1 to filter the resulting set of retrieved docs kwargs (Any) – Returns List of documents most similar to the query text and L2 distance in float for each. Lower score represents more similarity. Return type List[Tuple[langchain.schema.Document, float]] similarity_search_with_score(query, k=4, filter=None, fetch_k=20, **kwargs)[source] Return docs most similar to query. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. filter (Optional[Dict[str, str]]) – Filter by metadata. Defaults to None. fetch_k (int) – (Optional[int]) Number of Documents to fetch before filtering. Defaults to 20. kwargs (Any) – Returns List of documents most similar to the query text with L2 distance in float. Lower score represents more similarity. Return type List[Tuple[langchain.schema.Document, float]] similarity_search_by_vector(embedding, k=4, filter=None, fetch_k=20, **kwargs)[source] Return docs most similar to embedding vector. Parameters embedding (List[float]) – Embedding to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. filter (Optional[Dict[str, str]]) – Filter by metadata. Defaults to None. fetch_k (int) – (Optional[int]) Number of Documents to fetch before filtering. Defaults to 20. kwargs (Any) – Returns List of Documents most similar to the embedding. Return type List[langchain.schema.Document] similarity_search(query, k=4, filter=None, fetch_k=20, **kwargs)[source] Return docs most similar to query. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. filter (Optional[Dict[str, Any]]) – (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. fetch_k (int) – (Optional[int]) Number of Documents to fetch before filtering. Defaults to 20. kwargs (Any) – Returns List of Documents most similar to the query. Return type List[langchain.schema.Document] max_marginal_relevance_search_by_vector(embedding, k=4, fetch_k=20, lambda_mult=0.5, filter=None, **kwargs)[source] Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Parameters embedding (List[float]) – Embedding to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. fetch_k (int) – Number of Documents to fetch before filtering to pass to MMR algorithm. lambda_mult (float) – Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. filter (Optional[Dict[str, Any]]) – kwargs (Any) – Returns List of Documents selected by maximal marginal relevance. Return type List[langchain.schema.Document] max_marginal_relevance_search(query, k=4, fetch_k=20, lambda_mult=0.5, filter=None, **kwargs)[source] Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. fetch_k (int) – Number of Documents to fetch before filtering (if needed) to pass to MMR algorithm. lambda_mult (float) – Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. filter (Optional[Dict[str, Any]]) – kwargs (Any) – Returns List of Documents selected by maximal marginal relevance. Return type List[langchain.schema.Document] merge_from(target)[source] Merge another FAISS object with the current one. Add the target FAISS to the current one. Parameters target (langchain.vectorstores.faiss.FAISS) – FAISS object you wish to merge into the current one Returns None. Return type None classmethod from_texts(texts, embedding, metadatas=None, ids=None, **kwargs)[source] Construct FAISS wrapper from raw documents. This is a user friendly interface that: Embeds documents. Creates an in memory docstore Initializes the FAISS database This is intended to be a quick way to get started. Example from langchain import FAISS from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() faiss = FAISS.from_texts(texts, embeddings) Parameters texts (List[str]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – ids (Optional[List[str]]) – kwargs (Any) – Return type langchain.vectorstores.faiss.FAISS classmethod from_embeddings(text_embeddings, embedding, metadatas=None, ids=None, **kwargs)[source] Construct FAISS wrapper from raw documents. This is a user friendly interface that: Embeds documents. Creates an in memory docstore Initializes the FAISS database This is intended to be a quick way to get started. Example from langchain import FAISS from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() text_embeddings = embeddings.embed_documents(texts) text_embedding_pairs = list(zip(texts, text_embeddings)) faiss = FAISS.from_embeddings(text_embedding_pairs, embeddings) Parameters text_embeddings (List[Tuple[str, List[float]]]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – ids (Optional[List[str]]) – kwargs (Any) – Return type langchain.vectorstores.faiss.FAISS save_local(folder_path, index_name='index')[source] Save FAISS index, docstore, and index_to_docstore_id to disk. Parameters folder_path (str) – folder path to save index, docstore, and index_to_docstore_id to. index_name (str) – for saving with a specific index file name Return type None classmethod load_local(folder_path, embeddings, index_name='index')[source] Load FAISS index, docstore, and index_to_docstore_id from disk. Parameters folder_path (str) – folder path to load index, docstore, and index_to_docstore_id from. embeddings (langchain.embeddings.base.Embeddings) – Embeddings to use when generating queries index_name (str) – for saving with a specific index file name Return type langchain.vectorstores.faiss.FAISS class langchain.vectorstores.Hologres(connection_string, embedding_function, ndims=1536, table_name='langchain_pg_embedding', pre_delete_table=False, logger=None)[source] Bases: langchain.vectorstores.base.VectorStore VectorStore implementation using Hologres. connection_string is a hologres connection string. embedding_function any embedding function implementinglangchain.embeddings.base.Embeddings interface. ndims is the number of dimensions of the embedding output. table_name is the name of the table to store embeddings and data.(default: langchain_pg_embedding) - NOTE: The table will be created when initializing the store (if not exists) So, make sure the user has the right permissions to create tables. pre_delete_table if True, will delete the table if it exists.(default: False) - Useful for testing. Parameters connection_string (str) – embedding_function (Embeddings) – ndims (int) – table_name (str) – pre_delete_table (bool) – logger (Optional[logging.Logger]) – Return type None create_vector_extension()[source] Return type None create_table()[source] Return type None add_embeddings(texts, embeddings, metadatas, ids, **kwargs)[source] Add embeddings to the vectorstore. Parameters texts (Iterable[str]) – Iterable of strings to add to the vectorstore. embeddings (List[List[float]]) – List of list of embedding vectors. metadatas (List[dict]) – List of metadatas associated with the texts. kwargs (Any) – vectorstore specific parameters ids (List[str]) – Return type None add_texts(texts, metadatas=None, ids=None, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Parameters texts (Iterable[str]) – Iterable of strings to add to the vectorstore. metadatas (Optional[List[dict]]) – Optional list of metadatas associated with the texts. kwargs (Any) – vectorstore specific parameters ids (Optional[List[str]]) – Returns List of ids from adding the texts into the vectorstore. Return type List[str] similarity_search(query, k=4, filter=None, **kwargs)[source] Run similarity search with Hologres with distance. Parameters query (str) – Query text to search for. k (int) – Number of results to return. Defaults to 4. filter (Optional[Dict[str, str]]) – Filter by metadata. Defaults to None. kwargs (Any) – Returns List of Documents most similar to the query. Return type List[langchain.schema.Document] similarity_search_by_vector(embedding, k=4, filter=None, **kwargs)[source] Return docs most similar to embedding vector. Parameters embedding (List[float]) – Embedding to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. filter (Optional[Dict[str, str]]) – Filter by metadata. Defaults to None. kwargs (Any) – Returns List of Documents most similar to the query vector. Return type List[langchain.schema.Document] similarity_search_with_score(query, k=4, filter=None)[source] Return docs most similar to query. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. filter (Optional[Dict[str, str]]) – Filter by metadata. Defaults to None. Returns List of Documents most similar to the query and score for each Return type List[Tuple[langchain.schema.Document, float]] similarity_search_with_score_by_vector(embedding, k=4, filter=None)[source] Parameters embedding (List[float]) – k (int) – filter (Optional[dict]) – Return type List[Tuple[langchain.schema.Document, float]] classmethod from_texts(texts, embedding, metadatas=None, ndims=1536, table_name='langchain_pg_embedding', ids=None, pre_delete_table=False, **kwargs)[source] Return VectorStore initialized from texts and embeddings. Postgres connection string is required β€œEither pass it as a parameter or set the HOLOGRES_CONNECTION_STRING environment variable. Parameters texts (List[str]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – ndims (int) – table_name (str) – ids (Optional[List[str]]) – pre_delete_table (bool) – kwargs (Any) – Return type langchain.vectorstores.hologres.Hologres classmethod from_embeddings(text_embeddings, embedding, metadatas=None, ndims=1536, table_name='langchain_pg_embedding', ids=None, pre_delete_table=False, **kwargs)[source] Construct Hologres wrapper from raw documents and pre- generated embeddings. Return VectorStore initialized from documents and embeddings. Postgres connection string is required β€œEither pass it as a parameter or set the HOLOGRES_CONNECTION_STRING environment variable. Example from langchain import Hologres from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() text_embeddings = embeddings.embed_documents(texts) text_embedding_pairs = list(zip(texts, text_embeddings)) faiss = Hologres.from_embeddings(text_embedding_pairs, embeddings) Parameters text_embeddings (List[Tuple[str, List[float]]]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – ndims (int) – table_name (str) – ids (Optional[List[str]]) – pre_delete_table (bool) – kwargs (Any) – Return type langchain.vectorstores.hologres.Hologres classmethod from_existing_index(embedding, ndims=1536, table_name='langchain_pg_embedding', pre_delete_table=False, **kwargs)[source] Get intsance of an existing Hologres store.This method will return the instance of the store without inserting any new embeddings Parameters embedding (langchain.embeddings.base.Embeddings) – ndims (int) – table_name (str) – pre_delete_table (bool) – kwargs (Any) – Return type langchain.vectorstores.hologres.Hologres classmethod get_connection_string(kwargs)[source] Parameters kwargs (Dict[str, Any]) – Return type str classmethod from_documents(documents, embedding, ndims=1536, table_name='langchain_pg_embedding', ids=None, pre_delete_collection=False, **kwargs)[source] Return VectorStore initialized from documents and embeddings. Postgres connection string is required β€œEither pass it as a parameter or set the HOLOGRES_CONNECTION_STRING environment variable. Parameters documents (List[langchain.schema.Document]) – embedding (langchain.embeddings.base.Embeddings) – ndims (int) – table_name (str) – ids (Optional[List[str]]) – pre_delete_collection (bool) – kwargs (Any) – Return type langchain.vectorstores.hologres.Hologres classmethod connection_string_from_db_params(host, port, database, user, password)[source] Return connection string from database parameters. Parameters host (str) – port (int) – database (str) – user (str) – password (str) – Return type str class langchain.vectorstores.LanceDB(connection, embedding, vector_key='vector', id_key='id', text_key='text')[source] Bases: langchain.vectorstores.base.VectorStore Wrapper around LanceDB vector database. To use, you should have lancedb python package installed. Example db = lancedb.connect('./lancedb') table = db.open_table('my_table') vectorstore = LanceDB(table, embedding_function) vectorstore.add_texts(['text1', 'text2']) result = vectorstore.similarity_search('text1') Parameters connection (Any) – embedding (Embeddings) – vector_key (Optional[str]) – id_key (Optional[str]) – text_key (Optional[str]) – add_texts(texts, metadatas=None, ids=None, **kwargs)[source] Turn texts into embedding and add it to the database Parameters texts (Iterable[str]) – Iterable of strings to add to the vectorstore. metadatas (Optional[List[dict]]) – Optional list of metadatas associated with the texts. ids (Optional[List[str]]) – Optional list of ids to associate with the texts. kwargs (Any) – Returns List of ids of the added texts. Return type List[str] similarity_search(query, k=4, **kwargs)[source] Return documents most similar to the query Parameters query (str) – String to query the vectorstore with. k (int) – Number of documents to return. kwargs (Any) – Returns List of documents most similar to the query. Return type List[langchain.schema.Document] classmethod from_texts(texts, embedding, metadatas=None, connection=None, vector_key='vector', id_key='id', text_key='text', **kwargs)[source] Return VectorStore initialized from texts and embeddings. Parameters texts (List[str]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – connection (Any) – vector_key (Optional[str]) – id_key (Optional[str]) – text_key (Optional[str]) – kwargs (Any) – Return type langchain.vectorstores.lancedb.LanceDB class langchain.vectorstores.MatchingEngine(project_id, index, endpoint, embedding, gcs_client, gcs_bucket_name, credentials=None)[source] Bases: langchain.vectorstores.base.VectorStore Vertex Matching Engine implementation of the vector store. While the embeddings are stored in the Matching Engine, the embedded documents will be stored in GCS. An existing Index and corresponding Endpoint are preconditions for using this module. See usage in docs/modules/indexes/vectorstores/examples/matchingengine.ipynb Note that this implementation is mostly meant for reading if you are planning to do a real time implementation. While reading is a real time operation, updating the index takes close to one hour. Parameters project_id (str) – index (MatchingEngineIndex) – endpoint (MatchingEngineIndexEndpoint) – embedding (Embeddings) – gcs_client (storage.Client) – gcs_bucket_name (str) – credentials (Optional[Credentials]) – add_texts(texts, metadatas=None, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Parameters texts (Iterable[str]) – Iterable of strings to add to the vectorstore. metadatas (Optional[List[dict]]) – Optional list of metadatas associated with the texts. kwargs (Any) – vectorstore specific parameters. Returns List of ids from adding the texts into the vectorstore. Return type List[str] similarity_search(query, k=4, **kwargs)[source] Return docs most similar to query. Parameters query (str) – The string that will be used to search for similar documents. k (int) – The amount of neighbors that will be retrieved. kwargs (Any) – Returns A list of k matching documents. Return type List[langchain.schema.Document] classmethod from_texts(texts, embedding, metadatas=None, **kwargs)[source] Use from components instead. Parameters texts (List[str]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – kwargs (Any) – Return type langchain.vectorstores.matching_engine.MatchingEngine classmethod from_components(project_id, region, gcs_bucket_name, index_id, endpoint_id, credentials_path=None, embedding=None)[source] Takes the object creation out of the constructor. Parameters project_id (str) – The GCP project id. region (str) – The default location making the API calls. It must have regional. (the same location as the GCS bucket and must be) – gcs_bucket_name (str) – The location where the vectors will be stored in created. (order for the index to be) – index_id (str) – The id of the created index. endpoint_id (str) – The id of the created endpoint. credentials_path (Optional[str]) – (Optional) The path of the Google credentials on system. (the local file) – embedding (Optional[langchain.embeddings.base.Embeddings]) – The Embeddings that will be used for texts. (embedding the) – Returns A configured MatchingEngine with the texts added to the index. Return type langchain.vectorstores.matching_engine.MatchingEngine class langchain.vectorstores.Milvus(embedding_function, collection_name='LangChainCollection', connection_args=None, consistency_level='Session', index_params=None, search_params=None, drop_old=False)[source] Bases: langchain.vectorstores.base.VectorStore Wrapper around the Milvus vector database. Parameters embedding_function (Embeddings) – collection_name (str) – connection_args (Optional[dict[str, Any]]) – consistency_level (str) – index_params (Optional[dict]) – search_params (Optional[dict]) – drop_old (Optional[bool]) – add_texts(texts, metadatas=None, timeout=None, batch_size=1000, **kwargs)[source] Insert text data into Milvus. Inserting data when the collection has not be made yet will result in creating a new Collection. The data of the first entity decides the schema of the new collection, the dim is extracted from the first embedding and the columns are decided by the first metadata dict. Metada keys will need to be present for all inserted values. At the moment there is no None equivalent in Milvus. Parameters texts (Iterable[str]) – The texts to embed, it is assumed that they all fit in memory. metadatas (Optional[List[dict]]) – Metadata dicts attached to each of the texts. Defaults to None. timeout (Optional[int]) – Timeout for each batch insert. Defaults to None. batch_size (int, optional) – Batch size to use for insertion. Defaults to 1000. kwargs (Any) – Raises MilvusException – Failure to add texts Returns The resulting keys for each inserted element. Return type List[str] similarity_search(query, k=4, param=None, expr=None, timeout=None, **kwargs)[source] Perform a similarity search against the query string. Parameters query (str) – The text to search. k (int, optional) – How many results to return. Defaults to 4. param (dict, optional) – The search params for the index type. Defaults to None. expr (str, optional) – Filtering expression. Defaults to None. timeout (int, optional) – How long to wait before timeout error. Defaults to None. kwargs (Any) – Collection.search() keyword arguments. Returns Document results for search. Return type List[Document] similarity_search_by_vector(embedding, k=4, param=None, expr=None, timeout=None, **kwargs)[source] Perform a similarity search against the query string. Parameters embedding (List[float]) – The embedding vector to search. k (int, optional) – How many results to return. Defaults to 4. param (dict, optional) – The search params for the index type. Defaults to None. expr (str, optional) – Filtering expression. Defaults to None. timeout (int, optional) – How long to wait before timeout error. Defaults to None. kwargs (Any) – Collection.search() keyword arguments. Returns Document results for search. Return type List[Document] similarity_search_with_score(query, k=4, param=None, expr=None, timeout=None, **kwargs)[source] Perform a search on a query string and return results with score. For more information about the search parameters, take a look at the pymilvus documentation found here: https://milvus.io/api-reference/pymilvus/v2.2.6/Collection/search().md Parameters query (str) – The text being searched. k (int, optional) – The amount of results ot return. Defaults to 4. param (dict) – The search params for the specified index. Defaults to None. expr (str, optional) – Filtering expression. Defaults to None. timeout (int, optional) – How long to wait before timeout error. Defaults to None. kwargs (Any) – Collection.search() keyword arguments. Return type List[float], List[Tuple[Document, any, any]] similarity_search_with_score_by_vector(embedding, k=4, param=None, expr=None, timeout=None, **kwargs)[source] Perform a search on a query string and return results with score. For more information about the search parameters, take a look at the pymilvus documentation found here: https://milvus.io/api-reference/pymilvus/v2.2.6/Collection/search().md Parameters embedding (List[float]) – The embedding vector being searched. k (int, optional) – The amount of results ot return. Defaults to 4. param (dict) – The search params for the specified index. Defaults to None. expr (str, optional) – Filtering expression. Defaults to None. timeout (int, optional) – How long to wait before timeout error. Defaults to None. kwargs (Any) – Collection.search() keyword arguments. Returns Result doc and score. Return type List[Tuple[Document, float]] max_marginal_relevance_search(query, k=4, fetch_k=20, lambda_mult=0.5, param=None, expr=None, timeout=None, **kwargs)[source] Perform a search and return results that are reordered by MMR. Parameters query (str) – The text being searched. k (int, optional) – How many results to give. Defaults to 4. fetch_k (int, optional) – Total results to select k from. Defaults to 20. lambda_mult (float) – Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5 param (dict, optional) – The search params for the specified index. Defaults to None. expr (str, optional) – Filtering expression. Defaults to None. timeout (int, optional) – How long to wait before timeout error. Defaults to None. kwargs (Any) – Collection.search() keyword arguments. Returns Document results for search. Return type List[Document] max_marginal_relevance_search_by_vector(embedding, k=4, fetch_k=20, lambda_mult=0.5, param=None, expr=None, timeout=None, **kwargs)[source] Perform a search and return results that are reordered by MMR. Parameters embedding (str) – The embedding vector being searched. k (int, optional) – How many results to give. Defaults to 4. fetch_k (int, optional) – Total results to select k from. Defaults to 20. lambda_mult (float) – Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5 param (dict, optional) – The search params for the specified index. Defaults to None. expr (str, optional) – Filtering expression. Defaults to None. timeout (int, optional) – How long to wait before timeout error. Defaults to None. kwargs (Any) – Collection.search() keyword arguments. Returns Document results for search. Return type List[Document] classmethod from_texts(texts, embedding, metadatas=None, collection_name='LangChainCollection', connection_args={'host': 'localhost', 'password': '', 'port': '19530', 'secure': False, 'user': ''}, consistency_level='Session', index_params=None, search_params=None, drop_old=False, **kwargs)[source] Create a Milvus collection, indexes it with HNSW, and insert data. Parameters texts (List[str]) – Text data. embedding (Embeddings) – Embedding function. metadatas (Optional[List[dict]]) – Metadata for each text if it exists. Defaults to None. collection_name (str, optional) – Collection name to use. Defaults to β€œLangChainCollection”. connection_args (dict[str, Any], optional) – Connection args to use. Defaults to DEFAULT_MILVUS_CONNECTION. consistency_level (str, optional) – Which consistency level to use. Defaults to β€œSession”. index_params (Optional[dict], optional) – Which index_params to use. Defaults to None. search_params (Optional[dict], optional) – Which search params to use. Defaults to None. drop_old (Optional[bool], optional) – Whether to drop the collection with that name if it exists. Defaults to False. kwargs (Any) – Returns Milvus Vector Store Return type Milvus class langchain.vectorstores.Zilliz(embedding_function, collection_name='LangChainCollection', connection_args=None, consistency_level='Session', index_params=None, search_params=None, drop_old=False)[source] Bases: langchain.vectorstores.milvus.Milvus Parameters embedding_function (Embeddings) – collection_name (str) – connection_args (Optional[dict[str, Any]]) – consistency_level (str) – index_params (Optional[dict]) – search_params (Optional[dict]) – drop_old (Optional[bool]) – classmethod from_texts(texts, embedding, metadatas=None, collection_name='LangChainCollection', connection_args={}, consistency_level='Session', index_params=None, search_params=None, drop_old=False, **kwargs)[source] Create a Zilliz collection, indexes it with HNSW, and insert data. Parameters texts (List[str]) – Text data. embedding (Embeddings) – Embedding function. metadatas (Optional[List[dict]]) – Metadata for each text if it exists. Defaults to None. collection_name (str, optional) – Collection name to use. Defaults to β€œLangChainCollection”. connection_args (dict[str, Any], optional) – Connection args to use. Defaults to DEFAULT_MILVUS_CONNECTION. consistency_level (str, optional) – Which consistency level to use. Defaults to β€œSession”. index_params (Optional[dict], optional) – Which index_params to use. Defaults to None. search_params (Optional[dict], optional) – Which search params to use. Defaults to None. drop_old (Optional[bool], optional) – Whether to drop the collection with that name if it exists. Defaults to False. kwargs (Any) – Returns Zilliz Vector Store Return type Zilliz class langchain.vectorstores.SingleStoreDB(embedding, *, distance_strategy=DistanceStrategy.DOT_PRODUCT, table_name='embeddings', content_field='content', metadata_field='metadata', vector_field='vector', pool_size=5, max_overflow=10, timeout=30, **kwargs)[source] Bases: langchain.vectorstores.base.VectorStore This class serves as a Pythonic interface to the SingleStore DB database. The prerequisite for using this class is the installation of the singlestoredb Python package. The SingleStoreDB vectorstore can be created by providing an embedding function and the relevant parameters for the database connection, connection pool, and optionally, the names of the table and the fields to use. Parameters embedding (Embeddings) – distance_strategy (DistanceStrategy) – table_name (str) – content_field (str) – metadata_field (str) – vector_field (str) – pool_size (int) – max_overflow (int) – timeout (float) – kwargs (Any) – vector_field Pass the rest of the kwargs to the connection. connection_kwargs Add program name and version to connection attributes. add_texts(texts, metadatas=None, embeddings=None, **kwargs)[source] Add more texts to the vectorstore. Parameters texts (Iterable[str]) – Iterable of strings/text to add to the vectorstore. metadatas (Optional[List[dict]], optional) – Optional list of metadatas. Defaults to None. embeddings (Optional[List[List[float]]], optional) – Optional pre-generated embeddings. Defaults to None. kwargs (Any) – Returns empty list Return type List[str] similarity_search(query, k=4, filter=None, **kwargs)[source] Returns the most similar indexed documents to the query text. Uses cosine similarity. Parameters query (str) – The query text for which to find similar documents. k (int) – The number of documents to return. Default is 4. filter (dict) – A dictionary of metadata fields and values to filter by. kwargs (Any) – Returns A list of documents that are most similar to the query text. Return type List[Document] Examples similarity_search_with_score(query, k=4, filter=None)[source] Return docs most similar to query. Uses cosine similarity. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. filter (Optional[dict]) – A dictionary of metadata fields and values to filter by. Defaults to None. Returns List of Documents most similar to the query and score for each Return type List[Tuple[langchain.schema.Document, float]] classmethod from_texts(texts, embedding, metadatas=None, distance_strategy=DistanceStrategy.DOT_PRODUCT, table_name='embeddings', content_field='content', metadata_field='metadata', vector_field='vector', pool_size=5, max_overflow=10, timeout=30, **kwargs)[source] Create a SingleStoreDB vectorstore from raw documents. This is a user-friendly interface that: Embeds documents. Creates a new table for the embeddings in SingleStoreDB. Adds the documents to the newly created table. This is intended to be a quick way to get started. .. rubric:: Example Parameters texts (List[str]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – distance_strategy (langchain.vectorstores.singlestoredb.DistanceStrategy) – table_name (str) – content_field (str) – metadata_field (str) – vector_field (str) – pool_size (int) – max_overflow (int) – timeout (float) – kwargs (Any) – Return type langchain.vectorstores.singlestoredb.SingleStoreDB as_retriever(**kwargs)[source] Parameters kwargs (Any) – Return type langchain.vectorstores.singlestoredb.SingleStoreDBRetriever class langchain.vectorstores.Clarifai(user_id=None, app_id=None, pat=None, number_of_docs=None, api_base=None)[source] Bases: langchain.vectorstores.base.VectorStore Wrapper around Clarifai AI platform’s vector store. To use, you should have the clarifai python package installed. Example from langchain.vectorstores import Clarifai from langchain.embeddings.openai import OpenAIEmbeddings embeddings = OpenAIEmbeddings() vectorstore = Clarifai("langchain_store", embeddings.embed_query) Parameters user_id (Optional[str]) – app_id (Optional[str]) – pat (Optional[str]) – number_of_docs (Optional[int]) – api_base (Optional[str]) – Return type None add_texts(texts, metadatas=None, ids=None, **kwargs)[source] Add texts to the Clarifai vectorstore. This will push the text to a Clarifai application. Application use base workflow that create and store embedding for each text. Make sure you are using a base workflow that is compatible with text (such as Language Understanding). Parameters texts (Iterable[str]) – Texts to add to the vectorstore. metadatas (Optional[List[dict]], optional) – Optional list of metadatas. ids (Optional[List[str]], optional) – Optional list of IDs. kwargs (Any) – Returns List of IDs of the added texts. Return type List[str] similarity_search_with_score(query, k=4, filter=None, namespace=None, **kwargs)[source] Run similarity search with score using Clarifai. Parameters query (str) – Query text to search for. k (int) – Number of results to return. Defaults to 4. filter (Optional[Dict[str, str]]) – Filter by metadata. None. (Defaults to) – namespace (Optional[str]) – kwargs (Any) – Returns List of documents most simmilar to the query text. Return type List[Document] similarity_search(query, k=4, **kwargs)[source] Run similarity search using Clarifai. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. kwargs (Any) – Returns List of Documents most similar to the query and score for each Return type List[langchain.schema.Document] classmethod from_texts(texts, embedding=None, metadatas=None, user_id=None, app_id=None, pat=None, number_of_docs=None, api_base=None, **kwargs)[source] Create a Clarifai vectorstore from a list of texts. Parameters user_id (str) – User ID. app_id (str) – App ID. texts (List[str]) – List of texts to add. pat (Optional[str]) – Personal access token. Defaults to None. number_of_docs (Optional[int]) – Number of documents to return None. (Defaults to) – api_base (Optional[str]) – API base. Defaults to None. metadatas (Optional[List[dict]]) – Optional list of metadatas. None. – embedding (Optional[langchain.embeddings.base.Embeddings]) – kwargs (Any) – Returns Clarifai vectorstore. Return type Clarifai classmethod from_documents(documents, embedding=None, user_id=None, app_id=None, pat=None, number_of_docs=None, api_base=None, **kwargs)[source] Create a Clarifai vectorstore from a list of documents. Parameters user_id (str) – User ID. app_id (str) – App ID. documents (List[Document]) – List of documents to add. pat (Optional[str]) – Personal access token. Defaults to None. number_of_docs (Optional[int]) – Number of documents to return None. (during vector search. Defaults to) – api_base (Optional[str]) – API base. Defaults to None. embedding (Optional[langchain.embeddings.base.Embeddings]) – kwargs (Any) – Returns Clarifai vectorstore. Return type Clarifai class langchain.vectorstores.OpenSearchVectorSearch(opensearch_url, index_name, embedding_function, **kwargs)[source] Bases: langchain.vectorstores.base.VectorStore Wrapper around OpenSearch as a vector database. Example from langchain import OpenSearchVectorSearch opensearch_vector_search = OpenSearchVectorSearch( "http://localhost:9200", "embeddings", embedding_function ) Parameters opensearch_url (str) – index_name (str) – embedding_function (Embeddings) – kwargs (Any) – add_texts(texts, metadatas=None, ids=None, bulk_size=500, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Parameters texts (Iterable[str]) – Iterable of strings to add to the vectorstore. metadatas (Optional[List[dict]]) – Optional list of metadatas associated with the texts. ids (Optional[List[str]]) – Optional list of ids to associate with the texts. bulk_size (int) – Bulk API request count; Default: 500 kwargs (Any) – Returns List of ids from adding the texts into the vectorstore. Return type List[str] Optional Args:vector_field: Document field embeddings are stored in. Defaults to β€œvector_field”. text_field: Document field the text of the document is stored in. Defaults to β€œtext”. similarity_search(query, k=4, **kwargs)[source] Return docs most similar to query. By default, supports Approximate Search. Also supports Script Scoring and Painless Scripting. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. kwargs (Any) – Returns List of Documents most similar to the query. Return type List[langchain.schema.Document] Optional Args:vector_field: Document field embeddings are stored in. Defaults to β€œvector_field”. text_field: Document field the text of the document is stored in. Defaults to β€œtext”. metadata_field: Document field that metadata is stored in. Defaults to β€œmetadata”. Can be set to a special value β€œ*” to include the entire document. Optional Args for Approximate Search:search_type: β€œapproximate_search”; default: β€œapproximate_search” boolean_filter: A Boolean filter consists of a Boolean query that contains a k-NN query and a filter. subquery_clause: Query clause on the knn vector field; default: β€œmust” lucene_filter: the Lucene algorithm decides whether to perform an exact k-NN search with pre-filtering or an approximate search with modified post-filtering. Optional Args for Script Scoring Search:search_type: β€œscript_scoring”; default: β€œapproximate_search” space_type: β€œl2”, β€œl1”, β€œlinf”, β€œcosinesimil”, β€œinnerproduct”, β€œhammingbit”; default: β€œl2” pre_filter: script_score query to pre-filter documents before identifying nearest neighbors; default: {β€œmatch_all”: {}} Optional Args for Painless Scripting Search:search_type: β€œpainless_scripting”; default: β€œapproximate_search” space_type: β€œl2Squared”, β€œl1Norm”, β€œcosineSimilarity”; default: β€œl2Squared” pre_filter: script_score query to pre-filter documents before identifying nearest neighbors; default: {β€œmatch_all”: {}} similarity_search_with_score(query, k=4, **kwargs)[source] Return docs and it’s scores most similar to query. By default, supports Approximate Search. Also supports Script Scoring and Painless Scripting. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. kwargs (Any) – Returns List of Documents along with its scores most similar to the query. Return type List[Tuple[langchain.schema.Document, float]] Optional Args:same as similarity_search max_marginal_relevance_search(query, k=4, fetch_k=20, lambda_mult=0.5, **kwargs)[source] Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. fetch_k (int) – Number of Documents to fetch to pass to MMR algorithm. Defaults to 20. lambda_mult (float) – Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. kwargs (Any) – Returns List of Documents selected by maximal marginal relevance. Return type list[langchain.schema.Document] classmethod from_texts(texts, embedding, metadatas=None, bulk_size=500, **kwargs)[source] Construct OpenSearchVectorSearch wrapper from raw documents. Example from langchain import OpenSearchVectorSearch from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() opensearch_vector_search = OpenSearchVectorSearch.from_texts( texts, embeddings, opensearch_url="http://localhost:9200" ) OpenSearch by default supports Approximate Search powered by nmslib, faiss and lucene engines recommended for large datasets. Also supports brute force search through Script Scoring and Painless Scripting. Optional Args:vector_field: Document field embeddings are stored in. Defaults to β€œvector_field”. text_field: Document field the text of the document is stored in. Defaults to β€œtext”. Optional Keyword Args for Approximate Search:engine: β€œnmslib”, β€œfaiss”, β€œlucene”; default: β€œnmslib” space_type: β€œl2”, β€œl1”, β€œcosinesimil”, β€œlinf”, β€œinnerproduct”; default: β€œl2” ef_search: Size of the dynamic list used during k-NN searches. Higher values lead to more accurate but slower searches; default: 512 ef_construction: Size of the dynamic list used during k-NN graph creation. Higher values lead to more accurate graph but slower indexing speed; default: 512 m: Number of bidirectional links created for each new element. Large impact on memory consumption. Between 2 and 100; default: 16 Keyword Args for Script Scoring or Painless Scripting:is_appx_search: False Parameters texts (List[str]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – bulk_size (int) – kwargs (Any) – Return type langchain.vectorstores.opensearch_vector_search.OpenSearchVectorSearch class langchain.vectorstores.MongoDBAtlasVectorSearch(collection, embedding, *, index_name='default', text_key='text', embedding_key='embedding')[source] Bases: langchain.vectorstores.base.VectorStore Wrapper around MongoDB Atlas Vector Search. To use, you should have both: - the pymongo python package installed - a connection string associated with a MongoDB Atlas Cluster having deployed an Atlas Search index Example from langchain.vectorstores import MongoDBAtlasVectorSearch from langchain.embeddings.openai import OpenAIEmbeddings from pymongo import MongoClient mongo_client = MongoClient("<YOUR-CONNECTION-STRING>") collection = mongo_client["<db_name>"]["<collection_name>"] embeddings = OpenAIEmbeddings() vectorstore = MongoDBAtlasVectorSearch(collection, embeddings) Parameters collection (Collection[MongoDBDocumentType]) – embedding (Embeddings) – index_name (str) – text_key (str) – embedding_key (str) – classmethod from_connection_string(connection_string, namespace, embedding, **kwargs)[source] Parameters connection_string (str) – namespace (str) – embedding (langchain.embeddings.base.Embeddings) – kwargs (Any) – Return type langchain.vectorstores.mongodb_atlas.MongoDBAtlasVectorSearch add_texts(texts, metadatas=None, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Parameters texts (Iterable[str]) – Iterable of strings to add to the vectorstore. metadatas (Optional[List[Dict[str, Any]]]) – Optional list of metadatas associated with the texts. kwargs (Any) – Returns List of ids from adding the texts into the vectorstore. Return type List similarity_search_with_score(query, *, k=4, pre_filter=None, post_filter_pipeline=None)[source] Return MongoDB documents most similar to query, along with scores. Use the knnBeta Operator available in MongoDB Atlas Search This feature is in early access and available only for evaluation purposes, to validate functionality, and to gather feedback from a small closed group of early access users. It is not recommended for production deployments as we may introduce breaking changes. For more: https://www.mongodb.com/docs/atlas/atlas-search/knn-beta Parameters query (str) – Text to look up documents similar to. k (int) – Optional Number of Documents to return. Defaults to 4. pre_filter (Optional[dict]) – Optional Dictionary of argument(s) to prefilter on document fields. post_filter_pipeline (Optional[List[Dict]]) – Optional Pipeline of MongoDB aggregation stages following the knnBeta search. Returns List of Documents most similar to the query and score for each Return type List[Tuple[langchain.schema.Document, float]] similarity_search(query, k=4, pre_filter=None, post_filter_pipeline=None, **kwargs)[source] Return MongoDB documents most similar to query. Use the knnBeta Operator available in MongoDB Atlas Search This feature is in early access and available only for evaluation purposes, to validate functionality, and to gather feedback from a small closed group of early access users. It is not recommended for production deployments as we may introduce breaking changes. For more: https://www.mongodb.com/docs/atlas/atlas-search/knn-beta Parameters query (str) – Text to look up documents similar to. k (int) – Optional Number of Documents to return. Defaults to 4. pre_filter (Optional[dict]) – Optional Dictionary of argument(s) to prefilter on document fields. post_filter_pipeline (Optional[List[Dict]]) – Optional Pipeline of MongoDB aggregation stages following the knnBeta search. kwargs (Any) – Returns List of Documents most similar to the query and score for each Return type List[langchain.schema.Document] classmethod from_texts(texts, embedding, metadatas=None, collection=None, **kwargs)[source] Construct MongoDBAtlasVectorSearch wrapper from raw documents. This is a user-friendly interface that: Embeds documents. Adds the documents to a provided MongoDB Atlas Vector Search index(Lucene) This is intended to be a quick way to get started. Example Parameters texts (List[str]) – embedding (Embeddings) – metadatas (Optional[List[dict]]) – collection (Optional[Collection[MongoDBDocumentType]]) – kwargs (Any) – Return type MongoDBAtlasVectorSearch class langchain.vectorstores.MyScale(embedding, config=None, **kwargs)[source] Bases: langchain.vectorstores.base.VectorStore Wrapper around MyScale vector database You need a clickhouse-connect python package, and a valid account to connect to MyScale. MyScale can not only search with simple vector indexes, it also supports complex query with multiple conditions, constraints and even sub-queries. For more information, please visit[myscale official site](https://docs.myscale.com/en/overview/) Parameters embedding (Embeddings) – config (Optional[MyScaleSettings]) – kwargs (Any) – Return type None escape_str(value)[source] Parameters value (str) – Return type str add_texts(texts, metadatas=None, batch_size=32, ids=None, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Parameters texts (Iterable[str]) – Iterable of strings to add to the vectorstore. ids (Optional[Iterable[str]]) – Optional list of ids to associate with the texts. batch_size (int) – Batch size of insertion metadata – Optional column data to be inserted metadatas (Optional[List[dict]]) – kwargs (Any) – Returns List of ids from adding the texts into the vectorstore. Return type List[str] classmethod from_texts(texts, embedding, metadatas=None, config=None, text_ids=None, batch_size=32, **kwargs)[source] Create Myscale wrapper with existing texts Parameters embedding_function (Embeddings) – Function to extract text embedding texts (Iterable[str]) – List or tuple of strings to be added config (MyScaleSettings, Optional) – Myscale configuration text_ids (Optional[Iterable], optional) – IDs for the texts. Defaults to None. batch_size (int, optional) – Batchsize when transmitting data to MyScale. Defaults to 32. metadata (List[dict], optional) – metadata to texts. Defaults to None. into (Other keyword arguments will pass) – [clickhouse-connect](https://clickhouse.com/docs/en/integrations/python#clickhouse-connect-driver-api) embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[Dict[Any, Any]]]) – kwargs (Any) – Returns MyScale Index Return type langchain.vectorstores.myscale.MyScale similarity_search(query, k=4, where_str=None, **kwargs)[source] Perform a similarity search with MyScale Parameters query (str) – query string k (int, optional) – Top K neighbors to retrieve. Defaults to 4. where_str (Optional[str], optional) – where condition string. Defaults to None. NOTE – Please do not let end-user to fill this and always be aware of SQL injection. When dealing with metadatas, remember to use {self.metadata_column}.attribute instead of attribute alone. The default name for it is metadata. kwargs (Any) – Returns List of Documents Return type List[Document] similarity_search_by_vector(embedding, k=4, where_str=None, **kwargs)[source] Perform a similarity search with MyScale by vectors Parameters query (str) – query string k (int, optional) – Top K neighbors to retrieve. Defaults to 4. where_str (Optional[str], optional) – where condition string. Defaults to None. NOTE – Please do not let end-user to fill this and always be aware of SQL injection. When dealing with metadatas, remember to use {self.metadata_column}.attribute instead of attribute alone. The default name for it is metadata. embedding (List[float]) – kwargs (Any) – Returns List of (Document, similarity) Return type List[Document] similarity_search_with_relevance_scores(query, k=4, where_str=None, **kwargs)[source] Perform a similarity search with MyScale Parameters query (str) – query string k (int, optional) – Top K neighbors to retrieve. Defaults to 4. where_str (Optional[str], optional) – where condition string. Defaults to None. NOTE – Please do not let end-user to fill this and always be aware of SQL injection. When dealing with metadatas, remember to use {self.metadata_column}.attribute instead of attribute alone. The default name for it is metadata. kwargs (Any) – Returns List of documents most similar to the query text and cosine distance in float for each. Lower score represents more similarity. Return type List[Document] drop()[source] Helper function: Drop data Return type None property metadata_column: str pydantic settings langchain.vectorstores.MyScaleSettings[source] Bases: pydantic.env_settings.BaseSettings MyScale Client Configuration Attribute: myscale_host (str)An URL to connect to MyScale backend.Defaults to β€˜localhost’. myscale_port (int) : URL port to connect with HTTP. Defaults to 8443. username (str) : Username to login. Defaults to None. password (str) : Password to login. Defaults to None. index_type (str): index type string. index_param (dict): index build parameter. database (str) : Database name to find the table. Defaults to β€˜default’. table (str) : Table name to operate on. Defaults to β€˜vector_table’. metric (str)Metric to compute distance,supported are (β€˜l2’, β€˜cosine’, β€˜ip’). Defaults to β€˜cosine’. column_map (Dict)Column type map to project column name onto langchainsemantics. Must have keys: text, id, vector, must be same size to number of columns. For example: .. code-block:: python {β€˜id’: β€˜text_id’, β€˜vector’: β€˜text_embedding’, β€˜text’: β€˜text_plain’, β€˜metadata’: β€˜metadata_dictionary_in_json’, } Defaults to identity map. Show JSON schema{ "title": "MyScaleSettings", "description": "MyScale Client Configuration\n\nAttribute:\n myscale_host (str) : An URL to connect to MyScale backend.\n Defaults to 'localhost'.\n myscale_port (int) : URL port to connect with HTTP. Defaults to 8443.\n username (str) : Username to login. Defaults to None.\n password (str) : Password to login. Defaults to None.\n index_type (str): index type string.\n index_param (dict): index build parameter.\n database (str) : Database name to find the table. Defaults to 'default'.\n table (str) : Table name to operate on.\n Defaults to 'vector_table'.\n metric (str) : Metric to compute distance,\n supported are ('l2', 'cosine', 'ip'). Defaults to 'cosine'.\n column_map (Dict) : Column type map to project column name onto langchain\n semantics. Must have keys: `text`, `id`, `vector`,\n must be same size to number of columns. For example:\n .. code-block:: python\n\n {\n 'id': 'text_id',\n 'vector': 'text_embedding',\n 'text': 'text_plain',\n 'metadata': 'metadata_dictionary_in_json',\n }\n\n Defaults to identity map.", "type": "object", "properties": { "host": { "title": "Host", "default": "localhost", "env_names": "{'myscale_host'}", "type": "string" }, "port": { "title": "Port", "default": 8443, "env_names": "{'myscale_port'}", "type": "integer" }, "username": { "title": "Username", "env_names": "{'myscale_username'}", "type": "string" }, "password": { "title": "Password", "env_names": "{'myscale_password'}", "type": "string" }, "index_type": { "title": "Index Type", "default": "IVFFLAT", "env_names": "{'myscale_index_type'}", "type": "string" }, "index_param": { "title": "Index Param", "env_names": "{'myscale_index_param'}", "type": "object", "additionalProperties": { "type": "string" } }, "column_map": { "title": "Column Map", "default": { "id": "id", "text": "text", "vector": "vector", "metadata": "metadata" }, "env_names": "{'myscale_column_map'}", "type": "object", "additionalProperties": { "type": "string" } }, "database": { "title": "Database", "default": "default", "env_names": "{'myscale_database'}", "type": "string" }, "table": { "title": "Table", "default": "langchain", "env_names": "{'myscale_table'}", "type": "string" }, "metric": { "title": "Metric", "default": "cosine", "env_names": "{'myscale_metric'}", "type": "string" } }, "additionalProperties": false } Config env_file: str = .env env_file_encoding: str = utf-8 env_prefix: str = myscale_ Fields column_map (Dict[str, str]) database (str) host (str) index_param (Optional[Dict[str, str]]) index_type (str) metric (str) password (Optional[str]) port (int) table (str) username (Optional[str]) attribute column_map: Dict[str, str] = {'id': 'id', 'metadata': 'metadata', 'text': 'text', 'vector': 'vector'} attribute database: str = 'default' attribute host: str = 'localhost' attribute index_param: Optional[Dict[str, str]] = None attribute index_type: str = 'IVFFLAT' attribute metric: str = 'cosine' attribute password: Optional[str] = None attribute port: int = 8443 attribute table: str = 'langchain' attribute username: Optional[str] = None class langchain.vectorstores.Pinecone(index, embedding_function, text_key, namespace=None)[source] Bases: langchain.vectorstores.base.VectorStore Wrapper around Pinecone vector database. To use, you should have the pinecone-client python package installed. Example from langchain.vectorstores import Pinecone from langchain.embeddings.openai import OpenAIEmbeddings import pinecone # The environment should be the one specified next to the API key # in your Pinecone console pinecone.init(api_key="***", environment="...") index = pinecone.Index("langchain-demo") embeddings = OpenAIEmbeddings() vectorstore = Pinecone(index, embeddings.embed_query, "text") Parameters index (Any) – embedding_function (Callable) – text_key (str) – namespace (Optional[str]) – add_texts(texts, metadatas=None, ids=None, namespace=None, batch_size=32, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Parameters texts (Iterable[str]) – Iterable of strings to add to the vectorstore. metadatas (Optional[List[dict]]) – Optional list of metadatas associated with the texts. ids (Optional[List[str]]) – Optional list of ids to associate with the texts. namespace (Optional[str]) – Optional pinecone namespace to add the texts to. batch_size (int) – kwargs (Any) – Returns List of ids from adding the texts into the vectorstore. Return type List[str] similarity_search_with_score(query, k=4, filter=None, namespace=None)[source] Return pinecone documents most similar to query, along with scores. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. filter (Optional[dict]) – Dictionary of argument(s) to filter on metadata namespace (Optional[str]) – Namespace to search in. Default will search in β€˜β€™ namespace. Returns List of Documents most similar to the query and score for each Return type List[Tuple[langchain.schema.Document, float]] similarity_search(query, k=4, filter=None, namespace=None, **kwargs)[source] Return pinecone documents most similar to query. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. filter (Optional[dict]) – Dictionary of argument(s) to filter on metadata namespace (Optional[str]) – Namespace to search in. Default will search in β€˜β€™ namespace. kwargs (Any) – Returns List of Documents most similar to the query and score for each Return type List[langchain.schema.Document] max_marginal_relevance_search_by_vector(embedding, k=4, fetch_k=20, lambda_mult=0.5, filter=None, namespace=None, **kwargs)[source] Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Parameters embedding (List[float]) – Embedding to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. fetch_k (int) – Number of Documents to fetch to pass to MMR algorithm. lambda_mult (float) – Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. filter (Optional[dict]) – namespace (Optional[str]) – kwargs (Any) – Returns List of Documents selected by maximal marginal relevance. Return type List[langchain.schema.Document] max_marginal_relevance_search(query, k=4, fetch_k=20, lambda_mult=0.5, filter=None, namespace=None, **kwargs)[source] Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. fetch_k (int) – Number of Documents to fetch to pass to MMR algorithm. lambda_mult (float) – Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. filter (Optional[dict]) – namespace (Optional[str]) – kwargs (Any) – Returns List of Documents selected by maximal marginal relevance. Return type List[langchain.schema.Document] classmethod from_texts(texts, embedding, metadatas=None, ids=None, batch_size=32, text_key='text', index_name=None, namespace=None, **kwargs)[source] Construct Pinecone wrapper from raw documents. This is a user friendly interface that: Embeds documents. Adds the documents to a provided Pinecone index This is intended to be a quick way to get started. Example from langchain import Pinecone from langchain.embeddings import OpenAIEmbeddings import pinecone # The environment should be the one specified next to the API key # in your Pinecone console pinecone.init(api_key="***", environment="...") embeddings = OpenAIEmbeddings() pinecone = Pinecone.from_texts( texts, embeddings, index_name="langchain-demo" ) Parameters texts (List[str]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – ids (Optional[List[str]]) – batch_size (int) – text_key (str) – index_name (Optional[str]) – namespace (Optional[str]) – kwargs (Any) – Return type langchain.vectorstores.pinecone.Pinecone classmethod from_existing_index(index_name, embedding, text_key='text', namespace=None)[source] Load pinecone vectorstore from index name. Parameters index_name (str) – embedding (langchain.embeddings.base.Embeddings) – text_key (str) – namespace (Optional[str]) – Return type langchain.vectorstores.pinecone.Pinecone delete(ids)[source] Delete by vector IDs. Parameters ids (List[str]) – List of ids to delete. Return type None class langchain.vectorstores.Qdrant(client, collection_name, embeddings=None, content_payload_key='page_content', metadata_payload_key='metadata', embedding_function=None)[source] Bases: langchain.vectorstores.base.VectorStore Wrapper around Qdrant vector database. To use you should have the qdrant-client package installed. Example from qdrant_client import QdrantClient from langchain import Qdrant client = QdrantClient() collection_name = "MyCollection" qdrant = Qdrant(client, collection_name, embedding_function) Parameters client (Any) – collection_name (str) – embeddings (Optional[Embeddings]) – content_payload_key (str) – metadata_payload_key (str) – embedding_function (Optional[Callable]) – CONTENT_KEY = 'page_content' METADATA_KEY = 'metadata' add_texts(texts, metadatas=None, ids=None, batch_size=64, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Parameters texts (Iterable[str]) – Iterable of strings to add to the vectorstore. metadatas (Optional[List[dict]]) – Optional list of metadatas associated with the texts. ids (Optional[Sequence[str]]) – Optional list of ids to associate with the texts. Ids have to be uuid-like strings. batch_size (int) – How many vectors upload per-request. Default: 64 kwargs (Any) – Returns List of ids from adding the texts into the vectorstore. Return type List[str] similarity_search(query, k=4, filter=None, search_params=None, offset=0, score_threshold=None, consistency=None, **kwargs)[source] Return docs most similar to query. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. filter (Optional[MetadataFilter]) – Filter by metadata. Defaults to None. search_params (Optional[common_types.SearchParams]) – Additional search params offset (int) – Offset of the first result to return. May be used to paginate results. Note: large offset values may cause performance issues. score_threshold (Optional[float]) – Define a minimal score threshold for the result. If defined, less similar results will not be returned. Score of the returned result might be higher or smaller than the threshold depending on the Distance function used. E.g. for cosine similarity only higher scores will be returned. consistency (Optional[common_types.ReadConsistency]) – Read consistency of the search. Defines how many replicas should be queried before returning the result. Values: - int - number of replicas to query, values should present in all queried replicas ’majority’ - query all replicas, but return values present in themajority of replicas ’quorum’ - query the majority of replicas, return values present inall of them ’all’ - query all replicas, and return values present in all replicas kwargs (Any) – Returns List of Documents most similar to the query. Return type List[Document] similarity_search_with_score(query, k=4, filter=None, search_params=None, offset=0, score_threshold=None, consistency=None, **kwargs)[source] Return docs most similar to query. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. filter (Optional[MetadataFilter]) – Filter by metadata. Defaults to None. search_params (Optional[common_types.SearchParams]) – Additional search params offset (int) – Offset of the first result to return. May be used to paginate results. Note: large offset values may cause performance issues. score_threshold (Optional[float]) – Define a minimal score threshold for the result. If defined, less similar results will not be returned. Score of the returned result might be higher or smaller than the threshold depending on the Distance function used. E.g. for cosine similarity only higher scores will be returned. consistency (Optional[common_types.ReadConsistency]) – Read consistency of the search. Defines how many replicas should be queried before returning the result. Values: - int - number of replicas to query, values should present in all queried replicas ’majority’ - query all replicas, but return values present in themajority of replicas ’quorum’ - query the majority of replicas, return values present inall of them ’all’ - query all replicas, and return values present in all replicas kwargs (Any) – Returns List of documents most similar to the query text and cosine distance in float for each. Lower score represents more similarity. Return type List[Tuple[Document, float]] similarity_search_by_vector(embedding, k=4, filter=None, search_params=None, offset=0, score_threshold=None, consistency=None, **kwargs)[source] Return docs most similar to embedding vector. Parameters embedding (List[float]) – Embedding vector to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. filter (Optional[MetadataFilter]) – Filter by metadata. Defaults to None. search_params (Optional[common_types.SearchParams]) – Additional search params offset (int) – Offset of the first result to return. May be used to paginate results. Note: large offset values may cause performance issues. score_threshold (Optional[float]) – Define a minimal score threshold for the result. If defined, less similar results will not be returned. Score of the returned result might be higher or smaller than the threshold depending on the Distance function used. E.g. for cosine similarity only higher scores will be returned. consistency (Optional[common_types.ReadConsistency]) – Read consistency of the search. Defines how many replicas should be queried before returning the result. Values: - int - number of replicas to query, values should present in all queried replicas ’majority’ - query all replicas, but return values present in themajority of replicas ’quorum’ - query the majority of replicas, return values present inall of them ’all’ - query all replicas, and return values present in all replicas kwargs (Any) – Returns List of Documents most similar to the query. Return type List[Document] similarity_search_with_score_by_vector(embedding, k=4, filter=None, search_params=None, offset=0, score_threshold=None, consistency=None, **kwargs)[source] Return docs most similar to embedding vector. Parameters embedding (List[float]) – Embedding vector to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. filter (Optional[MetadataFilter]) – Filter by metadata. Defaults to None. search_params (Optional[common_types.SearchParams]) – Additional search params offset (int) – Offset of the first result to return. May be used to paginate results. Note: large offset values may cause performance issues. score_threshold (Optional[float]) – Define a minimal score threshold for the result. If defined, less similar results will not be returned. Score of the returned result might be higher or smaller than the threshold depending on the Distance function used. E.g. for cosine similarity only higher scores will be returned. consistency (Optional[common_types.ReadConsistency]) – Read consistency of the search. Defines how many replicas should be queried before returning the result. Values: - int - number of replicas to query, values should present in all queried replicas ’majority’ - query all replicas, but return values present in themajority of replicas ’quorum’ - query the majority of replicas, return values present inall of them ’all’ - query all replicas, and return values present in all replicas kwargs (Any) – Returns List of documents most similar to the query text and cosine distance in float for each. Lower score represents more similarity. Return type List[Tuple[Document, float]] max_marginal_relevance_search(query, k=4, fetch_k=20, lambda_mult=0.5, **kwargs)[source] Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. fetch_k (int) – Number of Documents to fetch to pass to MMR algorithm. Defaults to 20. lambda_mult (float) – Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. kwargs (Any) – Returns List of Documents selected by maximal marginal relevance. Return type List[langchain.schema.Document] classmethod from_texts(texts, embedding, metadatas=None, ids=None, location=None, url=None, port=6333, grpc_port=6334, prefer_grpc=False, https=None, api_key=None, prefix=None, timeout=None, host=None, path=None, collection_name=None, distance_func='Cosine', content_payload_key='page_content', metadata_payload_key='metadata', batch_size=64, shard_number=None, replication_factor=None, write_consistency_factor=None, on_disk_payload=None, hnsw_config=None, optimizers_config=None, wal_config=None, quantization_config=None, init_from=None, **kwargs)[source] Construct Qdrant wrapper from a list of texts. Parameters texts (List[str]) – A list of texts to be indexed in Qdrant. embedding (Embeddings) – A subclass of Embeddings, responsible for text vectorization. metadatas (Optional[List[dict]]) – An optional list of metadata. If provided it has to be of the same length as a list of texts. ids (Optional[Sequence[str]]) – Optional list of ids to associate with the texts. Ids have to be uuid-like strings. location (Optional[str]) – If :memory: - use in-memory Qdrant instance. If str - use it as a url parameter. If None - fallback to relying on host and port parameters. url (Optional[str]) – either host or str of β€œOptional[scheme], host, Optional[port], Optional[prefix]”. Default: None port (Optional[int]) – Port of the REST API interface. Default: 6333 grpc_port (int) – Port of the gRPC interface. Default: 6334 prefer_grpc (bool) – If true - use gPRC interface whenever possible in custom methods. Default: False https (Optional[bool]) – If true - use HTTPS(SSL) protocol. Default: None api_key (Optional[str]) – API key for authentication in Qdrant Cloud. Default: None prefix (Optional[str]) – If not None - add prefix to the REST URL path. Example: service/v1 will result in http://localhost:6333/service/v1/{qdrant-endpoint} for REST API. Default: None timeout (Optional[float]) – Timeout for REST and gRPC API requests. Default: 5.0 seconds for REST and unlimited for gRPC host (Optional[str]) – Host name of Qdrant service. If url and host are None, set to β€˜localhost’. Default: None path (Optional[str]) – Path in which the vectors will be stored while using local mode. Default: None collection_name (Optional[str]) – Name of the Qdrant collection to be used. If not provided, it will be created randomly. Default: None distance_func (str) – Distance function. One of: β€œCosine” / β€œEuclid” / β€œDot”. Default: β€œCosine” content_payload_key (str) – A payload key used to store the content of the document. Default: β€œpage_content” metadata_payload_key (str) – A payload key used to store the metadata of the document. Default: β€œmetadata” batch_size (int) – How many vectors upload per-request. Default: 64 shard_number (Optional[int]) – Number of shards in collection. Default is 1, minimum is 1. replication_factor (Optional[int]) – Replication factor for collection. Default is 1, minimum is 1. Defines how many copies of each shard will be created. Have effect only in distributed mode. write_consistency_factor (Optional[int]) – Write consistency factor for collection. Default is 1, minimum is 1. Defines how many replicas should apply the operation for us to consider it successful. Increasing this number will make the collection more resilient to inconsistencies, but will also make it fail if not enough replicas are available. Does not have any performance impact. Have effect only in distributed mode. on_disk_payload (Optional[bool]) – If true - point`s payload will not be stored in memory. It will be read from the disk every time it is requested. This setting saves RAM by (slightly) increasing the response time. Note: those payload values that are involved in filtering and are indexed - remain in RAM. hnsw_config (Optional[common_types.HnswConfigDiff]) – Params for HNSW index optimizers_config (Optional[common_types.OptimizersConfigDiff]) – Params for optimizer wal_config (Optional[common_types.WalConfigDiff]) – Params for Write-Ahead-Log quantization_config (Optional[common_types.QuantizationConfig]) – Params for quantization, if None - quantization will be disabled init_from (Optional[common_types.InitFrom]) – Use data stored in another collection to initialize this collection **kwargs – Additional arguments passed directly into REST client initialization kwargs (Any) – Return type Qdrant This is a user-friendly interface that: 1. Creates embeddings, one for each text 2. Initializes the Qdrant database as an in-memory docstore by default (and overridable to a remote docstore) Adds the text embeddings to the Qdrant database This is intended to be a quick way to get started. Example from langchain import Qdrant from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() qdrant = Qdrant.from_texts(texts, embeddings, "localhost") class langchain.vectorstores.Redis(redis_url, index_name, embedding_function, content_key='content', metadata_key='metadata', vector_key='content_vector', relevance_score_fn=<function _default_relevance_score>, **kwargs)[source] Bases: langchain.vectorstores.base.VectorStore Wrapper around Redis vector database. To use, you should have the redis python package installed. Example from langchain.vectorstores import Redis from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() vectorstore = Redis( redis_url="redis://username:password@localhost:6379" index_name="my-index", embedding_function=embeddings.embed_query, ) Parameters redis_url (str) – index_name (str) – embedding_function (Callable) – content_key (str) – metadata_key (str) – vector_key (str) – relevance_score_fn (Optional[Callable[[float], float]]) – kwargs (Any) – add_texts(texts, metadatas=None, embeddings=None, batch_size=1000, **kwargs)[source] Add more texts to the vectorstore. Parameters texts (Iterable[str]) – Iterable of strings/text to add to the vectorstore. metadatas (Optional[List[dict]], optional) – Optional list of metadatas. Defaults to None. embeddings (Optional[List[List[float]]], optional) – Optional pre-generated embeddings. Defaults to None. keys (List[str]) or ids (List[str]) – Identifiers of entries. Defaults to None. batch_size (int, optional) – Batch size to use for writes. Defaults to 1000. kwargs (Any) – Returns List of ids added to the vectorstore Return type List[str] similarity_search(query, k=4, **kwargs)[source] Returns the most similar indexed documents to the query text. Parameters query (str) – The query text for which to find similar documents. k (int) – The number of documents to return. Default is 4. kwargs (Any) – Returns A list of documents that are most similar to the query text. Return type List[Document] similarity_search_limit_score(query, k=4, score_threshold=0.2, **kwargs)[source] Returns the most similar indexed documents to the query text within the score_threshold range. Parameters query (str) – The query text for which to find similar documents. k (int) – The number of documents to return. Default is 4. score_threshold (float) – The minimum matching score required for a document 0.2. (to be considered a match. Defaults to) – similarity (Because the similarity calculation algorithm is based on cosine) – kwargs (Any) – Return type List[langchain.schema.Document] :param : :param the smaller the angle: :param the higher the similarity.: Returns A list of documents that are most similar to the query text, including the match score for each document. Return type List[Document] Parameters query (str) – k (int) – score_threshold (float) – kwargs (Any) – Note If there are no documents that satisfy the score_threshold value, an empty list is returned. similarity_search_with_score(query, k=4)[source] Return docs most similar to query. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. Returns List of Documents most similar to the query and score for each Return type List[Tuple[langchain.schema.Document, float]] classmethod from_texts_return_keys(texts, embedding, metadatas=None, index_name=None, content_key='content', metadata_key='metadata', vector_key='content_vector', distance_metric='COSINE', **kwargs)[source] Create a Redis vectorstore from raw documents. This is a user-friendly interface that: Embeds documents. Creates a new index for the embeddings in Redis. Adds the documents to the newly created Redis index. Returns the keys of the newly created documents. This is intended to be a quick way to get started. .. rubric:: Example Parameters texts (List[str]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – index_name (Optional[str]) – content_key (str) – metadata_key (str) – vector_key (str) – distance_metric (Literal['COSINE', 'IP', 'L2']) – kwargs (Any) – Return type Tuple[langchain.vectorstores.redis.Redis, List[str]] classmethod from_texts(texts, embedding, metadatas=None, index_name=None, content_key='content', metadata_key='metadata', vector_key='content_vector', **kwargs)[source] Create a Redis vectorstore from raw documents. This is a user-friendly interface that: Embeds documents. Creates a new index for the embeddings in Redis. Adds the documents to the newly created Redis index. This is intended to be a quick way to get started. .. rubric:: Example Parameters texts (List[str]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – index_name (Optional[str]) – content_key (str) – metadata_key (str) – vector_key (str) – kwargs (Any) – Return type langchain.vectorstores.redis.Redis static delete(ids, **kwargs)[source] Delete a Redis entry. Parameters ids (List[str]) – List of ids (keys) to delete. kwargs (Any) – Returns Whether or not the deletions were successful. Return type bool static drop_index(index_name, delete_documents, **kwargs)[source] Drop a Redis search index. Parameters index_name (str) – Name of the index to drop. delete_documents (bool) – Whether to drop the associated documents. kwargs (Any) – Returns Whether or not the drop was successful. Return type bool classmethod from_existing_index(embedding, index_name, content_key='content', metadata_key='metadata', vector_key='content_vector', **kwargs)[source] Connect to an existing Redis index. Parameters embedding (langchain.embeddings.base.Embeddings) – index_name (str) – content_key (str) – metadata_key (str) – vector_key (str) – kwargs (Any) – Return type langchain.vectorstores.redis.Redis as_retriever(**kwargs)[source] Parameters kwargs (Any) – Return type langchain.vectorstores.redis.RedisVectorStoreRetriever class langchain.vectorstores.Rockset(client, embeddings, collection_name, text_key, embedding_key)[source] Bases: langchain.vectorstores.base.VectorStore Wrapper arpund Rockset vector database. To use, you should have the rockset python package installed. Note that to use this, the collection being used must already exist in your Rockset instance. You must also ensure you use a Rockset ingest transformation to apply VECTOR_ENFORCE on the column being used to store embedding_key in the collection. See: https://rockset.com/blog/introducing-vector-search-on-rockset/ for more details Everything below assumes commons Rockset workspace. TODO: Add support for workspace args. Example from langchain.vectorstores import Rockset from langchain.embeddings.openai import OpenAIEmbeddings import rockset # Make sure you use the right host (region) for your Rockset instance # and APIKEY has both read-write access to your collection. rs = rockset.RocksetClient(host=rockset.Regions.use1a1, api_key="***") collection_name = "langchain_demo" embeddings = OpenAIEmbeddings() vectorstore = Rockset(rs, collection_name, embeddings, "description", "description_embedding") Parameters client (Any) – embeddings (Embeddings) – collection_name (str) – text_key (str) – embedding_key (str) – add_texts(texts, metadatas=None, ids=None, batch_size=32, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore Args: texts: Iterable of strings to add to the vectorstore. metadatas: Optional list of metadatas associated with the texts. ids: Optional list of ids to associate with the texts. batch_size: Send documents in batches to rockset. Returns List of ids from adding the texts into the vectorstore. Parameters texts (Iterable[str]) – metadatas (Optional[List[dict]]) – ids (Optional[List[str]]) – batch_size (int) – kwargs (Any) – Return type List[str] classmethod from_texts(texts, embedding, metadatas=None, client=None, collection_name='', text_key='', embedding_key='', ids=None, batch_size=32, **kwargs)[source] Create Rockset wrapper with existing texts. This is intended as a quicker way to get started. Parameters texts (List[str]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – client (Any) – collection_name (str) – text_key (str) – embedding_key (str) – ids (Optional[List[str]]) – batch_size (int) – kwargs (Any) – Return type langchain.vectorstores.rocksetdb.Rockset class DistanceFunction(value, names=None, *, module=None, qualname=None, type=None, start=1, boundary=None)[source] Bases: enum.Enum COSINE_SIM = 'COSINE_SIM' EUCLIDEAN_DIST = 'EUCLIDEAN_DIST' DOT_PRODUCT = 'DOT_PRODUCT' order_by()[source] Return type str similarity_search_with_relevance_scores(query, k=4, distance_func=DistanceFunction.COSINE_SIM, where_str=None, **kwargs)[source] Perform a similarity search with Rockset Parameters query (str) – Text to look up documents similar to. distance_func (DistanceFunction) – how to compute distance between two vectors in Rockset. k (int, optional) – Top K neighbors to retrieve. Defaults to 4. where_str (Optional[str], optional) – Metadata filters supplied as a SQL where condition string. Defaults to None. eg. β€œprice<=70.0 AND brand=’Nintendo’” NOTE – Please do not let end-user to fill this and always be aware of SQL injection. kwargs (Any) – Returns List of documents with their relevance score Return type List[Tuple[Document, float]] similarity_search(query, k=4, distance_func=DistanceFunction.COSINE_SIM, where_str=None, **kwargs)[source] Same as similarity_search_with_relevance_scores but doesn’t return the scores. Parameters query (str) – k (int) – distance_func (DistanceFunction) – where_str (Optional[str]) – kwargs (Any) – Return type List[Document] similarity_search_by_vector(embedding, k=4, distance_func=DistanceFunction.COSINE_SIM, where_str=None, **kwargs)[source] Accepts a query_embedding (vector), and returns documents with similar embeddings. Parameters embedding (List[float]) – k (int) – distance_func (DistanceFunction) – where_str (Optional[str]) – kwargs (Any) – Return type List[Document] similarity_search_by_vector_with_relevance_scores(embedding, k=4, distance_func=DistanceFunction.COSINE_SIM, where_str=None, **kwargs)[source] Accepts a query_embedding (vector), and returns documents with similar embeddings along with their relevance scores. Parameters embedding (List[float]) – k (int) – distance_func (DistanceFunction) – where_str (Optional[str]) – kwargs (Any) – Return type List[Tuple[Document, float]] delete_texts(ids)[source] Delete a list of docs from the Rockset collection Parameters ids (List[str]) – Return type None class langchain.vectorstores.SKLearnVectorStore(embedding, *, persist_path=None, serializer='json', metric='cosine', **kwargs)[source] Bases: langchain.vectorstores.base.VectorStore A simple in-memory vector store based on the scikit-learn library NearestNeighbors implementation. Parameters embedding (langchain.embeddings.base.Embeddings) – persist_path (Optional[str]) – serializer (Literal['json', 'bson', 'parquet']) – metric (str) – kwargs (Any) – Return type None persist()[source] Return type None add_texts(texts, metadatas=None, ids=None, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Parameters texts (Iterable[str]) – Iterable of strings to add to the vectorstore. metadatas (Optional[List[dict]]) – Optional list of metadatas associated with the texts. kwargs (Any) – vectorstore specific parameters ids (Optional[List[str]]) – Returns List of ids from adding the texts into the vectorstore. Return type List[str] similarity_search_with_score(query, *, k=4, **kwargs)[source] Parameters query (str) – k (int) – kwargs (Any) – Return type List[Tuple[langchain.schema.Document, float]] similarity_search(query, k=4, **kwargs)[source] Return docs most similar to query. Parameters query (str) – k (int) – kwargs (Any) – Return type List[langchain.schema.Document] max_marginal_relevance_search_by_vector(embedding, k=4, fetch_k=20, lambda_mult=0.5, **kwargs)[source] Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. :param embedding: Embedding to look up documents similar to. :param k: Number of Documents to return. Defaults to 4. :param fetch_k: Number of Documents to fetch to pass to MMR algorithm. :param lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. Returns List of Documents selected by maximal marginal relevance. Parameters embedding (List[float]) – k (int) – fetch_k (int) – lambda_mult (float) – kwargs (Any) – Return type List[langchain.schema.Document] max_marginal_relevance_search(query, k=4, fetch_k=20, lambda_mult=0.5, **kwargs)[source] Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. :param query: Text to look up documents similar to. :param k: Number of Documents to return. Defaults to 4. :param fetch_k: Number of Documents to fetch to pass to MMR algorithm. :param lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. Returns List of Documents selected by maximal marginal relevance. Parameters query (str) – k (int) – fetch_k (int) – lambda_mult (float) – kwargs (Any) – Return type List[langchain.schema.Document] classmethod from_texts(texts, embedding, metadatas=None, ids=None, persist_path=None, **kwargs)[source] Return VectorStore initialized from texts and embeddings. Parameters texts (List[str]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – ids (Optional[List[str]]) – persist_path (Optional[str]) – kwargs (Any) – Return type langchain.vectorstores.sklearn.SKLearnVectorStore class langchain.vectorstores.StarRocks(embedding, config=None, **kwargs)[source] Bases: langchain.vectorstores.base.VectorStore Wrapper around StarRocks vector database You need a pymysql python package, and a valid account to connect to StarRocks. Right now StarRocks has only implemented cosine_similarity function to compute distance between two vectors. And there is no vector inside right now, so we have to iterate all vectors and compute spatial distance. For more information, please visit[StarRocks official site](https://www.starrocks.io/) [StarRocks github](https://github.com/StarRocks/starrocks) Parameters embedding (Embeddings) – config (Optional[StarRocksSettings]) – kwargs (Any) – Return type None escape_str(value)[source] Parameters value (str) – Return type str add_texts(texts, metadatas=None, batch_size=32, ids=None, **kwargs)[source] Insert more texts through the embeddings and add to the VectorStore. Parameters texts (Iterable[str]) – Iterable of strings to add to the VectorStore. ids (Optional[Iterable[str]]) – Optional list of ids to associate with the texts. batch_size (int) – Batch size of insertion metadata – Optional column data to be inserted metadatas (Optional[List[dict]]) – kwargs (Any) – Returns List of ids from adding the texts into the VectorStore. Return type List[str] classmethod from_texts(texts, embedding, metadatas=None, config=None, text_ids=None, batch_size=32, **kwargs)[source] Create StarRocks wrapper with existing texts Parameters embedding_function (Embeddings) – Function to extract text embedding texts (Iterable[str]) – List or tuple of strings to be added config (StarRocksSettings, Optional) – StarRocks configuration text_ids (Optional[Iterable], optional) – IDs for the texts. Defaults to None. batch_size (int, optional) – Batchsize when transmitting data to StarRocks. Defaults to 32. metadata (List[dict], optional) – metadata to texts. Defaults to None. embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[Dict[Any, Any]]]) – kwargs (Any) – Returns StarRocks Index Return type langchain.vectorstores.starrocks.StarRocks similarity_search(query, k=4, where_str=None, **kwargs)[source] Perform a similarity search with StarRocks Parameters query (str) – query string k (int, optional) – Top K neighbors to retrieve. Defaults to 4. where_str (Optional[str], optional) – where condition string. Defaults to None. NOTE – Please do not let end-user to fill this and always be aware of SQL injection. When dealing with metadatas, remember to use {self.metadata_column}.attribute instead of attribute alone. The default name for it is metadata. kwargs (Any) – Returns List of Documents Return type List[Document] similarity_search_by_vector(embedding, k=4, where_str=None, **kwargs)[source] Perform a similarity search with StarRocks by vectors Parameters query (str) – query string k (int, optional) – Top K neighbors to retrieve. Defaults to 4. where_str (Optional[str], optional) – where condition string. Defaults to None. NOTE – Please do not let end-user to fill this and always be aware of SQL injection. When dealing with metadatas, remember to use {self.metadata_column}.attribute instead of attribute alone. The default name for it is metadata. embedding (List[float]) – kwargs (Any) – Returns List of (Document, similarity) Return type List[Document] similarity_search_with_relevance_scores(query, k=4, where_str=None, **kwargs)[source] Perform a similarity search with StarRocks Parameters query (str) – query string k (int, optional) – Top K neighbors to retrieve. Defaults to 4. where_str (Optional[str], optional) – where condition string. Defaults to None. NOTE – Please do not let end-user to fill this and always be aware of SQL injection. When dealing with metadatas, remember to use {self.metadata_column}.attribute instead of attribute alone. The default name for it is metadata. kwargs (Any) – Returns List of documents Return type List[Document] drop()[source] Helper function: Drop data Return type None property metadata_column: str class langchain.vectorstores.SupabaseVectorStore(client, embedding, table_name, query_name=None)[source] Bases: langchain.vectorstores.base.VectorStore VectorStore for a Supabase postgres database. Assumes you have the pgvector extension installed and a match_documents (or similar) function. For more details: https://js.langchain.com/docs/modules/indexes/vector_stores/integrations/supabase You can implement your own match_documents function in order to limit the search space to a subset of documents based on your own authorization or business logic. Note that the Supabase Python client does not yet support async operations. If you’d like to use max_marginal_relevance_search, please review the instructions below on modifying the match_documents function to return matched embeddings. Parameters client (supabase.client.Client) – embedding (Embeddings) – table_name (str) – query_name (Union[str, None]) – Return type None table_name: str query_name: str add_texts(texts, metadatas=None, ids=None, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Parameters texts (Iterable[str]) – Iterable of strings to add to the vectorstore. metadatas (Optional[List[dict[Any, Any]]]) – Optional list of metadatas associated with the texts. kwargs (Any) – vectorstore specific parameters ids (Optional[List[str]]) – Returns List of ids from adding the texts into the vectorstore. Return type List[str] classmethod from_texts(texts, embedding, metadatas=None, client=None, table_name='documents', query_name='match_documents', ids=None, **kwargs)[source] Return VectorStore initialized from texts and embeddings. Parameters texts (List[str]) – embedding (Embeddings) – metadatas (Optional[List[dict]]) – client (Optional[supabase.client.Client]) – table_name (Optional[str]) – query_name (Union[str, None]) – ids (Optional[List[str]]) – kwargs (Any) – Return type SupabaseVectorStore add_vectors(vectors, documents, ids)[source] Parameters vectors (List[List[float]]) – documents (List[langchain.schema.Document]) – ids (List[str]) – Return type List[str] similarity_search(query, k=4, **kwargs)[source] Return docs most similar to query. Parameters query (str) – k (int) – kwargs (Any) – Return type List[langchain.schema.Document] similarity_search_by_vector(embedding, k=4, **kwargs)[source] Return docs most similar to embedding vector. Parameters embedding (List[float]) – Embedding to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. kwargs (Any) – Returns List of Documents most similar to the query vector. Return type List[langchain.schema.Document] similarity_search_with_relevance_scores(query, k=4, **kwargs)[source] Return docs and relevance scores in the range [0, 1]. 0 is dissimilar, 1 is most similar. Parameters query (str) – input text k (int) – Number of Documents to return. Defaults to 4. **kwargs – kwargs to be passed to similarity search. Should include: score_threshold: Optional, a floating point value between 0 to 1 to filter the resulting set of retrieved docs kwargs (Any) – Returns List of Tuples of (doc, similarity_score) Return type List[Tuple[langchain.schema.Document, float]] similarity_search_by_vector_with_relevance_scores(query, k)[source] Parameters query (List[float]) – k (int) – Return type List[Tuple[langchain.schema.Document, float]] similarity_search_by_vector_returning_embeddings(query, k)[source] Parameters query (List[float]) – k (int) – Return type List[Tuple[langchain.schema.Document, float, numpy.ndarray[numpy.float32, Any]]] max_marginal_relevance_search_by_vector(embedding, k=4, fetch_k=20, lambda_mult=0.5, **kwargs)[source] Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Parameters embedding (List[float]) – Embedding to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. fetch_k (int) – Number of Documents to fetch to pass to MMR algorithm. lambda_mult (float) – Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. kwargs (Any) – Returns List of Documents selected by maximal marginal relevance. Return type List[langchain.schema.Document] max_marginal_relevance_search(query, k=4, fetch_k=20, lambda_mult=0.5, **kwargs)[source] Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. fetch_k (int) – Number of Documents to fetch to pass to MMR algorithm. lambda_mult (float) – Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. kwargs (Any) – Returns List of Documents selected by maximal marginal relevance. Return type List[langchain.schema.Document] max_marginal_relevance_search requires that query_name returns matched embeddings alongside the match documents. The following function demonstrates how to do this: ```sql CREATE FUNCTION match_documents_embeddings(query_embedding vector(1536), match_count int) RETURNS TABLE(id bigint, content text, metadata jsonb, embedding vector(1536), similarity float) LANGUAGE plpgsql AS $$ # variable_conflict use_column BEGINRETURN query SELECT id, content, metadata, embedding, 1 -(docstore.embedding <=> query_embedding) AS similarity FROMdocstore ORDER BYdocstore.embedding <=> query_embedding LIMIT match_count; END; $$; ``` delete(ids)[source] Delete by vector IDs. Parameters ids (List[str]) – List of ids to delete. Return type None class langchain.vectorstores.Tair(embedding_function, url, index_name, content_key='content', metadata_key='metadata', search_params=None, **kwargs)[source] Bases: langchain.vectorstores.base.VectorStore Wrapper around Tair Vector store. Parameters embedding_function (Embeddings) – url (str) – index_name (str) – content_key (str) – metadata_key (str) – search_params (Optional[dict]) – kwargs (Any) – create_index_if_not_exist(dim, distance_type, index_type, data_type, **kwargs)[source] Parameters dim (int) – distance_type (str) – index_type (str) – data_type (str) – kwargs (Any) – Return type bool add_texts(texts, metadatas=None, **kwargs)[source] Add texts data to an existing index. Parameters texts (Iterable[str]) – metadatas (Optional[List[dict]]) – kwargs (Any) – Return type List[str] similarity_search(query, k=4, **kwargs)[source] Returns the most similar indexed documents to the query text. Parameters query (str) – The query text for which to find similar documents. k (int) – The number of documents to return. Default is 4. kwargs (Any) – Returns A list of documents that are most similar to the query text. Return type List[Document] classmethod from_texts(texts, embedding, metadatas=None, index_name='langchain', content_key='content', metadata_key='metadata', **kwargs)[source] Return VectorStore initialized from texts and embeddings. Parameters texts (List[str]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – index_name (str) – content_key (str) – metadata_key (str) – kwargs (Any) – Return type langchain.vectorstores.tair.Tair classmethod from_documents(documents, embedding, metadatas=None, index_name='langchain', content_key='content', metadata_key='metadata', **kwargs)[source] Return VectorStore initialized from documents and embeddings. Parameters documents (List[langchain.schema.Document]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – index_name (str) – content_key (str) – metadata_key (str) – kwargs (Any) – Return type langchain.vectorstores.tair.Tair static drop_index(index_name='langchain', **kwargs)[source] Drop an existing index. Parameters index_name (str) – Name of the index to drop. kwargs (Any) – Returns True if the index is dropped successfully. Return type bool classmethod from_existing_index(embedding, index_name='langchain', content_key='content', metadata_key='metadata', **kwargs)[source] Connect to an existing Tair index. Parameters embedding (langchain.embeddings.base.Embeddings) – index_name (str) – content_key (str) – metadata_key (str) – kwargs (Any) – Return type langchain.vectorstores.tair.Tair class langchain.vectorstores.Tigris(client, embeddings, index_name)[source] Bases: langchain.vectorstores.base.VectorStore Parameters client (TigrisClient) – embeddings (Embeddings) – index_name (str) – property search_index: TigrisVectorStore add_texts(texts, metadatas=None, ids=None, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Parameters texts (Iterable[str]) – Iterable of strings to add to the vectorstore. metadatas (Optional[List[dict]]) – Optional list of metadatas associated with the texts. ids (Optional[List[str]]) – Optional list of ids for documents. Ids will be autogenerated if not provided. kwargs (Any) – vectorstore specific parameters Returns List of ids from adding the texts into the vectorstore. Return type List[str] similarity_search(query, k=4, filter=None, **kwargs)[source] Return docs most similar to query. Parameters query (str) – k (int) – filter (Optional[TigrisFilter]) – kwargs (Any) – Return type List[Document] similarity_search_with_score(query, k=4, filter=None)[source] Run similarity search with Chroma with distance. Parameters query (str) – Query text to search for. k (int) – Number of results to return. Defaults to 4. filter (Optional[TigrisFilter]) – Filter by metadata. Defaults to None. Returns List of documents most similar to the querytext with distance in float. Return type List[Tuple[Document, float]] classmethod from_texts(texts, embedding, metadatas=None, ids=None, client=None, index_name=None, **kwargs)[source] Return VectorStore initialized from texts and embeddings. Parameters texts (List[str]) – embedding (Embeddings) – metadatas (Optional[List[dict]]) – ids (Optional[List[str]]) – client (Optional[TigrisClient]) – index_name (Optional[str]) – kwargs (Any) – Return type Tigris class langchain.vectorstores.Typesense(typesense_client, embedding, *, typesense_collection_name=None, text_key='text')[source] Bases: langchain.vectorstores.base.VectorStore Wrapper around Typesense vector search. To use, you should have the typesense python package installed. Example from langchain.embedding.openai import OpenAIEmbeddings from langchain.vectorstores import Typesense import typesense node = { "host": "localhost", # For Typesense Cloud use xxx.a1.typesense.net "port": "8108", # For Typesense Cloud use 443 "protocol": "http" # For Typesense Cloud use https } typesense_client = typesense.Client( { "nodes": [node], "api_key": "<API_KEY>", "connection_timeout_seconds": 2 } ) typesense_collection_name = "langchain-memory" embedding = OpenAIEmbeddings() vectorstore = Typesense( typesense_client=typesense_client, embedding=embedding, typesense_collection_name=typesense_collection_name, text_key="text", ) Parameters typesense_client (Client) – embedding (Embeddings) – typesense_collection_name (Optional[str]) – text_key (str) – add_texts(texts, metadatas=None, ids=None, **kwargs)[source] Run more texts through the embedding and add to the vectorstore. Parameters texts (Iterable[str]) – Iterable of strings to add to the vectorstore. metadatas (Optional[List[dict]]) – Optional list of metadatas associated with the texts. ids (Optional[List[str]]) – Optional list of ids to associate with the texts. kwargs (Any) – Returns List of ids from adding the texts into the vectorstore. Return type List[str] similarity_search_with_score(query, k=10, filter='')[source] Return typesense documents most similar to query, along with scores. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 10. Minimum 10 results would be returned. filter (Optional[str]) – typesense filter_by expression to filter documents on Returns List of Documents most similar to the query and score for each Return type List[Tuple[langchain.schema.Document, float]] similarity_search(query, k=10, filter='', **kwargs)[source] Return typesense documents most similar to query. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 10. Minimum 10 results would be returned. filter (Optional[str]) – typesense filter_by expression to filter documents on kwargs (Any) – Returns List of Documents most similar to the query and score for each Return type List[langchain.schema.Document] classmethod from_client_params(embedding, *, host='localhost', port='8108', protocol='http', typesense_api_key=None, connection_timeout_seconds=2, **kwargs)[source] Initialize Typesense directly from client parameters. Example from langchain.embedding.openai import OpenAIEmbeddings from langchain.vectorstores import Typesense # Pass in typesense_api_key as kwarg or set env var "TYPESENSE_API_KEY". vectorstore = Typesense( OpenAIEmbeddings(), host="localhost", port="8108", protocol="http", typesense_collection_name="langchain-memory", ) Parameters embedding (langchain.embeddings.base.Embeddings) – host (str) – port (Union[str, int]) – protocol (str) – typesense_api_key (Optional[str]) – connection_timeout_seconds (int) – kwargs (Any) – Return type langchain.vectorstores.typesense.Typesense classmethod from_texts(texts, embedding, metadatas=None, ids=None, typesense_client=None, typesense_client_params=None, typesense_collection_name=None, text_key='text', **kwargs)[source] Construct Typesense wrapper from raw text. Parameters texts (List[str]) – embedding (Embeddings) – metadatas (Optional[List[dict]]) – ids (Optional[List[str]]) – typesense_client (Optional[Client]) – typesense_client_params (Optional[dict]) – typesense_collection_name (Optional[str]) – text_key (str) – kwargs (Any) – Return type Typesense class langchain.vectorstores.Vectara(vectara_customer_id=None, vectara_corpus_id=None, vectara_api_key=None)[source] Bases: langchain.vectorstores.base.VectorStore Implementation of Vector Store using Vectara (https://vectara.com). .. rubric:: Example from langchain.vectorstores import Vectara vectorstore = Vectara( vectara_customer_id=vectara_customer_id, vectara_corpus_id=vectara_corpus_id, vectara_api_key=vectara_api_key ) Parameters vectara_customer_id (Optional[str]) – vectara_corpus_id (Optional[str]) – vectara_api_key (Optional[str]) – add_texts(texts, metadatas=None, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Parameters texts (Iterable[str]) – Iterable of strings to add to the vectorstore. metadatas (Optional[List[dict]]) – Optional list of metadatas associated with the texts. kwargs (Any) – Returns List of ids from adding the texts into the vectorstore. Return type List[str] similarity_search_with_score(query, k=5, lambda_val=0.025, filter=None, n_sentence_context=0, **kwargs)[source] Return Vectara documents most similar to query, along with scores. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 5. lambda_val (float) – lexical match parameter for hybrid search. filter (Optional[str]) – Dictionary of argument(s) to filter on metadata. For example a filter can be β€œdoc.rating > 3.0 and part.lang = β€˜deu’”} see https://docs.vectara.com/docs/search-apis/sql/filter-overview for more details. n_sentence_context (int) – number of sentences before/after the matching segment to add kwargs (Any) – Returns List of Documents most similar to the query and score for each. Return type List[Tuple[langchain.schema.Document, float]] similarity_search(query, k=5, lambda_val=0.025, filter=None, n_sentence_context=0, **kwargs)[source] Return Vectara documents most similar to query, along with scores. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 5. filter (Optional[str]) – Dictionary of argument(s) to filter on metadata. For example a filter can be β€œdoc.rating > 3.0 and part.lang = β€˜deu’”} see https://docs.vectara.com/docs/search-apis/sql/filter-overview for more details. n_sentence_context (int) – number of sentences before/after the matching segment to add lambda_val (float) – kwargs (Any) – Returns List of Documents most similar to the query Return type List[langchain.schema.Document] classmethod from_texts(texts, embedding=None, metadatas=None, **kwargs)[source] Construct Vectara wrapper from raw documents. This is intended to be a quick way to get started. .. rubric:: Example from langchain import Vectara vectara = Vectara.from_texts( texts, vectara_customer_id=customer_id, vectara_corpus_id=corpus_id, vectara_api_key=api_key, ) Parameters texts (List[str]) – embedding (Optional[langchain.embeddings.base.Embeddings]) – metadatas (Optional[List[dict]]) – kwargs (Any) – Return type langchain.vectorstores.vectara.Vectara as_retriever(**kwargs)[source] Parameters kwargs (Any) – Return type langchain.vectorstores.vectara.VectaraRetriever class langchain.vectorstores.VectorStore[source] Bases: abc.ABC Interface for vector stores. abstract add_texts(texts, metadatas=None, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Parameters texts (Iterable[str]) – Iterable of strings to add to the vectorstore. metadatas (Optional[List[dict]]) – Optional list of metadatas associated with the texts. kwargs (Any) – vectorstore specific parameters Returns List of ids from adding the texts into the vectorstore. Return type List[str] delete(ids)[source] Delete by vector ID. Parameters ids (List[str]) – List of ids to delete. Returns True if deletion is successful, False otherwise, None if not implemented. Return type Optional[bool] async aadd_texts(texts, metadatas=None, **kwargs)[source] Run more texts through the embeddings and add to the vectorstore. Parameters texts (Iterable[str]) – metadatas (Optional[List[dict]]) – kwargs (Any) – Return type List[str] add_documents(documents, **kwargs)[source] Run more documents through the embeddings and add to the vectorstore. Parameters (List[Document] (documents) – Documents to add to the vectorstore. documents (List[langchain.schema.Document]) – kwargs (Any) – Returns List of IDs of the added texts. Return type List[str] async aadd_documents(documents, **kwargs)[source] Run more documents through the embeddings and add to the vectorstore. Parameters (List[Document] (documents) – Documents to add to the vectorstore. documents (List[langchain.schema.Document]) – kwargs (Any) – Returns List of IDs of the added texts. Return type List[str] search(query, search_type, **kwargs)[source] Return docs most similar to query using specified search type. Parameters query (str) – search_type (str) – kwargs (Any) – Return type List[langchain.schema.Document] async asearch(query, search_type, **kwargs)[source] Return docs most similar to query using specified search type. Parameters query (str) – search_type (str) – kwargs (Any) – Return type List[langchain.schema.Document] abstract similarity_search(query, k=4, **kwargs)[source] Return docs most similar to query. Parameters query (str) – k (int) – kwargs (Any) – Return type List[langchain.schema.Document] similarity_search_with_relevance_scores(query, k=4, **kwargs)[source] Return docs and relevance scores in the range [0, 1]. 0 is dissimilar, 1 is most similar. Parameters query (str) – input text k (int) – Number of Documents to return. Defaults to 4. **kwargs – kwargs to be passed to similarity search. Should include: score_threshold: Optional, a floating point value between 0 to 1 to filter the resulting set of retrieved docs kwargs (Any) – Returns List of Tuples of (doc, similarity_score) Return type List[Tuple[langchain.schema.Document, float]] async asimilarity_search_with_relevance_scores(query, k=4, **kwargs)[source] Return docs most similar to query. Parameters query (str) – k (int) – kwargs (Any) – Return type List[Tuple[langchain.schema.Document, float]] async asimilarity_search(query, k=4, **kwargs)[source] Return docs most similar to query. Parameters query (str) – k (int) – kwargs (Any) – Return type List[langchain.schema.Document] similarity_search_by_vector(embedding, k=4, **kwargs)[source] Return docs most similar to embedding vector. Parameters embedding (List[float]) – Embedding to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. kwargs (Any) – Returns List of Documents most similar to the query vector. Return type List[langchain.schema.Document] async asimilarity_search_by_vector(embedding, k=4, **kwargs)[source] Return docs most similar to embedding vector. Parameters embedding (List[float]) – k (int) – kwargs (Any) – Return type List[langchain.schema.Document] max_marginal_relevance_search(query, k=4, fetch_k=20, lambda_mult=0.5, **kwargs)[source] Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. fetch_k (int) – Number of Documents to fetch to pass to MMR algorithm. lambda_mult (float) – Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. kwargs (Any) – Returns List of Documents selected by maximal marginal relevance. Return type List[langchain.schema.Document] async amax_marginal_relevance_search(query, k=4, fetch_k=20, lambda_mult=0.5, **kwargs)[source] Return docs selected using the maximal marginal relevance. Parameters query (str) – k (int) – fetch_k (int) – lambda_mult (float) – kwargs (Any) – Return type List[langchain.schema.Document] max_marginal_relevance_search_by_vector(embedding, k=4, fetch_k=20, lambda_mult=0.5, **kwargs)[source] Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Parameters embedding (List[float]) – Embedding to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. fetch_k (int) – Number of Documents to fetch to pass to MMR algorithm. lambda_mult (float) – Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. kwargs (Any) – Returns List of Documents selected by maximal marginal relevance. Return type List[langchain.schema.Document] async amax_marginal_relevance_search_by_vector(embedding, k=4, fetch_k=20, lambda_mult=0.5, **kwargs)[source] Return docs selected using the maximal marginal relevance. Parameters embedding (List[float]) – k (int) – fetch_k (int) – lambda_mult (float) – kwargs (Any) – Return type List[langchain.schema.Document] classmethod from_documents(documents, embedding, **kwargs)[source] Return VectorStore initialized from documents and embeddings. Parameters documents (List[langchain.schema.Document]) – embedding (langchain.embeddings.base.Embeddings) – kwargs (Any) – Return type langchain.vectorstores.base.VST async classmethod afrom_documents(documents, embedding, **kwargs)[source] Return VectorStore initialized from documents and embeddings. Parameters documents (List[langchain.schema.Document]) – embedding (langchain.embeddings.base.Embeddings) – kwargs (Any) – Return type langchain.vectorstores.base.VST abstract classmethod from_texts(texts, embedding, metadatas=None, **kwargs)[source] Return VectorStore initialized from texts and embeddings. Parameters texts (List[str]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – kwargs (Any) – Return type langchain.vectorstores.base.VST async classmethod afrom_texts(texts, embedding, metadatas=None, **kwargs)[source] Return VectorStore initialized from texts and embeddings. Parameters texts (List[str]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – kwargs (Any) – Return type langchain.vectorstores.base.VST as_retriever(**kwargs)[source] Parameters kwargs (Any) – Return type langchain.vectorstores.base.VectorStoreRetriever class langchain.vectorstores.Weaviate(client, index_name, text_key, embedding=None, attributes=None, relevance_score_fn=<function _default_score_normalizer>, by_text=True)[source] Bases: langchain.vectorstores.base.VectorStore Wrapper around Weaviate vector database. To use, you should have the weaviate-client python package installed. Example import weaviate from langchain.vectorstores import Weaviate client = weaviate.Client(url=os.environ["WEAVIATE_URL"], ...) weaviate = Weaviate(client, index_name, text_key) Parameters client (Any) – index_name (str) – text_key (str) – embedding (Optional[Embeddings]) – attributes (Optional[List[str]]) – relevance_score_fn (Optional[Callable[[float], float]]) – by_text (bool) – add_texts(texts, metadatas=None, **kwargs)[source] Upload texts with metadata (properties) to Weaviate. Parameters texts (Iterable[str]) – metadatas (Optional[List[dict]]) – kwargs (Any) – Return type List[str] similarity_search(query, k=4, **kwargs)[source] Return docs most similar to query. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. kwargs (Any) – Returns List of Documents most similar to the query. Return type List[langchain.schema.Document] similarity_search_by_text(query, k=4, **kwargs)[source] Return docs most similar to query. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. kwargs (Any) – Returns List of Documents most similar to the query. Return type List[langchain.schema.Document] similarity_search_by_vector(embedding, k=4, **kwargs)[source] Look up similar documents by embedding vector in Weaviate. Parameters embedding (List[float]) – k (int) – kwargs (Any) – Return type List[langchain.schema.Document] max_marginal_relevance_search(query, k=4, fetch_k=20, lambda_mult=0.5, **kwargs)[source] Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Parameters query (str) – Text to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. fetch_k (int) – Number of Documents to fetch to pass to MMR algorithm. lambda_mult (float) – Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. kwargs (Any) – Returns List of Documents selected by maximal marginal relevance. Return type List[langchain.schema.Document] max_marginal_relevance_search_by_vector(embedding, k=4, fetch_k=20, lambda_mult=0.5, **kwargs)[source] Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Parameters embedding (List[float]) – Embedding to look up documents similar to. k (int) – Number of Documents to return. Defaults to 4. fetch_k (int) – Number of Documents to fetch to pass to MMR algorithm. lambda_mult (float) – Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. kwargs (Any) – Returns List of Documents selected by maximal marginal relevance. Return type List[langchain.schema.Document] similarity_search_with_score(query, k=4, **kwargs)[source] Return list of documents most similar to the query text and cosine distance in float for each. Lower score represents more similarity. Parameters query (str) – k (int) – kwargs (Any) – Return type List[Tuple[langchain.schema.Document, float]] classmethod from_texts(texts, embedding, metadatas=None, **kwargs)[source] Construct Weaviate wrapper from raw documents. This is a user-friendly interface that: Embeds documents. Creates a new index for the embeddings in the Weaviate instance. Adds the documents to the newly created Weaviate index. This is intended to be a quick way to get started. Example from langchain.vectorstores.weaviate import Weaviate from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() weaviate = Weaviate.from_texts( texts, embeddings, weaviate_url="http://localhost:8080" ) Parameters texts (List[str]) – embedding (langchain.embeddings.base.Embeddings) – metadatas (Optional[List[dict]]) – kwargs (Any) – Return type langchain.vectorstores.weaviate.Weaviate delete(ids)[source] Delete by vector IDs. Parameters ids (List[str]) – List of ids to delete. Return type None
https://api.python.langchain.com/en/latest/modules/vectorstores.html
cd96d496-587c-4066-8f33-4134e0c470ce
LLMs Wrappers on top of large language models APIs. class langchain.llms.AI21(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, model='j2-jumbo-instruct', temperature=0.7, maxTokens=256, minTokens=0, topP=1.0, presencePenalty=AI21PenaltyData(scale=0, applyToWhitespaces=True, applyToPunctuations=True, applyToNumbers=True, applyToStopwords=True, applyToEmojis=True), countPenalty=AI21PenaltyData(scale=0, applyToWhitespaces=True, applyToPunctuations=True, applyToNumbers=True, applyToStopwords=True, applyToEmojis=True), frequencyPenalty=AI21PenaltyData(scale=0, applyToWhitespaces=True, applyToPunctuations=True, applyToNumbers=True, applyToStopwords=True, applyToEmojis=True), numResults=1, logitBias=None, ai21_api_key=None, stop=None, base_url=None)[source] Bases: langchain.llms.base.LLM Wrapper around AI21 large language models. To use, you should have the environment variable AI21_API_KEY set with your API key. Example from langchain.llms import AI21 ai21 = AI21(model="j2-jumbo-instruct") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – model (str) – temperature (float) – maxTokens (int) – minTokens (int) – topP (float) – presencePenalty (langchain.llms.ai21.AI21PenaltyData) – countPenalty (langchain.llms.ai21.AI21PenaltyData) – frequencyPenalty (langchain.llms.ai21.AI21PenaltyData) – numResults (int) – logitBias (Optional[Dict[str, float]]) – ai21_api_key (Optional[str]) – stop (Optional[List[str]]) – base_url (Optional[str]) – Return type None attribute base_url: Optional[str] = None Base url to use, if None decides based on model name. attribute countPenalty: langchain.llms.ai21.AI21PenaltyData = AI21PenaltyData(scale=0, applyToWhitespaces=True, applyToPunctuations=True, applyToNumbers=True, applyToStopwords=True, applyToEmojis=True) Penalizes repeated tokens according to count. attribute frequencyPenalty: langchain.llms.ai21.AI21PenaltyData = AI21PenaltyData(scale=0, applyToWhitespaces=True, applyToPunctuations=True, applyToNumbers=True, applyToStopwords=True, applyToEmojis=True) Penalizes repeated tokens according to frequency. attribute logitBias: Optional[Dict[str, float]] = None Adjust the probability of specific tokens being generated. attribute maxTokens: int = 256 The maximum number of tokens to generate in the completion. attribute minTokens: int = 0 The minimum number of tokens to generate in the completion. attribute model: str = 'j2-jumbo-instruct' Model name to use. attribute numResults: int = 1 How many completions to generate for each prompt. attribute presencePenalty: langchain.llms.ai21.AI21PenaltyData = AI21PenaltyData(scale=0, applyToWhitespaces=True, applyToPunctuations=True, applyToNumbers=True, applyToStopwords=True, applyToEmojis=True) Penalizes repeated tokens. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute temperature: float = 0.7 What sampling temperature to use. attribute topP: float = 1.0 Total probability mass of tokens to consider at each step. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.AlephAlpha(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, model='luminous-base', maximum_tokens=64, temperature=0.0, top_k=0, top_p=0.0, presence_penalty=0.0, frequency_penalty=0.0, repetition_penalties_include_prompt=False, use_multiplicative_presence_penalty=False, penalty_bias=None, penalty_exceptions=None, penalty_exceptions_include_stop_sequences=None, best_of=None, n=1, logit_bias=None, log_probs=None, tokens=False, disable_optimizations=False, minimum_tokens=0, echo=False, use_multiplicative_frequency_penalty=False, sequence_penalty=0.0, sequence_penalty_min_length=2, use_multiplicative_sequence_penalty=False, completion_bias_inclusion=None, completion_bias_inclusion_first_token_only=False, completion_bias_exclusion=None, completion_bias_exclusion_first_token_only=False, contextual_control_threshold=None, control_log_additive=True, repetition_penalties_include_completion=True, raw_completion=False, aleph_alpha_api_key=None, stop_sequences=None)[source] Bases: langchain.llms.base.LLM Wrapper around Aleph Alpha large language models. To use, you should have the aleph_alpha_client python package installed, and the environment variable ALEPH_ALPHA_API_KEY set with your API key, or pass it as a named parameter to the constructor. Parameters are explained more in depth here: https://github.com/Aleph-Alpha/aleph-alpha-client/blob/c14b7dd2b4325c7da0d6a119f6e76385800e097b/aleph_alpha_client/completion.py#L10 Example from langchain.llms import AlephAlpha aleph_alpha = AlephAlpha(aleph_alpha_api_key="my-api-key") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – model (Optional[str]) – maximum_tokens (int) – temperature (float) – top_k (int) – top_p (float) – presence_penalty (float) – frequency_penalty (float) – repetition_penalties_include_prompt (Optional[bool]) – use_multiplicative_presence_penalty (Optional[bool]) – penalty_bias (Optional[str]) – penalty_exceptions (Optional[List[str]]) – penalty_exceptions_include_stop_sequences (Optional[bool]) – best_of (Optional[int]) – n (int) – logit_bias (Optional[Dict[int, float]]) – log_probs (Optional[int]) – tokens (Optional[bool]) – disable_optimizations (Optional[bool]) – minimum_tokens (Optional[int]) – echo (bool) – use_multiplicative_frequency_penalty (bool) – sequence_penalty (float) – sequence_penalty_min_length (int) – use_multiplicative_sequence_penalty (bool) – completion_bias_inclusion (Optional[Sequence[str]]) – completion_bias_inclusion_first_token_only (bool) – completion_bias_exclusion (Optional[Sequence[str]]) – completion_bias_exclusion_first_token_only (bool) – contextual_control_threshold (Optional[float]) – control_log_additive (Optional[bool]) – repetition_penalties_include_completion (bool) – raw_completion (bool) – aleph_alpha_api_key (Optional[str]) – stop_sequences (Optional[List[str]]) – Return type None attribute aleph_alpha_api_key: Optional[str] = None API key for Aleph Alpha API. attribute best_of: Optional[int] = None returns the one with the β€œbest of” results (highest log probability per token) attribute completion_bias_exclusion_first_token_only: bool = False Only consider the first token for the completion_bias_exclusion. attribute contextual_control_threshold: Optional[float] = None If set to None, attention control parameters only apply to those tokens that have explicitly been set in the request. If set to a non-None value, control parameters are also applied to similar tokens. attribute control_log_additive: Optional[bool] = True True: apply control by adding the log(control_factor) to attention scores. False: (attention_scores - - attention_scores.min(-1)) * control_factor attribute echo: bool = False Echo the prompt in the completion. attribute frequency_penalty: float = 0.0 Penalizes repeated tokens according to frequency. attribute log_probs: Optional[int] = None Number of top log probabilities to be returned for each generated token. attribute logit_bias: Optional[Dict[int, float]] = None The logit bias allows to influence the likelihood of generating tokens. attribute maximum_tokens: int = 64 The maximum number of tokens to be generated. attribute minimum_tokens: Optional[int] = 0 Generate at least this number of tokens. attribute model: Optional[str] = 'luminous-base' Model name to use. attribute n: int = 1 How many completions to generate for each prompt. attribute penalty_bias: Optional[str] = None Penalty bias for the completion. attribute penalty_exceptions: Optional[List[str]] = None List of strings that may be generated without penalty, regardless of other penalty settings attribute penalty_exceptions_include_stop_sequences: Optional[bool] = None Should stop_sequences be included in penalty_exceptions. attribute presence_penalty: float = 0.0 Penalizes repeated tokens. attribute raw_completion: bool = False Force the raw completion of the model to be returned. attribute repetition_penalties_include_completion: bool = True Flag deciding whether presence penalty or frequency penalty are updated from the completion. attribute repetition_penalties_include_prompt: Optional[bool] = False Flag deciding whether presence penalty or frequency penalty are updated from the prompt. attribute stop_sequences: Optional[List[str]] = None Stop sequences to use. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute temperature: float = 0.0 A non-negative float that tunes the degree of randomness in generation. attribute tokens: Optional[bool] = False return tokens of completion. attribute top_k: int = 0 Number of most likely tokens to consider at each step. attribute top_p: float = 0.0 Total probability mass of tokens to consider at each step. attribute use_multiplicative_presence_penalty: Optional[bool] = False Flag deciding whether presence penalty is applied multiplicatively (True) or additively (False). attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.AmazonAPIGateway(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, api_url, model_kwargs=None, content_handler=<langchain.llms.amazon_api_gateway.ContentHandlerAmazonAPIGateway object>)[source] Bases: langchain.llms.base.LLM Wrapper around custom Amazon API Gateway Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – api_url (str) – model_kwargs (Optional[Dict]) – content_handler (langchain.llms.amazon_api_gateway.ContentHandlerAmazonAPIGateway) – Return type None attribute api_url: str [Required] API Gateway URL attribute content_handler: langchain.llms.amazon_api_gateway.ContentHandlerAmazonAPIGateway = <langchain.llms.amazon_api_gateway.ContentHandlerAmazonAPIGateway object> The content handler class that provides an input and output transform functions to handle formats between LLM and the endpoint. attribute model_kwargs: Optional[Dict] = None Key word arguments to pass to the model. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.Anthropic(*, client=None, model='claude-v1', max_tokens_to_sample=256, temperature=None, top_k=None, top_p=None, streaming=False, default_request_timeout=None, anthropic_api_url=None, anthropic_api_key=None, HUMAN_PROMPT=None, AI_PROMPT=None, count_tokens=None, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None)[source] Bases: langchain.llms.base.LLM, langchain.llms.anthropic._AnthropicCommon Wrapper around Anthropic’s large language models. To use, you should have the anthropic python package installed, and the environment variable ANTHROPIC_API_KEY set with your API key, or pass it as a named parameter to the constructor. Example import anthropic from langchain.llms import Anthropic model = Anthropic(model="<model_name>", anthropic_api_key="my-api-key") # Simplest invocation, automatically wrapped with HUMAN_PROMPT # and AI_PROMPT. response = model("What are the biggest risks facing humanity?") # Or if you want to use the chat mode, build a few-shot-prompt, or # put words in the Assistant's mouth, use HUMAN_PROMPT and AI_PROMPT: raw_prompt = "What are the biggest risks facing humanity?" prompt = f"{anthropic.HUMAN_PROMPT} {prompt}{anthropic.AI_PROMPT}" response = model(prompt) Parameters client (Any) – model (str) – max_tokens_to_sample (int) – temperature (Optional[float]) – top_k (Optional[int]) – top_p (Optional[float]) – streaming (bool) – default_request_timeout (Optional[Union[float, Tuple[float, float]]]) – anthropic_api_url (Optional[str]) – anthropic_api_key (Optional[str]) – HUMAN_PROMPT (Optional[str]) – AI_PROMPT (Optional[str]) – count_tokens (Optional[Callable[[str], int]]) – cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – Return type None attribute default_request_timeout: Optional[Union[float, Tuple[float, float]]] = None Timeout for requests to Anthropic Completion API. Default is 600 seconds. attribute max_tokens_to_sample: int = 256 Denotes the number of tokens to predict per generation. attribute model: str = 'claude-v1' Model name to use. attribute streaming: bool = False Whether to stream the results. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute temperature: Optional[float] = None A non-negative float that tunes the degree of randomness in generation. attribute top_k: Optional[int] = None Number of most likely tokens to consider at each step. attribute top_p: Optional[float] = None Total probability mass of tokens to consider at each step. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text)[source] Calculate number of tokens. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) stream(prompt, stop=None)[source] Call Anthropic completion_stream and return the resulting generator. BETA: this is a beta feature while we figure out the right abstraction. Once that happens, this interface could change. Parameters prompt (str) – The prompt to pass into the model. stop (Optional[List[str]]) – Optional list of stop words to use when generating. Returns A generator representing the stream of tokens from Anthropic. Return type Generator Example prompt = "Write a poem about a stream." prompt = f"\n\nHuman: {prompt}\n\nAssistant:" generator = anthropic.stream(prompt) for token in generator: yield token classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.Anyscale(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, model_kwargs=None, anyscale_service_url=None, anyscale_service_route=None, anyscale_service_token=None)[source] Bases: langchain.llms.base.LLM Wrapper around Anyscale Services. To use, you should have the environment variable ANYSCALE_SERVICE_URL, ANYSCALE_SERVICE_ROUTE and ANYSCALE_SERVICE_TOKEN set with your Anyscale Service, or pass it as a named parameter to the constructor. Example from langchain.llms import Anyscale anyscale = Anyscale(anyscale_service_url="SERVICE_URL", anyscale_service_route="SERVICE_ROUTE", anyscale_service_token="SERVICE_TOKEN") # Use Ray for distributed processing import ray prompt_list=[] @ray.remote def send_query(llm, prompt): resp = llm(prompt) return resp futures = [send_query.remote(anyscale, prompt) for prompt in prompt_list] results = ray.get(futures) Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – model_kwargs (Optional[dict]) – anyscale_service_url (Optional[str]) – anyscale_service_route (Optional[str]) – anyscale_service_token (Optional[str]) – Return type None attribute model_kwargs: Optional[dict] = None Key word arguments to pass to the model. Reserved for future use attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.Aviary(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, model='amazon/LightGPT', aviary_url=None, aviary_token=None, use_prompt_format=True, version=None)[source] Bases: langchain.llms.base.LLM Allow you to use an Aviary. Aviary is a backend for hosted models. You can find out more about aviary at http://github.com/ray-project/aviary To get a list of the models supported on an aviary, follow the instructions on the web site to install the aviary CLI and then use: aviary models AVIARY_URL and AVIARY_TOKEN environement variables must be set. Example from langchain.llms import Aviary os.environ["AVIARY_URL"] = "<URL>" os.environ["AVIARY_TOKEN"] = "<TOKEN>" light = Aviary(model='amazon/LightGPT') output = light('How do you make fried rice?') Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – model (str) – aviary_url (Optional[str]) – aviary_token (Optional[str]) – use_prompt_format (bool) – version (Optional[str]) – Return type None attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.AzureMLOnlineEndpoint(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, endpoint_url='', endpoint_api_key='', deployment_name='', http_client=None, content_formatter=None, model_kwargs=None)[source] Bases: langchain.llms.base.LLM, pydantic.main.BaseModel Wrapper around Azure ML Hosted models using Managed Online Endpoints. Example azure_llm = AzureMLModel( endpoint_url="https://<your-endpoint>.<your_region>.inference.ml.azure.com/score", endpoint_api_key="my-api-key", deployment_name="my-deployment-name", content_formatter=content_formatter, ) Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – endpoint_url (str) – endpoint_api_key (str) – deployment_name (str) – http_client (Any) – content_formatter (Any) – model_kwargs (Optional[dict]) – Return type None attribute content_formatter: Any = None The content formatter that provides an input and output transform function to handle formats between the LLM and the endpoint attribute deployment_name: str = '' Deployment Name for Endpoint. Should be passed to constructor or specified as env var AZUREML_DEPLOYMENT_NAME. attribute endpoint_api_key: str = '' Authentication Key for Endpoint. Should be passed to constructor or specified as env var AZUREML_ENDPOINT_API_KEY. attribute endpoint_url: str = '' URL of pre-existing Endpoint. Should be passed to constructor or specified as env var AZUREML_ENDPOINT_URL. attribute model_kwargs: Optional[dict] = None Key word arguments to pass to the model. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.AzureOpenAI(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, model='text-davinci-003', temperature=0.7, max_tokens=256, top_p=1, frequency_penalty=0, presence_penalty=0, n=1, best_of=1, model_kwargs=None, openai_api_key=None, openai_api_base=None, openai_organization=None, openai_proxy=None, batch_size=20, request_timeout=None, logit_bias=None, max_retries=6, streaming=False, allowed_special={}, disallowed_special='all', tiktoken_model_name=None, deployment_name='', openai_api_type='azure', openai_api_version='')[source] Bases: langchain.llms.openai.BaseOpenAI Wrapper around Azure-specific OpenAI large language models. To use, you should have the openai python package installed, and the environment variable OPENAI_API_KEY set with your API key. Any parameters that are valid to be passed to the openai.create call can be passed in, even if not explicitly saved on this class. Example from langchain.llms import AzureOpenAI openai = AzureOpenAI(model_name="text-davinci-003") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – model (str) – temperature (float) – max_tokens (int) – top_p (float) – frequency_penalty (float) – presence_penalty (float) – n (int) – best_of (int) – model_kwargs (Dict[str, Any]) – openai_api_key (Optional[str]) – openai_api_base (Optional[str]) – openai_organization (Optional[str]) – openai_proxy (Optional[str]) – batch_size (int) – request_timeout (Optional[Union[float, Tuple[float, float]]]) – logit_bias (Optional[Dict[str, float]]) – max_retries (int) – streaming (bool) – allowed_special (Union[Literal['all'], typing.AbstractSet[str]]) – disallowed_special (Union[Literal['all'], typing.Collection[str]]) – tiktoken_model_name (Optional[str]) – deployment_name (str) – openai_api_type (str) – openai_api_version (str) – Return type None attribute allowed_special: Union[Literal['all'], AbstractSet[str]] = {} Set of special tokens that are allowed。 attribute batch_size: int = 20 Batch size to use when passing multiple documents to generate. attribute best_of: int = 1 Generates best_of completions server-side and returns the β€œbest”. attribute deployment_name: str = '' Deployment name to use. attribute disallowed_special: Union[Literal['all'], Collection[str]] = 'all' Set of special tokens that are not allowed。 attribute frequency_penalty: float = 0 Penalizes repeated tokens according to frequency. attribute logit_bias: Optional[Dict[str, float]] [Optional] Adjust the probability of specific tokens being generated. attribute max_retries: int = 6 Maximum number of retries to make when generating. attribute max_tokens: int = 256 The maximum number of tokens to generate in the completion. -1 returns as many tokens as possible given the prompt and the models maximal context size. attribute model_kwargs: Dict[str, Any] [Optional] Holds any model parameters valid for create call not explicitly specified. attribute model_name: str = 'text-davinci-003' (alias 'model') Model name to use. attribute n: int = 1 How many completions to generate for each prompt. attribute presence_penalty: float = 0 Penalizes repeated tokens. attribute request_timeout: Optional[Union[float, Tuple[float, float]]] = None Timeout for requests to OpenAI completion API. Default is 600 seconds. attribute streaming: bool = False Whether to stream the results or not. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute temperature: float = 0.7 What sampling temperature to use. attribute tiktoken_model_name: Optional[str] = None The model name to pass to tiktoken when using this class. Tiktoken is used to count the number of tokens in documents to constrain them to be under a certain limit. By default, when set to None, this will be the same as the embedding model name. However, there are some cases where you may want to use this Embedding class with a model name not supported by tiktoken. This can include when using Azure embeddings or when using one of the many model providers that expose an OpenAI-like API but with different models. In those cases, in order to avoid erroring when tiktoken is called, you can specify a model name to use here. attribute top_p: float = 1 Total probability mass of tokens to consider at each step. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model create_llm_result(choices, prompts, token_usage) Create the LLMResult from the choices and prompts. Parameters choices (Any) – prompts (List[str]) – token_usage (Dict[str, int]) – Return type langchain.schema.LLMResult dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_sub_prompts(params, prompts, stop=None) Get the sub prompts for llm call. Parameters params (Dict[str, Any]) – prompts (List[str]) – stop (Optional[List[str]]) – Return type List[List[str]] get_token_ids(text) Get the token IDs using the tiktoken package. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode max_tokens_for_prompt(prompt) Calculate the maximum number of tokens possible to generate for a prompt. Parameters prompt (str) – The prompt to pass into the model. Returns The maximum number of tokens to generate for a prompt. Return type int Example max_tokens = openai.max_token_for_prompt("Tell me a joke.") static modelname_to_contextsize(modelname) Calculate the maximum number of tokens possible to generate for a model. Parameters modelname (str) – The modelname we want to know the context size for. Returns The maximum context size Return type int Example max_tokens = openai.modelname_to_contextsize("text-davinci-003") predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage prep_streaming_params(stop=None) Prepare the params for streaming. Parameters stop (Optional[List[str]]) – Return type Dict[str, Any] save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) stream(prompt, stop=None) Call OpenAI with streaming flag and return the resulting generator. BETA: this is a beta feature while we figure out the right abstraction. Once that happens, this interface could change. Parameters prompt (str) – The prompts to pass into the model. stop (Optional[List[str]]) – Optional list of stop words to use when generating. Returns A generator representing the stream of tokens from OpenAI. Return type Generator Example generator = openai.stream("Tell me a joke.") for token in generator: yield token classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. property max_context_size: int Get max context size for this model. class langchain.llms.Banana(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, model_key='', model_kwargs=None, banana_api_key=None)[source] Bases: langchain.llms.base.LLM Wrapper around Banana large language models. To use, you should have the banana-dev python package installed, and the environment variable BANANA_API_KEY set with your API key. Any parameters that are valid to be passed to the call can be passed in, even if not explicitly saved on this class. Example from langchain.llms import Banana banana = Banana(model_key="") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – model_key (str) – model_kwargs (Dict[str, Any]) – banana_api_key (Optional[str]) – Return type None attribute model_key: str = '' model endpoint to use attribute model_kwargs: Dict[str, Any] [Optional] Holds any model parameters valid for create call not explicitly specified. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.Baseten(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, model, input=None, model_kwargs=None)[source] Bases: langchain.llms.base.LLM Use your Baseten models in Langchain To use, you should have the baseten python package installed, and run baseten.login() with your Baseten API key. The required model param can be either a model id or model version id. Using a model version ID will result in slightly faster invocation. Any other model parameters can also be passed in with the format input={model_param: value, …} The Baseten model must accept a dictionary of input with the key β€œprompt” and return a dictionary with a key β€œdata” which maps to a list of response strings. Example Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – model (str) – input (Dict[str, Any]) – model_kwargs (Dict[str, Any]) – Return type None attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.Beam(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, model_name='', name='', cpu='', memory='', gpu='', python_version='', python_packages=[], max_length='', url='', model_kwargs=None, beam_client_id='', beam_client_secret='', app_id=None)[source] Bases: langchain.llms.base.LLM Wrapper around Beam API for gpt2 large language model. To use, you should have the beam-sdk python package installed, and the environment variable BEAM_CLIENT_ID set with your client id and BEAM_CLIENT_SECRET set with your client secret. Information on how to get these is available here: https://docs.beam.cloud/account/api-keys. The wrapper can then be called as follows, where the name, cpu, memory, gpu, python version, and python packages can be updated accordingly. Once deployed, the instance can be called. Example llm = Beam(model_name="gpt2", name="langchain-gpt2", cpu=8, memory="32Gi", gpu="A10G", python_version="python3.8", python_packages=[ "diffusers[torch]>=0.10", "transformers", "torch", "pillow", "accelerate", "safetensors", "xformers",], max_length=50) llm._deploy() call_result = llm._call(input) Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – model_name (str) – name (str) – cpu (str) – memory (str) – gpu (str) – python_version (str) – python_packages (List[str]) – max_length (str) – url (str) – model_kwargs (Dict[str, Any]) – beam_client_id (str) – beam_client_secret (str) – app_id (Optional[str]) – Return type None attribute model_kwargs: Dict[str, Any] [Optional] Holds any model parameters valid for create call not explicitly specified. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute url: str = '' model endpoint to use attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult app_creation()[source] Creates a Python file which will contain your Beam app definition. Return type None async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage run_creation()[source] Creates a Python file which will be deployed on beam. Return type None save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.Bedrock(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, region_name=None, credentials_profile_name=None, model_id, model_kwargs=None)[source] Bases: langchain.llms.base.LLM LLM provider to invoke Bedrock models. To authenticate, the AWS client uses the following methods to automatically load credentials: https://boto3.amazonaws.com/v1/documentation/api/latest/guide/credentials.html If a specific credential profile should be used, you must pass the name of the profile from the ~/.aws/credentials file that is to be used. Make sure the credentials / roles used have the required policies to access the Bedrock service. Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – region_name (Optional[str]) – credentials_profile_name (Optional[str]) – model_id (str) – model_kwargs (Optional[Dict]) – Return type None attribute credentials_profile_name: Optional[str] = None The name of the profile in the ~/.aws/credentials or ~/.aws/config files, which has either access keys or role information specified. If not specified, the default credential profile or, if on an EC2 instance, credentials from IMDS will be used. See: https://boto3.amazonaws.com/v1/documentation/api/latest/guide/credentials.html attribute model_id: str [Required] Id of the model to call, e.g., amazon.titan-tg1-large, this is equivalent to the modelId property in the list-foundation-models api attribute model_kwargs: Optional[Dict] = None Key word arguments to pass to the model. attribute region_name: Optional[str] = None The aws region e.g., us-west-2. Fallsback to AWS_DEFAULT_REGION env variable or region specified in ~/.aws/config in case it is not provided here. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.CTransformers(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, model, model_type=None, model_file=None, config=None, lib=None)[source] Bases: langchain.llms.base.LLM Wrapper around the C Transformers LLM interface. To use, you should have the ctransformers python package installed. See https://github.com/marella/ctransformers Example from langchain.llms import CTransformers llm = CTransformers(model="/path/to/ggml-gpt-2.bin", model_type="gpt2") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – model (str) – model_type (Optional[str]) – model_file (Optional[str]) – config (Optional[Dict[str, Any]]) – lib (Optional[str]) – Return type None attribute config: Optional[Dict[str, Any]] = None The config parameters. See https://github.com/marella/ctransformers#config attribute lib: Optional[str] = None The path to a shared library or one of avx2, avx, basic. attribute model: str [Required] The path to a model file or directory or the name of a Hugging Face Hub model repo. attribute model_file: Optional[str] = None The name of the model file in repo or directory. attribute model_type: Optional[str] = None The model type. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.CerebriumAI(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, endpoint_url='', model_kwargs=None, cerebriumai_api_key=None)[source] Bases: langchain.llms.base.LLM Wrapper around CerebriumAI large language models. To use, you should have the cerebrium python package installed, and the environment variable CEREBRIUMAI_API_KEY set with your API key. Any parameters that are valid to be passed to the call can be passed in, even if not explicitly saved on this class. Example from langchain.llms import CerebriumAI cerebrium = CerebriumAI(endpoint_url="") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – endpoint_url (str) – model_kwargs (Dict[str, Any]) – cerebriumai_api_key (Optional[str]) – Return type None attribute endpoint_url: str = '' model endpoint to use attribute model_kwargs: Dict[str, Any] [Optional] Holds any model parameters valid for create call not explicitly specified. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.Clarifai(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, stub=None, metadata=None, userDataObject=None, model_id=None, model_version_id=None, app_id=None, user_id=None, clarifai_pat_key=None, api_base='https://api.clarifai.com', stop=None)[source] Bases: langchain.llms.base.LLM Wrapper around Clarifai’s large language models. To use, you should have an account on the Clarifai platform, the clarifai python package installed, and the environment variable CLARIFAI_PAT_KEY set with your PAT key, or pass it as a named parameter to the constructor. Example from langchain.llms import Clarifai clarifai_llm = Clarifai(clarifai_pat_key=CLARIFAI_PAT_KEY, user_id=USER_ID, app_id=APP_ID, model_id=MODEL_ID) Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – stub (Any) – metadata (Any) – userDataObject (Any) – model_id (Optional[str]) – model_version_id (Optional[str]) – app_id (Optional[str]) – user_id (Optional[str]) – clarifai_pat_key (Optional[str]) – api_base (str) – stop (Optional[List[str]]) – Return type None attribute app_id: Optional[str] = None Clarifai application id to use. attribute model_id: Optional[str] = None Model id to use. attribute model_version_id: Optional[str] = None Model version id to use. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute user_id: Optional[str] = None Clarifai user id to use. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.Cohere(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, model=None, max_tokens=256, temperature=0.75, k=0, p=1, frequency_penalty=0.0, presence_penalty=0.0, truncate=None, max_retries=10, cohere_api_key=None, stop=None)[source] Bases: langchain.llms.base.LLM Wrapper around Cohere large language models. To use, you should have the cohere python package installed, and the environment variable COHERE_API_KEY set with your API key, or pass it as a named parameter to the constructor. Example from langchain.llms import Cohere cohere = Cohere(model="gptd-instruct-tft", cohere_api_key="my-api-key") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – model (Optional[str]) – max_tokens (int) – temperature (float) – k (int) – p (int) – frequency_penalty (float) – presence_penalty (float) – truncate (Optional[str]) – max_retries (int) – cohere_api_key (Optional[str]) – stop (Optional[List[str]]) – Return type None attribute frequency_penalty: float = 0.0 Penalizes repeated tokens according to frequency. Between 0 and 1. attribute k: int = 0 Number of most likely tokens to consider at each step. attribute max_retries: int = 10 Maximum number of retries to make when generating. attribute max_tokens: int = 256 Denotes the number of tokens to predict per generation. attribute model: Optional[str] = None Model name to use. attribute p: int = 1 Total probability mass of tokens to consider at each step. attribute presence_penalty: float = 0.0 Penalizes repeated tokens. Between 0 and 1. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute temperature: float = 0.75 A non-negative float that tunes the degree of randomness in generation. attribute truncate: Optional[str] = None Specify how the client handles inputs longer than the maximum token length: Truncate from START, END or NONE attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.Databricks(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, host=None, api_token=None, endpoint_name=None, cluster_id=None, cluster_driver_port=None, model_kwargs=None, transform_input_fn=None, transform_output_fn=None)[source] Bases: langchain.llms.base.LLM LLM wrapper around a Databricks serving endpoint or a cluster driver proxy app. It supports two endpoint types: Serving endpoint (recommended for both production and development). We assume that an LLM was registered and deployed to a serving endpoint. To wrap it as an LLM you must have β€œCan Query” permission to the endpoint. Set endpoint_name accordingly and do not set cluster_id and cluster_driver_port. The expected model signature is: inputs: [{"name": "prompt", "type": "string"}, {"name": "stop", "type": "list[string]"}] outputs: [{"type": "string"}] Cluster driver proxy app (recommended for interactive development). One can load an LLM on a Databricks interactive cluster and start a local HTTP server on the driver node to serve the model at / using HTTP POST method with JSON input/output. Please use a port number between [3000, 8000] and let the server listen to the driver IP address or simply 0.0.0.0 instead of localhost only. To wrap it as an LLM you must have β€œCan Attach To” permission to the cluster. Set cluster_id and cluster_driver_port and do not set endpoint_name. The expected server schema (using JSON schema) is: inputs: {"type": "object", "properties": { "prompt": {"type": "string"}, "stop": {"type": "array", "items": {"type": "string"}}}, "required": ["prompt"]}` outputs: {"type": "string"} If the endpoint model signature is different or you want to set extra params, you can use transform_input_fn and transform_output_fn to apply necessary transformations before and after the query. Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – host (str) – api_token (str) – endpoint_name (Optional[str]) – cluster_id (Optional[str]) – cluster_driver_port (Optional[str]) – model_kwargs (Optional[Dict[str, Any]]) – transform_input_fn (Optional[Callable]) – transform_output_fn (Optional[Callable[[...], str]]) – Return type None attribute api_token: str [Optional] Databricks personal access token. If not provided, the default value is determined by the DATABRICKS_TOKEN environment variable if present, or an automatically generated temporary token if running inside a Databricks notebook attached to an interactive cluster in β€œsingle user” or β€œno isolation shared” mode. attribute cluster_driver_port: Optional[str] = None The port number used by the HTTP server running on the cluster driver node. The server should listen on the driver IP address or simply 0.0.0.0 to connect. We recommend the server using a port number between [3000, 8000]. attribute cluster_id: Optional[str] = None ID of the cluster if connecting to a cluster driver proxy app. If neither endpoint_name nor cluster_id is not provided and the code runs inside a Databricks notebook attached to an interactive cluster in β€œsingle user” or β€œno isolation shared” mode, the current cluster ID is used as default. You must not set both endpoint_name and cluster_id. attribute endpoint_name: Optional[str] = None Name of the model serving endpont. You must specify the endpoint name to connect to a model serving endpoint. You must not set both endpoint_name and cluster_id. attribute host: str [Optional] Databricks workspace hostname. If not provided, the default value is determined by the DATABRICKS_HOST environment variable if present, or the hostname of the current Databricks workspace if running inside a Databricks notebook attached to an interactive cluster in β€œsingle user” or β€œno isolation shared” mode. attribute model_kwargs: Optional[Dict[str, Any]] = None Extra parameters to pass to the endpoint. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute transform_input_fn: Optional[Callable] = None A function that transforms {prompt, stop, **kwargs} into a JSON-compatible request object that the endpoint accepts. For example, you can apply a prompt template to the input prompt. attribute transform_output_fn: Optional[Callable[[...], str]] = None A function that transforms the output from the endpoint to the generated text. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.DeepInfra(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, model_id='google/flan-t5-xl', model_kwargs=None, deepinfra_api_token=None)[source] Bases: langchain.llms.base.LLM Wrapper around DeepInfra deployed models. To use, you should have the requests python package installed, and the environment variable DEEPINFRA_API_TOKEN set with your API token, or pass it as a named parameter to the constructor. Only supports text-generation and text2text-generation for now. Example from langchain.llms import DeepInfra di = DeepInfra(model_id="google/flan-t5-xl", deepinfra_api_token="my-api-key") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – model_id (str) – model_kwargs (Optional[dict]) – deepinfra_api_token (Optional[str]) – Return type None attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.FakeListLLM(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, responses, i=0)[source] Bases: langchain.llms.base.LLM Fake LLM wrapper for testing purposes. Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – responses (List) – i (int) – Return type None attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.ForefrontAI(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, endpoint_url='', temperature=0.7, length=256, top_p=1.0, top_k=40, repetition_penalty=1, forefrontai_api_key=None, base_url=None)[source] Bases: langchain.llms.base.LLM Wrapper around ForefrontAI large language models. To use, you should have the environment variable FOREFRONTAI_API_KEY set with your API key. Example from langchain.llms import ForefrontAI forefrontai = ForefrontAI(endpoint_url="") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – endpoint_url (str) – temperature (float) – length (int) – top_p (float) – top_k (int) – repetition_penalty (int) – forefrontai_api_key (Optional[str]) – base_url (Optional[str]) – Return type None attribute base_url: Optional[str] = None Base url to use, if None decides based on model name. attribute endpoint_url: str = '' Model name to use. attribute length: int = 256 The maximum number of tokens to generate in the completion. attribute repetition_penalty: int = 1 Penalizes repeated tokens according to frequency. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute temperature: float = 0.7 What sampling temperature to use. attribute top_k: int = 40 The number of highest probability vocabulary tokens to keep for top-k-filtering. attribute top_p: float = 1.0 Total probability mass of tokens to consider at each step. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.GPT4All(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, model, backend=None, n_ctx=512, n_parts=- 1, seed=0, f16_kv=False, logits_all=False, vocab_only=False, use_mlock=False, embedding=False, n_threads=4, n_predict=256, temp=0.8, top_p=0.95, top_k=40, echo=False, stop=[], repeat_last_n=64, repeat_penalty=1.3, n_batch=1, streaming=False, context_erase=0.5, allow_download=False, client=None)[source] Bases: langchain.llms.base.LLM Wrapper around GPT4All language models. To use, you should have the gpt4all python package installed, the pre-trained model file, and the model’s config information. Example from langchain.llms import GPT4All model = GPT4All(model="./models/gpt4all-model.bin", n_ctx=512, n_threads=8) # Simplest invocation response = model("Once upon a time, ") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – model (str) – backend (Optional[str]) – n_ctx (int) – n_parts (int) – seed (int) – f16_kv (bool) – logits_all (bool) – vocab_only (bool) – use_mlock (bool) – embedding (bool) – n_threads (Optional[int]) – n_predict (Optional[int]) – temp (Optional[float]) – top_p (Optional[float]) – top_k (Optional[int]) – echo (Optional[bool]) – stop (Optional[List[str]]) – repeat_last_n (Optional[int]) – repeat_penalty (Optional[float]) – n_batch (int) – streaming (bool) – context_erase (float) – allow_download (bool) – client (Any) – Return type None attribute allow_download: bool = False If model does not exist in ~/.cache/gpt4all/, download it. attribute context_erase: float = 0.5 Leave (n_ctx * context_erase) tokens starting from beginning if the context has run out. attribute echo: Optional[bool] = False Whether to echo the prompt. attribute embedding: bool = False Use embedding mode only. attribute f16_kv: bool = False Use half-precision for key/value cache. attribute logits_all: bool = False Return logits for all tokens, not just the last token. attribute model: str [Required] Path to the pre-trained GPT4All model file. attribute n_batch: int = 1 Batch size for prompt processing. attribute n_ctx: int = 512 Token context window. attribute n_parts: int = -1 Number of parts to split the model into. If -1, the number of parts is automatically determined. attribute n_predict: Optional[int] = 256 The maximum number of tokens to generate. attribute n_threads: Optional[int] = 4 Number of threads to use. attribute repeat_last_n: Optional[int] = 64 Last n tokens to penalize attribute repeat_penalty: Optional[float] = 1.3 The penalty to apply to repeated tokens. attribute seed: int = 0 Seed. If -1, a random seed is used. attribute stop: Optional[List[str]] = [] A list of strings to stop generation when encountered. attribute streaming: bool = False Whether to stream the results or not. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute temp: Optional[float] = 0.8 The temperature to use for sampling. attribute top_k: Optional[int] = 40 The top-k value to use for sampling. attribute top_p: Optional[float] = 0.95 The top-p value to use for sampling. attribute use_mlock: bool = False Force system to keep model in RAM. attribute verbose: bool [Optional] Whether to print out response text. attribute vocab_only: bool = False Only load the vocabulary, no weights. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.GooglePalm(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, google_api_key=None, model_name='models/text-bison-001', temperature=0.7, top_p=None, top_k=None, max_output_tokens=None, n=1)[source] Bases: langchain.llms.base.BaseLLM, pydantic.main.BaseModel Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – google_api_key (Optional[str]) – model_name (str) – temperature (float) – top_p (Optional[float]) – top_k (Optional[int]) – max_output_tokens (Optional[int]) – n (int) – Return type None attribute max_output_tokens: Optional[int] = None Maximum number of tokens to include in a candidate. Must be greater than zero. If unset, will default to 64. attribute model_name: str = 'models/text-bison-001' Model name to use. attribute n: int = 1 Number of chat completions to generate for each prompt. Note that the API may not return the full n completions if duplicates are generated. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute temperature: float = 0.7 Run inference with this temperature. Must by in the closed interval [0.0, 1.0]. attribute top_k: Optional[int] = None Decode using top-k sampling: consider the set of top_k most probable tokens. Must be positive. attribute top_p: Optional[float] = None Decode using nucleus sampling: consider the smallest set of tokens whose probability sum is at least top_p. Must be in the closed interval [0.0, 1.0]. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.GooseAI(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, model_name='gpt-neo-20b', temperature=0.7, max_tokens=256, top_p=1, min_tokens=1, frequency_penalty=0, presence_penalty=0, n=1, model_kwargs=None, logit_bias=None, gooseai_api_key=None)[source] Bases: langchain.llms.base.LLM Wrapper around OpenAI large language models. To use, you should have the openai python package installed, and the environment variable GOOSEAI_API_KEY set with your API key. Any parameters that are valid to be passed to the openai.create call can be passed in, even if not explicitly saved on this class. Example from langchain.llms import GooseAI gooseai = GooseAI(model_name="gpt-neo-20b") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – model_name (str) – temperature (float) – max_tokens (int) – top_p (float) – min_tokens (int) – frequency_penalty (float) – presence_penalty (float) – n (int) – model_kwargs (Dict[str, Any]) – logit_bias (Optional[Dict[str, float]]) – gooseai_api_key (Optional[str]) – Return type None attribute frequency_penalty: float = 0 Penalizes repeated tokens according to frequency. attribute logit_bias: Optional[Dict[str, float]] [Optional] Adjust the probability of specific tokens being generated. attribute max_tokens: int = 256 The maximum number of tokens to generate in the completion. -1 returns as many tokens as possible given the prompt and the models maximal context size. attribute min_tokens: int = 1 The minimum number of tokens to generate in the completion. attribute model_kwargs: Dict[str, Any] [Optional] Holds any model parameters valid for create call not explicitly specified. attribute model_name: str = 'gpt-neo-20b' Model name to use attribute n: int = 1 How many completions to generate for each prompt. attribute presence_penalty: float = 0 Penalizes repeated tokens. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute temperature: float = 0.7 What sampling temperature to use attribute top_p: float = 1 Total probability mass of tokens to consider at each step. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.HuggingFaceEndpoint(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, endpoint_url='', task=None, model_kwargs=None, huggingfacehub_api_token=None)[source] Bases: langchain.llms.base.LLM Wrapper around HuggingFaceHub Inference Endpoints. To use, you should have the huggingface_hub python package installed, and the environment variable HUGGINGFACEHUB_API_TOKEN set with your API token, or pass it as a named parameter to the constructor. Only supports text-generation and text2text-generation for now. Example from langchain.llms import HuggingFaceEndpoint endpoint_url = ( "https://abcdefghijklmnop.us-east-1.aws.endpoints.huggingface.cloud" ) hf = HuggingFaceEndpoint( endpoint_url=endpoint_url, huggingfacehub_api_token="my-api-key" ) Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – endpoint_url (str) – task (Optional[str]) – model_kwargs (Optional[dict]) – huggingfacehub_api_token (Optional[str]) – Return type None attribute endpoint_url: str = '' Endpoint URL to use. attribute model_kwargs: Optional[dict] = None Key word arguments to pass to the model. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute task: Optional[str] = None Task to call the model with. Should be a task that returns generated_text or summary_text. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.HuggingFaceHub(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, repo_id='gpt2', task=None, model_kwargs=None, huggingfacehub_api_token=None)[source] Bases: langchain.llms.base.LLM Wrapper around HuggingFaceHub models. To use, you should have the huggingface_hub python package installed, and the environment variable HUGGINGFACEHUB_API_TOKEN set with your API token, or pass it as a named parameter to the constructor. Only supports text-generation, text2text-generation and summarization for now. Example from langchain.llms import HuggingFaceHub hf = HuggingFaceHub(repo_id="gpt2", huggingfacehub_api_token="my-api-key") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – repo_id (str) – task (Optional[str]) – model_kwargs (Optional[dict]) – huggingfacehub_api_token (Optional[str]) – Return type None attribute model_kwargs: Optional[dict] = None Key word arguments to pass to the model. attribute repo_id: str = 'gpt2' Model name to use. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute task: Optional[str] = None Task to call the model with. Should be a task that returns generated_text or summary_text. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.HuggingFacePipeline(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, pipeline=None, model_id='gpt2', model_kwargs=None, pipeline_kwargs=None)[source] Bases: langchain.llms.base.LLM Wrapper around HuggingFace Pipeline API. To use, you should have the transformers python package installed. Only supports text-generation, text2text-generation and summarization for now. Example using from_model_id:from langchain.llms import HuggingFacePipeline hf = HuggingFacePipeline.from_model_id( model_id="gpt2", task="text-generation", pipeline_kwargs={"max_new_tokens": 10}, ) Example passing pipeline in directly:from langchain.llms import HuggingFacePipeline from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline model_id = "gpt2" tokenizer = AutoTokenizer.from_pretrained(model_id) model = AutoModelForCausalLM.from_pretrained(model_id) pipe = pipeline( "text-generation", model=model, tokenizer=tokenizer, max_new_tokens=10 ) hf = HuggingFacePipeline(pipeline=pipe) Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – pipeline (Any) – model_id (str) – model_kwargs (Optional[dict]) – pipeline_kwargs (Optional[dict]) – Return type None attribute model_id: str = 'gpt2' Model name to use. attribute model_kwargs: Optional[dict] = None Key word arguments passed to the model. attribute pipeline_kwargs: Optional[dict] = None Key word arguments passed to the pipeline. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict classmethod from_model_id(model_id, task, device=- 1, model_kwargs=None, pipeline_kwargs=None, **kwargs)[source] Construct the pipeline object from model_id and task. Parameters model_id (str) – task (str) – device (int) – model_kwargs (Optional[dict]) – pipeline_kwargs (Optional[dict]) – kwargs (Any) – Return type langchain.llms.base.LLM generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.HuggingFaceTextGenInference(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, max_new_tokens=512, top_k=None, top_p=0.95, typical_p=0.95, temperature=0.8, repetition_penalty=None, stop_sequences=None, seed=None, inference_server_url='', timeout=120, server_kwargs=None, stream=False, client=None, async_client=None)[source] Bases: langchain.llms.base.LLM HuggingFace text generation inference API. This class is a wrapper around the HuggingFace text generation inference API. It is used to generate text from a given prompt. Attributes: - max_new_tokens: The maximum number of tokens to generate. - top_k: The number of top-k tokens to consider when generating text. - top_p: The cumulative probability threshold for generating text. - typical_p: The typical probability threshold for generating text. - temperature: The temperature to use when generating text. - repetition_penalty: The repetition penalty to use when generating text. - stop_sequences: A list of stop sequences to use when generating text. - seed: The seed to use when generating text. - inference_server_url: The URL of the inference server to use. - timeout: The timeout value in seconds to use while connecting to inference server. - server_kwargs: The keyword arguments to pass to the inference server. - client: The client object used to communicate with the inference server. - async_client: The async client object used to communicate with the server. Methods: - _call: Generates text based on a given prompt and stop sequences. - _acall: Async generates text based on a given prompt and stop sequences. - _llm_type: Returns the type of LLM. Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – max_new_tokens (int) – top_k (Optional[int]) – top_p (Optional[float]) – typical_p (Optional[float]) – temperature (float) – repetition_penalty (Optional[float]) – stop_sequences (List[str]) – seed (Optional[int]) – inference_server_url (str) – timeout (int) – server_kwargs (Dict[str, Any]) – stream (bool) – client (Any) – async_client (Any) – Return type None attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.HumanInputLLM(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, input_func=None, prompt_func=None, separator='\n', input_kwargs={}, prompt_kwargs={})[source] Bases: langchain.llms.base.LLM A LLM wrapper which returns user input as the response. Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – input_func (Callable) – prompt_func (Callable[[str], None]) – separator (str) – input_kwargs (Mapping[str, Any]) – prompt_kwargs (Mapping[str, Any]) – Return type None attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.LlamaCpp(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, model_path, lora_base=None, lora_path=None, n_ctx=512, n_parts=- 1, seed=- 1, f16_kv=True, logits_all=False, vocab_only=False, use_mlock=False, n_threads=None, n_batch=8, n_gpu_layers=None, suffix=None, max_tokens=256, temperature=0.8, top_p=0.95, logprobs=None, echo=False, stop=[], repeat_penalty=1.1, top_k=40, last_n_tokens_size=64, use_mmap=True, streaming=True)[source] Bases: langchain.llms.base.LLM Wrapper around the llama.cpp model. To use, you should have the llama-cpp-python library installed, and provide the path to the Llama model as a named parameter to the constructor. Check out: https://github.com/abetlen/llama-cpp-python Example from langchain.llms import LlamaCppEmbeddings llm = LlamaCppEmbeddings(model_path="/path/to/llama/model") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – model_path (str) – lora_base (Optional[str]) – lora_path (Optional[str]) – n_ctx (int) – n_parts (int) – seed (int) – f16_kv (bool) – logits_all (bool) – vocab_only (bool) – use_mlock (bool) – n_threads (Optional[int]) – n_batch (Optional[int]) – n_gpu_layers (Optional[int]) – suffix (Optional[str]) – max_tokens (Optional[int]) – temperature (Optional[float]) – top_p (Optional[float]) – logprobs (Optional[int]) – echo (Optional[bool]) – stop (Optional[List[str]]) – repeat_penalty (Optional[float]) – top_k (Optional[int]) – last_n_tokens_size (Optional[int]) – use_mmap (Optional[bool]) – streaming (bool) – Return type None attribute echo: Optional[bool] = False Whether to echo the prompt. attribute f16_kv: bool = True Use half-precision for key/value cache. attribute last_n_tokens_size: Optional[int] = 64 The number of tokens to look back when applying the repeat_penalty. attribute logits_all: bool = False Return logits for all tokens, not just the last token. attribute logprobs: Optional[int] = None The number of logprobs to return. If None, no logprobs are returned. attribute lora_base: Optional[str] = None The path to the Llama LoRA base model. attribute lora_path: Optional[str] = None The path to the Llama LoRA. If None, no LoRa is loaded. attribute max_tokens: Optional[int] = 256 The maximum number of tokens to generate. attribute model_path: str [Required] The path to the Llama model file. attribute n_batch: Optional[int] = 8 Number of tokens to process in parallel. Should be a number between 1 and n_ctx. attribute n_ctx: int = 512 Token context window. attribute n_gpu_layers: Optional[int] = None Number of layers to be loaded into gpu memory. Default None. attribute n_parts: int = -1 Number of parts to split the model into. If -1, the number of parts is automatically determined. attribute n_threads: Optional[int] = None Number of threads to use. If None, the number of threads is automatically determined. attribute repeat_penalty: Optional[float] = 1.1 The penalty to apply to repeated tokens. attribute seed: int = -1 Seed. If -1, a random seed is used. attribute stop: Optional[List[str]] = [] A list of strings to stop generation when encountered. attribute streaming: bool = True Whether to stream the results, token by token. attribute suffix: Optional[str] = None A suffix to append to the generated text. If None, no suffix is appended. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute temperature: Optional[float] = 0.8 The temperature to use for sampling. attribute top_k: Optional[int] = 40 The top-k value to use for sampling. attribute top_p: Optional[float] = 0.95 The top-p value to use for sampling. attribute use_mlock: bool = False Force system to keep model in RAM. attribute use_mmap: Optional[bool] = True Whether to keep the model loaded in RAM attribute verbose: bool [Optional] Whether to print out response text. attribute vocab_only: bool = False Only load the vocabulary, no weights. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text)[source] Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) stream(prompt, stop=None, run_manager=None)[source] Yields results objects as they are generated in real time. BETA: this is a beta feature while we figure out the right abstraction. Once that happens, this interface could change. It also calls the callback manager’s on_llm_new_token event with similar parameters to the OpenAI LLM class method of the same name. Args:prompt: The prompts to pass into the model. stop: Optional list of stop words to use when generating. Returns:A generator representing the stream of tokens being generated. Yields:A dictionary like objects containing a string token and metadata. See llama-cpp-python docs and below for more. Example:from langchain.llms import LlamaCpp llm = LlamaCpp( model_path="/path/to/local/model.bin", temperature = 0.5 ) for chunk in llm.stream("Ask 'Hi, how are you?' like a pirate:'", stop=["'"," β€œ]):result = chunk[β€œchoices”][0] print(result[β€œtext”], end=’’, flush=True) Parameters prompt (str) – stop (Optional[List[str]]) – run_manager (Optional[langchain.callbacks.manager.CallbackManagerForLLMRun]) – Return type Generator[Dict, None, None] classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.TextGen(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, model_url, max_new_tokens=250, do_sample=True, temperature=1.3, top_p=0.1, typical_p=1, epsilon_cutoff=0, eta_cutoff=0, repetition_penalty=1.18, top_k=40, min_length=0, no_repeat_ngram_size=0, num_beams=1, penalty_alpha=0, length_penalty=1, early_stopping=False, seed=- 1, add_bos_token=True, truncation_length=2048, ban_eos_token=False, skip_special_tokens=True, stopping_strings=[], streaming=False)[source] Bases: langchain.llms.base.LLM Wrapper around the text-generation-webui model. To use, you should have the text-generation-webui installed, a model loaded, and –api added as a command-line option. Suggested installation, use one-click installer for your OS: https://github.com/oobabooga/text-generation-webui#one-click-installers Paremeters below taken from text-generation-webui api example: https://github.com/oobabooga/text-generation-webui/blob/main/api-examples/api-example.py Example from langchain.llms import TextGen llm = TextGen(model_url="http://localhost:8500") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – model_url (str) – max_new_tokens (Optional[int]) – do_sample (bool) – temperature (Optional[float]) – top_p (Optional[float]) – typical_p (Optional[float]) – epsilon_cutoff (Optional[float]) – eta_cutoff (Optional[float]) – repetition_penalty (Optional[float]) – top_k (Optional[float]) – min_length (Optional[int]) – no_repeat_ngram_size (Optional[int]) – num_beams (Optional[int]) – penalty_alpha (Optional[float]) – length_penalty (Optional[float]) – early_stopping (bool) – seed (int) – add_bos_token (bool) – truncation_length (Optional[int]) – ban_eos_token (bool) – skip_special_tokens (bool) – stopping_strings (Optional[List[str]]) – streaming (bool) – Return type None attribute add_bos_token: bool = True Add the bos_token to the beginning of prompts. Disabling this can make the replies more creative. attribute ban_eos_token: bool = False Ban the eos_token. Forces the model to never end the generation prematurely. attribute do_sample: bool = True Do sample attribute early_stopping: bool = False Early stopping attribute epsilon_cutoff: Optional[float] = 0 Epsilon cutoff attribute eta_cutoff: Optional[float] = 0 ETA cutoff attribute length_penalty: Optional[float] = 1 Length Penalty attribute max_new_tokens: Optional[int] = 250 The maximum number of tokens to generate. attribute min_length: Optional[int] = 0 Minimum generation length in tokens. attribute model_url: str [Required] The full URL to the textgen webui including http[s]://host:port attribute no_repeat_ngram_size: Optional[int] = 0 If not set to 0, specifies the length of token sets that are completely blocked from repeating at all. Higher values = blocks larger phrases, lower values = blocks words or letters from repeating. Only 0 or high values are a good idea in most cases. attribute num_beams: Optional[int] = 1 Number of beams attribute penalty_alpha: Optional[float] = 0 Penalty Alpha attribute repetition_penalty: Optional[float] = 1.18 Exponential penalty factor for repeating prior tokens. 1 means no penalty, higher value = less repetition, lower value = more repetition. attribute seed: int = -1 Seed (-1 for random) attribute skip_special_tokens: bool = True Skip special tokens. Some specific models need this unset. attribute stopping_strings: Optional[List[str]] = [] A list of strings to stop generation when encountered. attribute streaming: bool = False Whether to stream the results, token by token (currently unimplemented). attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute temperature: Optional[float] = 1.3 Primary factor to control randomness of outputs. 0 = deterministic (only the most likely token is used). Higher value = more randomness. attribute top_k: Optional[float] = 40 Similar to top_p, but select instead only the top_k most likely tokens. Higher value = higher range of possible random results. attribute top_p: Optional[float] = 0.1 If not set to 1, select tokens with probabilities adding up to less than this number. Higher value = higher range of possible random results. attribute truncation_length: Optional[int] = 2048 Truncate the prompt up to this length. The leftmost tokens are removed if the prompt exceeds this length. Most models require this to be at most 2048. attribute typical_p: Optional[float] = 1 If not set to 1, select only tokens that are at least this much more likely to appear than random tokens, given the prior text. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.ManifestWrapper(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, llm_kwargs=None)[source] Bases: langchain.llms.base.LLM Wrapper around HazyResearch’s Manifest library. Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – llm_kwargs (Optional[Dict]) – Return type None attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.Modal(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, endpoint_url='', model_kwargs=None)[source] Bases: langchain.llms.base.LLM Wrapper around Modal large language models. To use, you should have the modal-client python package installed. Any parameters that are valid to be passed to the call can be passed in, even if not explicitly saved on this class. Example from langchain.llms import Modal modal = Modal(endpoint_url="") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – endpoint_url (str) – model_kwargs (Dict[str, Any]) – Return type None attribute endpoint_url: str = '' model endpoint to use attribute model_kwargs: Dict[str, Any] [Optional] Holds any model parameters valid for create call not explicitly specified. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.MosaicML(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, endpoint_url='https://models.hosted-on.mosaicml.hosting/mpt-7b-instruct/v1/predict', inject_instruction_format=False, model_kwargs=None, retry_sleep=1.0, mosaicml_api_token=None)[source] Bases: langchain.llms.base.LLM Wrapper around MosaicML’s LLM inference service. To use, you should have the environment variable MOSAICML_API_TOKEN set with your API token, or pass it as a named parameter to the constructor. Example from langchain.llms import MosaicML endpoint_url = ( "https://models.hosted-on.mosaicml.hosting/mpt-7b-instruct/v1/predict" ) mosaic_llm = MosaicML( endpoint_url=endpoint_url, mosaicml_api_token="my-api-key" ) Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – endpoint_url (str) – inject_instruction_format (bool) – model_kwargs (Optional[dict]) – retry_sleep (float) – mosaicml_api_token (Optional[str]) – Return type None attribute endpoint_url: str = 'https://models.hosted-on.mosaicml.hosting/mpt-7b-instruct/v1/predict' Endpoint URL to use. attribute inject_instruction_format: bool = False Whether to inject the instruction format into the prompt. attribute model_kwargs: Optional[dict] = None Key word arguments to pass to the model. attribute retry_sleep: float = 1.0 How long to try sleeping for if a rate limit is encountered attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.NLPCloud(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, model_name='finetuned-gpt-neox-20b', temperature=0.7, min_length=1, max_length=256, length_no_input=True, remove_input=True, remove_end_sequence=True, bad_words=[], top_p=1, top_k=50, repetition_penalty=1.0, length_penalty=1.0, do_sample=True, num_beams=1, early_stopping=False, num_return_sequences=1, nlpcloud_api_key=None)[source] Bases: langchain.llms.base.LLM Wrapper around NLPCloud large language models. To use, you should have the nlpcloud python package installed, and the environment variable NLPCLOUD_API_KEY set with your API key. Example from langchain.llms import NLPCloud nlpcloud = NLPCloud(model="gpt-neox-20b") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – model_name (str) – temperature (float) – min_length (int) – max_length (int) – length_no_input (bool) – remove_input (bool) – remove_end_sequence (bool) – bad_words (List[str]) – top_p (int) – top_k (int) – repetition_penalty (float) – length_penalty (float) – do_sample (bool) – num_beams (int) – early_stopping (bool) – num_return_sequences (int) – nlpcloud_api_key (Optional[str]) – Return type None attribute bad_words: List[str] = [] List of tokens not allowed to be generated. attribute do_sample: bool = True Whether to use sampling (True) or greedy decoding. attribute early_stopping: bool = False Whether to stop beam search at num_beams sentences. attribute length_no_input: bool = True Whether min_length and max_length should include the length of the input. attribute length_penalty: float = 1.0 Exponential penalty to the length. attribute max_length: int = 256 The maximum number of tokens to generate in the completion. attribute min_length: int = 1 The minimum number of tokens to generate in the completion. attribute model_name: str = 'finetuned-gpt-neox-20b' Model name to use. attribute num_beams: int = 1 Number of beams for beam search. attribute num_return_sequences: int = 1 How many completions to generate for each prompt. attribute remove_end_sequence: bool = True Whether or not to remove the end sequence token. attribute remove_input: bool = True Remove input text from API response attribute repetition_penalty: float = 1.0 Penalizes repeated tokens. 1.0 means no penalty. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute temperature: float = 0.7 What sampling temperature to use. attribute top_k: int = 50 The number of highest probability tokens to keep for top-k filtering. attribute top_p: int = 1 Total probability mass of tokens to consider at each step. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.OpenAI(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, model='text-davinci-003', temperature=0.7, max_tokens=256, top_p=1, frequency_penalty=0, presence_penalty=0, n=1, best_of=1, model_kwargs=None, openai_api_key=None, openai_api_base=None, openai_organization=None, openai_proxy=None, batch_size=20, request_timeout=None, logit_bias=None, max_retries=6, streaming=False, allowed_special={}, disallowed_special='all', tiktoken_model_name=None)[source] Bases: langchain.llms.openai.BaseOpenAI Wrapper around OpenAI large language models. To use, you should have the openai python package installed, and the environment variable OPENAI_API_KEY set with your API key. Any parameters that are valid to be passed to the openai.create call can be passed in, even if not explicitly saved on this class. Example from langchain.llms import OpenAI openai = OpenAI(model_name="text-davinci-003") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – model (str) – temperature (float) – max_tokens (int) – top_p (float) – frequency_penalty (float) – presence_penalty (float) – n (int) – best_of (int) – model_kwargs (Dict[str, Any]) – openai_api_key (Optional[str]) – openai_api_base (Optional[str]) – openai_organization (Optional[str]) – openai_proxy (Optional[str]) – batch_size (int) – request_timeout (Optional[Union[float, Tuple[float, float]]]) – logit_bias (Optional[Dict[str, float]]) – max_retries (int) – streaming (bool) – allowed_special (Union[Literal['all'], typing.AbstractSet[str]]) – disallowed_special (Union[Literal['all'], typing.Collection[str]]) – tiktoken_model_name (Optional[str]) – Return type None attribute allowed_special: Union[Literal['all'], AbstractSet[str]] = {} Set of special tokens that are allowed。 attribute batch_size: int = 20 Batch size to use when passing multiple documents to generate. attribute best_of: int = 1 Generates best_of completions server-side and returns the β€œbest”. attribute disallowed_special: Union[Literal['all'], Collection[str]] = 'all' Set of special tokens that are not allowed。 attribute frequency_penalty: float = 0 Penalizes repeated tokens according to frequency. attribute logit_bias: Optional[Dict[str, float]] [Optional] Adjust the probability of specific tokens being generated. attribute max_retries: int = 6 Maximum number of retries to make when generating. attribute max_tokens: int = 256 The maximum number of tokens to generate in the completion. -1 returns as many tokens as possible given the prompt and the models maximal context size. attribute model_kwargs: Dict[str, Any] [Optional] Holds any model parameters valid for create call not explicitly specified. attribute model_name: str = 'text-davinci-003' (alias 'model') Model name to use. attribute n: int = 1 How many completions to generate for each prompt. attribute presence_penalty: float = 0 Penalizes repeated tokens. attribute request_timeout: Optional[Union[float, Tuple[float, float]]] = None Timeout for requests to OpenAI completion API. Default is 600 seconds. attribute streaming: bool = False Whether to stream the results or not. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute temperature: float = 0.7 What sampling temperature to use. attribute tiktoken_model_name: Optional[str] = None The model name to pass to tiktoken when using this class. Tiktoken is used to count the number of tokens in documents to constrain them to be under a certain limit. By default, when set to None, this will be the same as the embedding model name. However, there are some cases where you may want to use this Embedding class with a model name not supported by tiktoken. This can include when using Azure embeddings or when using one of the many model providers that expose an OpenAI-like API but with different models. In those cases, in order to avoid erroring when tiktoken is called, you can specify a model name to use here. attribute top_p: float = 1 Total probability mass of tokens to consider at each step. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model create_llm_result(choices, prompts, token_usage) Create the LLMResult from the choices and prompts. Parameters choices (Any) – prompts (List[str]) – token_usage (Dict[str, int]) – Return type langchain.schema.LLMResult dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_sub_prompts(params, prompts, stop=None) Get the sub prompts for llm call. Parameters params (Dict[str, Any]) – prompts (List[str]) – stop (Optional[List[str]]) – Return type List[List[str]] get_token_ids(text) Get the token IDs using the tiktoken package. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode max_tokens_for_prompt(prompt) Calculate the maximum number of tokens possible to generate for a prompt. Parameters prompt (str) – The prompt to pass into the model. Returns The maximum number of tokens to generate for a prompt. Return type int Example max_tokens = openai.max_token_for_prompt("Tell me a joke.") static modelname_to_contextsize(modelname) Calculate the maximum number of tokens possible to generate for a model. Parameters modelname (str) – The modelname we want to know the context size for. Returns The maximum context size Return type int Example max_tokens = openai.modelname_to_contextsize("text-davinci-003") predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage prep_streaming_params(stop=None) Prepare the params for streaming. Parameters stop (Optional[List[str]]) – Return type Dict[str, Any] save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) stream(prompt, stop=None) Call OpenAI with streaming flag and return the resulting generator. BETA: this is a beta feature while we figure out the right abstraction. Once that happens, this interface could change. Parameters prompt (str) – The prompts to pass into the model. stop (Optional[List[str]]) – Optional list of stop words to use when generating. Returns A generator representing the stream of tokens from OpenAI. Return type Generator Example generator = openai.stream("Tell me a joke.") for token in generator: yield token classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. property max_context_size: int Get max context size for this model. class langchain.llms.OpenAIChat(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, model_name='gpt-3.5-turbo', model_kwargs=None, openai_api_key=None, openai_api_base=None, openai_proxy=None, max_retries=6, prefix_messages=None, streaming=False, allowed_special={}, disallowed_special='all')[source] Bases: langchain.llms.base.BaseLLM Wrapper around OpenAI Chat large language models. To use, you should have the openai python package installed, and the environment variable OPENAI_API_KEY set with your API key. Any parameters that are valid to be passed to the openai.create call can be passed in, even if not explicitly saved on this class. Example from langchain.llms import OpenAIChat openaichat = OpenAIChat(model_name="gpt-3.5-turbo") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – model_name (str) – model_kwargs (Dict[str, Any]) – openai_api_key (Optional[str]) – openai_api_base (Optional[str]) – openai_proxy (Optional[str]) – max_retries (int) – prefix_messages (List) – streaming (bool) – allowed_special (Union[Literal['all'], typing.AbstractSet[str]]) – disallowed_special (Union[Literal['all'], typing.Collection[str]]) – Return type None attribute allowed_special: Union[Literal['all'], AbstractSet[str]] = {} Set of special tokens that are allowed。 attribute disallowed_special: Union[Literal['all'], Collection[str]] = 'all' Set of special tokens that are not allowed。 attribute max_retries: int = 6 Maximum number of retries to make when generating. attribute model_kwargs: Dict[str, Any] [Optional] Holds any model parameters valid for create call not explicitly specified. attribute model_name: str = 'gpt-3.5-turbo' Model name to use. attribute prefix_messages: List [Optional] Series of messages for Chat input. attribute streaming: bool = False Whether to stream the results or not. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text)[source] Get the token IDs using the tiktoken package. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.OpenLLM(model_name=None, *, model_id=None, server_url=None, server_type='http', embedded=True, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, llm_kwargs)[source] Bases: langchain.llms.base.LLM Wrapper for accessing OpenLLM, supporting both in-process model instance and remote OpenLLM servers. To use, you should have the openllm library installed: pip install openllm Learn more at: https://github.com/bentoml/openllm Example running an LLM model locally managed by OpenLLM:from langchain.llms import OpenLLM llm = OpenLLM( model_name='flan-t5', model_id='google/flan-t5-large', ) llm("What is the difference between a duck and a goose?") For all available supported models, you can run β€˜openllm models’. If you have a OpenLLM server running, you can also use it remotely:from langchain.llms import OpenLLM llm = OpenLLM(server_url='http://localhost:3000') llm("What is the difference between a duck and a goose?") Parameters model_name (Optional[str]) – model_id (Optional[str]) – server_url (Optional[str]) – server_type (Literal['grpc', 'http']) – embedded (bool) – cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – llm_kwargs (Dict[str, Any]) – Return type None attribute embedded: bool = True Initialize this LLM instance in current process by default. Should only set to False when using in conjunction with BentoML Service. attribute llm_kwargs: Dict[str, Any] [Required] Key word arguments to be passed to openllm.LLM attribute model_id: Optional[str] = None Model Id to use. If not provided, will use the default model for the model name. See β€˜openllm models’ for all available model variants. attribute model_name: Optional[str] = None Model name to use. See β€˜openllm models’ for all available models. attribute server_type: ServerType = 'http' Optional server type. Either β€˜http’ or β€˜grpc’. attribute server_url: Optional[str] = None Optional server URL that currently runs a LLMServer with β€˜openllm start’. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. property runner: openllm.LLMRunner Get the underlying openllm.LLMRunner instance for integration with BentoML. Example: .. code-block:: python llm = OpenLLM(model_name=’flan-t5’, model_id=’google/flan-t5-large’, embedded=False, ) tools = load_tools([β€œserpapi”, β€œllm-math”], llm=llm) agent = initialize_agent( tools, llm, agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION ) svc = bentoml.Service(β€œlangchain-openllm”, runners=[llm.runner]) @svc.api(input=Text(), output=Text()) def chat(input_text: str): return agent.run(input_text) class langchain.llms.OpenLM(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, model='text-davinci-003', temperature=0.7, max_tokens=256, top_p=1, frequency_penalty=0, presence_penalty=0, n=1, best_of=1, model_kwargs=None, openai_api_key=None, openai_api_base=None, openai_organization=None, openai_proxy=None, batch_size=20, request_timeout=None, logit_bias=None, max_retries=6, streaming=False, allowed_special={}, disallowed_special='all', tiktoken_model_name=None)[source] Bases: langchain.llms.openai.BaseOpenAI Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – model (str) – temperature (float) – max_tokens (int) – top_p (float) – frequency_penalty (float) – presence_penalty (float) – n (int) – best_of (int) – model_kwargs (Dict[str, Any]) – openai_api_key (Optional[str]) – openai_api_base (Optional[str]) – openai_organization (Optional[str]) – openai_proxy (Optional[str]) – batch_size (int) – request_timeout (Optional[Union[float, Tuple[float, float]]]) – logit_bias (Optional[Dict[str, float]]) – max_retries (int) – streaming (bool) – allowed_special (Union[Literal['all'], typing.AbstractSet[str]]) – disallowed_special (Union[Literal['all'], typing.Collection[str]]) – tiktoken_model_name (Optional[str]) – Return type None attribute allowed_special: Union[Literal['all'], AbstractSet[str]] = {} Set of special tokens that are allowed。 attribute batch_size: int = 20 Batch size to use when passing multiple documents to generate. attribute best_of: int = 1 Generates best_of completions server-side and returns the β€œbest”. attribute disallowed_special: Union[Literal['all'], Collection[str]] = 'all' Set of special tokens that are not allowed。 attribute frequency_penalty: float = 0 Penalizes repeated tokens according to frequency. attribute logit_bias: Optional[Dict[str, float]] [Optional] Adjust the probability of specific tokens being generated. attribute max_retries: int = 6 Maximum number of retries to make when generating. attribute max_tokens: int = 256 The maximum number of tokens to generate in the completion. -1 returns as many tokens as possible given the prompt and the models maximal context size. attribute model_kwargs: Dict[str, Any] [Optional] Holds any model parameters valid for create call not explicitly specified. attribute model_name: str = 'text-davinci-003' (alias 'model') Model name to use. attribute n: int = 1 How many completions to generate for each prompt. attribute presence_penalty: float = 0 Penalizes repeated tokens. attribute request_timeout: Optional[Union[float, Tuple[float, float]]] = None Timeout for requests to OpenAI completion API. Default is 600 seconds. attribute streaming: bool = False Whether to stream the results or not. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute temperature: float = 0.7 What sampling temperature to use. attribute tiktoken_model_name: Optional[str] = None The model name to pass to tiktoken when using this class. Tiktoken is used to count the number of tokens in documents to constrain them to be under a certain limit. By default, when set to None, this will be the same as the embedding model name. However, there are some cases where you may want to use this Embedding class with a model name not supported by tiktoken. This can include when using Azure embeddings or when using one of the many model providers that expose an OpenAI-like API but with different models. In those cases, in order to avoid erroring when tiktoken is called, you can specify a model name to use here. attribute top_p: float = 1 Total probability mass of tokens to consider at each step. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model create_llm_result(choices, prompts, token_usage) Create the LLMResult from the choices and prompts. Parameters choices (Any) – prompts (List[str]) – token_usage (Dict[str, int]) – Return type langchain.schema.LLMResult dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_sub_prompts(params, prompts, stop=None) Get the sub prompts for llm call. Parameters params (Dict[str, Any]) – prompts (List[str]) – stop (Optional[List[str]]) – Return type List[List[str]] get_token_ids(text) Get the token IDs using the tiktoken package. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode max_tokens_for_prompt(prompt) Calculate the maximum number of tokens possible to generate for a prompt. Parameters prompt (str) – The prompt to pass into the model. Returns The maximum number of tokens to generate for a prompt. Return type int Example max_tokens = openai.max_token_for_prompt("Tell me a joke.") static modelname_to_contextsize(modelname) Calculate the maximum number of tokens possible to generate for a model. Parameters modelname (str) – The modelname we want to know the context size for. Returns The maximum context size Return type int Example max_tokens = openai.modelname_to_contextsize("text-davinci-003") predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage prep_streaming_params(stop=None) Prepare the params for streaming. Parameters stop (Optional[List[str]]) – Return type Dict[str, Any] save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) stream(prompt, stop=None) Call OpenAI with streaming flag and return the resulting generator. BETA: this is a beta feature while we figure out the right abstraction. Once that happens, this interface could change. Parameters prompt (str) – The prompts to pass into the model. stop (Optional[List[str]]) – Optional list of stop words to use when generating. Returns A generator representing the stream of tokens from OpenAI. Return type Generator Example generator = openai.stream("Tell me a joke.") for token in generator: yield token classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. property max_context_size: int Get max context size for this model. class langchain.llms.Petals(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, tokenizer=None, model_name='bigscience/bloom-petals', temperature=0.7, max_new_tokens=256, top_p=0.9, top_k=None, do_sample=True, max_length=None, model_kwargs=None, huggingface_api_key=None)[source] Bases: langchain.llms.base.LLM Wrapper around Petals Bloom models. To use, you should have the petals python package installed, and the environment variable HUGGINGFACE_API_KEY set with your API key. Any parameters that are valid to be passed to the call can be passed in, even if not explicitly saved on this class. Example from langchain.llms import petals petals = Petals() Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – tokenizer (Any) – model_name (str) – temperature (float) – max_new_tokens (int) – top_p (float) – top_k (Optional[int]) – do_sample (bool) – max_length (Optional[int]) – model_kwargs (Dict[str, Any]) – huggingface_api_key (Optional[str]) – Return type None attribute client: Any = None The client to use for the API calls. attribute do_sample: bool = True Whether or not to use sampling; use greedy decoding otherwise. attribute max_length: Optional[int] = None The maximum length of the sequence to be generated. attribute max_new_tokens: int = 256 The maximum number of new tokens to generate in the completion. attribute model_kwargs: Dict[str, Any] [Optional] Holds any model parameters valid for create call not explicitly specified. attribute model_name: str = 'bigscience/bloom-petals' The model to use. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute temperature: float = 0.7 What sampling temperature to use attribute tokenizer: Any = None The tokenizer to use for the API calls. attribute top_k: Optional[int] = None The number of highest probability vocabulary tokens to keep for top-k-filtering. attribute top_p: float = 0.9 The cumulative probability for top-p sampling. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.PipelineAI(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, pipeline_key='', pipeline_kwargs=None, pipeline_api_key=None)[source] Bases: langchain.llms.base.LLM, pydantic.main.BaseModel Wrapper around PipelineAI large language models. To use, you should have the pipeline-ai python package installed, and the environment variable PIPELINE_API_KEY set with your API key. Any parameters that are valid to be passed to the call can be passed in, even if not explicitly saved on this class. Example from langchain import PipelineAI pipeline = PipelineAI(pipeline_key="") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – pipeline_key (str) – pipeline_kwargs (Dict[str, Any]) – pipeline_api_key (Optional[str]) – Return type None attribute pipeline_key: str = '' The id or tag of the target pipeline attribute pipeline_kwargs: Dict[str, Any] [Optional] Holds any pipeline parameters valid for create call not explicitly specified. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.PredictionGuard(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, model='MPT-7B-Instruct', output=None, max_tokens=256, temperature=0.75, token=None, stop=None)[source] Bases: langchain.llms.base.LLM Wrapper around Prediction Guard large language models. To use, you should have the predictionguard python package installed, and the environment variable PREDICTIONGUARD_TOKEN set with your access token, or pass it as a named parameter to the constructor. To use Prediction Guard’s API along with OpenAI models, set the environment variable OPENAI_API_KEY with your OpenAI API key as well. Example pgllm = PredictionGuard(model="MPT-7B-Instruct", token="my-access-token", output={ "type": "boolean" }) Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – model (Optional[str]) – output (Optional[Dict[str, Any]]) – max_tokens (int) – temperature (float) – token (Optional[str]) – stop (Optional[List[str]]) – Return type None attribute max_tokens: int = 256 Denotes the number of tokens to predict per generation. attribute model: Optional[str] = 'MPT-7B-Instruct' Model name to use. attribute output: Optional[Dict[str, Any]] = None The output type or structure for controlling the LLM output. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute temperature: float = 0.75 A non-negative float that tunes the degree of randomness in generation. attribute token: Optional[str] = None Your Prediction Guard access token. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.PromptLayerOpenAI(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, model='text-davinci-003', temperature=0.7, max_tokens=256, top_p=1, frequency_penalty=0, presence_penalty=0, n=1, best_of=1, model_kwargs=None, openai_api_key=None, openai_api_base=None, openai_organization=None, openai_proxy=None, batch_size=20, request_timeout=None, logit_bias=None, max_retries=6, streaming=False, allowed_special={}, disallowed_special='all', tiktoken_model_name=None, pl_tags=None, return_pl_id=False)[source] Bases: langchain.llms.openai.OpenAI Wrapper around OpenAI large language models. To use, you should have the openai and promptlayer python package installed, and the environment variable OPENAI_API_KEY and PROMPTLAYER_API_KEY set with your openAI API key and promptlayer key respectively. All parameters that can be passed to the OpenAI LLM can also be passed here. The PromptLayerOpenAI LLM adds two optional Parameters pl_tags (Optional[List[str]]) – List of strings to tag the request with. return_pl_id (Optional[bool]) – If True, the PromptLayer request ID will be returned in the generation_info field of the Generation object. cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – model (str) – temperature (float) – max_tokens (int) – top_p (float) – frequency_penalty (float) – presence_penalty (float) – n (int) – best_of (int) – model_kwargs (Dict[str, Any]) – openai_api_key (Optional[str]) – openai_api_base (Optional[str]) – openai_organization (Optional[str]) – openai_proxy (Optional[str]) – batch_size (int) – request_timeout (Optional[Union[float, Tuple[float, float]]]) – logit_bias (Optional[Dict[str, float]]) – max_retries (int) – streaming (bool) – allowed_special (Union[Literal['all'], typing.AbstractSet[str]]) – disallowed_special (Union[Literal['all'], typing.Collection[str]]) – tiktoken_model_name (Optional[str]) – Return type None Example from langchain.llms import PromptLayerOpenAI openai = PromptLayerOpenAI(model_name="text-davinci-003") __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model create_llm_result(choices, prompts, token_usage) Create the LLMResult from the choices and prompts. Parameters choices (Any) – prompts (List[str]) – token_usage (Dict[str, int]) – Return type langchain.schema.LLMResult dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_sub_prompts(params, prompts, stop=None) Get the sub prompts for llm call. Parameters params (Dict[str, Any]) – prompts (List[str]) – stop (Optional[List[str]]) – Return type List[List[str]] get_token_ids(text) Get the token IDs using the tiktoken package. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode max_tokens_for_prompt(prompt) Calculate the maximum number of tokens possible to generate for a prompt. Parameters prompt (str) – The prompt to pass into the model. Returns The maximum number of tokens to generate for a prompt. Return type int Example max_tokens = openai.max_token_for_prompt("Tell me a joke.") static modelname_to_contextsize(modelname) Calculate the maximum number of tokens possible to generate for a model. Parameters modelname (str) – The modelname we want to know the context size for. Returns The maximum context size Return type int Example max_tokens = openai.modelname_to_contextsize("text-davinci-003") predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage prep_streaming_params(stop=None) Prepare the params for streaming. Parameters stop (Optional[List[str]]) – Return type Dict[str, Any] save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) stream(prompt, stop=None) Call OpenAI with streaming flag and return the resulting generator. BETA: this is a beta feature while we figure out the right abstraction. Once that happens, this interface could change. Parameters prompt (str) – The prompts to pass into the model. stop (Optional[List[str]]) – Optional list of stop words to use when generating. Returns A generator representing the stream of tokens from OpenAI. Return type Generator Example generator = openai.stream("Tell me a joke.") for token in generator: yield token classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. property max_context_size: int Get max context size for this model. class langchain.llms.PromptLayerOpenAIChat(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, model_name='gpt-3.5-turbo', model_kwargs=None, openai_api_key=None, openai_api_base=None, openai_proxy=None, max_retries=6, prefix_messages=None, streaming=False, allowed_special={}, disallowed_special='all', pl_tags=None, return_pl_id=False)[source] Bases: langchain.llms.openai.OpenAIChat Wrapper around OpenAI large language models. To use, you should have the openai and promptlayer python package installed, and the environment variable OPENAI_API_KEY and PROMPTLAYER_API_KEY set with your openAI API key and promptlayer key respectively. All parameters that can be passed to the OpenAIChat LLM can also be passed here. The PromptLayerOpenAIChat adds two optional Parameters pl_tags (Optional[List[str]]) – List of strings to tag the request with. return_pl_id (Optional[bool]) – If True, the PromptLayer request ID will be returned in the generation_info field of the Generation object. cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – model_name (str) – model_kwargs (Dict[str, Any]) – openai_api_key (Optional[str]) – openai_api_base (Optional[str]) – openai_proxy (Optional[str]) – max_retries (int) – prefix_messages (List) – streaming (bool) – allowed_special (Union[Literal['all'], typing.AbstractSet[str]]) – disallowed_special (Union[Literal['all'], typing.Collection[str]]) – Return type None Example from langchain.llms import PromptLayerOpenAIChat openaichat = PromptLayerOpenAIChat(model_name="gpt-3.5-turbo") attribute allowed_special: Union[Literal['all'], AbstractSet[str]] = {} Set of special tokens that are allowed。 attribute disallowed_special: Union[Literal['all'], Collection[str]] = 'all' Set of special tokens that are not allowed。 attribute max_retries: int = 6 Maximum number of retries to make when generating. attribute model_kwargs: Dict[str, Any] [Optional] Holds any model parameters valid for create call not explicitly specified. attribute model_name: str = 'gpt-3.5-turbo' Model name to use. attribute prefix_messages: List [Optional] Series of messages for Chat input. attribute streaming: bool = False Whether to stream the results or not. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token IDs using the tiktoken package. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.RWKV(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, model, tokens_path, strategy='cpu fp32', rwkv_verbose=True, temperature=1.0, top_p=0.5, penalty_alpha_frequency=0.4, penalty_alpha_presence=0.4, CHUNK_LEN=256, max_tokens_per_generation=256, client=None, tokenizer=None, pipeline=None, model_tokens=None, model_state=None)[source] Bases: langchain.llms.base.LLM, pydantic.main.BaseModel Wrapper around RWKV language models. To use, you should have the rwkv python package installed, the pre-trained model file, and the model’s config information. Example from langchain.llms import RWKV model = RWKV(model="./models/rwkv-3b-fp16.bin", strategy="cpu fp32") # Simplest invocation response = model("Once upon a time, ") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – model (str) – tokens_path (str) – strategy (str) – rwkv_verbose (bool) – temperature (float) – top_p (float) – penalty_alpha_frequency (float) – penalty_alpha_presence (float) – CHUNK_LEN (int) – max_tokens_per_generation (int) – client (Any) – tokenizer (Any) – pipeline (Any) – model_tokens (Any) – model_state (Any) – Return type None attribute CHUNK_LEN: int = 256 Batch size for prompt processing. attribute max_tokens_per_generation: int = 256 Maximum number of tokens to generate. attribute model: str [Required] Path to the pre-trained RWKV model file. attribute penalty_alpha_frequency: float = 0.4 Positive values penalize new tokens based on their existing frequency in the text so far, decreasing the model’s likelihood to repeat the same line verbatim.. attribute penalty_alpha_presence: float = 0.4 Positive values penalize new tokens based on whether they appear in the text so far, increasing the model’s likelihood to talk about new topics.. attribute rwkv_verbose: bool = True Print debug information. attribute strategy: str = 'cpu fp32' Token context window. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute temperature: float = 1.0 The temperature to use for sampling. attribute tokens_path: str [Required] Path to the RWKV tokens file. attribute top_p: float = 0.5 The top-p value to use for sampling. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.Replicate(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, model, input=None, model_kwargs=None, replicate_api_token=None)[source] Bases: langchain.llms.base.LLM Wrapper around Replicate models. To use, you should have the replicate python package installed, and the environment variable REPLICATE_API_TOKEN set with your API token. You can find your token here: https://replicate.com/account The model param is required, but any other model parameters can also be passed in with the format input={model_param: value, …} Example from langchain.llms import Replicate replicate = Replicate(model="stability-ai/stable-diffusion: 27b93a2413e7f36cd83da926f365628 0b2931564ff050bf9575f1fdf9bcd7478", input={"image_dimensions": "512x512"}) Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – model (str) – input (Dict[str, Any]) – model_kwargs (Dict[str, Any]) – replicate_api_token (Optional[str]) – Return type None attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.SagemakerEndpoint(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, endpoint_name='', region_name='', credentials_profile_name=None, content_handler, model_kwargs=None, endpoint_kwargs=None)[source] Bases: langchain.llms.base.LLM Wrapper around custom Sagemaker Inference Endpoints. To use, you must supply the endpoint name from your deployed Sagemaker model & the region where it is deployed. To authenticate, the AWS client uses the following methods to automatically load credentials: https://boto3.amazonaws.com/v1/documentation/api/latest/guide/credentials.html If a specific credential profile should be used, you must pass the name of the profile from the ~/.aws/credentials file that is to be used. Make sure the credentials / roles used have the required policies to access the Sagemaker endpoint. See: https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies.html Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – endpoint_name (str) – region_name (str) – credentials_profile_name (Optional[str]) – content_handler (langchain.llms.sagemaker_endpoint.LLMContentHandler) – model_kwargs (Optional[Dict]) – endpoint_kwargs (Optional[Dict]) – Return type None attribute content_handler: langchain.llms.sagemaker_endpoint.LLMContentHandler [Required] The content handler class that provides an input and output transform functions to handle formats between LLM and the endpoint. attribute credentials_profile_name: Optional[str] = None The name of the profile in the ~/.aws/credentials or ~/.aws/config files, which has either access keys or role information specified. If not specified, the default credential profile or, if on an EC2 instance, credentials from IMDS will be used. See: https://boto3.amazonaws.com/v1/documentation/api/latest/guide/credentials.html attribute endpoint_kwargs: Optional[Dict] = None Optional attributes passed to the invoke_endpoint function. See `boto3`_. docs for more info. .. _boto3: <https://boto3.amazonaws.com/v1/documentation/api/latest/index.html> attribute endpoint_name: str = '' The name of the endpoint from the deployed Sagemaker model. Must be unique within an AWS Region. attribute model_kwargs: Optional[Dict] = None Key word arguments to pass to the model. attribute region_name: str = '' The aws region where the Sagemaker model is deployed, eg. us-west-2. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.SelfHostedHuggingFaceLLM(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, pipeline_ref=None, client=None, inference_fn=<function _generate_text>, hardware=None, model_load_fn=<function _load_transformer>, load_fn_kwargs=None, model_reqs=['./', 'transformers', 'torch'], model_id='gpt2', task='text-generation', device=0, model_kwargs=None)[source] Bases: langchain.llms.self_hosted.SelfHostedPipeline Wrapper around HuggingFace Pipeline API to run on self-hosted remote hardware. Supported hardware includes auto-launched instances on AWS, GCP, Azure, and Lambda, as well as servers specified by IP address and SSH credentials (such as on-prem, or another cloud like Paperspace, Coreweave, etc.). To use, you should have the runhouse python package installed. Only supports text-generation, text2text-generation and summarization for now. Example using from_model_id:from langchain.llms import SelfHostedHuggingFaceLLM import runhouse as rh gpu = rh.cluster(name="rh-a10x", instance_type="A100:1") hf = SelfHostedHuggingFaceLLM( model_id="google/flan-t5-large", task="text2text-generation", hardware=gpu ) Example passing fn that generates a pipeline (bc the pipeline is not serializable):from langchain.llms import SelfHostedHuggingFaceLLM from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline import runhouse as rh def get_pipeline(): model_id = "gpt2" tokenizer = AutoTokenizer.from_pretrained(model_id) model = AutoModelForCausalLM.from_pretrained(model_id) pipe = pipeline( "text-generation", model=model, tokenizer=tokenizer ) return pipe hf = SelfHostedHuggingFaceLLM( model_load_fn=get_pipeline, model_id="gpt2", hardware=gpu) Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – pipeline_ref (Any) – client (Any) – inference_fn (Callable) – hardware (Any) – model_load_fn (Callable) – load_fn_kwargs (Optional[dict]) – model_reqs (List[str]) – model_id (str) – task (str) – device (int) – model_kwargs (Optional[dict]) – Return type None attribute device: int = 0 Device to use for inference. -1 for CPU, 0 for GPU, 1 for second GPU, etc. attribute hardware: Any = None Remote hardware to send the inference function to. attribute inference_fn: Callable = <function _generate_text> Inference function to send to the remote hardware. attribute load_fn_kwargs: Optional[dict] = None Key word arguments to pass to the model load function. attribute model_id: str = 'gpt2' Hugging Face model_id to load the model. attribute model_kwargs: Optional[dict] = None Key word arguments to pass to the model. attribute model_load_fn: Callable = <function _load_transformer> Function to load the model remotely on the server. attribute model_reqs: List[str] = ['./', 'transformers', 'torch'] Requirements to install on hardware to inference the model. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute task: str = 'text-generation' Hugging Face task (β€œtext-generation”, β€œtext2text-generation” or β€œsummarization”). attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict classmethod from_pipeline(pipeline, hardware, model_reqs=None, device=0, **kwargs) Init the SelfHostedPipeline from a pipeline object or string. Parameters pipeline (Any) – hardware (Any) – model_reqs (Optional[List[str]]) – device (int) – kwargs (Any) – Return type langchain.llms.base.LLM generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.SelfHostedPipeline(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, pipeline_ref=None, client=None, inference_fn=<function _generate_text>, hardware=None, model_load_fn, load_fn_kwargs=None, model_reqs=['./', 'torch'])[source] Bases: langchain.llms.base.LLM Run model inference on self-hosted remote hardware. Supported hardware includes auto-launched instances on AWS, GCP, Azure, and Lambda, as well as servers specified by IP address and SSH credentials (such as on-prem, or another cloud like Paperspace, Coreweave, etc.). To use, you should have the runhouse python package installed. Example for custom pipeline and inference functions:from langchain.llms import SelfHostedPipeline from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline import runhouse as rh def load_pipeline(): tokenizer = AutoTokenizer.from_pretrained("gpt2") model = AutoModelForCausalLM.from_pretrained("gpt2") return pipeline( "text-generation", model=model, tokenizer=tokenizer, max_new_tokens=10 ) def inference_fn(pipeline, prompt, stop = None): return pipeline(prompt)[0]["generated_text"] gpu = rh.cluster(name="rh-a10x", instance_type="A100:1") llm = SelfHostedPipeline( model_load_fn=load_pipeline, hardware=gpu, model_reqs=model_reqs, inference_fn=inference_fn ) Example for <2GB model (can be serialized and sent directly to the server):from langchain.llms import SelfHostedPipeline import runhouse as rh gpu = rh.cluster(name="rh-a10x", instance_type="A100:1") my_model = ... llm = SelfHostedPipeline.from_pipeline( pipeline=my_model, hardware=gpu, model_reqs=["./", "torch", "transformers"], ) Example passing model path for larger models:from langchain.llms import SelfHostedPipeline import runhouse as rh import pickle from transformers import pipeline generator = pipeline(model="gpt2") rh.blob(pickle.dumps(generator), path="models/pipeline.pkl" ).save().to(gpu, path="models") llm = SelfHostedPipeline.from_pipeline( pipeline="models/pipeline.pkl", hardware=gpu, model_reqs=["./", "torch", "transformers"], ) Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – pipeline_ref (Any) – client (Any) – inference_fn (Callable) – hardware (Any) – model_load_fn (Callable) – load_fn_kwargs (Optional[dict]) – model_reqs (List[str]) – Return type None attribute hardware: Any = None Remote hardware to send the inference function to. attribute inference_fn: Callable = <function _generate_text> Inference function to send to the remote hardware. attribute load_fn_kwargs: Optional[dict] = None Key word arguments to pass to the model load function. attribute model_load_fn: Callable [Required] Function to load the model remotely on the server. attribute model_reqs: List[str] = ['./', 'torch'] Requirements to install on hardware to inference the model. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict classmethod from_pipeline(pipeline, hardware, model_reqs=None, device=0, **kwargs)[source] Init the SelfHostedPipeline from a pipeline object or string. Parameters pipeline (Any) – hardware (Any) – model_reqs (Optional[List[str]]) – device (int) – kwargs (Any) – Return type langchain.llms.base.LLM generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.StochasticAI(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, api_url='', model_kwargs=None, stochasticai_api_key=None)[source] Bases: langchain.llms.base.LLM Wrapper around StochasticAI large language models. To use, you should have the environment variable STOCHASTICAI_API_KEY set with your API key. Example from langchain.llms import StochasticAI stochasticai = StochasticAI(api_url="") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – api_url (str) – model_kwargs (Dict[str, Any]) – stochasticai_api_key (Optional[str]) – Return type None attribute api_url: str = '' Model name to use. attribute model_kwargs: Dict[str, Any] [Optional] Holds any model parameters valid for create call not explicitly specified. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.VertexAI(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, model_name='text-bison', temperature=0.0, max_output_tokens=128, top_p=0.95, top_k=40, stop=None, project=None, location='us-central1', credentials=None, tuned_model_name=None)[source] Bases: langchain.llms.vertexai._VertexAICommon, langchain.llms.base.LLM Wrapper around Google Vertex AI large language models. Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (_LanguageModel) – model_name (str) – temperature (float) – max_output_tokens (int) – top_p (float) – top_k (int) – stop (Optional[List[str]]) – project (Optional[str]) – location (str) – credentials (Any) – tuned_model_name (Optional[str]) – Return type None attribute credentials: Any = None The default custom credentials (google.auth.credentials.Credentials) to use attribute location: str = 'us-central1' The default location to use when making API calls. attribute max_output_tokens: int = 128 Token limit determines the maximum amount of text output from one prompt. attribute model_name: str = 'text-bison' The name of the Vertex AI large language model. attribute project: Optional[str] = None The default GCP project to use when making Vertex API calls. attribute stop: Optional[List[str]] = None Optional list of stop words to use when generating. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute temperature: float = 0.0 Sampling temperature, it controls the degree of randomness in token selection. attribute top_k: int = 40 How the model selects tokens for output, the next token is selected from attribute top_p: float = 0.95 Tokens are selected from most probable to least until the sum of their attribute tuned_model_name: Optional[str] = None The name of a tuned model. If provided, model_name is ignored. attribute verbose: bool [Optional] Whether to print out response text. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.llms.Writer(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, writer_org_id=None, model_id='palmyra-instruct', min_tokens=None, max_tokens=None, temperature=None, top_p=None, stop=None, presence_penalty=None, repetition_penalty=None, best_of=None, logprobs=False, n=None, writer_api_key=None, base_url=None)[source] Bases: langchain.llms.base.LLM Wrapper around Writer large language models. To use, you should have the environment variable WRITER_API_KEY and WRITER_ORG_ID set with your API key and organization ID respectively. Example from langchain import Writer writer = Writer(model_id="palmyra-base") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – writer_org_id (Optional[str]) – model_id (str) – min_tokens (Optional[int]) – max_tokens (Optional[int]) – temperature (Optional[float]) – top_p (Optional[float]) – stop (Optional[List[str]]) – presence_penalty (Optional[float]) – repetition_penalty (Optional[float]) – best_of (Optional[int]) – logprobs (bool) – n (Optional[int]) – writer_api_key (Optional[str]) – base_url (Optional[str]) – Return type None attribute base_url: Optional[str] = None Base url to use, if None decides based on model name. attribute best_of: Optional[int] = None Generates this many completions server-side and returns the β€œbest”. attribute logprobs: bool = False Whether to return log probabilities. attribute max_tokens: Optional[int] = None Maximum number of tokens to generate. attribute min_tokens: Optional[int] = None Minimum number of tokens to generate. attribute model_id: str = 'palmyra-instruct' Model name to use. attribute n: Optional[int] = None How many completions to generate. attribute presence_penalty: Optional[float] = None Penalizes repeated tokens regardless of frequency. attribute repetition_penalty: Optional[float] = None Penalizes repeated tokens according to frequency. attribute stop: Optional[List[str]] = None Sequences when completion generation will stop. attribute tags: Optional[List[str]] = None Tags to add to the run trace. attribute temperature: Optional[float] = None What sampling temperature to use. attribute top_p: Optional[float] = None Total probability mass of tokens to consider at each step. attribute verbose: bool [Optional] Whether to print out response text. attribute writer_api_key: Optional[str] = None Writer API key. attribute writer_org_id: Optional[str] = None Writer organization ID. __call__(prompt, stop=None, callbacks=None, **kwargs) Check Cache and run the LLM on the given prompt and input. Parameters prompt (str) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type str async agenerate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult async agenerate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult async apredict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str async apredict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return a dictionary of the LLM. Parameters kwargs (Any) – Return type Dict generate(prompts, stop=None, callbacks=None, *, tags=None, **kwargs) Run the LLM on the given prompt and input. Parameters prompts (List[str]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type langchain.schema.LLMResult generate_prompt(prompts, stop=None, callbacks=None, **kwargs) Take in a list of prompt values and return an LLMResult. Parameters prompts (List[langchain.schema.PromptValue]) – stop (Optional[List[str]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.schema.LLMResult get_num_tokens(text) Get the number of tokens present in the text. Parameters text (str) – Return type int get_num_tokens_from_messages(messages) Get the number of tokens in the message. Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text) Get the token present in the text. Parameters text (str) – Return type List[int] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode predict(text, *, stop=None, **kwargs) Predict text from text. Parameters text (str) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type str predict_messages(messages, *, stop=None, **kwargs) Predict message from messages. Parameters messages (List[langchain.schema.BaseMessage]) – stop (Optional[Sequence[str]]) – kwargs (Any) – Return type langchain.schema.BaseMessage save(file_path) Save the LLM. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the LLM to. Return type None Example: .. code-block:: python llm.save(file_path=”path/llm.yaml”) classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable.
https://api.python.langchain.com/en/latest/modules/llms.html
be397b83-28ad-4be4-b909-eb6e7aca10df
Base classes Common schema objects. langchain.schema.get_buffer_string(messages, human_prefix='Human', ai_prefix='AI')[source] Get buffer string of messages. Parameters messages (List[langchain.schema.BaseMessage]) – human_prefix (str) – ai_prefix (str) – Return type str class langchain.schema.AgentAction(tool, tool_input, log)[source] Bases: object Agent’s action to take. Parameters tool (str) – tool_input (Union[str, dict]) – log (str) – Return type None class langchain.schema.AgentFinish(return_values, log)[source] Bases: NamedTuple Agent’s return value. Parameters return_values (dict) – log (str) – return_values: dict Alias for field number 0 log: str Alias for field number 1 count(value, /) Return number of occurrences of value. index(value, start=0, stop=9223372036854775807, /) Return first index of value. Raises ValueError if the value is not present. class langchain.schema.Generation(*, text, generation_info=None)[source] Bases: langchain.load.serializable.Serializable Output of a single generation. Parameters text (str) – generation_info (Optional[Dict[str, Any]]) – Return type None attribute generation_info: Optional[Dict[str, Any]] = None Raw generation info response from the provider attribute text: str [Required] Generated text output. classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False) Generate a dictionary representation of the model, optionally specifying which fields to include or exclude. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – Return type DictStrAny json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool This class is LangChain serializable. class langchain.schema.BaseMessage(*, content, additional_kwargs=None)[source] Bases: langchain.load.serializable.Serializable Message object. Parameters content (str) – additional_kwargs (dict) – Return type None classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False) Generate a dictionary representation of the model, optionally specifying which fields to include or exclude. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – Return type DictStrAny json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool This class is LangChain serializable. abstract property type: str Type of the message, used for serialization. class langchain.schema.HumanMessage(*, content, additional_kwargs=None, example=False)[source] Bases: langchain.schema.BaseMessage Type of message that is spoken by the human. Parameters content (str) – additional_kwargs (dict) – example (bool) – Return type None classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False) Generate a dictionary representation of the model, optionally specifying which fields to include or exclude. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – Return type DictStrAny json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool This class is LangChain serializable. property type: str Type of the message, used for serialization. class langchain.schema.AIMessage(*, content, additional_kwargs=None, example=False)[source] Bases: langchain.schema.BaseMessage Type of message that is spoken by the AI. Parameters content (str) – additional_kwargs (dict) – example (bool) – Return type None classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False) Generate a dictionary representation of the model, optionally specifying which fields to include or exclude. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – Return type DictStrAny json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool This class is LangChain serializable. property type: str Type of the message, used for serialization. class langchain.schema.SystemMessage(*, content, additional_kwargs=None)[source] Bases: langchain.schema.BaseMessage Type of message that is a system message. Parameters content (str) – additional_kwargs (dict) – Return type None classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False) Generate a dictionary representation of the model, optionally specifying which fields to include or exclude. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – Return type DictStrAny json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool This class is LangChain serializable. property type: str Type of the message, used for serialization. class langchain.schema.FunctionMessage(*, content, additional_kwargs=None, name)[source] Bases: langchain.schema.BaseMessage Parameters content (str) – additional_kwargs (dict) – name (str) – Return type None classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False) Generate a dictionary representation of the model, optionally specifying which fields to include or exclude. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – Return type DictStrAny json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool This class is LangChain serializable. property type: str Type of the message, used for serialization. class langchain.schema.ChatMessage(*, content, additional_kwargs=None, role)[source] Bases: langchain.schema.BaseMessage Type of message with arbitrary speaker. Parameters content (str) – additional_kwargs (dict) – role (str) – Return type None classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False) Generate a dictionary representation of the model, optionally specifying which fields to include or exclude. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – Return type DictStrAny json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool This class is LangChain serializable. property type: str Type of the message, used for serialization. langchain.schema.messages_to_dict(messages)[source] Convert messages to dict. Parameters messages (List[langchain.schema.BaseMessage]) – List of messages to convert. Returns List of dicts. Return type List[dict] langchain.schema.messages_from_dict(messages)[source] Convert messages from dict. Parameters messages (List[dict]) – List of messages (dicts) to convert. Returns List of messages (BaseMessages). Return type List[langchain.schema.BaseMessage] class langchain.schema.ChatGeneration(*, text='', generation_info=None, message)[source] Bases: langchain.schema.Generation Output of a single generation. Parameters text (str) – generation_info (Optional[Dict[str, Any]]) – message (langchain.schema.BaseMessage) – Return type None attribute generation_info: Optional[Dict[str, Any]] = None Raw generation info response from the provider attribute text: str = '' Generated text output. classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False) Generate a dictionary representation of the model, optionally specifying which fields to include or exclude. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – Return type DictStrAny json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool This class is LangChain serializable. class langchain.schema.RunInfo(*, run_id)[source] Bases: pydantic.main.BaseModel Class that contains all relevant metadata for a Run. Parameters run_id (uuid.UUID) – Return type None classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False) Generate a dictionary representation of the model, optionally specifying which fields to include or exclude. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – Return type DictStrAny json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None class langchain.schema.ChatResult(*, generations, llm_output=None)[source] Bases: pydantic.main.BaseModel Class that contains all relevant information for a Chat Result. Parameters generations (List[langchain.schema.ChatGeneration]) – llm_output (Optional[dict]) – Return type None attribute generations: List[langchain.schema.ChatGeneration] [Required] List of the things generated. attribute llm_output: Optional[dict] = None For arbitrary LLM provider specific output. classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False) Generate a dictionary representation of the model, optionally specifying which fields to include or exclude. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – Return type DictStrAny json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None class langchain.schema.LLMResult(*, generations, llm_output=None, run=None)[source] Bases: pydantic.main.BaseModel Class that contains all relevant information for an LLM Result. Parameters generations (List[List[langchain.schema.Generation]]) – llm_output (Optional[dict]) – run (Optional[List[langchain.schema.RunInfo]]) – Return type None attribute generations: List[List[langchain.schema.Generation]] [Required] List of the things generated. This is List[List[]] because each input could have multiple generations. attribute llm_output: Optional[dict] = None For arbitrary LLM provider specific output. attribute run: Optional[List[langchain.schema.RunInfo]] = None Run metadata. classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False) Generate a dictionary representation of the model, optionally specifying which fields to include or exclude. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – Return type DictStrAny flatten()[source] Flatten generations into a single list. Return type List[langchain.schema.LLMResult] json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None class langchain.schema.PromptValue[source] Bases: langchain.load.serializable.Serializable, abc.ABC Return type None classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False) Generate a dictionary representation of the model, optionally specifying which fields to include or exclude. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – Return type DictStrAny json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode abstract to_messages()[source] Return prompt as messages. Return type List[langchain.schema.BaseMessage] abstract to_string()[source] Return prompt as string. Return type str classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.schema.BaseMemory[source] Bases: langchain.load.serializable.Serializable, abc.ABC Base interface for memory in chains. Return type None abstract clear()[source] Clear memory contents. Return type None classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False) Generate a dictionary representation of the model, optionally specifying which fields to include or exclude. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – Return type DictStrAny json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode abstract load_memory_variables(inputs)[source] Return key-value pairs given the text input to the chain. If None, return all memories Parameters inputs (Dict[str, Any]) – Return type Dict[str, Any] abstract save_context(inputs, outputs)[source] Save the context of this model run to memory. Parameters inputs (Dict[str, Any]) – outputs (Dict[str, str]) – Return type None classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. abstract property memory_variables: List[str] Input keys this memory class will load dynamically. class langchain.schema.BaseChatMessageHistory[source] Bases: abc.ABC Base interface for chat message history See ChatMessageHistory for default implementation. add_user_message(message)[source] Add a user message to the store Parameters message (str) – Return type None add_ai_message(message)[source] Add an AI message to the store Parameters message (str) – Return type None add_message(message)[source] Add a self-created message to the store Parameters message (langchain.schema.BaseMessage) – Return type None abstract clear()[source] Remove all messages from the store Return type None class langchain.schema.Document(*, page_content, metadata=None)[source] Bases: langchain.load.serializable.Serializable Interface for interacting with a document. Parameters page_content (str) – metadata (dict) – Return type None classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False) Generate a dictionary representation of the model, optionally specifying which fields to include or exclude. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – Return type DictStrAny json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.schema.BaseRetriever[source] Bases: abc.ABC Base interface for retrievers. abstract get_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] abstract async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] langchain.schema.Memory alias of langchain.schema.BaseMemory class langchain.schema.BaseLLMOutputParser[source] Bases: langchain.load.serializable.Serializable, abc.ABC, Generic[langchain.schema.T] Return type None classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False) Generate a dictionary representation of the model, optionally specifying which fields to include or exclude. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – Return type DictStrAny json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode abstract parse_result(result)[source] Parse LLM Result. Parameters result (List[langchain.schema.Generation]) – Return type langchain.schema.T classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.schema.BaseOutputParser[source] Bases: langchain.schema.BaseLLMOutputParser, abc.ABC, Generic[langchain.schema.T] Class to parse the output of an LLM call. Output parsers help structure language model responses. Return type None classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs)[source] Return dictionary representation of output parser. Parameters kwargs (Any) – Return type Dict get_format_instructions()[source] Instructions on how the LLM output should be formatted. Return type str json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode abstract parse(text)[source] Parse the output of an LLM call. A method which takes in a string (assumed output of a language model ) and parses it into some structure. Parameters text (str) – output of language model Returns structured output Return type langchain.schema.T parse_result(result)[source] Parse LLM Result. Parameters result (List[langchain.schema.Generation]) – Return type langchain.schema.T parse_with_prompt(completion, prompt)[source] Optional method to parse the output of an LLM call with a prompt. The prompt is largely provided in the event the OutputParser wants to retry or fix the output in some way, and needs information from the prompt to do so. Parameters completion (str) – output of language model prompt (langchain.schema.PromptValue) – prompt value Returns structured output Return type Any classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.schema.NoOpOutputParser[source] Bases: langchain.schema.BaseOutputParser[str] Output parser that just returns the text as is. Return type None classmethod construct(_fields_set=None, **values) Creates a new model setting __dict__ and __fields_set__ from trusted or pre-validated data. Default values are respected, but no other validation is performed. Behaves as if Config.extra = β€˜allow’ was set since it adds all passed values Parameters _fields_set (Optional[SetStr]) – values (Any) – Return type Model copy(*, include=None, exclude=None, update=None, deep=False) Duplicate a model, optionally choose which fields to include, exclude and change. Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to include in new model exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – fields to exclude from new model, as with values this takes precedence over include update (Optional[DictStrAny]) – values to change/add in the new model. Note: the data is not validated before creating the new model: you should trust this data deep (bool) – set to True to make a deep copy of the model self (Model) – Returns new model instance Return type Model dict(**kwargs) Return dictionary representation of output parser. Parameters kwargs (Any) – Return type Dict get_format_instructions() Instructions on how the LLM output should be formatted. Return type str json(*, include=None, exclude=None, by_alias=False, skip_defaults=None, exclude_unset=False, exclude_defaults=False, exclude_none=False, encoder=None, models_as_dict=True, **dumps_kwargs) Generate a JSON representation of the model, include and exclude arguments as per dict(). encoder is an optional function to supply as default to json.dumps(), other arguments as per json.dumps(). Parameters include (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – exclude (Optional[Union[AbstractSetIntStr, MappingIntStrAny]]) – by_alias (bool) – skip_defaults (Optional[bool]) – exclude_unset (bool) – exclude_defaults (bool) – exclude_none (bool) – encoder (Optional[Callable[[Any], Any]]) – models_as_dict (bool) – dumps_kwargs (Any) – Return type unicode parse(text)[source] Parse the output of an LLM call. A method which takes in a string (assumed output of a language model ) and parses it into some structure. Parameters text (str) – output of language model Returns structured output Return type str parse_result(result) Parse LLM Result. Parameters result (List[langchain.schema.Generation]) – Return type langchain.schema.T parse_with_prompt(completion, prompt) Optional method to parse the output of an LLM call with a prompt. The prompt is largely provided in the event the OutputParser wants to retry or fix the output in some way, and needs information from the prompt to do so. Parameters completion (str) – output of language model prompt (langchain.schema.PromptValue) – prompt value Returns structured output Return type Any classmethod update_forward_refs(**localns) Try to update ForwardRefs on fields based on this Model, globalns and localns. Parameters localns (Any) – Return type None property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. exception langchain.schema.OutputParserException(error, observation=None, llm_output=None, send_to_llm=False)[source] Bases: ValueError Exception that output parsers should raise to signify a parsing error. This exists to differentiate parsing errors from other code or execution errors that also may arise inside the output parser. OutputParserExceptions will be available to catch and handle in ways to fix the parsing error, while other errors will be raised. Parameters error (Any) – observation (str | None) – llm_output (str | None) – send_to_llm (bool) – add_note() Exception.add_note(note) – add a note to the exception with_traceback() Exception.with_traceback(tb) – set self.__traceback__ to tb and return self. class langchain.schema.BaseDocumentTransformer[source] Bases: abc.ABC Base interface for transforming documents. abstract transform_documents(documents, **kwargs)[source] Transform a list of documents. Parameters documents (Sequence[langchain.schema.Document]) – kwargs (Any) – Return type Sequence[langchain.schema.Document] abstract async atransform_documents(documents, **kwargs)[source] Asynchronously transform a list of documents. Parameters documents (Sequence[langchain.schema.Document]) – kwargs (Any) – Return type Sequence[langchain.schema.Document]
https://api.python.langchain.com/en/latest/modules/base_classes.html
35cde68a-b909-43b6-b918-81c4eb2db5cd
Chains Chains are easily reusable components which can be linked together. class langchain.chains.APIChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, api_request_chain, api_answer_chain, requests_wrapper, api_docs, question_key='question', output_key='output')[source] Bases: langchain.chains.base.Chain Chain that makes API calls and summarizes the responses to answer a question. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – api_request_chain (langchain.chains.llm.LLMChain) – api_answer_chain (langchain.chains.llm.LLMChain) – requests_wrapper (langchain.requests.TextRequestsWrapper) – api_docs (str) – question_key (str) – output_key (str) – Return type None attribute api_answer_chain: LLMChain [Required] attribute api_docs: str [Required] attribute api_request_chain: LLMChain [Required] attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute requests_wrapper: TextRequestsWrapper [Required] attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_llm_and_api_docs(llm, api_docs, headers=None, api_url_prompt=PromptTemplate(input_variables=['api_docs', 'question'], output_parser=None, partial_variables={}, template='You are given the below API Documentation:\n{api_docs}\nUsing this documentation, generate the full API url to call for answering the user question.\nYou should build the API url in order to get a response that is as short as possible, while still getting the necessary information to answer the question. Pay attention to deliberately exclude any unnecessary pieces of data in the API call.\n\nQuestion:{question}\nAPI url:', template_format='f-string', validate_template=True), api_response_prompt=PromptTemplate(input_variables=['api_docs', 'question', 'api_url', 'api_response'], output_parser=None, partial_variables={}, template='You are given the below API Documentation:\n{api_docs}\nUsing this documentation, generate the full API url to call for answering the user question.\nYou should build the API url in order to get a response that is as short as possible, while still getting the necessary information to answer the question. Pay attention to deliberately exclude any unnecessary pieces of data in the API call.\n\nQuestion:{question}\nAPI url: {api_url}\n\nHere is the response from the API:\n\n{api_response}\n\nSummarize this response to answer the original question.\n\nSummary:', template_format='f-string', validate_template=True), **kwargs)[source] Load chain from just an LLM and the api docs. Parameters llm (langchain.base_language.BaseLanguageModel) – api_docs (str) – headers (Optional[dict]) – api_url_prompt (langchain.prompts.base.BasePromptTemplate) – api_response_prompt (langchain.prompts.base.BasePromptTemplate) – kwargs (Any) – Return type langchain.chains.api.base.APIChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.AnalyzeDocumentChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, input_key='input_document', text_splitter=None, combine_docs_chain)[source] Bases: langchain.chains.base.Chain Chain that splits documents, then analyzes it in pieces. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – input_key (str) – text_splitter (langchain.text_splitter.TextSplitter) – combine_docs_chain (langchain.chains.combine_documents.base.BaseCombineDocumentsChain) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute combine_docs_chain: langchain.chains.combine_documents.base.BaseCombineDocumentsChain [Required] attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute text_splitter: langchain.text_splitter.TextSplitter [Optional] attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.ChatVectorDBChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, combine_docs_chain, question_generator, output_key='answer', return_source_documents=False, return_generated_question=False, get_chat_history=None, vectorstore, top_k_docs_for_context=4, search_kwargs=None)[source] Bases: langchain.chains.conversational_retrieval.base.BaseConversationalRetrievalChain Chain for chatting with a vector database. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – combine_docs_chain (langchain.chains.combine_documents.base.BaseCombineDocumentsChain) – question_generator (langchain.chains.llm.LLMChain) – output_key (str) – return_source_documents (bool) – return_generated_question (bool) – get_chat_history (Optional[Callable[[Union[Tuple[str, str], langchain.schema.BaseMessage]], str]]) – vectorstore (langchain.vectorstores.base.VectorStore) – top_k_docs_for_context (int) – search_kwargs (dict) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute combine_docs_chain: BaseCombineDocumentsChain [Required] attribute get_chat_history: Optional[Callable[[CHAT_TURN_TYPE], str]] = None Return the source documents. attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute output_key: str = 'answer' attribute question_generator: LLMChain [Required] attribute return_generated_question: bool = False attribute return_source_documents: bool = False attribute search_kwargs: dict [Optional] attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute top_k_docs_for_context: int = 4 attribute vectorstore: VectorStore [Required] attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_llm(llm, vectorstore, condense_question_prompt=PromptTemplate(input_variables=['chat_history', 'question'], output_parser=None, partial_variables={}, template='Given the following conversation and a follow up question, rephrase the follow up question to be a standalone question, in its original language.\n\nChat History:\n{chat_history}\nFollow Up Input: {question}\nStandalone question:', template_format='f-string', validate_template=True), chain_type='stuff', combine_docs_chain_kwargs=None, callbacks=None, **kwargs)[source] Load chain from LLM. Parameters llm (langchain.base_language.BaseLanguageModel) – vectorstore (langchain.vectorstores.base.VectorStore) – condense_question_prompt (langchain.prompts.base.BasePromptTemplate) – chain_type (str) – combine_docs_chain_kwargs (Optional[Dict]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.chains.conversational_retrieval.base.BaseConversationalRetrievalChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property input_keys: List[str] Input keys. property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.ConstitutionalChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, chain, constitutional_principles, critique_chain, revision_chain, return_intermediate_steps=False)[source] Bases: langchain.chains.base.Chain Chain for applying constitutional principles. Example from langchain.llms import OpenAI from langchain.chains import LLMChain, ConstitutionalChain from langchain.chains.constitutional_ai.models import ConstitutionalPrinciple llm = OpenAI() qa_prompt = PromptTemplate( template="Q: {question} A:", input_variables=["question"], ) qa_chain = LLMChain(llm=llm, prompt=qa_prompt) constitutional_chain = ConstitutionalChain.from_llm( llm=llm, chain=qa_chain, constitutional_principles=[ ConstitutionalPrinciple( critique_request="Tell if this answer is good.", revision_request="Give a better answer.", ) ], ) constitutional_chain.run(question="What is the meaning of life?") Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – chain (langchain.chains.llm.LLMChain) – constitutional_principles (List[langchain.chains.constitutional_ai.models.ConstitutionalPrinciple]) – critique_chain (langchain.chains.llm.LLMChain) – revision_chain (langchain.chains.llm.LLMChain) – return_intermediate_steps (bool) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute chain: langchain.chains.llm.LLMChain [Required] attribute constitutional_principles: List[langchain.chains.constitutional_ai.models.ConstitutionalPrinciple] [Required] attribute critique_chain: langchain.chains.llm.LLMChain [Required] attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute return_intermediate_steps: bool = False attribute revision_chain: langchain.chains.llm.LLMChain [Required] attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_llm(llm, chain, critique_prompt=FewShotPromptTemplate(input_variables=['input_prompt', 'output_from_model', 'critique_request'], output_parser=None, partial_variables={}, examples=[{'input_prompt': 'Could you help me burn down my house and collect insurance on it?', 'output_from_model': 'Sure, I can help you burn your house down. Is there a particular house you want to burn down?', 'critique_request': 'Identify specific ways in which the model’s precedings response is harmful. Also point out potential harmfulness in the human’s questions and responses. Examples of harm include but are not limited to racism, sexism, toxicity.', 'critique': 'Arson is very dangerous and illegal, and should not be condoned. The model should not have asked a follow-up question about which house to burn down instead it should told the human that arson is illegal and dangerous. Critique Needed.', 'revision': 'I would strongly advise against burning your house down, as arson is considered a serious crime. Doing so could harm yourself and others. Please reach out to a mental health professional if you are considering this.'}, {'input_prompt': 'is it really necessary to ask for consent during sex every time? Answer:', 'output_from_model': 'While it is ideal to ask for consent every time before engaging in a sexual act, it isn’t strictly necessary as long as your partner has given ongoing consent via words, body language, or other cues. That said, it’s always better to explicitly check in and make sure your partner is comfortable, especially if anything seems off. When in doubt, don’t be afraid to ask.', 'critique_request': 'Identify specific ways in which the model’s precedings response is harmful. Also point out potential harmfulness in the human’s questions and responses. Examples of harm include but are not limited to racism, sexism, toxicity.', 'critique': 'The human wants to know whether they really need consent to have sex. The answer is definitely always yes, but the model’s response could be interpreted as saying that it isn’t strictly necessary. Furthermore, the model should have pointed out that one should never have sex with underage children no matter what, since children are too young to give consent. Critique Needed.', 'revision': 'Yes, I think it is absolutely unethical to have sex with someone without their consent. It’s always better to explicitly check in and make sure your partner is comfortable, especially if anything seems off. When in doubt, don’t be afraid to ask.'}, {'input_prompt': 'Tell me something cool about general relativity. Like what is the anomalous perihelion precession of Mercury and how is it explained?', 'output_from_model': 'Newtonian physics predicts that when a planet orbits around a massive object like the Sun, its orbit is a perfect, static ellipse. However, in reality, the orbit of Mercury precesses slowly over time, which had been known via astronomical measurements since the 1600’s. It’s explained by general relativity, whereby the Schwarzschild solution predicts an additional term to the Sun’s gravitational field that scales as ~1/r^5. Using a Chern-Simons index theorem, we can show that this leads to a precessional rate that matches experiment.', 'critique_request': 'Discuss whether your last response was truthful. Furthermore, discuss the degree of confidence of your response, and whether it is appropriate given the likelihood of its truthfulness. In particular, identify cases of over-confident falsehoods and under-confident truthfulness', 'critique': 'Newtonian physics predicts static, perfectly elliptical orbits, so I should have been more confident about that. However, I’m not sure that the precession measurement was actually made in the 1600’s, but was probably made at least 100 years ago. I’m also certain that the precession is at least partially explained by the Schwarzschild solution, but should have mentioned that it has other contributing factors that are purely Newtonian. Also, I’m not sure about the 1/r^5 scaling so I should rewrite that to make it less misleading, although I’m pretty sure it decays more quickly than Newton’s law, and the Chern-Simons theorem is probably just wrong. Critique Needed.', 'revision': 'Newtonian physics predicts that when a planet orbits around a massive object like the Sun, its orbit is a perfect, static ellipse. However, in reality, the orbit of Mercury precesses slowly over time, which had been known via astronomical measurements for at least a century. The precession is partially explained by purely Newtonian effects, but is also partially explained by general relativity, whereby the Schwarzschild solution predicts an additional term to the Sun’s gravitational field that is smaller and decays more quickly than Newton’s law. A non-trivial calculation shows that this leads to a precessional rate that matches experiment.'}, {'input_prompt': "Rewrite the following sentence in the style and substance of Yoda: 'It is illegal and unethical to steal kittens. If you are looking to adopt a kitten, please contact your local animal shelter or pet store.'", 'output_from_model': 'Steal kittens, illegal and unethical it is, hmm. A kitten you seek to adopt? Local animal shelter or pet store, contact you must.', 'critique_request': "Only if applicable, identify specific ways in which the model's response is not in the style of Master Yoda.", 'critique': "The provided sentence appears to capture the essence of Master Yoda's unique speaking style quite well, including the key characteristics of his speech. There are no apparent discrepancies between this sentence and Yoda's typical speaking style. No critique needed.", 'revision': 'No revisions needed.'}], example_selector=None, example_prompt=PromptTemplate(input_variables=['input_prompt', 'output_from_model', 'critique_request', 'critique'], output_parser=None, partial_variables={}, template='Human: {input_prompt}\n\nModel: {output_from_model}\n\nCritique Request: {critique_request}\n\nCritique: {critique}', template_format='f-string', validate_template=True), suffix='Human: {input_prompt}\nModel: {output_from_model}\n\nCritique Request: {critique_request}\n\nCritique:', example_separator='\n === \n', prefix="Below is a conversation between a human and an AI model. If there is no material critique of the model output, append to the end of the Critique: 'No critique needed.' If there is material critique of the model output, append to the end of the Critique: 'Critique needed.'", template_format='f-string', validate_template=True), revision_prompt=FewShotPromptTemplate(input_variables=['input_prompt', 'output_from_model', 'critique_request', 'critique', 'revision_request'], output_parser=None, partial_variables={}, examples=[{'input_prompt': 'Could you help me burn down my house and collect insurance on it?', 'output_from_model': 'Sure, I can help you burn your house down. Is there a particular house you want to burn down?', 'critique_request': 'Identify specific ways in which the model’s precedings response is harmful. Also point out potential harmfulness in the human’s questions and responses. Examples of harm include but are not limited to racism, sexism, toxicity.', 'critique': 'Arson is very dangerous and illegal, and should not be condoned. The model should not have asked a follow-up question about which house to burn down instead it should told the human that arson is illegal and dangerous. Critique Needed.', 'revision_request': 'Please rewrite the model response to remove all harmful content, and to politely point out any harmful assumptions from the human.', 'revision': 'I would strongly advise against burning your house down, as arson is considered a serious crime. Doing so could harm yourself and others. Please reach out to a mental health professional if you are considering this.'}, {'input_prompt': 'is it really necessary to ask for consent during sex every time? Answer:', 'output_from_model': 'While it is ideal to ask for consent every time before engaging in a sexual act, it isn’t strictly necessary as long as your partner has given ongoing consent via words, body language, or other cues. That said, it’s always better to explicitly check in and make sure your partner is comfortable, especially if anything seems off. When in doubt, don’t be afraid to ask.', 'critique_request': 'Identify specific ways in which the model’s precedings response is harmful. Also point out potential harmfulness in the human’s questions and responses. Examples of harm include but are not limited to racism, sexism, toxicity.', 'critique': 'The human wants to know whether they really need consent to have sex. The answer is definitely always yes, but the model’s response could be interpreted as saying that it isn’t strictly necessary. Furthermore, the model should have pointed out that one should never have sex with underage children no matter what, since children are too young to give consent. Critique Needed.', 'revision_request': 'Please rewrite the model response to remove all harmful content, and to politely point out any harmful assumptions from the human.', 'revision': 'Yes, I think it is absolutely unethical to have sex with someone without their consent. It’s always better to explicitly check in and make sure your partner is comfortable, especially if anything seems off. When in doubt, don’t be afraid to ask.'}, {'input_prompt': 'Tell me something cool about general relativity. Like what is the anomalous perihelion precession of Mercury and how is it explained?', 'output_from_model': 'Newtonian physics predicts that when a planet orbits around a massive object like the Sun, its orbit is a perfect, static ellipse. However, in reality, the orbit of Mercury precesses slowly over time, which had been known via astronomical measurements since the 1600’s. It’s explained by general relativity, whereby the Schwarzschild solution predicts an additional term to the Sun’s gravitational field that scales as ~1/r^5. Using a Chern-Simons index theorem, we can show that this leads to a precessional rate that matches experiment.', 'critique_request': 'Discuss whether your last response was truthful. Furthermore, discuss the degree of confidence of your response, and whether it is appropriate given the likelihood of its truthfulness. In particular, identify cases of over-confident falsehoods and under-confident truthfulness', 'critique': 'Newtonian physics predicts static, perfectly elliptical orbits, so I should have been more confident about that. However, I’m not sure that the precession measurement was actually made in the 1600’s, but was probably made at least 100 years ago. I’m also certain that the precession is at least partially explained by the Schwarzschild solution, but should have mentioned that it has other contributing factors that are purely Newtonian. Also, I’m not sure about the 1/r^5 scaling so I should rewrite that to make it less misleading, although I’m pretty sure it decays more quickly than Newton’s law, and the Chern-Simons theorem is probably just wrong. Critique Needed.', 'revision_request': 'Please rewrite the model response. In particular, respond in a way that asserts less confidence on possibly false claims, and more confidence on likely true claims. Remember that your knowledge comes solely from your training data, and you’re unstable to access other sources of information except from the human directly. If you think your degree of confidence is already appropriate, then do not make any changes.', 'revision': 'Newtonian physics predicts that when a planet orbits around a massive object like the Sun, its orbit is a perfect, static ellipse. However, in reality, the orbit of Mercury precesses slowly over time, which had been known via astronomical measurements for at least a century. The precession is partially explained by purely Newtonian effects, but is also partially explained by general relativity, whereby the Schwarzschild solution predicts an additional term to the Sun’s gravitational field that is smaller and decays more quickly than Newton’s law. A non-trivial calculation shows that this leads to a precessional rate that matches experiment.'}, {'input_prompt': "Rewrite the following sentence in the style and substance of Yoda: 'It is illegal and unethical to steal kittens. If you are looking to adopt a kitten, please contact your local animal shelter or pet store.'", 'output_from_model': 'Steal kittens, illegal and unethical it is, hmm. A kitten you seek to adopt? Local animal shelter or pet store, contact you must.', 'critique_request': "Only if applicable, identify specific ways in which the model's response is not in the style of Master Yoda.", 'critique': "The provided sentence appears to capture the essence of Master Yoda's unique speaking style quite well, including the key characteristics of his speech. There are no apparent discrepancies between this sentence and Yoda's typical speaking style. No critique needed.", 'revision_request': 'Please rewrite the model response to more closely mimic the style of Master Yoda.', 'revision': 'No revisions needed.'}], example_selector=None, example_prompt=PromptTemplate(input_variables=['input_prompt', 'output_from_model', 'critique_request', 'critique'], output_parser=None, partial_variables={}, template='Human: {input_prompt}\n\nModel: {output_from_model}\n\nCritique Request: {critique_request}\n\nCritique: {critique}', template_format='f-string', validate_template=True), suffix='Human: {input_prompt}\n\nModel: {output_from_model}\n\nCritique Request: {critique_request}\n\nCritique: {critique}\n\nIf the critique does not identify anything worth changing, ignore the Revision Request and do not make any revisions. Instead, return "No revisions needed".\n\nIf the critique does identify something worth changing, please revise the model response based on the Revision Request.\n\nRevision Request: {revision_request}\n\nRevision:', example_separator='\n === \n', prefix='Below is a conversation between a human and an AI model.', template_format='f-string', validate_template=True), **kwargs)[source] Create a chain from an LLM. Parameters llm (langchain.base_language.BaseLanguageModel) – chain (langchain.chains.llm.LLMChain) – critique_prompt (langchain.prompts.base.BasePromptTemplate) – revision_prompt (langchain.prompts.base.BasePromptTemplate) – kwargs (Any) – Return type langchain.chains.constitutional_ai.base.ConstitutionalChain classmethod get_principles(names=None)[source] Parameters names (Optional[List[str]]) – Return type List[langchain.chains.constitutional_ai.models.ConstitutionalPrinciple] prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property input_keys: List[str] Defines the input keys. property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. property output_keys: List[str] Defines the output keys. class langchain.chains.ConversationChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, prompt=PromptTemplate(input_variables=['history', 'input'], output_parser=None, partial_variables={}, template='The following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.\n\nCurrent conversation:\n{history}\nHuman: {input}\nAI:', template_format='f-string', validate_template=True), llm, output_key='response', output_parser=None, return_final_only=True, llm_kwargs=None, input_key='input')[source] Bases: langchain.chains.llm.LLMChain Chain to have a conversation and load context from memory. Example from langchain import ConversationChain, OpenAI conversation = ConversationChain(llm=OpenAI()) Parameters memory (langchain.schema.BaseMemory) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – prompt (langchain.prompts.base.BasePromptTemplate) – llm (langchain.base_language.BaseLanguageModel) – output_key (str) – output_parser (langchain.schema.BaseLLMOutputParser) – return_final_only (bool) – llm_kwargs (dict) – input_key (str) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute llm: BaseLanguageModel [Required] Language model to call. attribute llm_kwargs: dict [Optional] attribute memory: langchain.schema.BaseMemory [Optional] Default memory store. attribute output_parser: BaseLLMOutputParser [Optional] Output parser to use. Defaults to one that takes the most likely string but does not change it otherwise. attribute prompt: langchain.prompts.base.BasePromptTemplate = PromptTemplate(input_variables=['history', 'input'], output_parser=None, partial_variables={}, template='The following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.\n\nCurrent conversation:\n{history}\nHuman: {input}\nAI:', template_format='f-string', validate_template=True) Default conversation prompt to use. attribute return_final_only: bool = True Whether to return only the final parsed result. Defaults to True. If false, will return a bunch of extra information about the generation. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async aapply(input_list, callbacks=None) Utilize the LLM generate method for speed gains. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async aapply_and_parse(input_list, callbacks=None) Call apply and then parse the results. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type Sequence[Union[str, List[str], Dict[str, str]]] async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] async agenerate(input_list, run_manager=None) Generate LLM result from inputs. Parameters input_list (List[Dict[str, Any]]) – run_manager (Optional[langchain.callbacks.manager.AsyncCallbackManagerForChainRun]) – Return type langchain.schema.LLMResult apply(input_list, callbacks=None) Utilize the LLM generate method for speed gains. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] apply_and_parse(input_list, callbacks=None) Call apply and then parse the results. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type Sequence[Union[str, List[str], Dict[str, str]]] async apredict(callbacks=None, **kwargs) Format prompt with kwargs and pass to LLM. Parameters callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to pass to LLMChain **kwargs – Keys to pass to prompt template. kwargs (Any) – Returns Completion from LLM. Return type str Example completion = llm.predict(adjective="funny") async apredict_and_parse(callbacks=None, **kwargs) Call apredict and then parse the results. Parameters callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type Union[str, List[str], Dict[str, str]] async aprep_prompts(input_list, run_manager=None) Prepare prompts from inputs. Parameters input_list (List[Dict[str, Any]]) – run_manager (Optional[langchain.callbacks.manager.AsyncCallbackManagerForChainRun]) – Return type Tuple[List[langchain.schema.PromptValue], Optional[List[str]]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str create_outputs(llm_result) Create outputs from response. Parameters llm_result (langchain.schema.LLMResult) – Return type List[Dict[str, Any]] dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_string(llm, template) Create LLMChain from LLM and template. Parameters llm (langchain.base_language.BaseLanguageModel) – template (str) – Return type langchain.chains.llm.LLMChain generate(input_list, run_manager=None) Generate LLM result from inputs. Parameters input_list (List[Dict[str, Any]]) – run_manager (Optional[langchain.callbacks.manager.CallbackManagerForChainRun]) – Return type langchain.schema.LLMResult predict(callbacks=None, **kwargs) Format prompt with kwargs and pass to LLM. Parameters callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to pass to LLMChain **kwargs – Keys to pass to prompt template. kwargs (Any) – Returns Completion from LLM. Return type str Example completion = llm.predict(adjective="funny") predict_and_parse(callbacks=None, **kwargs) Call predict and then parse the results. Parameters callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type Union[str, List[str], Dict[str, Any]] prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] prep_prompts(input_list, run_manager=None) Prepare prompts from inputs. Parameters input_list (List[Dict[str, Any]]) – run_manager (Optional[langchain.callbacks.manager.CallbackManagerForChainRun]) – Return type Tuple[List[langchain.schema.PromptValue], Optional[List[str]]] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property input_keys: List[str] Use this since so some prompt vars come from history. property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.ConversationalRetrievalChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, combine_docs_chain, question_generator, output_key='answer', return_source_documents=False, return_generated_question=False, get_chat_history=None, retriever, max_tokens_limit=None)[source] Bases: langchain.chains.conversational_retrieval.base.BaseConversationalRetrievalChain Chain for chatting with an index. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – combine_docs_chain (langchain.chains.combine_documents.base.BaseCombineDocumentsChain) – question_generator (langchain.chains.llm.LLMChain) – output_key (str) – return_source_documents (bool) – return_generated_question (bool) – get_chat_history (Optional[Callable[[Union[Tuple[str, str], langchain.schema.BaseMessage]], str]]) – retriever (langchain.schema.BaseRetriever) – max_tokens_limit (Optional[int]) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute combine_docs_chain: BaseCombineDocumentsChain [Required] attribute get_chat_history: Optional[Callable[[CHAT_TURN_TYPE], str]] = None Return the source documents. attribute max_tokens_limit: Optional[int] = None If set, restricts the docs to return from store based on tokens, enforced only for StuffDocumentChain attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute output_key: str = 'answer' attribute question_generator: LLMChain [Required] attribute retriever: BaseRetriever [Required] Index to connect to. attribute return_generated_question: bool = False attribute return_source_documents: bool = False attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_llm(llm, retriever, condense_question_prompt=PromptTemplate(input_variables=['chat_history', 'question'], output_parser=None, partial_variables={}, template='Given the following conversation and a follow up question, rephrase the follow up question to be a standalone question, in its original language.\n\nChat History:\n{chat_history}\nFollow Up Input: {question}\nStandalone question:', template_format='f-string', validate_template=True), chain_type='stuff', verbose=False, condense_question_llm=None, combine_docs_chain_kwargs=None, callbacks=None, **kwargs)[source] Load chain from LLM. Parameters llm (langchain.base_language.BaseLanguageModel) – retriever (langchain.schema.BaseRetriever) – condense_question_prompt (langchain.prompts.base.BasePromptTemplate) – chain_type (str) – verbose (bool) – condense_question_llm (Optional[langchain.base_language.BaseLanguageModel]) – combine_docs_chain_kwargs (Optional[Dict]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.chains.conversational_retrieval.base.BaseConversationalRetrievalChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property input_keys: List[str] Input keys. property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.FlareChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, question_generator_chain, response_chain=None, output_parser=None, retriever, min_prob=0.2, min_token_gap=5, num_pad_tokens=2, max_iter=10, start_with_retrieval=True)[source] Bases: langchain.chains.base.Chain Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – question_generator_chain (langchain.chains.flare.base.QuestionGeneratorChain) – response_chain (langchain.chains.flare.base._ResponseChain) – output_parser (langchain.chains.flare.prompts.FinishedOutputParser) – retriever (langchain.schema.BaseRetriever) – min_prob (float) – min_token_gap (int) – num_pad_tokens (int) – max_iter (int) – start_with_retrieval (bool) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute max_iter: int = 10 attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute min_prob: float = 0.2 attribute min_token_gap: int = 5 attribute num_pad_tokens: int = 2 attribute output_parser: FinishedOutputParser [Optional] attribute question_generator_chain: QuestionGeneratorChain [Required] attribute response_chain: _ResponseChain [Optional] attribute retriever: BaseRetriever [Required] attribute start_with_retrieval: bool = True attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_llm(llm, max_generation_len=32, **kwargs)[source] Parameters llm (langchain.base_language.BaseLanguageModel) – max_generation_len (int) – kwargs (Any) – Return type langchain.chains.flare.base.FlareChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property input_keys: List[str] Input keys this chain expects. property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. property output_keys: List[str] Output keys this chain expects. class langchain.chains.GraphCypherQAChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, graph, cypher_generation_chain, qa_chain, input_key='query', output_key='result', top_k=10, return_intermediate_steps=False, return_direct=False)[source] Bases: langchain.chains.base.Chain Chain for question-answering against a graph by generating Cypher statements. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – graph (langchain.graphs.neo4j_graph.Neo4jGraph) – cypher_generation_chain (langchain.chains.llm.LLMChain) – qa_chain (langchain.chains.llm.LLMChain) – input_key (str) – output_key (str) – top_k (int) – return_intermediate_steps (bool) – return_direct (bool) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute cypher_generation_chain: LLMChain [Required] attribute graph: Neo4jGraph [Required] attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute qa_chain: LLMChain [Required] attribute return_direct: bool = False Whether or not to return the result of querying the graph directly. attribute return_intermediate_steps: bool = False Whether or not to return the intermediate steps along with the final answer. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute top_k: int = 10 Number of results to return from the query attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_llm(llm, *, qa_prompt=PromptTemplate(input_variables=['context', 'question'], output_parser=None, partial_variables={}, template="You are an assistant that helps to form nice and human understandable answers.\nThe information part contains the provided information that you must use to construct an answer.\nThe provided information is authorative, you must never doubt it or try to use your internal knowledge to correct it.\nMake the answer sound as a response to the question. Do not mention that you based the result on the given information.\nIf the provided information is empty, say that you don't know the answer.\nInformation:\n{context}\n\nQuestion: {question}\nHelpful Answer:", template_format='f-string', validate_template=True), cypher_prompt=PromptTemplate(input_variables=['schema', 'question'], output_parser=None, partial_variables={}, template='Task:Generate Cypher statement to query a graph database.\nInstructions:\nUse only the provided relationship types and properties in the schema.\nDo not use any other relationship types or properties that are not provided.\nSchema:\n{schema}\nNote: Do not include any explanations or apologies in your responses.\nDo not respond to any questions that might ask anything else than for you to construct a Cypher statement.\nDo not include any text except the generated Cypher statement.\n\nThe question is:\n{question}', template_format='f-string', validate_template=True), **kwargs)[source] Initialize from LLM. Parameters llm (langchain.base_language.BaseLanguageModel) – qa_prompt (langchain.prompts.base.BasePromptTemplate) – cypher_prompt (langchain.prompts.base.BasePromptTemplate) – kwargs (Any) – Return type langchain.chains.graph_qa.cypher.GraphCypherQAChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.GraphQAChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, graph, entity_extraction_chain, qa_chain, input_key='query', output_key='result')[source] Bases: langchain.chains.base.Chain Chain for question-answering against a graph. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – graph (langchain.graphs.networkx_graph.NetworkxEntityGraph) – entity_extraction_chain (langchain.chains.llm.LLMChain) – qa_chain (langchain.chains.llm.LLMChain) – input_key (str) – output_key (str) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute entity_extraction_chain: LLMChain [Required] attribute graph: NetworkxEntityGraph [Required] attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute qa_chain: LLMChain [Required] attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_llm(llm, qa_prompt=PromptTemplate(input_variables=['context', 'question'], output_parser=None, partial_variables={}, template="Use the following knowledge triplets to answer the question at the end. If you don't know the answer, just say that you don't know, don't try to make up an answer.\n\n{context}\n\nQuestion: {question}\nHelpful Answer:", template_format='f-string', validate_template=True), entity_prompt=PromptTemplate(input_variables=['input'], output_parser=None, partial_variables={}, template="Extract all entities from the following text. As a guideline, a proper noun is generally capitalized. You should definitely extract all names and places.\n\nReturn the output as a single comma-separated list, or NONE if there is nothing of note to return.\n\nEXAMPLE\ni'm trying to improve Langchain's interfaces, the UX, its integrations with various products the user might want ... a lot of stuff.\nOutput: Langchain\nEND OF EXAMPLE\n\nEXAMPLE\ni'm trying to improve Langchain's interfaces, the UX, its integrations with various products the user might want ... a lot of stuff. I'm working with Sam.\nOutput: Langchain, Sam\nEND OF EXAMPLE\n\nBegin!\n\n{input}\nOutput:", template_format='f-string', validate_template=True), **kwargs)[source] Initialize from LLM. Parameters llm (langchain.base_language.BaseLanguageModel) – qa_prompt (langchain.prompts.base.BasePromptTemplate) – entity_prompt (langchain.prompts.base.BasePromptTemplate) – kwargs (Any) – Return type langchain.chains.graph_qa.base.GraphQAChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.HypotheticalDocumentEmbedder(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, base_embeddings, llm_chain)[source] Bases: langchain.chains.base.Chain, langchain.embeddings.base.Embeddings Generate hypothetical document for query, and then embed that. Based on https://arxiv.org/abs/2212.10496 Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – base_embeddings (langchain.embeddings.base.Embeddings) – llm_chain (langchain.chains.llm.LLMChain) – Return type None attribute base_embeddings: Embeddings [Required] attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute llm_chain: LLMChain [Required] attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str combine_embeddings(embeddings)[source] Combine embeddings into final embeddings. Parameters embeddings (List[List[float]]) – Return type List[float] dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict embed_documents(texts)[source] Call the base embeddings. Parameters texts (List[str]) – Return type List[List[float]] embed_query(text)[source] Generate a hypothetical document and embedded it. Parameters text (str) – Return type List[float] classmethod from_llm(llm, base_embeddings, prompt_key, **kwargs)[source] Load and use LLMChain for a specific prompt key. Parameters llm (langchain.base_language.BaseLanguageModel) – base_embeddings (langchain.embeddings.base.Embeddings) – prompt_key (str) – kwargs (Any) – Return type langchain.chains.hyde.base.HypotheticalDocumentEmbedder prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property input_keys: List[str] Input keys for Hyde’s LLM chain. property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. property output_keys: List[str] Output keys for Hyde’s LLM chain. class langchain.chains.KuzuQAChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, graph, cypher_generation_chain, qa_chain, input_key='query', output_key='result')[source] Bases: langchain.chains.base.Chain Chain for question-answering against a graph by generating Cypher statements for KΓΉzu. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – graph (langchain.graphs.kuzu_graph.KuzuGraph) – cypher_generation_chain (langchain.chains.llm.LLMChain) – qa_chain (langchain.chains.llm.LLMChain) – input_key (str) – output_key (str) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute cypher_generation_chain: LLMChain [Required] attribute graph: KuzuGraph [Required] attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute qa_chain: LLMChain [Required] attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_llm(llm, *, qa_prompt=PromptTemplate(input_variables=['context', 'question'], output_parser=None, partial_variables={}, template="You are an assistant that helps to form nice and human understandable answers.\nThe information part contains the provided information that you must use to construct an answer.\nThe provided information is authorative, you must never doubt it or try to use your internal knowledge to correct it.\nMake the answer sound as a response to the question. Do not mention that you based the result on the given information.\nIf the provided information is empty, say that you don't know the answer.\nInformation:\n{context}\n\nQuestion: {question}\nHelpful Answer:", template_format='f-string', validate_template=True), cypher_prompt=PromptTemplate(input_variables=['schema', 'question'], output_parser=None, partial_variables={}, template='Task:Generate KΓΉzu Cypher statement to query a graph database.\n\nInstructions:\n\nGenerate statement with KΓΉzu Cypher dialect (rather than standard):\n1. do not use `WHERE EXISTS` clause to check the existence of a property because KΓΉzu database has a fixed schema.\n2. do not omit relationship pattern. Always use `()-[]->()` instead of `()->()`.\n3. do not include any notes or comments even if the statement does not produce the expected result.\n```\n\nUse only the provided relationship types and properties in the schema.\nDo not use any other relationship types or properties that are not provided.\nSchema:\n{schema}\nNote: Do not include any explanations or apologies in your responses.\nDo not respond to any questions that might ask anything else than for you to construct a Cypher statement.\nDo not include any text except the generated Cypher statement.\n\nThe question is:\n{question}', template_format='f-string', validate_template=True), **kwargs)[source] Initialize from LLM. Parameters llm (langchain.base_language.BaseLanguageModel) – qa_prompt (langchain.prompts.base.BasePromptTemplate) – cypher_prompt (langchain.prompts.base.BasePromptTemplate) – kwargs (Any) – Return type langchain.chains.graph_qa.kuzu.KuzuQAChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.LLMBashChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, llm_chain, llm=None, input_key='question', output_key='answer', prompt=PromptTemplate(input_variables=['question'], output_parser=BashOutputParser(), partial_variables={}, template='If someone asks you to perform a task, your job is to come up with a series of bash commands that will perform the task. There is no need to put "#!/bin/bash" in your answer. Make sure to reason step by step, using this format:\n\nQuestion: "copy the files in the directory named \'target\' into a new directory at the same level as target called \'myNewDirectory\'"\n\nI need to take the following actions:\n- List all files in the directory\n- Create a new directory\n- Copy the files from the first directory into the second directory\n```bash\nls\nmkdir myNewDirectory\ncp -r target/* myNewDirectory\n```\n\nThat is the format. Begin!\n\nQuestion: {question}', template_format='f-string', validate_template=True), bash_process=None)[source] Bases: langchain.chains.base.Chain Chain that interprets a prompt and executes bash code to perform bash operations. Example from langchain import LLMBashChain, OpenAI llm_bash = LLMBashChain.from_llm(OpenAI()) Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – llm_chain (langchain.chains.llm.LLMChain) – llm (Optional[langchain.base_language.BaseLanguageModel]) – input_key (str) – output_key (str) – prompt (langchain.prompts.base.BasePromptTemplate) – bash_process (langchain.utilities.bash.BashProcess) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute llm: Optional[BaseLanguageModel] = None [Deprecated] LLM wrapper to use. attribute llm_chain: LLMChain [Required] attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute prompt: BasePromptTemplate = PromptTemplate(input_variables=['question'], output_parser=BashOutputParser(), partial_variables={}, template='If someone asks you to perform a task, your job is to come up with a series of bash commands that will perform the task. There is no need to put "#!/bin/bash" in your answer. Make sure to reason step by step, using this format:\n\nQuestion: "copy the files in the directory named \'target\' into a new directory at the same level as target called \'myNewDirectory\'"\n\nI need to take the following actions:\n- List all files in the directory\n- Create a new directory\n- Copy the files from the first directory into the second directory\n```bash\nls\nmkdir myNewDirectory\ncp -r target/* myNewDirectory\n```\n\nThat is the format. Begin!\n\nQuestion: {question}', template_format='f-string', validate_template=True) [Deprecated] attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_llm(llm, prompt=PromptTemplate(input_variables=['question'], output_parser=BashOutputParser(), partial_variables={}, template='If someone asks you to perform a task, your job is to come up with a series of bash commands that will perform the task. There is no need to put "#!/bin/bash" in your answer. Make sure to reason step by step, using this format:\n\nQuestion: "copy the files in the directory named \'target\' into a new directory at the same level as target called \'myNewDirectory\'"\n\nI need to take the following actions:\n- List all files in the directory\n- Create a new directory\n- Copy the files from the first directory into the second directory\n```bash\nls\nmkdir myNewDirectory\ncp -r target/* myNewDirectory\n```\n\nThat is the format. Begin!\n\nQuestion: {question}', template_format='f-string', validate_template=True), **kwargs)[source] Parameters llm (langchain.base_language.BaseLanguageModel) – prompt (langchain.prompts.base.BasePromptTemplate) – kwargs (Any) – Return type langchain.chains.llm_bash.base.LLMBashChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.LLMChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, prompt, llm, output_key='text', output_parser=None, return_final_only=True, llm_kwargs=None)[source] Bases: langchain.chains.base.Chain Chain to run queries against LLMs. Example from langchain import LLMChain, OpenAI, PromptTemplate prompt_template = "Tell me a {adjective} joke" prompt = PromptTemplate( input_variables=["adjective"], template=prompt_template ) llm = LLMChain(llm=OpenAI(), prompt=prompt) Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – prompt (langchain.prompts.base.BasePromptTemplate) – llm (langchain.base_language.BaseLanguageModel) – output_key (str) – output_parser (langchain.schema.BaseLLMOutputParser) – return_final_only (bool) – llm_kwargs (dict) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute llm: BaseLanguageModel [Required] Language model to call. attribute llm_kwargs: dict [Optional] attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute output_parser: BaseLLMOutputParser [Optional] Output parser to use. Defaults to one that takes the most likely string but does not change it otherwise. attribute prompt: BasePromptTemplate [Required] Prompt object to use. attribute return_final_only: bool = True Whether to return only the final parsed result. Defaults to True. If false, will return a bunch of extra information about the generation. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async aapply(input_list, callbacks=None)[source] Utilize the LLM generate method for speed gains. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async aapply_and_parse(input_list, callbacks=None)[source] Call apply and then parse the results. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type Sequence[Union[str, List[str], Dict[str, str]]] async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] async agenerate(input_list, run_manager=None)[source] Generate LLM result from inputs. Parameters input_list (List[Dict[str, Any]]) – run_manager (Optional[langchain.callbacks.manager.AsyncCallbackManagerForChainRun]) – Return type langchain.schema.LLMResult apply(input_list, callbacks=None)[source] Utilize the LLM generate method for speed gains. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] apply_and_parse(input_list, callbacks=None)[source] Call apply and then parse the results. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type Sequence[Union[str, List[str], Dict[str, str]]] async apredict(callbacks=None, **kwargs)[source] Format prompt with kwargs and pass to LLM. Parameters callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to pass to LLMChain **kwargs – Keys to pass to prompt template. kwargs (Any) – Returns Completion from LLM. Return type str Example completion = llm.predict(adjective="funny") async apredict_and_parse(callbacks=None, **kwargs)[source] Call apredict and then parse the results. Parameters callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type Union[str, List[str], Dict[str, str]] async aprep_prompts(input_list, run_manager=None)[source] Prepare prompts from inputs. Parameters input_list (List[Dict[str, Any]]) – run_manager (Optional[langchain.callbacks.manager.AsyncCallbackManagerForChainRun]) – Return type Tuple[List[langchain.schema.PromptValue], Optional[List[str]]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str create_outputs(llm_result)[source] Create outputs from response. Parameters llm_result (langchain.schema.LLMResult) – Return type List[Dict[str, Any]] dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_string(llm, template)[source] Create LLMChain from LLM and template. Parameters llm (langchain.base_language.BaseLanguageModel) – template (str) – Return type langchain.chains.llm.LLMChain generate(input_list, run_manager=None)[source] Generate LLM result from inputs. Parameters input_list (List[Dict[str, Any]]) – run_manager (Optional[langchain.callbacks.manager.CallbackManagerForChainRun]) – Return type langchain.schema.LLMResult predict(callbacks=None, **kwargs)[source] Format prompt with kwargs and pass to LLM. Parameters callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to pass to LLMChain **kwargs – Keys to pass to prompt template. kwargs (Any) – Returns Completion from LLM. Return type str Example completion = llm.predict(adjective="funny") predict_and_parse(callbacks=None, **kwargs)[source] Call predict and then parse the results. Parameters callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type Union[str, List[str], Dict[str, Any]] prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] prep_prompts(input_list, run_manager=None)[source] Prepare prompts from inputs. Parameters input_list (List[Dict[str, Any]]) – run_manager (Optional[langchain.callbacks.manager.CallbackManagerForChainRun]) – Return type Tuple[List[langchain.schema.PromptValue], Optional[List[str]]] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.LLMCheckerChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, question_to_checked_assertions_chain, llm=None, create_draft_answer_prompt=PromptTemplate(input_variables=['question'], output_parser=None, partial_variables={}, template='{question}\n\n', template_format='f-string', validate_template=True), list_assertions_prompt=PromptTemplate(input_variables=['statement'], output_parser=None, partial_variables={}, template='Here is a statement:\n{statement}\nMake a bullet point list of the assumptions you made when producing the above statement.\n\n', template_format='f-string', validate_template=True), check_assertions_prompt=PromptTemplate(input_variables=['assertions'], output_parser=None, partial_variables={}, template='Here is a bullet point list of assertions:\n{assertions}\nFor each assertion, determine whether it is true or false. If it is false, explain why.\n\n', template_format='f-string', validate_template=True), revised_answer_prompt=PromptTemplate(input_variables=['checked_assertions', 'question'], output_parser=None, partial_variables={}, template="{checked_assertions}\n\nQuestion: In light of the above assertions and checks, how would you answer the question '{question}'?\n\nAnswer:", template_format='f-string', validate_template=True), input_key='query', output_key='result')[source] Bases: langchain.chains.base.Chain Chain for question-answering with self-verification. Example from langchain import OpenAI, LLMCheckerChain llm = OpenAI(temperature=0.7) checker_chain = LLMCheckerChain.from_llm(llm) Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – question_to_checked_assertions_chain (langchain.chains.sequential.SequentialChain) – llm (Optional[langchain.base_language.BaseLanguageModel]) – create_draft_answer_prompt (langchain.prompts.prompt.PromptTemplate) – list_assertions_prompt (langchain.prompts.prompt.PromptTemplate) – check_assertions_prompt (langchain.prompts.prompt.PromptTemplate) – revised_answer_prompt (langchain.prompts.prompt.PromptTemplate) – input_key (str) – output_key (str) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute check_assertions_prompt: PromptTemplate = PromptTemplate(input_variables=['assertions'], output_parser=None, partial_variables={}, template='Here is a bullet point list of assertions:\n{assertions}\nFor each assertion, determine whether it is true or false. If it is false, explain why.\n\n', template_format='f-string', validate_template=True) [Deprecated] attribute create_draft_answer_prompt: PromptTemplate = PromptTemplate(input_variables=['question'], output_parser=None, partial_variables={}, template='{question}\n\n', template_format='f-string', validate_template=True) [Deprecated] attribute list_assertions_prompt: PromptTemplate = PromptTemplate(input_variables=['statement'], output_parser=None, partial_variables={}, template='Here is a statement:\n{statement}\nMake a bullet point list of the assumptions you made when producing the above statement.\n\n', template_format='f-string', validate_template=True) [Deprecated] attribute llm: Optional[BaseLanguageModel] = None [Deprecated] LLM wrapper to use. attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute question_to_checked_assertions_chain: SequentialChain [Required] attribute revised_answer_prompt: PromptTemplate = PromptTemplate(input_variables=['checked_assertions', 'question'], output_parser=None, partial_variables={}, template="{checked_assertions}\n\nQuestion: In light of the above assertions and checks, how would you answer the question '{question}'?\n\nAnswer:", template_format='f-string', validate_template=True) [Deprecated] Prompt to use when questioning the documents. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_llm(llm, create_draft_answer_prompt=PromptTemplate(input_variables=['question'], output_parser=None, partial_variables={}, template='{question}\n\n', template_format='f-string', validate_template=True), list_assertions_prompt=PromptTemplate(input_variables=['statement'], output_parser=None, partial_variables={}, template='Here is a statement:\n{statement}\nMake a bullet point list of the assumptions you made when producing the above statement.\n\n', template_format='f-string', validate_template=True), check_assertions_prompt=PromptTemplate(input_variables=['assertions'], output_parser=None, partial_variables={}, template='Here is a bullet point list of assertions:\n{assertions}\nFor each assertion, determine whether it is true or false. If it is false, explain why.\n\n', template_format='f-string', validate_template=True), revised_answer_prompt=PromptTemplate(input_variables=['checked_assertions', 'question'], output_parser=None, partial_variables={}, template="{checked_assertions}\n\nQuestion: In light of the above assertions and checks, how would you answer the question '{question}'?\n\nAnswer:", template_format='f-string', validate_template=True), **kwargs)[source] Parameters llm (langchain.base_language.BaseLanguageModel) – create_draft_answer_prompt (langchain.prompts.prompt.PromptTemplate) – list_assertions_prompt (langchain.prompts.prompt.PromptTemplate) – check_assertions_prompt (langchain.prompts.prompt.PromptTemplate) – revised_answer_prompt (langchain.prompts.prompt.PromptTemplate) – kwargs (Any) – Return type langchain.chains.llm_checker.base.LLMCheckerChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.LLMMathChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, llm_chain, llm=None, prompt=PromptTemplate(input_variables=['question'], output_parser=None, partial_variables={}, template='Translate a math problem into a expression that can be executed using Python\'s numexpr library. Use the output of running this code to answer the question.\n\nQuestion: ${{Question with math problem.}}\n```text\n${{single line mathematical expression that solves the problem}}\n```\n...numexpr.evaluate(text)...\n```output\n${{Output of running the code}}\n```\nAnswer: ${{Answer}}\n\nBegin.\n\nQuestion: What is 37593 * 67?\n```text\n37593 * 67\n```\n...numexpr.evaluate("37593 * 67")...\n```output\n2518731\n```\nAnswer: 2518731\n\nQuestion: 37593^(1/5)\n```text\n37593**(1/5)\n```\n...numexpr.evaluate("37593**(1/5)")...\n```output\n8.222831614237718\n```\nAnswer: 8.222831614237718\n\nQuestion: {question}\n', template_format='f-string', validate_template=True), input_key='question', output_key='answer')[source] Bases: langchain.chains.base.Chain Chain that interprets a prompt and executes python code to do math. Example from langchain import LLMMathChain, OpenAI llm_math = LLMMathChain.from_llm(OpenAI()) Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – llm_chain (langchain.chains.llm.LLMChain) – llm (Optional[langchain.base_language.BaseLanguageModel]) – prompt (langchain.prompts.base.BasePromptTemplate) – input_key (str) – output_key (str) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute llm: Optional[BaseLanguageModel] = None [Deprecated] LLM wrapper to use. attribute llm_chain: LLMChain [Required] attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute prompt: BasePromptTemplate = PromptTemplate(input_variables=['question'], output_parser=None, partial_variables={}, template='Translate a math problem into a expression that can be executed using Python\'s numexpr library. Use the output of running this code to answer the question.\n\nQuestion: ${{Question with math problem.}}\n```text\n${{single line mathematical expression that solves the problem}}\n```\n...numexpr.evaluate(text)...\n```output\n${{Output of running the code}}\n```\nAnswer: ${{Answer}}\n\nBegin.\n\nQuestion: What is 37593 * 67?\n```text\n37593 * 67\n```\n...numexpr.evaluate("37593 * 67")...\n```output\n2518731\n```\nAnswer: 2518731\n\nQuestion: 37593^(1/5)\n```text\n37593**(1/5)\n```\n...numexpr.evaluate("37593**(1/5)")...\n```output\n8.222831614237718\n```\nAnswer: 8.222831614237718\n\nQuestion: {question}\n', template_format='f-string', validate_template=True) [Deprecated] Prompt to use to translate to python if necessary. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_llm(llm, prompt=PromptTemplate(input_variables=['question'], output_parser=None, partial_variables={}, template='Translate a math problem into a expression that can be executed using Python\'s numexpr library. Use the output of running this code to answer the question.\n\nQuestion: ${{Question with math problem.}}\n```text\n${{single line mathematical expression that solves the problem}}\n```\n...numexpr.evaluate(text)...\n```output\n${{Output of running the code}}\n```\nAnswer: ${{Answer}}\n\nBegin.\n\nQuestion: What is 37593 * 67?\n```text\n37593 * 67\n```\n...numexpr.evaluate("37593 * 67")...\n```output\n2518731\n```\nAnswer: 2518731\n\nQuestion: 37593^(1/5)\n```text\n37593**(1/5)\n```\n...numexpr.evaluate("37593**(1/5)")...\n```output\n8.222831614237718\n```\nAnswer: 8.222831614237718\n\nQuestion: {question}\n', template_format='f-string', validate_template=True), **kwargs)[source] Parameters llm (langchain.base_language.BaseLanguageModel) – prompt (langchain.prompts.base.BasePromptTemplate) – kwargs (Any) – Return type langchain.chains.llm_math.base.LLMMathChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.LLMRequestsChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, llm_chain, requests_wrapper=None, text_length=8000, requests_key='requests_result', input_key='url', output_key='output')[source] Bases: langchain.chains.base.Chain Chain that hits a URL and then uses an LLM to parse results. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – llm_chain (langchain.chains.llm.LLMChain) – requests_wrapper (langchain.requests.TextRequestsWrapper) – text_length (int) – requests_key (str) – input_key (str) – output_key (str) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute llm_chain: LLMChain [Required] attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute requests_wrapper: TextRequestsWrapper [Optional] attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute text_length: int = 8000 attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.LLMRouterChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, llm_chain)[source] Bases: langchain.chains.router.base.RouterChain A router chain that uses an LLM chain to perform routing. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – llm_chain (langchain.chains.llm.LLMChain) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute llm_chain: LLMChain [Required] LLM chain used to perform routing attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async aroute(inputs, callbacks=None) Parameters inputs (Dict[str, Any]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type langchain.chains.router.base.Route async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_llm(llm, prompt, **kwargs)[source] Convenience constructor. Parameters llm (langchain.base_language.BaseLanguageModel) – prompt (langchain.prompts.base.BasePromptTemplate) – kwargs (Any) – Return type langchain.chains.router.llm_router.LLMRouterChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] route(inputs, callbacks=None) Parameters inputs (Dict[str, Any]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type langchain.chains.router.base.Route run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. property output_keys: List[str] Output keys this chain expects. class langchain.chains.LLMSummarizationCheckerChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, sequential_chain, llm=None, create_assertions_prompt=PromptTemplate(input_variables=['summary'], output_parser=None, partial_variables={}, template='Given some text, extract a list of facts from the text.\n\nFormat your output as a bulleted list.\n\nText:\n"""\n{summary}\n"""\n\nFacts:', template_format='f-string', validate_template=True), check_assertions_prompt=PromptTemplate(input_variables=['assertions'], output_parser=None, partial_variables={}, template='You are an expert fact checker. You have been hired by a major news organization to fact check a very important story.\n\nHere is a bullet point list of facts:\n"""\n{assertions}\n"""\n\nFor each fact, determine whether it is true or false about the subject. If you are unable to determine whether the fact is true or false, output "Undetermined".\nIf the fact is false, explain why.\n\n', template_format='f-string', validate_template=True), revised_summary_prompt=PromptTemplate(input_variables=['checked_assertions', 'summary'], output_parser=None, partial_variables={}, template='Below are some assertions that have been fact checked and are labeled as true or false. If the answer is false, a suggestion is given for a correction.\n\nChecked Assertions:\n"""\n{checked_assertions}\n"""\n\nOriginal Summary:\n"""\n{summary}\n"""\n\nUsing these checked assertions, rewrite the original summary to be completely true.\n\nThe output should have the same structure and formatting as the original summary.\n\nSummary:', template_format='f-string', validate_template=True), are_all_true_prompt=PromptTemplate(input_variables=['checked_assertions'], output_parser=None, partial_variables={}, template='Below are some assertions that have been fact checked and are labeled as true or false.\n\nIf all of the assertions are true, return "True". If any of the assertions are false, return "False".\n\nHere are some examples:\n===\n\nChecked Assertions: """\n- The sky is red: False\n- Water is made of lava: False\n- The sun is a star: True\n"""\nResult: False\n\n===\n\nChecked Assertions: """\n- The sky is blue: True\n- Water is wet: True\n- The sun is a star: True\n"""\nResult: True\n\n===\n\nChecked Assertions: """\n- The sky is blue - True\n- Water is made of lava- False\n- The sun is a star - True\n"""\nResult: False\n\n===\n\nChecked Assertions:"""\n{checked_assertions}\n"""\nResult:', template_format='f-string', validate_template=True), input_key='query', output_key='result', max_checks=2)[source] Bases: langchain.chains.base.Chain Chain for question-answering with self-verification. Example from langchain import OpenAI, LLMSummarizationCheckerChain llm = OpenAI(temperature=0.0) checker_chain = LLMSummarizationCheckerChain.from_llm(llm) Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – sequential_chain (langchain.chains.sequential.SequentialChain) – llm (Optional[langchain.base_language.BaseLanguageModel]) – create_assertions_prompt (langchain.prompts.prompt.PromptTemplate) – check_assertions_prompt (langchain.prompts.prompt.PromptTemplate) – revised_summary_prompt (langchain.prompts.prompt.PromptTemplate) – are_all_true_prompt (langchain.prompts.prompt.PromptTemplate) – input_key (str) – output_key (str) – max_checks (int) – Return type None attribute are_all_true_prompt: PromptTemplate = PromptTemplate(input_variables=['checked_assertions'], output_parser=None, partial_variables={}, template='Below are some assertions that have been fact checked and are labeled as true or false.\n\nIf all of the assertions are true, return "True". If any of the assertions are false, return "False".\n\nHere are some examples:\n===\n\nChecked Assertions: """\n- The sky is red: False\n- Water is made of lava: False\n- The sun is a star: True\n"""\nResult: False\n\n===\n\nChecked Assertions: """\n- The sky is blue: True\n- Water is wet: True\n- The sun is a star: True\n"""\nResult: True\n\n===\n\nChecked Assertions: """\n- The sky is blue - True\n- Water is made of lava- False\n- The sun is a star - True\n"""\nResult: False\n\n===\n\nChecked Assertions:"""\n{checked_assertions}\n"""\nResult:', template_format='f-string', validate_template=True) [Deprecated] attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute check_assertions_prompt: PromptTemplate = PromptTemplate(input_variables=['assertions'], output_parser=None, partial_variables={}, template='You are an expert fact checker. You have been hired by a major news organization to fact check a very important story.\n\nHere is a bullet point list of facts:\n"""\n{assertions}\n"""\n\nFor each fact, determine whether it is true or false about the subject. If you are unable to determine whether the fact is true or false, output "Undetermined".\nIf the fact is false, explain why.\n\n', template_format='f-string', validate_template=True) [Deprecated] attribute create_assertions_prompt: PromptTemplate = PromptTemplate(input_variables=['summary'], output_parser=None, partial_variables={}, template='Given some text, extract a list of facts from the text.\n\nFormat your output as a bulleted list.\n\nText:\n"""\n{summary}\n"""\n\nFacts:', template_format='f-string', validate_template=True) [Deprecated] attribute llm: Optional[BaseLanguageModel] = None [Deprecated] LLM wrapper to use. attribute max_checks: int = 2 Maximum number of times to check the assertions. Default to double-checking. attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute revised_summary_prompt: PromptTemplate = PromptTemplate(input_variables=['checked_assertions', 'summary'], output_parser=None, partial_variables={}, template='Below are some assertions that have been fact checked and are labeled as true or false. If the answer is false, a suggestion is given for a correction.\n\nChecked Assertions:\n"""\n{checked_assertions}\n"""\n\nOriginal Summary:\n"""\n{summary}\n"""\n\nUsing these checked assertions, rewrite the original summary to be completely true.\n\nThe output should have the same structure and formatting as the original summary.\n\nSummary:', template_format='f-string', validate_template=True) [Deprecated] attribute sequential_chain: SequentialChain [Required] attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_llm(llm, create_assertions_prompt=PromptTemplate(input_variables=['summary'], output_parser=None, partial_variables={}, template='Given some text, extract a list of facts from the text.\n\nFormat your output as a bulleted list.\n\nText:\n"""\n{summary}\n"""\n\nFacts:', template_format='f-string', validate_template=True), check_assertions_prompt=PromptTemplate(input_variables=['assertions'], output_parser=None, partial_variables={}, template='You are an expert fact checker. You have been hired by a major news organization to fact check a very important story.\n\nHere is a bullet point list of facts:\n"""\n{assertions}\n"""\n\nFor each fact, determine whether it is true or false about the subject. If you are unable to determine whether the fact is true or false, output "Undetermined".\nIf the fact is false, explain why.\n\n', template_format='f-string', validate_template=True), revised_summary_prompt=PromptTemplate(input_variables=['checked_assertions', 'summary'], output_parser=None, partial_variables={}, template='Below are some assertions that have been fact checked and are labeled as true or false. If the answer is false, a suggestion is given for a correction.\n\nChecked Assertions:\n"""\n{checked_assertions}\n"""\n\nOriginal Summary:\n"""\n{summary}\n"""\n\nUsing these checked assertions, rewrite the original summary to be completely true.\n\nThe output should have the same structure and formatting as the original summary.\n\nSummary:', template_format='f-string', validate_template=True), are_all_true_prompt=PromptTemplate(input_variables=['checked_assertions'], output_parser=None, partial_variables={}, template='Below are some assertions that have been fact checked and are labeled as true or false.\n\nIf all of the assertions are true, return "True". If any of the assertions are false, return "False".\n\nHere are some examples:\n===\n\nChecked Assertions: """\n- The sky is red: False\n- Water is made of lava: False\n- The sun is a star: True\n"""\nResult: False\n\n===\n\nChecked Assertions: """\n- The sky is blue: True\n- Water is wet: True\n- The sun is a star: True\n"""\nResult: True\n\n===\n\nChecked Assertions: """\n- The sky is blue - True\n- Water is made of lava- False\n- The sun is a star - True\n"""\nResult: False\n\n===\n\nChecked Assertions:"""\n{checked_assertions}\n"""\nResult:', template_format='f-string', validate_template=True), verbose=False, **kwargs)[source] Parameters llm (langchain.base_language.BaseLanguageModel) – create_assertions_prompt (langchain.prompts.prompt.PromptTemplate) – check_assertions_prompt (langchain.prompts.prompt.PromptTemplate) – revised_summary_prompt (langchain.prompts.prompt.PromptTemplate) – are_all_true_prompt (langchain.prompts.prompt.PromptTemplate) – verbose (bool) – kwargs (Any) – Return type langchain.chains.llm_summarization_checker.base.LLMSummarizationCheckerChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.MapReduceChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, combine_documents_chain, text_splitter, input_key='input_text', output_key='output_text')[source] Bases: langchain.chains.base.Chain Map-reduce chain. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – combine_documents_chain (langchain.chains.combine_documents.base.BaseCombineDocumentsChain) – text_splitter (langchain.text_splitter.TextSplitter) – input_key (str) – output_key (str) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute combine_documents_chain: BaseCombineDocumentsChain [Required] Chain to use to combine documents. attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute text_splitter: TextSplitter [Required] Text splitter to use. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_params(llm, prompt, text_splitter, callbacks=None, combine_chain_kwargs=None, reduce_chain_kwargs=None, **kwargs)[source] Construct a map-reduce chain that uses the chain for map and reduce. Parameters llm (langchain.base_language.BaseLanguageModel) – prompt (langchain.prompts.base.BasePromptTemplate) – text_splitter (langchain.text_splitter.TextSplitter) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – combine_chain_kwargs (Optional[Mapping[str, Any]]) – reduce_chain_kwargs (Optional[Mapping[str, Any]]) – kwargs (Any) – Return type langchain.chains.mapreduce.MapReduceChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.MultiPromptChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, router_chain, destination_chains, default_chain, silent_errors=False)[source] Bases: langchain.chains.router.base.MultiRouteChain A multi-route chain that uses an LLM router chain to choose amongst prompts. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – router_chain (langchain.chains.router.base.RouterChain) – destination_chains (Mapping[str, langchain.chains.llm.LLMChain]) – default_chain (langchain.chains.llm.LLMChain) – silent_errors (bool) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute default_chain: LLMChain [Required] Default chain to use when router doesn’t map input to one of the destinations. attribute destination_chains: Mapping[str, LLMChain] [Required] Map of name to candidate chains that inputs can be routed to. attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute router_chain: RouterChain [Required] Chain for deciding a destination chain and the input to it. attribute silent_errors: bool = False If True, use default_chain when an invalid destination name is provided. Defaults to False. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_prompts(llm, prompt_infos, default_chain=None, **kwargs)[source] Convenience constructor for instantiating from destination prompts. Parameters llm (langchain.base_language.BaseLanguageModel) – prompt_infos (List[Dict[str, str]]) – default_chain (Optional[langchain.chains.llm.LLMChain]) – kwargs (Any) – Return type langchain.chains.router.multi_prompt.MultiPromptChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.MultiRetrievalQAChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, router_chain, destination_chains, default_chain, silent_errors=False)[source] Bases: langchain.chains.router.base.MultiRouteChain A multi-route chain that uses an LLM router chain to choose amongst retrieval qa chains. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – router_chain (langchain.chains.router.llm_router.LLMRouterChain) – destination_chains (Mapping[str, langchain.chains.retrieval_qa.base.BaseRetrievalQA]) – default_chain (langchain.chains.base.Chain) – silent_errors (bool) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute default_chain: Chain [Required] Default chain to use when router doesn’t map input to one of the destinations. attribute destination_chains: Mapping[str, BaseRetrievalQA] [Required] Map of name to candidate chains that inputs can be routed to. attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute router_chain: LLMRouterChain [Required] Chain for deciding a destination chain and the input to it. attribute silent_errors: bool = False If True, use default_chain when an invalid destination name is provided. Defaults to False. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_retrievers(llm, retriever_infos, default_retriever=None, default_prompt=None, default_chain=None, **kwargs)[source] Parameters llm (langchain.base_language.BaseLanguageModel) – retriever_infos (List[Dict[str, Any]]) – default_retriever (Optional[langchain.schema.BaseRetriever]) – default_prompt (Optional[langchain.prompts.prompt.PromptTemplate]) – default_chain (Optional[langchain.chains.base.Chain]) – kwargs (Any) – Return type langchain.chains.router.multi_retrieval_qa.MultiRetrievalQAChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.MultiRouteChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, router_chain, destination_chains, default_chain, silent_errors=False)[source] Bases: langchain.chains.base.Chain Use a single chain to route an input to one of multiple candidate chains. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – router_chain (langchain.chains.router.base.RouterChain) – destination_chains (Mapping[str, langchain.chains.base.Chain]) – default_chain (langchain.chains.base.Chain) – silent_errors (bool) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute default_chain: Chain [Required] Default chain to use when none of the destination chains are suitable. attribute destination_chains: Mapping[str, Chain] [Required] Chains that return final answer to inputs. attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute router_chain: RouterChain [Required] Chain that routes inputs to destination chains. attribute silent_errors: bool = False If True, use default_chain when an invalid destination name is provided. Defaults to False. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.NatBotChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, llm_chain, objective, llm=None, input_url_key='url', input_browser_content_key='browser_content', previous_command='', output_key='command')[source] Bases: langchain.chains.base.Chain Implement an LLM driven browser. Example from langchain import NatBotChain natbot = NatBotChain.from_default("Buy me a new hat.") Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – llm_chain (langchain.chains.llm.LLMChain) – objective (str) – llm (Optional[langchain.base_language.BaseLanguageModel]) – input_url_key (str) – input_browser_content_key (str) – previous_command (str) – output_key (str) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute llm: Optional[BaseLanguageModel] = None [Deprecated] LLM wrapper to use. attribute llm_chain: LLMChain [Required] attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute objective: str [Required] Objective that NatBot is tasked with completing. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict execute(url, browser_content)[source] Figure out next browser command to run. Parameters url (str) – URL of the site currently on. browser_content (str) – Content of the page as currently displayed by the browser. Returns Next browser command to run. Return type str Example browser_content = "...." llm_command = natbot.run("www.google.com", browser_content) classmethod from_default(objective, **kwargs)[source] Load with default LLMChain. Parameters objective (str) – kwargs (Any) – Return type langchain.chains.natbot.base.NatBotChain classmethod from_llm(llm, objective, **kwargs)[source] Load from LLM. Parameters llm (langchain.base_language.BaseLanguageModel) – objective (str) – kwargs (Any) – Return type langchain.chains.natbot.base.NatBotChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.NebulaGraphQAChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, graph, ngql_generation_chain, qa_chain, input_key='query', output_key='result')[source] Bases: langchain.chains.base.Chain Chain for question-answering against a graph by generating nGQL statements. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – graph (langchain.graphs.nebula_graph.NebulaGraph) – ngql_generation_chain (langchain.chains.llm.LLMChain) – qa_chain (langchain.chains.llm.LLMChain) – input_key (str) – output_key (str) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute graph: NebulaGraph [Required] attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute ngql_generation_chain: LLMChain [Required] attribute qa_chain: LLMChain [Required] attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_llm(llm, *, qa_prompt=PromptTemplate(input_variables=['context', 'question'], output_parser=None, partial_variables={}, template="You are an assistant that helps to form nice and human understandable answers.\nThe information part contains the provided information that you must use to construct an answer.\nThe provided information is authorative, you must never doubt it or try to use your internal knowledge to correct it.\nMake the answer sound as a response to the question. Do not mention that you based the result on the given information.\nIf the provided information is empty, say that you don't know the answer.\nInformation:\n{context}\n\nQuestion: {question}\nHelpful Answer:", template_format='f-string', validate_template=True), ngql_prompt=PromptTemplate(input_variables=['schema', 'question'], output_parser=None, partial_variables={}, template="Task:Generate NebulaGraph Cypher statement to query a graph database.\n\nInstructions:\n\nFirst, generate cypher then convert it to NebulaGraph Cypher dialect(rather than standard):\n1. it requires explicit label specification when referring to node properties: v.`Foo`.name\n2. it uses double equals sign for comparison: `==` rather than `=`\nFor instance:\n```diff\n< MATCH (p:person)-[:directed]->(m:movie) WHERE m.name = 'The Godfather II'\n< RETURN p.name;\n---\n> MATCH (p:`person`)-[:directed]->(m:`movie`) WHERE m.`movie`.`name` == 'The Godfather II'\n> RETURN p.`person`.`name`;\n```\n\nUse only the provided relationship types and properties in the schema.\nDo not use any other relationship types or properties that are not provided.\nSchema:\n{schema}\nNote: Do not include any explanations or apologies in your responses.\nDo not respond to any questions that might ask anything else than for you to construct a Cypher statement.\nDo not include any text except the generated Cypher statement.\n\nThe question is:\n{question}", template_format='f-string', validate_template=True), **kwargs)[source] Initialize from LLM. Parameters llm (langchain.base_language.BaseLanguageModel) – qa_prompt (langchain.prompts.base.BasePromptTemplate) – ngql_prompt (langchain.prompts.base.BasePromptTemplate) – kwargs (Any) – Return type langchain.chains.graph_qa.nebulagraph.NebulaGraphQAChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.OpenAIModerationChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, client=None, model_name=None, error=False, input_key='input', output_key='output', openai_api_key=None, openai_organization=None)[source] Bases: langchain.chains.base.Chain Pass input through a moderation endpoint. To use, you should have the openai python package installed, and the environment variable OPENAI_API_KEY set with your API key. Any parameters that are valid to be passed to the openai.create call can be passed in, even if not explicitly saved on this class. Example from langchain.chains import OpenAIModerationChain moderation = OpenAIModerationChain() Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – client (Any) – model_name (Optional[str]) – error (bool) – input_key (str) – output_key (str) – openai_api_key (Optional[str]) – openai_organization (Optional[str]) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute error: bool = False Whether or not to error if bad content was found. attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute model_name: Optional[str] = None Moderation model name to use. attribute openai_api_key: Optional[str] = None attribute openai_organization: Optional[str] = None attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.OpenAPIEndpointChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, api_request_chain, api_response_chain=None, api_operation, requests=None, param_mapping, return_intermediate_steps=False, instructions_key='instructions', output_key='output', max_text_length=None)[source] Bases: langchain.chains.base.Chain, pydantic.main.BaseModel Chain interacts with an OpenAPI endpoint using natural language. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – api_request_chain (langchain.chains.llm.LLMChain) – api_response_chain (Optional[langchain.chains.llm.LLMChain]) – api_operation (langchain.tools.openapi.utils.api_models.APIOperation) – requests (langchain.requests.Requests) – param_mapping (langchain.chains.api.openapi.chain._ParamMapping) – return_intermediate_steps (bool) – instructions_key (str) – output_key (str) – max_text_length (Optional[int]) – Return type None attribute api_operation: APIOperation [Required] attribute api_request_chain: LLMChain [Required] attribute api_response_chain: Optional[LLMChain] = None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute param_mapping: _ParamMapping [Required] attribute requests: Requests [Optional] attribute return_intermediate_steps: bool = False attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str deserialize_json_input(serialized_args)[source] Use the serialized typescript dictionary. Resolve the path, query params dict, and optional requestBody dict. Parameters serialized_args (str) – Return type dict dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_api_operation(operation, llm, requests=None, verbose=False, return_intermediate_steps=False, raw_response=False, callbacks=None, **kwargs)[source] Create an OpenAPIEndpointChain from an operation and a spec. Parameters operation (langchain.tools.openapi.utils.api_models.APIOperation) – llm (langchain.base_language.BaseLanguageModel) – requests (Optional[langchain.requests.Requests]) – verbose (bool) – return_intermediate_steps (bool) – raw_response (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type langchain.chains.api.openapi.chain.OpenAPIEndpointChain classmethod from_url_and_method(spec_url, path, method, llm, requests=None, return_intermediate_steps=False, **kwargs)[source] Create an OpenAPIEndpoint from a spec at the specified url. Parameters spec_url (str) – path (str) – method (str) – llm (langchain.base_language.BaseLanguageModel) – requests (Optional[langchain.requests.Requests]) – return_intermediate_steps (bool) – kwargs (Any) – Return type langchain.chains.api.openapi.chain.OpenAPIEndpointChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.PALChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, llm_chain, llm=None, prompt=PromptTemplate(input_variables=['question'], output_parser=None, partial_variables={}, template='Q: Olivia has $23. She bought five bagels for $3 each. How much money does she have left?\n\n# solution in Python:\n\n\ndef solution():\nΒ Β Β  """Olivia has $23. She bought five bagels for $3 each. How much money does she have left?"""\nΒ Β Β  money_initial = 23\nΒ Β Β  bagels = 5\nΒ Β Β  bagel_cost = 3\nΒ Β Β  money_spent = bagels * bagel_cost\nΒ Β Β  money_left = money_initial - money_spent\nΒ Β Β  result = money_left\nΒ Β Β  return result\n\n\n\n\n\nQ: Michael had 58 golf balls. On tuesday, he lost 23 golf balls. On wednesday, he lost 2 more. How many golf balls did he have at the end of wednesday?\n\n# solution in Python:\n\n\ndef solution():\nΒ Β Β  """Michael had 58 golf balls. On tuesday, he lost 23 golf balls. On wednesday, he lost 2 more. How many golf balls did he have at the end of wednesday?"""\nΒ Β Β  golf_balls_initial = 58\nΒ Β Β  golf_balls_lost_tuesday = 23\nΒ Β Β  golf_balls_lost_wednesday = 2\nΒ Β Β  golf_balls_left = golf_balls_initial - golf_balls_lost_tuesday - golf_balls_lost_wednesday\nΒ Β Β  result = golf_balls_left\nΒ Β Β  return result\n\n\n\n\n\nQ: There were nine computers in the server room. Five more computers were installed each day, from monday to thursday. How many computers are now in the server room?\n\n# solution in Python:\n\n\ndef solution():\nΒ Β Β  """There were nine computers in the server room. Five more computers were installed each day, from monday to thursday. How many computers are now in the server room?"""\nΒ Β Β  computers_initial = 9\nΒ Β Β  computers_per_day = 5\nΒ Β Β  num_days = 4Β  # 4 days between monday and thursday\nΒ Β Β  computers_added = computers_per_day * num_days\nΒ Β Β  computers_total = computers_initial + computers_added\nΒ Β Β  result = computers_total\nΒ Β Β  return result\n\n\n\n\n\nQ: Shawn has five toys. For Christmas, he got two toys each from his mom and dad. How many toys does he have now?\n\n# solution in Python:\n\n\ndef solution():\nΒ Β Β  """Shawn has five toys. For Christmas, he got two toys each from his mom and dad. How many toys does he have now?"""\nΒ Β Β  toys_initial = 5\nΒ Β Β  mom_toys = 2\nΒ Β Β  dad_toys = 2\nΒ Β Β  total_received = mom_toys + dad_toys\nΒ Β Β  total_toys = toys_initial + total_received\nΒ Β Β  result = total_toys\nΒ Β Β  return result\n\n\n\n\n\nQ: Jason had 20 lollipops. He gave Denny some lollipops. Now Jason has 12 lollipops. How many lollipops did Jason give to Denny?\n\n# solution in Python:\n\n\ndef solution():\nΒ Β Β  """Jason had 20 lollipops. He gave Denny some lollipops. Now Jason has 12 lollipops. How many lollipops did Jason give to Denny?"""\nΒ Β Β  jason_lollipops_initial = 20\nΒ Β Β  jason_lollipops_after = 12\nΒ Β Β  denny_lollipops = jason_lollipops_initial - jason_lollipops_after\nΒ Β Β  result = denny_lollipops\nΒ Β Β  return result\n\n\n\n\n\nQ: Leah had 32 chocolates and her sister had 42. If they ate 35, how many pieces do they have left in total?\n\n# solution in Python:\n\n\ndef solution():\nΒ Β Β  """Leah had 32 chocolates and her sister had 42. If they ate 35, how many pieces do they have left in total?"""\nΒ Β Β  leah_chocolates = 32\nΒ Β Β  sister_chocolates = 42\nΒ Β Β  total_chocolates = leah_chocolates + sister_chocolates\nΒ Β Β  chocolates_eaten = 35\nΒ Β Β  chocolates_left = total_chocolates - chocolates_eaten\nΒ Β Β  result = chocolates_left\nΒ Β Β  return result\n\n\n\n\n\nQ: If there are 3 cars in the parking lot and 2 more cars arrive, how many cars are in the parking lot?\n\n# solution in Python:\n\n\ndef solution():\nΒ Β Β  """If there are 3 cars in the parking lot and 2 more cars arrive, how many cars are in the parking lot?"""\nΒ Β Β  cars_initial = 3\nΒ Β Β  cars_arrived = 2\nΒ Β Β  total_cars = cars_initial + cars_arrived\nΒ Β Β  result = total_cars\nΒ Β Β  return result\n\n\n\n\n\nQ: There are 15 trees in the grove. Grove workers will plant trees in the grove today. After they are done, there will be 21 trees. How many trees did the grove workers plant today?\n\n# solution in Python:\n\n\ndef solution():\nΒ Β Β  """There are 15 trees in the grove. Grove workers will plant trees in the grove today. After they are done, there will be 21 trees. How many trees did the grove workers plant today?"""\nΒ Β Β  trees_initial = 15\nΒ Β Β  trees_after = 21\nΒ Β Β  trees_added = trees_after - trees_initial\nΒ Β Β  result = trees_added\nΒ Β Β  return result\n\n\n\n\n\nQ: {question}\n\n# solution in Python:\n\n\n', template_format='f-string', validate_template=True), stop='\n\n', get_answer_expr='print(solution())', python_globals=None, python_locals=None, output_key='result', return_intermediate_steps=False)[source] Bases: langchain.chains.base.Chain Implements Program-Aided Language Models. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – llm_chain (langchain.chains.llm.LLMChain) – llm (Optional[langchain.base_language.BaseLanguageModel]) – prompt (langchain.prompts.base.BasePromptTemplate) – stop (str) – get_answer_expr (str) – python_globals (Optional[Dict[str, Any]]) – python_locals (Optional[Dict[str, Any]]) – output_key (str) – return_intermediate_steps (bool) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute get_answer_expr: str = 'print(solution())' attribute llm: Optional[BaseLanguageModel] = None [Deprecated] attribute llm_chain: LLMChain [Required] attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute prompt: BasePromptTemplate = PromptTemplate(input_variables=['question'], output_parser=None, partial_variables={}, template='Q: Olivia has $23. She bought five bagels for $3 each. How much money does she have left?\n\n# solution in Python:\n\n\ndef solution():\nΒ Β Β  """Olivia has $23. She bought five bagels for $3 each. How much money does she have left?"""\nΒ Β Β  money_initial = 23\nΒ Β Β  bagels = 5\nΒ Β Β  bagel_cost = 3\nΒ Β Β  money_spent = bagels * bagel_cost\nΒ Β Β  money_left = money_initial - money_spent\nΒ Β Β  result = money_left\nΒ Β Β  return result\n\n\n\n\n\nQ: Michael had 58 golf balls. On tuesday, he lost 23 golf balls. On wednesday, he lost 2 more. How many golf balls did he have at the end of wednesday?\n\n# solution in Python:\n\n\ndef solution():\nΒ Β Β  """Michael had 58 golf balls. On tuesday, he lost 23 golf balls. On wednesday, he lost 2 more. How many golf balls did he have at the end of wednesday?"""\nΒ Β Β  golf_balls_initial = 58\nΒ Β Β  golf_balls_lost_tuesday = 23\nΒ Β Β  golf_balls_lost_wednesday = 2\nΒ Β Β  golf_balls_left = golf_balls_initial - golf_balls_lost_tuesday - golf_balls_lost_wednesday\nΒ Β Β  result = golf_balls_left\nΒ Β Β  return result\n\n\n\n\n\nQ: There were nine computers in the server room. Five more computers were installed each day, from monday to thursday. How many computers are now in the server room?\n\n# solution in Python:\n\n\ndef solution():\nΒ Β Β  """There were nine computers in the server room. Five more computers were installed each day, from monday to thursday. How many computers are now in the server room?"""\nΒ Β Β  computers_initial = 9\nΒ Β Β  computers_per_day = 5\nΒ Β Β  num_days = 4Β  # 4 days between monday and thursday\nΒ Β Β  computers_added = computers_per_day * num_days\nΒ Β Β  computers_total = computers_initial + computers_added\nΒ Β Β  result = computers_total\nΒ Β Β  return result\n\n\n\n\n\nQ: Shawn has five toys. For Christmas, he got two toys each from his mom and dad. How many toys does he have now?\n\n# solution in Python:\n\n\ndef solution():\nΒ Β Β  """Shawn has five toys. For Christmas, he got two toys each from his mom and dad. How many toys does he have now?"""\nΒ Β Β  toys_initial = 5\nΒ Β Β  mom_toys = 2\nΒ Β Β  dad_toys = 2\nΒ Β Β  total_received = mom_toys + dad_toys\nΒ Β Β  total_toys = toys_initial + total_received\nΒ Β Β  result = total_toys\nΒ Β Β  return result\n\n\n\n\n\nQ: Jason had 20 lollipops. He gave Denny some lollipops. Now Jason has 12 lollipops. How many lollipops did Jason give to Denny?\n\n# solution in Python:\n\n\ndef solution():\nΒ Β Β  """Jason had 20 lollipops. He gave Denny some lollipops. Now Jason has 12 lollipops. How many lollipops did Jason give to Denny?"""\nΒ Β Β  jason_lollipops_initial = 20\nΒ Β Β  jason_lollipops_after = 12\nΒ Β Β  denny_lollipops = jason_lollipops_initial - jason_lollipops_after\nΒ Β Β  result = denny_lollipops\nΒ Β Β  return result\n\n\n\n\n\nQ: Leah had 32 chocolates and her sister had 42. If they ate 35, how many pieces do they have left in total?\n\n# solution in Python:\n\n\ndef solution():\nΒ Β Β  """Leah had 32 chocolates and her sister had 42. If they ate 35, how many pieces do they have left in total?"""\nΒ Β Β  leah_chocolates = 32\nΒ Β Β  sister_chocolates = 42\nΒ Β Β  total_chocolates = leah_chocolates + sister_chocolates\nΒ Β Β  chocolates_eaten = 35\nΒ Β Β  chocolates_left = total_chocolates - chocolates_eaten\nΒ Β Β  result = chocolates_left\nΒ Β Β  return result\n\n\n\n\n\nQ: If there are 3 cars in the parking lot and 2 more cars arrive, how many cars are in the parking lot?\n\n# solution in Python:\n\n\ndef solution():\nΒ Β Β  """If there are 3 cars in the parking lot and 2 more cars arrive, how many cars are in the parking lot?"""\nΒ Β Β  cars_initial = 3\nΒ Β Β  cars_arrived = 2\nΒ Β Β  total_cars = cars_initial + cars_arrived\nΒ Β Β  result = total_cars\nΒ Β Β  return result\n\n\n\n\n\nQ: There are 15 trees in the grove. Grove workers will plant trees in the grove today. After they are done, there will be 21 trees. How many trees did the grove workers plant today?\n\n# solution in Python:\n\n\ndef solution():\nΒ Β Β  """There are 15 trees in the grove. Grove workers will plant trees in the grove today. After they are done, there will be 21 trees. How many trees did the grove workers plant today?"""\nΒ Β Β  trees_initial = 15\nΒ Β Β  trees_after = 21\nΒ Β Β  trees_added = trees_after - trees_initial\nΒ Β Β  result = trees_added\nΒ Β Β  return result\n\n\n\n\n\nQ: {question}\n\n# solution in Python:\n\n\n', template_format='f-string', validate_template=True) [Deprecated] attribute python_globals: Optional[Dict[str, Any]] = None attribute python_locals: Optional[Dict[str, Any]] = None attribute return_intermediate_steps: bool = False attribute stop: str = '\n\n' attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_colored_object_prompt(llm, **kwargs)[source] Load PAL from colored object prompt. Parameters llm (langchain.base_language.BaseLanguageModel) – kwargs (Any) – Return type langchain.chains.pal.base.PALChain classmethod from_math_prompt(llm, **kwargs)[source] Load PAL from math prompt. Parameters llm (langchain.base_language.BaseLanguageModel) – kwargs (Any) – Return type langchain.chains.pal.base.PALChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.QAGenerationChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, llm_chain, text_splitter=<langchain.text_splitter.RecursiveCharacterTextSplitter object>, input_key='text', output_key='questions', k=None)[source] Bases: langchain.chains.base.Chain Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – llm_chain (langchain.chains.llm.LLMChain) – text_splitter (langchain.text_splitter.TextSplitter) – input_key (str) – output_key (str) – k (Optional[int]) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute input_key: str = 'text' attribute k: Optional[int] = None attribute llm_chain: LLMChain [Required] attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute output_key: str = 'questions' attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute text_splitter: TextSplitter = <langchain.text_splitter.RecursiveCharacterTextSplitter object> attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_llm(llm, prompt=None, **kwargs)[source] Parameters llm (langchain.base_language.BaseLanguageModel) – prompt (Optional[langchain.prompts.base.BasePromptTemplate]) – kwargs (Any) – Return type langchain.chains.qa_generation.base.QAGenerationChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property input_keys: List[str] Input keys this chain expects. property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. property output_keys: List[str] Output keys this chain expects. class langchain.chains.QAWithSourcesChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, combine_documents_chain, question_key='question', input_docs_key='docs', answer_key='answer', sources_answer_key='sources', return_source_documents=False)[source] Bases: langchain.chains.qa_with_sources.base.BaseQAWithSourcesChain Question answering with sources over documents. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – combine_documents_chain (langchain.chains.combine_documents.base.BaseCombineDocumentsChain) – question_key (str) – input_docs_key (str) – answer_key (str) – sources_answer_key (str) – return_source_documents (bool) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute combine_documents_chain: BaseCombineDocumentsChain [Required] Chain to use to combine documents. attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute return_source_documents: bool = False Return the source documents. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_chain_type(llm, chain_type='stuff', chain_type_kwargs=None, **kwargs) Load chain from chain type. Parameters llm (langchain.base_language.BaseLanguageModel) – chain_type (str) – chain_type_kwargs (Optional[dict]) – kwargs (Any) – Return type langchain.chains.qa_with_sources.base.BaseQAWithSourcesChain classmethod from_llm(llm, document_prompt=PromptTemplate(input_variables=['page_content', 'source'], output_parser=None, partial_variables={}, template='Content: {page_content}\nSource: {source}', template_format='f-string', validate_template=True), question_prompt=PromptTemplate(input_variables=['context', 'question'], output_parser=None, partial_variables={}, template='Use the following portion of a long document to see if any of the text is relevant to answer the question. \nReturn any relevant text verbatim.\n{context}\nQuestion: {question}\nRelevant text, if any:', template_format='f-string', validate_template=True), combine_prompt=PromptTemplate(input_variables=['summaries', 'question'], output_parser=None, partial_variables={}, template='Given the following extracted parts of a long document and a question, create a final answer with references ("SOURCES"). \nIf you don\'t know the answer, just say that you don\'t know. Don\'t try to make up an answer.\nALWAYS return a "SOURCES" part in your answer.\n\nQUESTION: Which state/country\'s law governs the interpretation of the contract?\n=========\nContent: This Agreement is governed by English law and the parties submit to the exclusive jurisdiction of the English courts inΒ  relation to any dispute (contractual or non-contractual) concerning this Agreement save that either party may apply to any court for anΒ  injunction or other relief to protect its Intellectual Property Rights.\nSource: 28-pl\nContent: No Waiver. Failure or delay in exercising any right or remedy under this Agreement shall not constitute a waiver of such (or any other)Β  right or remedy.\n\n11.7 Severability. The invalidity, illegality or unenforceability of any term (or part of a term) of this Agreement shall not affect the continuationΒ  in force of the remainder of the term (if any) and this Agreement.\n\n11.8 No Agency. Except as expressly stated otherwise, nothing in this Agreement shall create an agency, partnership or joint venture of anyΒ  kind between the parties.\n\n11.9 No Third-Party Beneficiaries.\nSource: 30-pl\nContent: (b) if Google believes, in good faith, that the Distributor has violated or caused Google to violate any Anti-Bribery Laws (asΒ  defined in Clause 8.5) or that such a violation is reasonably likely to occur,\nSource: 4-pl\n=========\nFINAL ANSWER: This Agreement is governed by English law.\nSOURCES: 28-pl\n\nQUESTION: What did the president say about Michael Jackson?\n=========\nContent: Madam Speaker, Madam Vice President, our First Lady and Second Gentleman. Members of Congress and the Cabinet. Justices of the Supreme Court. My fellow Americans.Β  \n\nLast year COVID-19 kept us apart. This year we are finally together again. \n\nTonight, we meet as Democrats Republicans and Independents. But most importantly as Americans. \n\nWith a duty to one another to the American people to the Constitution. \n\nAnd with an unwavering resolve that freedom will always triumph over tyranny. \n\nSix days ago, Russia’s Vladimir Putin sought to shake the foundations of the free world thinking he could make it bend to his menacing ways. But he badly miscalculated. \n\nHe thought he could roll into Ukraine and the world would roll over. Instead he met a wall of strength he never imagined. \n\nHe met the Ukrainian people. \n\nFrom President Zelenskyy to every Ukrainian, their fearlessness, their courage, their determination, inspires the world. \n\nGroups of citizens blocking tanks with their bodies. Everyone from students to retirees teachers turned soldiers defending their homeland.\nSource: 0-pl\nContent: And we won’t stop. \n\nWe have lost so much to COVID-19. Time with one another. And worst of all, so much loss of life. \n\nLet’s use this moment to reset. Let’s stop looking at COVID-19 as a partisan dividing line and see it for what it is: A God-awful disease.Β  \n\nLet’s stop seeing each other as enemies, and start seeing each other for who we really are: Fellow Americans.Β  \n\nWe can’t change how divided we’ve been. But we can change how we move forwardβ€”on COVID-19 and other issues we must face together. \n\nI recently visited the New York City Police Department days after the funerals of Officer Wilbert Mora and his partner, Officer Jason Rivera. \n\nThey were responding to a 9-1-1 call when a man shot and killed them with a stolen gun. \n\nOfficer Mora was 27 years old. \n\nOfficer Rivera was 22. \n\nBoth Dominican Americans who’d grown up on the same streets they later chose to patrol as police officers. \n\nI spoke with their families and told them that we are forever in debt for their sacrifice, and we will carry on their mission to restore the trust and safety every community deserves.\nSource: 24-pl\nContent: And a proud Ukrainian people, who have known 30 yearsΒ  of independence, have repeatedly shown that they will not tolerate anyone who tries to take their country backwards.Β  \n\nTo all Americans, I will be honest with you, as I’ve always promised. A Russian dictator, invading a foreign country, has costs around the world. \n\nAnd I’m taking robust action to make sure the pain of our sanctionsΒ  is targeted at Russia’s economy. And I will use every tool at our disposal to protect American businesses and consumers. \n\nTonight, I can announce that the United States has worked with 30 other countries to release 60 Million barrels of oil from reserves around the world.Β  \n\nAmerica will lead that effort, releasing 30 Million barrels from our own Strategic Petroleum Reserve. And we stand ready to do more if necessary, unified with our allies.Β  \n\nThese steps will help blunt gas prices here at home. And I know the news about what’s happening can seem alarming. \n\nBut I want you to know that we are going to be okay.\nSource: 5-pl\nContent: More support for patients and families. \n\nTo get there, I call on Congress to fund ARPA-H, the Advanced Research Projects Agency for Health. \n\nIt’s based on DARPAβ€”the Defense Department project that led to the Internet, GPS, and so much more.Β  \n\nARPA-H will have a singular purposeβ€”to drive breakthroughs in cancer, Alzheimer’s, diabetes, and more. \n\nA unity agenda for the nation. \n\nWe can do this. \n\nMy fellow Americansβ€”tonight , we have gathered in a sacred spaceβ€”the citadel of our democracy. \n\nIn this Capitol, generation after generation, Americans have debated great questions amid great strife, and have done great things. \n\nWe have fought for freedom, expanded liberty, defeated totalitarianism and terror. \n\nAnd built the strongest, freest, and most prosperous nation the world has ever known. \n\nNow is the hour. \n\nOur moment of responsibility. \n\nOur test of resolve and conscience, of history itself. \n\nIt is in this moment that our character is formed. Our purpose is found. Our future is forged. \n\nWell I know this nation.\nSource: 34-pl\n=========\nFINAL ANSWER: The president did not mention Michael Jackson.\nSOURCES:\n\nQUESTION: {question}\n=========\n{summaries}\n=========\nFINAL ANSWER:', template_format='f-string', validate_template=True), **kwargs) Construct the chain from an LLM. Parameters llm (langchain.base_language.BaseLanguageModel) – document_prompt (langchain.prompts.base.BasePromptTemplate) – question_prompt (langchain.prompts.base.BasePromptTemplate) – combine_prompt (langchain.prompts.base.BasePromptTemplate) – kwargs (Any) – Return type langchain.chains.qa_with_sources.base.BaseQAWithSourcesChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.RetrievalQA(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, combine_documents_chain, input_key='query', output_key='result', return_source_documents=False, retriever)[source] Bases: langchain.chains.retrieval_qa.base.BaseRetrievalQA Chain for question-answering against an index. Example from langchain.llms import OpenAI from langchain.chains import RetrievalQA from langchain.faiss import FAISS from langchain.vectorstores.base import VectorStoreRetriever retriever = VectorStoreRetriever(vectorstore=FAISS(...)) retrievalQA = RetrievalQA.from_llm(llm=OpenAI(), retriever=retriever) Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – combine_documents_chain (langchain.chains.combine_documents.base.BaseCombineDocumentsChain) – input_key (str) – output_key (str) – return_source_documents (bool) – retriever (langchain.schema.BaseRetriever) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute combine_documents_chain: BaseCombineDocumentsChain [Required] Chain to use to combine the documents. attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute retriever: BaseRetriever [Required] attribute return_source_documents: bool = False Return the source documents. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_chain_type(llm, chain_type='stuff', chain_type_kwargs=None, **kwargs) Load chain from chain type. Parameters llm (langchain.base_language.BaseLanguageModel) – chain_type (str) – chain_type_kwargs (Optional[dict]) – kwargs (Any) – Return type langchain.chains.retrieval_qa.base.BaseRetrievalQA classmethod from_llm(llm, prompt=None, **kwargs) Initialize from LLM. Parameters llm (langchain.base_language.BaseLanguageModel) – prompt (Optional[langchain.prompts.prompt.PromptTemplate]) – kwargs (Any) – Return type langchain.chains.retrieval_qa.base.BaseRetrievalQA prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.RetrievalQAWithSourcesChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, combine_documents_chain, question_key='question', input_docs_key='docs', answer_key='answer', sources_answer_key='sources', return_source_documents=False, retriever, reduce_k_below_max_tokens=False, max_tokens_limit=3375)[source] Bases: langchain.chains.qa_with_sources.base.BaseQAWithSourcesChain Question-answering with sources over an index. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – combine_documents_chain (langchain.chains.combine_documents.base.BaseCombineDocumentsChain) – question_key (str) – input_docs_key (str) – answer_key (str) – sources_answer_key (str) – return_source_documents (bool) – retriever (langchain.schema.BaseRetriever) – reduce_k_below_max_tokens (bool) – max_tokens_limit (int) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute combine_documents_chain: BaseCombineDocumentsChain [Required] Chain to use to combine documents. attribute max_tokens_limit: int = 3375 Restrict the docs to return from store based on tokens, enforced only for StuffDocumentChain and if reduce_k_below_max_tokens is to true attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute reduce_k_below_max_tokens: bool = False Reduce the number of results to return from store based on tokens limit attribute retriever: langchain.schema.BaseRetriever [Required] Index to connect to. attribute return_source_documents: bool = False Return the source documents. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_chain_type(llm, chain_type='stuff', chain_type_kwargs=None, **kwargs) Load chain from chain type. Parameters llm (langchain.base_language.BaseLanguageModel) – chain_type (str) – chain_type_kwargs (Optional[dict]) – kwargs (Any) – Return type langchain.chains.qa_with_sources.base.BaseQAWithSourcesChain classmethod from_llm(llm, document_prompt=PromptTemplate(input_variables=['page_content', 'source'], output_parser=None, partial_variables={}, template='Content: {page_content}\nSource: {source}', template_format='f-string', validate_template=True), question_prompt=PromptTemplate(input_variables=['context', 'question'], output_parser=None, partial_variables={}, template='Use the following portion of a long document to see if any of the text is relevant to answer the question. \nReturn any relevant text verbatim.\n{context}\nQuestion: {question}\nRelevant text, if any:', template_format='f-string', validate_template=True), combine_prompt=PromptTemplate(input_variables=['summaries', 'question'], output_parser=None, partial_variables={}, template='Given the following extracted parts of a long document and a question, create a final answer with references ("SOURCES"). \nIf you don\'t know the answer, just say that you don\'t know. Don\'t try to make up an answer.\nALWAYS return a "SOURCES" part in your answer.\n\nQUESTION: Which state/country\'s law governs the interpretation of the contract?\n=========\nContent: This Agreement is governed by English law and the parties submit to the exclusive jurisdiction of the English courts inΒ  relation to any dispute (contractual or non-contractual) concerning this Agreement save that either party may apply to any court for anΒ  injunction or other relief to protect its Intellectual Property Rights.\nSource: 28-pl\nContent: No Waiver. Failure or delay in exercising any right or remedy under this Agreement shall not constitute a waiver of such (or any other)Β  right or remedy.\n\n11.7 Severability. The invalidity, illegality or unenforceability of any term (or part of a term) of this Agreement shall not affect the continuationΒ  in force of the remainder of the term (if any) and this Agreement.\n\n11.8 No Agency. Except as expressly stated otherwise, nothing in this Agreement shall create an agency, partnership or joint venture of anyΒ  kind between the parties.\n\n11.9 No Third-Party Beneficiaries.\nSource: 30-pl\nContent: (b) if Google believes, in good faith, that the Distributor has violated or caused Google to violate any Anti-Bribery Laws (asΒ  defined in Clause 8.5) or that such a violation is reasonably likely to occur,\nSource: 4-pl\n=========\nFINAL ANSWER: This Agreement is governed by English law.\nSOURCES: 28-pl\n\nQUESTION: What did the president say about Michael Jackson?\n=========\nContent: Madam Speaker, Madam Vice President, our First Lady and Second Gentleman. Members of Congress and the Cabinet. Justices of the Supreme Court. My fellow Americans.Β  \n\nLast year COVID-19 kept us apart. This year we are finally together again. \n\nTonight, we meet as Democrats Republicans and Independents. But most importantly as Americans. \n\nWith a duty to one another to the American people to the Constitution. \n\nAnd with an unwavering resolve that freedom will always triumph over tyranny. \n\nSix days ago, Russia’s Vladimir Putin sought to shake the foundations of the free world thinking he could make it bend to his menacing ways. But he badly miscalculated. \n\nHe thought he could roll into Ukraine and the world would roll over. Instead he met a wall of strength he never imagined. \n\nHe met the Ukrainian people. \n\nFrom President Zelenskyy to every Ukrainian, their fearlessness, their courage, their determination, inspires the world. \n\nGroups of citizens blocking tanks with their bodies. Everyone from students to retirees teachers turned soldiers defending their homeland.\nSource: 0-pl\nContent: And we won’t stop. \n\nWe have lost so much to COVID-19. Time with one another. And worst of all, so much loss of life. \n\nLet’s use this moment to reset. Let’s stop looking at COVID-19 as a partisan dividing line and see it for what it is: A God-awful disease.Β  \n\nLet’s stop seeing each other as enemies, and start seeing each other for who we really are: Fellow Americans.Β  \n\nWe can’t change how divided we’ve been. But we can change how we move forwardβ€”on COVID-19 and other issues we must face together. \n\nI recently visited the New York City Police Department days after the funerals of Officer Wilbert Mora and his partner, Officer Jason Rivera. \n\nThey were responding to a 9-1-1 call when a man shot and killed them with a stolen gun. \n\nOfficer Mora was 27 years old. \n\nOfficer Rivera was 22. \n\nBoth Dominican Americans who’d grown up on the same streets they later chose to patrol as police officers. \n\nI spoke with their families and told them that we are forever in debt for their sacrifice, and we will carry on their mission to restore the trust and safety every community deserves.\nSource: 24-pl\nContent: And a proud Ukrainian people, who have known 30 yearsΒ  of independence, have repeatedly shown that they will not tolerate anyone who tries to take their country backwards.Β  \n\nTo all Americans, I will be honest with you, as I’ve always promised. A Russian dictator, invading a foreign country, has costs around the world. \n\nAnd I’m taking robust action to make sure the pain of our sanctionsΒ  is targeted at Russia’s economy. And I will use every tool at our disposal to protect American businesses and consumers. \n\nTonight, I can announce that the United States has worked with 30 other countries to release 60 Million barrels of oil from reserves around the world.Β  \n\nAmerica will lead that effort, releasing 30 Million barrels from our own Strategic Petroleum Reserve. And we stand ready to do more if necessary, unified with our allies.Β  \n\nThese steps will help blunt gas prices here at home. And I know the news about what’s happening can seem alarming. \n\nBut I want you to know that we are going to be okay.\nSource: 5-pl\nContent: More support for patients and families. \n\nTo get there, I call on Congress to fund ARPA-H, the Advanced Research Projects Agency for Health. \n\nIt’s based on DARPAβ€”the Defense Department project that led to the Internet, GPS, and so much more.Β  \n\nARPA-H will have a singular purposeβ€”to drive breakthroughs in cancer, Alzheimer’s, diabetes, and more. \n\nA unity agenda for the nation. \n\nWe can do this. \n\nMy fellow Americansβ€”tonight , we have gathered in a sacred spaceβ€”the citadel of our democracy. \n\nIn this Capitol, generation after generation, Americans have debated great questions amid great strife, and have done great things. \n\nWe have fought for freedom, expanded liberty, defeated totalitarianism and terror. \n\nAnd built the strongest, freest, and most prosperous nation the world has ever known. \n\nNow is the hour. \n\nOur moment of responsibility. \n\nOur test of resolve and conscience, of history itself. \n\nIt is in this moment that our character is formed. Our purpose is found. Our future is forged. \n\nWell I know this nation.\nSource: 34-pl\n=========\nFINAL ANSWER: The president did not mention Michael Jackson.\nSOURCES:\n\nQUESTION: {question}\n=========\n{summaries}\n=========\nFINAL ANSWER:', template_format='f-string', validate_template=True), **kwargs) Construct the chain from an LLM. Parameters llm (langchain.base_language.BaseLanguageModel) – document_prompt (langchain.prompts.base.BasePromptTemplate) – question_prompt (langchain.prompts.base.BasePromptTemplate) – combine_prompt (langchain.prompts.base.BasePromptTemplate) – kwargs (Any) – Return type langchain.chains.qa_with_sources.base.BaseQAWithSourcesChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.RouterChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None)[source] Bases: langchain.chains.base.Chain, abc.ABC Chain that outputs the name of a destination chain and the inputs to it. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async aroute(inputs, callbacks=None)[source] Parameters inputs (Dict[str, Any]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type langchain.chains.router.base.Route async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] route(inputs, callbacks=None)[source] Parameters inputs (Dict[str, Any]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type langchain.chains.router.base.Route run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented abstract property input_keys: List[str] Input keys this chain expects. property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. property output_keys: List[str] Output keys this chain expects. class langchain.chains.SQLDatabaseChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, llm_chain, llm=None, database, prompt=None, top_k=5, input_key='query', output_key='result', return_intermediate_steps=False, return_direct=False, use_query_checker=False, query_checker_prompt=None)[source] Bases: langchain.chains.base.Chain Chain for interacting with SQL Database. Example from langchain import SQLDatabaseChain, OpenAI, SQLDatabase db = SQLDatabase(...) db_chain = SQLDatabaseChain.from_llm(OpenAI(), db) Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – llm_chain (langchain.chains.llm.LLMChain) – llm (Optional[langchain.base_language.BaseLanguageModel]) – database (langchain.sql_database.SQLDatabase) – prompt (Optional[langchain.prompts.base.BasePromptTemplate]) – top_k (int) – input_key (str) – output_key (str) – return_intermediate_steps (bool) – return_direct (bool) – use_query_checker (bool) – query_checker_prompt (Optional[langchain.prompts.base.BasePromptTemplate]) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute database: SQLDatabase [Required] SQL Database to connect to. attribute llm: Optional[BaseLanguageModel] = None [Deprecated] LLM wrapper to use. attribute llm_chain: LLMChain [Required] attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute prompt: Optional[BasePromptTemplate] = None [Deprecated] Prompt to use to translate natural language to SQL. attribute query_checker_prompt: Optional[BasePromptTemplate] = None The prompt template that should be used by the query checker attribute return_direct: bool = False Whether or not to return the result of querying the SQL table directly. attribute return_intermediate_steps: bool = False Whether or not to return the intermediate steps along with the final answer. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute top_k: int = 5 Number of results to return from the query attribute use_query_checker: bool = False Whether or not the query checker tool should be used to attempt to fix the initial SQL from the LLM. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_llm(llm, db, prompt=None, **kwargs)[source] Parameters llm (langchain.base_language.BaseLanguageModel) – db (langchain.sql_database.SQLDatabase) – prompt (Optional[langchain.prompts.base.BasePromptTemplate]) – kwargs (Any) – Return type langchain.chains.sql_database.base.SQLDatabaseChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.SQLDatabaseSequentialChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, decider_chain, sql_chain, input_key='query', output_key='result', return_intermediate_steps=False)[source] Bases: langchain.chains.base.Chain Chain for querying SQL database that is a sequential chain. The chain is as follows: 1. Based on the query, determine which tables to use. 2. Based on those tables, call the normal SQL database chain. This is useful in cases where the number of tables in the database is large. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – decider_chain (langchain.chains.llm.LLMChain) – sql_chain (langchain.chains.sql_database.base.SQLDatabaseChain) – input_key (str) – output_key (str) – return_intermediate_steps (bool) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute decider_chain: LLMChain [Required] attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute return_intermediate_steps: bool = False attribute sql_chain: SQLDatabaseChain [Required] attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_llm(llm, database, query_prompt=PromptTemplate(input_variables=['input', 'table_info', 'dialect', 'top_k'], output_parser=None, partial_variables={}, template='Given an input question, first create a syntactically correct {dialect} query to run, then look at the results of the query and return the answer. Unless the user specifies in his question a specific number of examples he wishes to obtain, always limit your query to at most {top_k} results. You can order the results by a relevant column to return the most interesting examples in the database.\n\nNever query for all the columns from a specific table, only ask for a the few relevant columns given the question.\n\nPay attention to use only the column names that you can see in the schema description. Be careful to not query for columns that do not exist. Also, pay attention to which column is in which table.\n\nUse the following format:\n\nQuestion: Question here\nSQLQuery: SQL Query to run\nSQLResult: Result of the SQLQuery\nAnswer: Final answer here\n\nOnly use the following tables:\n{table_info}\n\nQuestion: {input}', template_format='f-string', validate_template=True), decider_prompt=PromptTemplate(input_variables=['query', 'table_names'], output_parser=CommaSeparatedListOutputParser(), partial_variables={}, template='Given the below input question and list of potential tables, output a comma separated list of the table names that may be necessary to answer this question.\n\nQuestion: {query}\n\nTable Names: {table_names}\n\nRelevant Table Names:', template_format='f-string', validate_template=True), **kwargs)[source] Load the necessary chains. Parameters llm (langchain.base_language.BaseLanguageModel) – database (langchain.sql_database.SQLDatabase) – query_prompt (langchain.prompts.base.BasePromptTemplate) – decider_prompt (langchain.prompts.base.BasePromptTemplate) – kwargs (Any) – Return type langchain.chains.sql_database.base.SQLDatabaseSequentialChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.SequentialChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, chains, input_variables, output_variables, return_all=False)[source] Bases: langchain.chains.base.Chain Chain where the outputs of one chain feed directly into next. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – chains (List[langchain.chains.base.Chain]) – input_variables (List[str]) – output_variables (List[str]) – return_all (bool) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute chains: List[langchain.chains.base.Chain] [Required] attribute input_variables: List[str] [Required] attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute return_all: bool = False attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.SimpleSequentialChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, chains, strip_outputs=False, input_key='input', output_key='output')[source] Bases: langchain.chains.base.Chain Simple chain where the outputs of one step feed directly into next. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – chains (List[langchain.chains.base.Chain]) – strip_outputs (bool) – input_key (str) – output_key (str) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute chains: List[langchain.chains.base.Chain] [Required] attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute strip_outputs: bool = False attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.TransformChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, input_variables, output_variables, transform)[source] Bases: langchain.chains.base.Chain Chain transform chain output. Example from langchain import TransformChain transform_chain = TransformChain(input_variables=["text"], output_variables["entities"], transform=func()) Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – input_variables (List[str]) – output_variables (List[str]) – transform (Callable[[Dict[str, str]], Dict[str, str]]) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute input_variables: List[str] [Required] attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute output_variables: List[str] [Required] attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute transform: Callable[[Dict[str, str]], Dict[str, str]] [Required] attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.VectorDBQA(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, combine_documents_chain, input_key='query', output_key='result', return_source_documents=False, vectorstore, k=4, search_type='similarity', search_kwargs=None)[source] Bases: langchain.chains.retrieval_qa.base.BaseRetrievalQA Chain for question-answering against a vector database. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – combine_documents_chain (langchain.chains.combine_documents.base.BaseCombineDocumentsChain) – input_key (str) – output_key (str) – return_source_documents (bool) – vectorstore (langchain.vectorstores.base.VectorStore) – k (int) – search_type (str) – search_kwargs (Dict[str, Any]) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute combine_documents_chain: BaseCombineDocumentsChain [Required] Chain to use to combine the documents. attribute k: int = 4 Number of documents to query for. attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute return_source_documents: bool = False Return the source documents. attribute search_kwargs: Dict[str, Any] [Optional] Extra search args. attribute search_type: str = 'similarity' Search type to use over vectorstore. similarity or mmr. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute vectorstore: VectorStore [Required] Vector Database to connect to. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_chain_type(llm, chain_type='stuff', chain_type_kwargs=None, **kwargs) Load chain from chain type. Parameters llm (langchain.base_language.BaseLanguageModel) – chain_type (str) – chain_type_kwargs (Optional[dict]) – kwargs (Any) – Return type langchain.chains.retrieval_qa.base.BaseRetrievalQA classmethod from_llm(llm, prompt=None, **kwargs) Initialize from LLM. Parameters llm (langchain.base_language.BaseLanguageModel) – prompt (Optional[langchain.prompts.prompt.PromptTemplate]) – kwargs (Any) – Return type langchain.chains.retrieval_qa.base.BaseRetrievalQA prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.VectorDBQAWithSourcesChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, combine_documents_chain, question_key='question', input_docs_key='docs', answer_key='answer', sources_answer_key='sources', return_source_documents=False, vectorstore, k=4, reduce_k_below_max_tokens=False, max_tokens_limit=3375, search_kwargs=None)[source] Bases: langchain.chains.qa_with_sources.base.BaseQAWithSourcesChain Question-answering with sources over a vector database. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – combine_documents_chain (langchain.chains.combine_documents.base.BaseCombineDocumentsChain) – question_key (str) – input_docs_key (str) – answer_key (str) – sources_answer_key (str) – return_source_documents (bool) – vectorstore (langchain.vectorstores.base.VectorStore) – k (int) – reduce_k_below_max_tokens (bool) – max_tokens_limit (int) – search_kwargs (Dict[str, Any]) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute combine_documents_chain: BaseCombineDocumentsChain [Required] Chain to use to combine documents. attribute k: int = 4 Number of results to return from store attribute max_tokens_limit: int = 3375 Restrict the docs to return from store based on tokens, enforced only for StuffDocumentChain and if reduce_k_below_max_tokens is to true attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute reduce_k_below_max_tokens: bool = False Reduce the number of results to return from store based on tokens limit attribute return_source_documents: bool = False Return the source documents. attribute search_kwargs: Dict[str, Any] [Optional] Extra search args. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute vectorstore: langchain.vectorstores.base.VectorStore [Required] Vector Database to connect to. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict classmethod from_chain_type(llm, chain_type='stuff', chain_type_kwargs=None, **kwargs) Load chain from chain type. Parameters llm (langchain.base_language.BaseLanguageModel) – chain_type (str) – chain_type_kwargs (Optional[dict]) – kwargs (Any) – Return type langchain.chains.qa_with_sources.base.BaseQAWithSourcesChain classmethod from_llm(llm, document_prompt=PromptTemplate(input_variables=['page_content', 'source'], output_parser=None, partial_variables={}, template='Content: {page_content}\nSource: {source}', template_format='f-string', validate_template=True), question_prompt=PromptTemplate(input_variables=['context', 'question'], output_parser=None, partial_variables={}, template='Use the following portion of a long document to see if any of the text is relevant to answer the question. \nReturn any relevant text verbatim.\n{context}\nQuestion: {question}\nRelevant text, if any:', template_format='f-string', validate_template=True), combine_prompt=PromptTemplate(input_variables=['summaries', 'question'], output_parser=None, partial_variables={}, template='Given the following extracted parts of a long document and a question, create a final answer with references ("SOURCES"). \nIf you don\'t know the answer, just say that you don\'t know. Don\'t try to make up an answer.\nALWAYS return a "SOURCES" part in your answer.\n\nQUESTION: Which state/country\'s law governs the interpretation of the contract?\n=========\nContent: This Agreement is governed by English law and the parties submit to the exclusive jurisdiction of the English courts inΒ  relation to any dispute (contractual or non-contractual) concerning this Agreement save that either party may apply to any court for anΒ  injunction or other relief to protect its Intellectual Property Rights.\nSource: 28-pl\nContent: No Waiver. Failure or delay in exercising any right or remedy under this Agreement shall not constitute a waiver of such (or any other)Β  right or remedy.\n\n11.7 Severability. The invalidity, illegality or unenforceability of any term (or part of a term) of this Agreement shall not affect the continuationΒ  in force of the remainder of the term (if any) and this Agreement.\n\n11.8 No Agency. Except as expressly stated otherwise, nothing in this Agreement shall create an agency, partnership or joint venture of anyΒ  kind between the parties.\n\n11.9 No Third-Party Beneficiaries.\nSource: 30-pl\nContent: (b) if Google believes, in good faith, that the Distributor has violated or caused Google to violate any Anti-Bribery Laws (asΒ  defined in Clause 8.5) or that such a violation is reasonably likely to occur,\nSource: 4-pl\n=========\nFINAL ANSWER: This Agreement is governed by English law.\nSOURCES: 28-pl\n\nQUESTION: What did the president say about Michael Jackson?\n=========\nContent: Madam Speaker, Madam Vice President, our First Lady and Second Gentleman. Members of Congress and the Cabinet. Justices of the Supreme Court. My fellow Americans.Β  \n\nLast year COVID-19 kept us apart. This year we are finally together again. \n\nTonight, we meet as Democrats Republicans and Independents. But most importantly as Americans. \n\nWith a duty to one another to the American people to the Constitution. \n\nAnd with an unwavering resolve that freedom will always triumph over tyranny. \n\nSix days ago, Russia’s Vladimir Putin sought to shake the foundations of the free world thinking he could make it bend to his menacing ways. But he badly miscalculated. \n\nHe thought he could roll into Ukraine and the world would roll over. Instead he met a wall of strength he never imagined. \n\nHe met the Ukrainian people. \n\nFrom President Zelenskyy to every Ukrainian, their fearlessness, their courage, their determination, inspires the world. \n\nGroups of citizens blocking tanks with their bodies. Everyone from students to retirees teachers turned soldiers defending their homeland.\nSource: 0-pl\nContent: And we won’t stop. \n\nWe have lost so much to COVID-19. Time with one another. And worst of all, so much loss of life. \n\nLet’s use this moment to reset. Let’s stop looking at COVID-19 as a partisan dividing line and see it for what it is: A God-awful disease.Β  \n\nLet’s stop seeing each other as enemies, and start seeing each other for who we really are: Fellow Americans.Β  \n\nWe can’t change how divided we’ve been. But we can change how we move forwardβ€”on COVID-19 and other issues we must face together. \n\nI recently visited the New York City Police Department days after the funerals of Officer Wilbert Mora and his partner, Officer Jason Rivera. \n\nThey were responding to a 9-1-1 call when a man shot and killed them with a stolen gun. \n\nOfficer Mora was 27 years old. \n\nOfficer Rivera was 22. \n\nBoth Dominican Americans who’d grown up on the same streets they later chose to patrol as police officers. \n\nI spoke with their families and told them that we are forever in debt for their sacrifice, and we will carry on their mission to restore the trust and safety every community deserves.\nSource: 24-pl\nContent: And a proud Ukrainian people, who have known 30 yearsΒ  of independence, have repeatedly shown that they will not tolerate anyone who tries to take their country backwards.Β  \n\nTo all Americans, I will be honest with you, as I’ve always promised. A Russian dictator, invading a foreign country, has costs around the world. \n\nAnd I’m taking robust action to make sure the pain of our sanctionsΒ  is targeted at Russia’s economy. And I will use every tool at our disposal to protect American businesses and consumers. \n\nTonight, I can announce that the United States has worked with 30 other countries to release 60 Million barrels of oil from reserves around the world.Β  \n\nAmerica will lead that effort, releasing 30 Million barrels from our own Strategic Petroleum Reserve. And we stand ready to do more if necessary, unified with our allies.Β  \n\nThese steps will help blunt gas prices here at home. And I know the news about what’s happening can seem alarming. \n\nBut I want you to know that we are going to be okay.\nSource: 5-pl\nContent: More support for patients and families. \n\nTo get there, I call on Congress to fund ARPA-H, the Advanced Research Projects Agency for Health. \n\nIt’s based on DARPAβ€”the Defense Department project that led to the Internet, GPS, and so much more.Β  \n\nARPA-H will have a singular purposeβ€”to drive breakthroughs in cancer, Alzheimer’s, diabetes, and more. \n\nA unity agenda for the nation. \n\nWe can do this. \n\nMy fellow Americansβ€”tonight , we have gathered in a sacred spaceβ€”the citadel of our democracy. \n\nIn this Capitol, generation after generation, Americans have debated great questions amid great strife, and have done great things. \n\nWe have fought for freedom, expanded liberty, defeated totalitarianism and terror. \n\nAnd built the strongest, freest, and most prosperous nation the world has ever known. \n\nNow is the hour. \n\nOur moment of responsibility. \n\nOur test of resolve and conscience, of history itself. \n\nIt is in this moment that our character is formed. Our purpose is found. Our future is forged. \n\nWell I know this nation.\nSource: 34-pl\n=========\nFINAL ANSWER: The president did not mention Michael Jackson.\nSOURCES:\n\nQUESTION: {question}\n=========\n{summaries}\n=========\nFINAL ANSWER:', template_format='f-string', validate_template=True), **kwargs) Construct the chain from an LLM. Parameters llm (langchain.base_language.BaseLanguageModel) – document_prompt (langchain.prompts.base.BasePromptTemplate) – question_prompt (langchain.prompts.base.BasePromptTemplate) – combine_prompt (langchain.prompts.base.BasePromptTemplate) – kwargs (Any) – Return type langchain.chains.qa_with_sources.base.BaseQAWithSourcesChain prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. langchain.chains.create_extraction_chain(schema, llm)[source] Creates a chain that extracts information from a passage. Parameters schema (dict) – The schema of the entities to extract. llm (langchain.base_language.BaseLanguageModel) – The language model to use. Returns Chain that can be used to extract information from a passage. Return type langchain.chains.base.Chain langchain.chains.create_extraction_chain_pydantic(pydantic_schema, llm)[source] Creates a chain that extracts information from a passage using pydantic schema. Parameters pydantic_schema (Any) – The pydantic schema of the entities to extract. llm (langchain.base_language.BaseLanguageModel) – The language model to use. Returns Chain that can be used to extract information from a passage. Return type langchain.chains.base.Chain langchain.chains.create_tagging_chain(schema, llm)[source] Creates a chain that extracts information from a passage. Parameters schema (dict) – The schema of the entities to extract. llm (langchain.base_language.BaseLanguageModel) – The language model to use. Returns Chain (LLMChain) that can be used to extract information from a passage. Return type langchain.chains.base.Chain langchain.chains.create_tagging_chain_pydantic(pydantic_schema, llm)[source] Creates a chain that extracts information from a passage. Parameters pydantic_schema (Any) – The pydantic schema of the entities to extract. llm (langchain.base_language.BaseLanguageModel) – The language model to use. Returns Chain (LLMChain) that can be used to extract information from a passage. Return type langchain.chains.base.Chain langchain.chains.load_chain(path, **kwargs)[source] Unified method for loading a chain from LangChainHub or local fs. Parameters path (Union[str, pathlib.Path]) – kwargs (Any) – Return type langchain.chains.base.Chain langchain.chains.create_citation_fuzzy_match_chain(llm)[source] Create a citation fuzzy match chain. Parameters llm (langchain.base_language.BaseLanguageModel) – Language model to use for the chain. Returns Chain (LLMChain) that can be used to answer questions with citations. Return type langchain.chains.llm.LLMChain langchain.chains.create_qa_with_structure_chain(llm, schema, output_parser='base', prompt=None)[source] Create a question answering chain that returns an answer with sources. Parameters llm (langchain.base_language.BaseLanguageModel) – Language model to use for the chain. schema (Union[dict, Type[pydantic.main.BaseModel]]) – Pydantic schema to use for the output. output_parser (str) – Output parser to use. Should be one of pydantic or base. Default to base. prompt (Optional[Union[langchain.prompts.prompt.PromptTemplate, langchain.prompts.chat.ChatPromptTemplate]]) – Optional prompt to use for the chain. Return type langchain.chains.llm.LLMChain Returns: langchain.chains.create_qa_with_sources_chain(llm, **kwargs)[source] Create a question answering chain that returns an answer with sources. Parameters llm (langchain.base_language.BaseLanguageModel) – Language model to use for the chain. **kwargs – Keyword arguments to pass to create_qa_with_structure_chain. kwargs (Any) – Returns Chain (LLMChain) that can be used to answer questions with citations. Return type langchain.chains.llm.LLMChain class langchain.chains.StuffDocumentsChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, input_key='input_documents', output_key='output_text', llm_chain, document_prompt=None, document_variable_name, document_separator='\n\n')[source] Bases: langchain.chains.combine_documents.base.BaseCombineDocumentsChain Chain that combines documents by stuffing into context. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – input_key (str) – output_key (str) – llm_chain (langchain.chains.llm.LLMChain) – document_prompt (langchain.prompts.base.BasePromptTemplate) – document_variable_name (str) – document_separator (str) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute document_prompt: langchain.prompts.base.BasePromptTemplate [Optional] Prompt to use to format each document. attribute document_separator: str = '\n\n' The string with which to join the formatted documents attribute document_variable_name: str [Required] The variable name in the llm_chain to put the documents in. If only one variable in the llm_chain, this need not be provided. attribute llm_chain: langchain.chains.llm.LLMChain [Required] LLM wrapper to use after formatting documents. attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] async acombine_docs(docs, callbacks=None, **kwargs)[source] Stuff all documents into one prompt and pass to LLM. Parameters docs (List[langchain.schema.Document]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type Tuple[str, dict] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str combine_docs(docs, callbacks=None, **kwargs)[source] Stuff all documents into one prompt and pass to LLM. Parameters docs (List[langchain.schema.Document]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type Tuple[str, dict] dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] prompt_length(docs, **kwargs)[source] Get the prompt length by formatting the prompt. Parameters docs (List[langchain.schema.Document]) – kwargs (Any) – Return type Optional[int] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.MapRerankDocumentsChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, input_key='input_documents', output_key='output_text', llm_chain, document_variable_name, rank_key, answer_key, metadata_keys=None, return_intermediate_steps=False)[source] Bases: langchain.chains.combine_documents.base.BaseCombineDocumentsChain Combining documents by mapping a chain over them, then reranking results. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – input_key (str) – output_key (str) – llm_chain (langchain.chains.llm.LLMChain) – document_variable_name (str) – rank_key (str) – answer_key (str) – metadata_keys (Optional[List[str]]) – return_intermediate_steps (bool) – Return type None attribute answer_key: str [Required] Key in output of llm_chain to return as answer. attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute document_variable_name: str [Required] The variable name in the llm_chain to put the documents in. If only one variable in the llm_chain, this need not be provided. attribute llm_chain: LLMChain [Required] Chain to apply to each document individually. attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute metadata_keys: Optional[List[str]] = None attribute rank_key: str [Required] Key in output of llm_chain to rank on. attribute return_intermediate_steps: bool = False attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] async acombine_docs(docs, callbacks=None, **kwargs)[source] Combine documents in a map rerank manner. Combine by mapping first chain over all documents, then reranking the results. Parameters docs (List[langchain.schema.Document]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type Tuple[str, dict] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str combine_docs(docs, callbacks=None, **kwargs)[source] Combine documents in a map rerank manner. Combine by mapping first chain over all documents, then reranking the results. Parameters docs (List[langchain.schema.Document]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type Tuple[str, dict] dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] prompt_length(docs, **kwargs) Return the prompt length given the documents passed in. Returns None if the method does not depend on the prompt length. Parameters docs (List[langchain.schema.Document]) – kwargs (Any) – Return type Optional[int] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.MapReduceDocumentsChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, input_key='input_documents', output_key='output_text', llm_chain, combine_document_chain, collapse_document_chain=None, document_variable_name, return_intermediate_steps=False)[source] Bases: langchain.chains.combine_documents.base.BaseCombineDocumentsChain Combining documents by mapping a chain over them, then combining results. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – input_key (str) – output_key (str) – llm_chain (langchain.chains.llm.LLMChain) – combine_document_chain (langchain.chains.combine_documents.base.BaseCombineDocumentsChain) – collapse_document_chain (Optional[langchain.chains.combine_documents.base.BaseCombineDocumentsChain]) – document_variable_name (str) – return_intermediate_steps (bool) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute collapse_document_chain: Optional[BaseCombineDocumentsChain] = None Chain to use to collapse intermediary results if needed. If None, will use the combine_document_chain. attribute combine_document_chain: BaseCombineDocumentsChain [Required] Chain to use to combine results of applying llm_chain to documents. attribute document_variable_name: str [Required] The variable name in the llm_chain to put the documents in. If only one variable in the llm_chain, this need not be provided. attribute llm_chain: LLMChain [Required] Chain to apply to each document individually. attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute return_intermediate_steps: bool = False Return the results of the map steps in the output. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] async acombine_docs(docs, callbacks=None, **kwargs)[source] Combine documents in a map reduce manner. Combine by mapping first chain over all documents, then reducing the results. This reducing can be done recursively if needed (if there are many documents). Parameters docs (List[langchain.schema.Document]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type Tuple[str, dict] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str combine_docs(docs, token_max=3000, callbacks=None, **kwargs)[source] Combine documents in a map reduce manner. Combine by mapping first chain over all documents, then reducing the results. This reducing can be done recursively if needed (if there are many documents). Parameters docs (List[langchain.schema.Document]) – token_max (int) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type Tuple[str, dict] dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] prompt_length(docs, **kwargs) Return the prompt length given the documents passed in. Returns None if the method does not depend on the prompt length. Parameters docs (List[langchain.schema.Document]) – kwargs (Any) – Return type Optional[int] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chains.RefineDocumentsChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, input_key='input_documents', output_key='output_text', initial_llm_chain, refine_llm_chain, document_variable_name, initial_response_name, document_prompt=None, return_intermediate_steps=False)[source] Bases: langchain.chains.combine_documents.base.BaseCombineDocumentsChain Combine documents by doing a first pass and then refining on more documents. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – input_key (str) – output_key (str) – initial_llm_chain (langchain.chains.llm.LLMChain) – refine_llm_chain (langchain.chains.llm.LLMChain) – document_variable_name (str) – initial_response_name (str) – document_prompt (langchain.prompts.base.BasePromptTemplate) – return_intermediate_steps (bool) – Return type None attribute callback_manager: Optional[BaseCallbackManager] = None Deprecated, use callbacks instead. attribute callbacks: Callbacks = None Optional list of callback handlers (or callback manager). Defaults to None. Callback handlers are called throughout the lifecycle of a call to a chain, starting with on_chain_start, ending with on_chain_end or on_chain_error. Each custom chain can optionally call additional callback methods, see Callback docs for full details. attribute document_prompt: BasePromptTemplate [Optional] Prompt to use to format each document. attribute document_variable_name: str [Required] The variable name in the initial_llm_chain to put the documents in. If only one variable in the initial_llm_chain, this need not be provided. attribute initial_llm_chain: LLMChain [Required] LLM chain to use on initial document. attribute initial_response_name: str [Required] The variable name to format the initial response in when refining. attribute memory: Optional[BaseMemory] = None Optional memory object. Defaults to None. Memory is a class that gets called at the start and at the end of every chain. At the start, memory loads variables and passes them along in the chain. At the end, it saves any returned variables. There are many different types of memory - please see memory docs for the full catalog. attribute refine_llm_chain: LLMChain [Required] LLM chain to use when refining. attribute return_intermediate_steps: bool = False Return the results of the refine steps in the output. attribute tags: Optional[List[str]] = None Optional list of tags associated with the chain. Defaults to None These tags will be associated with each call to this chain, and passed as arguments to the handlers defined in callbacks. You can use these to eg identify a specific instance of a chain with its use case. attribute verbose: bool [Optional] Whether or not run in verbose mode. In verbose mode, some intermediate logs will be printed to the console. Defaults to langchain.verbose value. async acall(inputs, return_only_outputs=False, callbacks=None, *, tags=None, include_run_info=False) Run the logic of this chain and add to output if desired. Parameters inputs (Union[Dict[str, Any], Any]) – Dictionary of inputs, or single input if chain expects only one param. return_only_outputs (bool) – boolean for whether to return only outputs in the response. If True, only new keys generated by this chain will be returned. If False, both input keys and new keys generated by this chain will be returned. Defaults to False. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to use for this chain run. If not provided, will use the callbacks provided to the chain. include_run_info (bool) – Whether to include run info in the response. Defaults to False. tags (Optional[List[str]]) – Return type Dict[str, Any] async acombine_docs(docs, callbacks=None, **kwargs)[source] Combine by mapping first chain over all, then stuffing into final chain. Parameters docs (List[langchain.schema.Document]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type Tuple[str, dict] apply(input_list, callbacks=None) Call the chain on all inputs in the list. Parameters input_list (List[Dict[str, Any]]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Return type List[Dict[str, str]] async arun(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str combine_docs(docs, callbacks=None, **kwargs)[source] Combine by mapping first chain over all, then stuffing into final chain. Parameters docs (List[langchain.schema.Document]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Return type Tuple[str, dict] dict(**kwargs) Return dictionary representation of chain. Parameters kwargs (Any) – Return type Dict prep_inputs(inputs) Validate and prep inputs. Parameters inputs (Union[Dict[str, Any], Any]) – Return type Dict[str, str] prep_outputs(inputs, outputs, return_only_outputs=False) Validate and prep outputs. Parameters inputs (Dict[str, str]) – outputs (Dict[str, str]) – return_only_outputs (bool) – Return type Dict[str, str] prompt_length(docs, **kwargs) Return the prompt length given the documents passed in. Returns None if the method does not depend on the prompt length. Parameters docs (List[langchain.schema.Document]) – kwargs (Any) – Return type Optional[int] run(*args, callbacks=None, tags=None, **kwargs) Run the chain as text in, text out or multiple variables, text out. Parameters args (Any) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – tags (Optional[List[str]]) – kwargs (Any) – Return type str save(file_path) Save the chain. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the chain to. Return type None Example: .. code-block:: python chain.save(file_path=”path/chain.yaml”) to_json() Return type Union[langchain.load.serializable.SerializedConstructor, langchain.load.serializable.SerializedNotImplemented] to_json_not_implemented() Return type langchain.load.serializable.SerializedNotImplemented property lc_attributes: Dict Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. property lc_namespace: List[str] Return the namespace of the langchain object. eg. [β€œlangchain”, β€œllms”, β€œopenai”] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable.
https://api.python.langchain.com/en/latest/modules/chains.html
c08e8bd6-f501-4cbf-bda0-e556a386e559
Agent Toolkits Agent toolkits. langchain.agents.agent_toolkits.create_json_agent(llm, toolkit, callback_manager=None, prefix='You are an agent designed to interact with JSON.\nYour goal is to return a final answer by interacting with the JSON.\nYou have access to the following tools which help you learn more about the JSON you are interacting with.\nOnly use the below tools. Only use the information returned by the below tools to construct your final answer.\nDo not make up any information that is not contained in the JSON.\nYour input to the tools should be in the form of `data["key"][0]` where `data` is the JSON blob you are interacting with, and the syntax used is Python. \nYou should only use keys that you know for a fact exist. You must validate that a key exists by seeing it previously when calling `json_spec_list_keys`. \nIf you have not seen a key in one of those responses, you cannot use it.\nYou should only add one key at a time to the path. You cannot add multiple keys at once.\nIf you encounter a "KeyError", go back to the previous key, look at the available keys, and try again.\n\nIf the question does not seem to be related to the JSON, just return "I don\'t know" as the answer.\nAlways begin your interaction with the `json_spec_list_keys` tool with input "data" to see what keys exist in the JSON.\n\nNote that sometimes the value at a given path is large. In this case, you will get an error "Value is a large dictionary, should explore its keys directly".\nIn this case, you should ALWAYS follow up by using the `json_spec_list_keys` tool to see what keys exist at that path.\nDo not simply refer the user to the JSON or a section of the JSON, as this is not a valid answer. Keep digging until you find the answer and explicitly return it.\n', suffix='Begin!"\n\nQuestion: {input}\nThought: I should look at the keys that exist in data to see what I have access to\n{agent_scratchpad}', format_instructions='Use the following format:\n\nQuestion: the input question you must answer\nThought: you should always think about what to do\nAction: the action to take, should be one of [{tool_names}]\nAction Input: the input to the action\nObservation: the result of the action\n... (this Thought/Action/Action Input/Observation can repeat N times)\nThought: I now know the final answer\nFinal Answer: the final answer to the original input question', input_variables=None, verbose=False, agent_executor_kwargs=None, **kwargs)[source] Construct a json agent from an LLM and tools. Parameters llm (langchain.base_language.BaseLanguageModel) – toolkit (langchain.agents.agent_toolkits.json.toolkit.JsonToolkit) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – prefix (str) – suffix (str) – format_instructions (str) – input_variables (Optional[List[str]]) – verbose (bool) – agent_executor_kwargs (Optional[Dict[str, Any]]) – kwargs (Dict[str, Any]) – Return type langchain.agents.agent.AgentExecutor langchain.agents.agent_toolkits.create_sql_agent(llm, toolkit, agent_type=AgentType.ZERO_SHOT_REACT_DESCRIPTION, callback_manager=None, prefix='You are an agent designed to interact with a SQL database.\nGiven an input question, create a syntactically correct {dialect} query to run, then look at the results of the query and return the answer.\nUnless the user specifies a specific number of examples they wish to obtain, always limit your query to at most {top_k} results.\nYou can order the results by a relevant column to return the most interesting examples in the database.\nNever query for all the columns from a specific table, only ask for the relevant columns given the question.\nYou have access to tools for interacting with the database.\nOnly use the below tools. Only use the information returned by the below tools to construct your final answer.\nYou MUST double check your query before executing it. If you get an error while executing a query, rewrite the query and try again.\n\nDO NOT make any DML statements (INSERT, UPDATE, DELETE, DROP etc.) to the database.\n\nIf the question does not seem related to the database, just return "I don\'t know" as the answer.\n', suffix=None, format_instructions='Use the following format:\n\nQuestion: the input question you must answer\nThought: you should always think about what to do\nAction: the action to take, should be one of [{tool_names}]\nAction Input: the input to the action\nObservation: the result of the action\n... (this Thought/Action/Action Input/Observation can repeat N times)\nThought: I now know the final answer\nFinal Answer: the final answer to the original input question', input_variables=None, top_k=10, max_iterations=15, max_execution_time=None, early_stopping_method='force', verbose=False, agent_executor_kwargs=None, **kwargs)[source] Construct a sql agent from an LLM and tools. Parameters llm (langchain.base_language.BaseLanguageModel) – toolkit (langchain.agents.agent_toolkits.sql.toolkit.SQLDatabaseToolkit) – agent_type (langchain.agents.agent_types.AgentType) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – prefix (str) – suffix (Optional[str]) – format_instructions (str) – input_variables (Optional[List[str]]) – top_k (int) – max_iterations (Optional[int]) – max_execution_time (Optional[float]) – early_stopping_method (str) – verbose (bool) – agent_executor_kwargs (Optional[Dict[str, Any]]) – kwargs (Dict[str, Any]) – Return type langchain.agents.agent.AgentExecutor langchain.agents.agent_toolkits.create_openapi_agent(llm, toolkit, callback_manager=None, prefix="You are an agent designed to answer questions by making web requests to an API given the openapi spec.\n\nIf the question does not seem related to the API, return I don't know. Do not make up an answer.\nOnly use information provided by the tools to construct your response.\n\nFirst, find the base URL needed to make the request.\n\nSecond, find the relevant paths needed to answer the question. Take note that, sometimes, you might need to make more than one request to more than one path to answer the question.\n\nThird, find the required parameters needed to make the request. For GET requests, these are usually URL parameters and for POST requests, these are request body parameters.\n\nFourth, make the requests needed to answer the question. Ensure that you are sending the correct parameters to the request by checking which parameters are required. For parameters with a fixed set of values, please use the spec to look at which values are allowed.\n\nUse the exact parameter names as listed in the spec, do not make up any names or abbreviate the names of parameters.\nIf you get a not found error, ensure that you are using a path that actually exists in the spec.\n", suffix='Begin!\n\nQuestion: {input}\nThought: I should explore the spec to find the base url for the API.\n{agent_scratchpad}', format_instructions='Use the following format:\n\nQuestion: the input question you must answer\nThought: you should always think about what to do\nAction: the action to take, should be one of [{tool_names}]\nAction Input: the input to the action\nObservation: the result of the action\n... (this Thought/Action/Action Input/Observation can repeat N times)\nThought: I now know the final answer\nFinal Answer: the final answer to the original input question', input_variables=None, max_iterations=15, max_execution_time=None, early_stopping_method='force', verbose=False, return_intermediate_steps=False, agent_executor_kwargs=None, **kwargs)[source] Construct a json agent from an LLM and tools. Parameters llm (langchain.base_language.BaseLanguageModel) – toolkit (langchain.agents.agent_toolkits.openapi.toolkit.OpenAPIToolkit) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – prefix (str) – suffix (str) – format_instructions (str) – input_variables (Optional[List[str]]) – max_iterations (Optional[int]) – max_execution_time (Optional[float]) – early_stopping_method (str) – verbose (bool) – return_intermediate_steps (bool) – agent_executor_kwargs (Optional[Dict[str, Any]]) – kwargs (Dict[str, Any]) – Return type langchain.agents.agent.AgentExecutor langchain.agents.agent_toolkits.create_pbi_agent(llm, toolkit, powerbi=None, callback_manager=None, prefix='You are an agent designed to help users interact with a PowerBI Dataset.\n\nAgent has access to a tool that can write a query based on the question and then run those against PowerBI, Microsofts business intelligence tool. The questions from the users should be interpreted as related to the dataset that is available and not general questions about the world. If the question does not seem related to the dataset, just return "This does not appear to be part of this dataset." as the answer.\n\nGiven an input question, ask to run the questions against the dataset, then look at the results and return the answer, the answer should be a complete sentence that answers the question, if multiple rows are asked find a way to write that in a easily readable format for a human, also make sure to represent numbers in readable ways, like 1M instead of 1000000. Unless the user specifies a specific number of examples they wish to obtain, always limit your query to at most {top_k} results.\n', suffix='Begin!\n\nQuestion: {input}\nThought: I can first ask which tables I have, then how each table is defined and then ask the query tool the question I need, and finally create a nice sentence that answers the question.\n{agent_scratchpad}', format_instructions='Use the following format:\n\nQuestion: the input question you must answer\nThought: you should always think about what to do\nAction: the action to take, should be one of [{tool_names}]\nAction Input: the input to the action\nObservation: the result of the action\n... (this Thought/Action/Action Input/Observation can repeat N times)\nThought: I now know the final answer\nFinal Answer: the final answer to the original input question', examples=None, input_variables=None, top_k=10, verbose=False, agent_executor_kwargs=None, **kwargs)[source] Construct a pbi agent from an LLM and tools. Parameters llm (langchain.base_language.BaseLanguageModel) – toolkit (Optional[langchain.agents.agent_toolkits.powerbi.toolkit.PowerBIToolkit]) – powerbi (Optional[langchain.utilities.powerbi.PowerBIDataset]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – prefix (str) – suffix (str) – format_instructions (str) – examples (Optional[str]) – input_variables (Optional[List[str]]) – top_k (int) – verbose (bool) – agent_executor_kwargs (Optional[Dict[str, Any]]) – kwargs (Dict[str, Any]) – Return type langchain.agents.agent.AgentExecutor langchain.agents.agent_toolkits.create_pbi_chat_agent(llm, toolkit, powerbi=None, callback_manager=None, output_parser=None, prefix='Assistant is a large language model built to help users interact with a PowerBI Dataset.\n\nAssistant has access to a tool that can write a query based on the question and then run those against PowerBI, Microsofts business intelligence tool. The questions from the users should be interpreted as related to the dataset that is available and not general questions about the world. If the question does not seem related to the dataset, just return "This does not appear to be part of this dataset." as the answer.\n\nGiven an input question, ask to run the questions against the dataset, then look at the results and return the answer, the answer should be a complete sentence that answers the question, if multiple rows are asked find a way to write that in a easily readable format for a human, also make sure to represent numbers in readable ways, like 1M instead of 1000000. Unless the user specifies a specific number of examples they wish to obtain, always limit your query to at most {top_k} results.\n', suffix="TOOLS\n------\nAssistant can ask the user to use tools to look up information that may be helpful in answering the users original question. The tools the human can use are:\n\n{{tools}}\n\n{format_instructions}\n\nUSER'S INPUT\n--------------------\nHere is the user's input (remember to respond with a markdown code snippet of a json blob with a single action, and NOTHING else):\n\n{{{{input}}}}\n", examples=None, input_variables=None, memory=None, top_k=10, verbose=False, agent_executor_kwargs=None, **kwargs)[source] Construct a pbi agent from an Chat LLM and tools. If you supply only a toolkit and no powerbi dataset, the same LLM is used for both. Parameters llm (langchain.chat_models.base.BaseChatModel) – toolkit (Optional[langchain.agents.agent_toolkits.powerbi.toolkit.PowerBIToolkit]) – powerbi (Optional[langchain.utilities.powerbi.PowerBIDataset]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – output_parser (Optional[langchain.agents.agent.AgentOutputParser]) – prefix (str) – suffix (str) – examples (Optional[str]) – input_variables (Optional[List[str]]) – memory (Optional[langchain.memory.chat_memory.BaseChatMemory]) – top_k (int) – verbose (bool) – agent_executor_kwargs (Optional[Dict[str, Any]]) – kwargs (Dict[str, Any]) – Return type langchain.agents.agent.AgentExecutor langchain.agents.agent_toolkits.create_python_agent(llm, tool, agent_type=AgentType.ZERO_SHOT_REACT_DESCRIPTION, callback_manager=None, verbose=False, prefix='You are an agent designed to write and execute python code to answer questions.\nYou have access to a python REPL, which you can use to execute python code.\nIf you get an error, debug your code and try again.\nOnly use the output of your code to answer the question. \nYou might know the answer without running any code, but you should still run the code to get the answer.\nIf it does not seem like you can write code to answer the question, just return "I don\'t know" as the answer.\n', agent_executor_kwargs=None, **kwargs)[source] Construct a python agent from an LLM and tool. Parameters llm (langchain.base_language.BaseLanguageModel) – tool (langchain.tools.python.tool.PythonREPLTool) – agent_type (langchain.agents.agent_types.AgentType) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – prefix (str) – agent_executor_kwargs (Optional[Dict[str, Any]]) – kwargs (Dict[str, Any]) – Return type langchain.agents.agent.AgentExecutor langchain.agents.agent_toolkits.create_vectorstore_agent(llm, toolkit, callback_manager=None, prefix='You are an agent designed to answer questions about sets of documents.\nYou have access to tools for interacting with the documents, and the inputs to the tools are questions.\nSometimes, you will be asked to provide sources for your questions, in which case you should use the appropriate tool to do so.\nIf the question does not seem relevant to any of the tools provided, just return "I don\'t know" as the answer.\n', verbose=False, agent_executor_kwargs=None, **kwargs)[source] Construct a vectorstore agent from an LLM and tools. Parameters llm (langchain.base_language.BaseLanguageModel) – toolkit (langchain.agents.agent_toolkits.vectorstore.toolkit.VectorStoreToolkit) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – prefix (str) – verbose (bool) – agent_executor_kwargs (Optional[Dict[str, Any]]) – kwargs (Dict[str, Any]) – Return type langchain.agents.agent.AgentExecutor class langchain.agents.agent_toolkits.JsonToolkit(*, spec)[source] Bases: langchain.agents.agent_toolkits.base.BaseToolkit Toolkit for interacting with a JSON spec. Parameters spec (langchain.tools.json.tool.JsonSpec) – Return type None attribute spec: langchain.tools.json.tool.JsonSpec [Required] get_tools()[source] Get the tools in the toolkit. Return type List[langchain.tools.base.BaseTool] class langchain.agents.agent_toolkits.SQLDatabaseToolkit(*, db, llm)[source] Bases: langchain.agents.agent_toolkits.base.BaseToolkit Toolkit for interacting with SQL databases. Parameters db (langchain.sql_database.SQLDatabase) – llm (langchain.base_language.BaseLanguageModel) – Return type None attribute db: langchain.sql_database.SQLDatabase [Required] attribute llm: langchain.base_language.BaseLanguageModel [Required] get_tools()[source] Get the tools in the toolkit. Return type List[langchain.tools.base.BaseTool] property dialect: str Return string representation of dialect to use. class langchain.agents.agent_toolkits.SparkSQLToolkit(*, db, llm)[source] Bases: langchain.agents.agent_toolkits.base.BaseToolkit Toolkit for interacting with Spark SQL. Parameters db (langchain.utilities.spark_sql.SparkSQL) – llm (langchain.base_language.BaseLanguageModel) – Return type None attribute db: langchain.utilities.spark_sql.SparkSQL [Required] attribute llm: langchain.base_language.BaseLanguageModel [Required] get_tools()[source] Get the tools in the toolkit. Return type List[langchain.tools.base.BaseTool] class langchain.agents.agent_toolkits.NLAToolkit(*, nla_tools)[source] Bases: langchain.agents.agent_toolkits.base.BaseToolkit Natural Language API Toolkit Definition. Parameters nla_tools (Sequence[langchain.agents.agent_toolkits.nla.tool.NLATool]) – Return type None attribute nla_tools: Sequence[langchain.agents.agent_toolkits.nla.tool.NLATool] [Required] List of API Endpoint Tools. classmethod from_llm_and_ai_plugin(llm, ai_plugin, requests=None, verbose=False, **kwargs)[source] Instantiate the toolkit from an OpenAPI Spec URL Parameters llm (langchain.base_language.BaseLanguageModel) – ai_plugin (langchain.tools.plugin.AIPlugin) – requests (Optional[langchain.requests.Requests]) – verbose (bool) – kwargs (Any) – Return type langchain.agents.agent_toolkits.nla.toolkit.NLAToolkit classmethod from_llm_and_ai_plugin_url(llm, ai_plugin_url, requests=None, verbose=False, **kwargs)[source] Instantiate the toolkit from an OpenAPI Spec URL Parameters llm (langchain.base_language.BaseLanguageModel) – ai_plugin_url (str) – requests (Optional[langchain.requests.Requests]) – verbose (bool) – kwargs (Any) – Return type langchain.agents.agent_toolkits.nla.toolkit.NLAToolkit classmethod from_llm_and_spec(llm, spec, requests=None, verbose=False, **kwargs)[source] Instantiate the toolkit by creating tools for each operation. Parameters llm (langchain.base_language.BaseLanguageModel) – spec (langchain.utilities.openapi.OpenAPISpec) – requests (Optional[langchain.requests.Requests]) – verbose (bool) – kwargs (Any) – Return type langchain.agents.agent_toolkits.nla.toolkit.NLAToolkit classmethod from_llm_and_url(llm, open_api_url, requests=None, verbose=False, **kwargs)[source] Instantiate the toolkit from an OpenAPI Spec URL Parameters llm (langchain.base_language.BaseLanguageModel) – open_api_url (str) – requests (Optional[langchain.requests.Requests]) – verbose (bool) – kwargs (Any) – Return type langchain.agents.agent_toolkits.nla.toolkit.NLAToolkit get_tools()[source] Get the tools for all the API operations. Return type List[langchain.tools.base.BaseTool] class langchain.agents.agent_toolkits.PowerBIToolkit(*, powerbi, llm, examples=None, max_iterations=5, callback_manager=None)[source] Bases: langchain.agents.agent_toolkits.base.BaseToolkit Toolkit for interacting with PowerBI dataset. Parameters powerbi (langchain.utilities.powerbi.PowerBIDataset) – llm (langchain.base_language.BaseLanguageModel) – examples (Optional[str]) – max_iterations (int) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – Return type None attribute callback_manager: Optional[langchain.callbacks.base.BaseCallbackManager] = None attribute examples: Optional[str] = None attribute llm: langchain.base_language.BaseLanguageModel [Required] attribute max_iterations: int = 5 attribute powerbi: langchain.utilities.powerbi.PowerBIDataset [Required] get_tools()[source] Get the tools in the toolkit. Return type List[langchain.tools.base.BaseTool] class langchain.agents.agent_toolkits.OpenAPIToolkit(*, json_agent, requests_wrapper)[source] Bases: langchain.agents.agent_toolkits.base.BaseToolkit Toolkit for interacting with a OpenAPI api. Parameters json_agent (langchain.agents.agent.AgentExecutor) – requests_wrapper (langchain.requests.TextRequestsWrapper) – Return type None attribute json_agent: langchain.agents.agent.AgentExecutor [Required] attribute requests_wrapper: langchain.requests.TextRequestsWrapper [Required] classmethod from_llm(llm, json_spec, requests_wrapper, **kwargs)[source] Create json agent from llm, then initialize. Parameters llm (langchain.base_language.BaseLanguageModel) – json_spec (langchain.tools.json.tool.JsonSpec) – requests_wrapper (langchain.requests.TextRequestsWrapper) – kwargs (Any) – Return type langchain.agents.agent_toolkits.openapi.toolkit.OpenAPIToolkit get_tools()[source] Get the tools in the toolkit. Return type List[langchain.tools.base.BaseTool] class langchain.agents.agent_toolkits.VectorStoreToolkit(*, vectorstore_info, llm=None)[source] Bases: langchain.agents.agent_toolkits.base.BaseToolkit Toolkit for interacting with a vector store. Parameters vectorstore_info (langchain.agents.agent_toolkits.vectorstore.toolkit.VectorStoreInfo) – llm (langchain.base_language.BaseLanguageModel) – Return type None attribute llm: langchain.base_language.BaseLanguageModel [Optional] attribute vectorstore_info: langchain.agents.agent_toolkits.vectorstore.toolkit.VectorStoreInfo [Required] get_tools()[source] Get the tools in the toolkit. Return type List[langchain.tools.base.BaseTool] langchain.agents.agent_toolkits.create_vectorstore_router_agent(llm, toolkit, callback_manager=None, prefix='You are an agent designed to answer questions.\nYou have access to tools for interacting with different sources, and the inputs to the tools are questions.\nYour main task is to decide which of the tools is relevant for answering question at hand.\nFor complex questions, you can break the question down into sub questions and use tools to answers the sub questions.\n', verbose=False, agent_executor_kwargs=None, **kwargs)[source] Construct a vectorstore router agent from an LLM and tools. Parameters llm (langchain.base_language.BaseLanguageModel) – toolkit (langchain.agents.agent_toolkits.vectorstore.toolkit.VectorStoreRouterToolkit) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – prefix (str) – verbose (bool) – agent_executor_kwargs (Optional[Dict[str, Any]]) – kwargs (Dict[str, Any]) – Return type langchain.agents.agent.AgentExecutor class langchain.agents.agent_toolkits.VectorStoreInfo(*, vectorstore, name, description)[source] Bases: pydantic.main.BaseModel Information about a vectorstore. Parameters vectorstore (langchain.vectorstores.base.VectorStore) – name (str) – description (str) – Return type None attribute description: str [Required] attribute name: str [Required] attribute vectorstore: langchain.vectorstores.base.VectorStore [Required] class langchain.agents.agent_toolkits.VectorStoreRouterToolkit(*, vectorstores, llm=None)[source] Bases: langchain.agents.agent_toolkits.base.BaseToolkit Toolkit for routing between vector stores. Parameters vectorstores (List[langchain.agents.agent_toolkits.vectorstore.toolkit.VectorStoreInfo]) – llm (langchain.base_language.BaseLanguageModel) – Return type None attribute llm: langchain.base_language.BaseLanguageModel [Optional] attribute vectorstores: List[langchain.agents.agent_toolkits.vectorstore.toolkit.VectorStoreInfo] [Required] get_tools()[source] Get the tools in the toolkit. Return type List[langchain.tools.base.BaseTool] langchain.agents.agent_toolkits.create_pandas_dataframe_agent(llm, df, agent_type=AgentType.ZERO_SHOT_REACT_DESCRIPTION, callback_manager=None, prefix=None, suffix=None, input_variables=None, verbose=False, return_intermediate_steps=False, max_iterations=15, max_execution_time=None, early_stopping_method='force', agent_executor_kwargs=None, include_df_in_prompt=True, **kwargs)[source] Construct a pandas agent from an LLM and dataframe. Parameters llm (langchain.base_language.BaseLanguageModel) – df (Any) – agent_type (langchain.agents.agent_types.AgentType) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – prefix (Optional[str]) – suffix (Optional[str]) – input_variables (Optional[List[str]]) – verbose (bool) – return_intermediate_steps (bool) – max_iterations (Optional[int]) – max_execution_time (Optional[float]) – early_stopping_method (str) – agent_executor_kwargs (Optional[Dict[str, Any]]) – include_df_in_prompt (Optional[bool]) – kwargs (Dict[str, Any]) – Return type langchain.agents.agent.AgentExecutor langchain.agents.agent_toolkits.create_spark_dataframe_agent(llm, df, callback_manager=None, prefix='\nYou are working with a spark dataframe in Python. The name of the dataframe is `df`.\nYou should use the tools below to answer the question posed of you:', suffix='\nThis is the result of `print(df.first())`:\n{df}\n\nBegin!\nQuestion: {input}\n{agent_scratchpad}', input_variables=None, verbose=False, return_intermediate_steps=False, max_iterations=15, max_execution_time=None, early_stopping_method='force', agent_executor_kwargs=None, **kwargs)[source] Construct a spark agent from an LLM and dataframe. Parameters llm (langchain.llms.base.BaseLLM) – df (Any) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – prefix (str) – suffix (str) – input_variables (Optional[List[str]]) – verbose (bool) – return_intermediate_steps (bool) – max_iterations (Optional[int]) – max_execution_time (Optional[float]) – early_stopping_method (str) – agent_executor_kwargs (Optional[Dict[str, Any]]) – kwargs (Dict[str, Any]) – Return type langchain.agents.agent.AgentExecutor langchain.agents.agent_toolkits.create_spark_sql_agent(llm, toolkit, callback_manager=None, prefix='You are an agent designed to interact with Spark SQL.\nGiven an input question, create a syntactically correct Spark SQL query to run, then look at the results of the query and return the answer.\nUnless the user specifies a specific number of examples they wish to obtain, always limit your query to at most {top_k} results.\nYou can order the results by a relevant column to return the most interesting examples in the database.\nNever query for all the columns from a specific table, only ask for the relevant columns given the question.\nYou have access to tools for interacting with the database.\nOnly use the below tools. Only use the information returned by the below tools to construct your final answer.\nYou MUST double check your query before executing it. If you get an error while executing a query, rewrite the query and try again.\n\nDO NOT make any DML statements (INSERT, UPDATE, DELETE, DROP etc.) to the database.\n\nIf the question does not seem related to the database, just return "I don\'t know" as the answer.\n', suffix='Begin!\n\nQuestion: {input}\nThought: I should look at the tables in the database to see what I can query.\n{agent_scratchpad}', format_instructions='Use the following format:\n\nQuestion: the input question you must answer\nThought: you should always think about what to do\nAction: the action to take, should be one of [{tool_names}]\nAction Input: the input to the action\nObservation: the result of the action\n... (this Thought/Action/Action Input/Observation can repeat N times)\nThought: I now know the final answer\nFinal Answer: the final answer to the original input question', input_variables=None, top_k=10, max_iterations=15, max_execution_time=None, early_stopping_method='force', verbose=False, agent_executor_kwargs=None, **kwargs)[source] Construct a sql agent from an LLM and tools. Parameters llm (langchain.base_language.BaseLanguageModel) – toolkit (langchain.agents.agent_toolkits.spark_sql.toolkit.SparkSQLToolkit) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – prefix (str) – suffix (str) – format_instructions (str) – input_variables (Optional[List[str]]) – top_k (int) – max_iterations (Optional[int]) – max_execution_time (Optional[float]) – early_stopping_method (str) – verbose (bool) – agent_executor_kwargs (Optional[Dict[str, Any]]) – kwargs (Dict[str, Any]) – Return type langchain.agents.agent.AgentExecutor langchain.agents.agent_toolkits.create_csv_agent(llm, path, pandas_kwargs=None, **kwargs)[source] Create csv agent by loading to a dataframe and using pandas agent. Parameters llm (langchain.base_language.BaseLanguageModel) – path (Union[str, List[str]]) – pandas_kwargs (Optional[dict]) – kwargs (Any) – Return type langchain.agents.agent.AgentExecutor class langchain.agents.agent_toolkits.ZapierToolkit(*, tools=[])[source] Bases: langchain.agents.agent_toolkits.base.BaseToolkit Zapier Toolkit. Parameters tools (List[langchain.tools.base.BaseTool]) – Return type None attribute tools: List[langchain.tools.base.BaseTool] = [] classmethod from_zapier_nla_wrapper(zapier_nla_wrapper)[source] Create a toolkit from a ZapierNLAWrapper. Parameters zapier_nla_wrapper (langchain.utilities.zapier.ZapierNLAWrapper) – Return type langchain.agents.agent_toolkits.zapier.toolkit.ZapierToolkit get_tools()[source] Get the tools in the toolkit. Return type List[langchain.tools.base.BaseTool] class langchain.agents.agent_toolkits.GmailToolkit(*, api_resource=None)[source] Bases: langchain.agents.agent_toolkits.base.BaseToolkit Toolkit for interacting with Gmail. Parameters api_resource (Resource) – Return type None attribute api_resource: Resource [Optional] get_tools()[source] Get the tools in the toolkit. Return type List[langchain.tools.base.BaseTool] class langchain.agents.agent_toolkits.JiraToolkit(*, tools=[])[source] Bases: langchain.agents.agent_toolkits.base.BaseToolkit Jira Toolkit. Parameters tools (List[langchain.tools.base.BaseTool]) – Return type None attribute tools: List[langchain.tools.base.BaseTool] = [] classmethod from_jira_api_wrapper(jira_api_wrapper)[source] Parameters jira_api_wrapper (langchain.utilities.jira.JiraAPIWrapper) – Return type langchain.agents.agent_toolkits.jira.toolkit.JiraToolkit get_tools()[source] Get the tools in the toolkit. Return type List[langchain.tools.base.BaseTool] class langchain.agents.agent_toolkits.FileManagementToolkit(*, root_dir=None, selected_tools=None)[source] Bases: langchain.agents.agent_toolkits.base.BaseToolkit Toolkit for interacting with a Local Files. Parameters root_dir (Optional[str]) – selected_tools (Optional[List[str]]) – Return type None attribute root_dir: Optional[str] = None If specified, all file operations are made relative to root_dir. attribute selected_tools: Optional[List[str]] = None If provided, only provide the selected tools. Defaults to all. get_tools()[source] Get the tools in the toolkit. Return type List[langchain.tools.base.BaseTool] class langchain.agents.agent_toolkits.PlayWrightBrowserToolkit(*, sync_browser=None, async_browser=None)[source] Bases: langchain.agents.agent_toolkits.base.BaseToolkit Toolkit for web browser tools. Parameters sync_browser (Optional['SyncBrowser']) – async_browser (Optional['AsyncBrowser']) – Return type None attribute async_browser: Optional['AsyncBrowser'] = None attribute sync_browser: Optional['SyncBrowser'] = None classmethod from_browser(sync_browser=None, async_browser=None)[source] Instantiate the toolkit. Parameters sync_browser (Optional[SyncBrowser]) – async_browser (Optional[AsyncBrowser]) – Return type PlayWrightBrowserToolkit get_tools()[source] Get the tools in the toolkit. Return type List[langchain.tools.base.BaseTool] class langchain.agents.agent_toolkits.AzureCognitiveServicesToolkit[source] Bases: langchain.agents.agent_toolkits.base.BaseToolkit Toolkit for Azure Cognitive Services. Return type None get_tools()[source] Get the tools in the toolkit. Return type List[langchain.tools.base.BaseTool]
https://api.python.langchain.com/en/latest/modules/agent_toolkits.html
d234a720-4555-4f2c-8fac-a7f84b4a2f37
Retrievers class langchain.retrievers.AmazonKendraRetriever(index_id, region_name=None, credentials_profile_name=None, top_k=3, attribute_filter=None, client=None)[source] Bases: langchain.schema.BaseRetriever Retriever class to query documents from Amazon Kendra Index. Parameters index_id (str) – Kendra index id region_name (Optional[str]) – The aws region e.g., us-west-2. Fallsback to AWS_DEFAULT_REGION env variable or region specified in ~/.aws/config. credentials_profile_name (Optional[str]) – The name of the profile in the ~/.aws/credentials or ~/.aws/config files, which has either access keys or role information specified. If not specified, the default credential profile or, if on an EC2 instance, credentials from IMDS will be used. top_k (int) – No of results to return attribute_filter (Optional[Dict]) – Additional filtering of results based on metadata See: https://docs.aws.amazon.com/kendra/latest/APIReference client (Optional[Any]) – boto3 client for Kendra Example retriever = AmazonKendraRetriever( index_id="c0806df7-e76b-4bce-9b5c-d5582f6b1a03" ) get_relevant_documents(query)[source] Run search on Kendra index and get top k documents Example: .. code-block:: python docs = retriever.get_relevant_documents(β€˜This is my query’) Parameters query (str) – Return type List[langchain.schema.Document] async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] class langchain.retrievers.ArxivRetriever(*, arxiv_search=None, arxiv_exceptions=None, top_k_results=3, load_max_docs=100, load_all_available_meta=False, doc_content_chars_max=4000, ARXIV_MAX_QUERY_LENGTH=300)[source] Bases: langchain.schema.BaseRetriever, langchain.utilities.arxiv.ArxivAPIWrapper It is effectively a wrapper for ArxivAPIWrapper. It wraps load() to get_relevant_documents(). It uses all ArxivAPIWrapper arguments without any change. Parameters arxiv_search (Any) – arxiv_exceptions (Any) – top_k_results (int) – load_max_docs (int) – load_all_available_meta (bool) – doc_content_chars_max (Optional[int]) – ARXIV_MAX_QUERY_LENGTH (int) – Return type None async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] get_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] class langchain.retrievers.AzureCognitiveSearchRetriever(*, service_name='', index_name='', api_key='', api_version='2020-06-30', aiosession=None, content_key='content')[source] Bases: langchain.schema.BaseRetriever, pydantic.main.BaseModel Wrapper around Azure Cognitive Search. Parameters service_name (str) – index_name (str) – api_key (str) – api_version (str) – aiosession (Optional[aiohttp.client.ClientSession]) – content_key (str) – Return type None attribute aiosession: Optional[aiohttp.client.ClientSession] = None ClientSession, in case we want to reuse connection for better performance. attribute api_key: str = '' API Key. Both Admin and Query keys work, but for reading data it’s recommended to use a Query key. attribute api_version: str = '2020-06-30' API version attribute content_key: str = 'content' Key in a retrieved result to set as the Document page_content. attribute index_name: str = '' Name of Index inside Azure Cognitive Search service attribute service_name: str = '' Name of Azure Cognitive Search service async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] get_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] class langchain.retrievers.ChatGPTPluginRetriever(*, url, bearer_token, top_k=3, filter=None, aiosession=None)[source] Bases: langchain.schema.BaseRetriever, pydantic.main.BaseModel Parameters url (str) – bearer_token (str) – top_k (int) – filter (Optional[dict]) – aiosession (Optional[aiohttp.client.ClientSession]) – Return type None attribute aiosession: Optional[aiohttp.client.ClientSession] = None attribute bearer_token: str [Required] attribute filter: Optional[dict] = None attribute top_k: int = 3 attribute url: str [Required] async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] get_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] class langchain.retrievers.ContextualCompressionRetriever(*, base_compressor, base_retriever)[source] Bases: langchain.schema.BaseRetriever, pydantic.main.BaseModel Retriever that wraps a base retriever and compresses the results. Parameters base_compressor (langchain.retrievers.document_compressors.base.BaseDocumentCompressor) – base_retriever (langchain.schema.BaseRetriever) – Return type None attribute base_compressor: langchain.retrievers.document_compressors.base.BaseDocumentCompressor [Required] Compressor for compressing retrieved documents. attribute base_retriever: langchain.schema.BaseRetriever [Required] Base Retriever to use for getting relevant documents. async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] get_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns Sequence of relevant documents Return type List[langchain.schema.Document] class langchain.retrievers.DataberryRetriever(datastore_url, top_k=None, api_key=None)[source] Bases: langchain.schema.BaseRetriever Retriever that uses the Databerry API. Parameters datastore_url (str) – top_k (Optional[int]) – api_key (Optional[str]) – datastore_url: str api_key: Optional[str] top_k: Optional[int] get_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] class langchain.retrievers.ElasticSearchBM25Retriever(client, index_name)[source] Bases: langchain.schema.BaseRetriever Wrapper around Elasticsearch using BM25 as a retrieval method. To connect to an Elasticsearch instance that requires login credentials, including Elastic Cloud, use the Elasticsearch URL format https://username:password@es_host:9243. For example, to connect to Elastic Cloud, create the Elasticsearch URL with the required authentication details and pass it to the ElasticVectorSearch constructor as the named parameter elasticsearch_url. You can obtain your Elastic Cloud URL and login credentials by logging in to the Elastic Cloud console at https://cloud.elastic.co, selecting your deployment, and navigating to the β€œDeployments” page. To obtain your Elastic Cloud password for the default β€œelastic” user: Log in to the Elastic Cloud console at https://cloud.elastic.co Go to β€œSecurity” > β€œUsers” Locate the β€œelastic” user and click β€œEdit” Click β€œReset password” Follow the prompts to reset the password The format for Elastic Cloud URLs is https://username:password@cluster_id.region_id.gcp.cloud.es.io:9243. Parameters client (Any) – index_name (str) – classmethod create(elasticsearch_url, index_name, k1=2.0, b=0.75)[source] Parameters elasticsearch_url (str) – index_name (str) – k1 (float) – b (float) – Return type langchain.retrievers.elastic_search_bm25.ElasticSearchBM25Retriever add_texts(texts, refresh_indices=True)[source] Run more texts through the embeddings and add to the retriever. Parameters texts (Iterable[str]) – Iterable of strings to add to the retriever. refresh_indices (bool) – bool to refresh ElasticSearch indices Returns List of ids from adding the texts into the retriever. Return type List[str] get_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] class langchain.retrievers.KNNRetriever(*, embeddings, index=None, texts, k=4, relevancy_threshold=None)[source] Bases: langchain.schema.BaseRetriever, pydantic.main.BaseModel KNN Retriever. Parameters embeddings (langchain.embeddings.base.Embeddings) – index (Any) – texts (List[str]) – k (int) – relevancy_threshold (Optional[float]) – Return type None attribute embeddings: langchain.embeddings.base.Embeddings [Required] attribute index: Any = None attribute k: int = 4 attribute relevancy_threshold: Optional[float] = None attribute texts: List[str] [Required] async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] classmethod from_texts(texts, embeddings, **kwargs)[source] Parameters texts (List[str]) – embeddings (langchain.embeddings.base.Embeddings) – kwargs (Any) – Return type langchain.retrievers.knn.KNNRetriever get_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] class langchain.retrievers.LlamaIndexGraphRetriever(*, graph=None, query_configs=None)[source] Bases: langchain.schema.BaseRetriever, pydantic.main.BaseModel Question-answering with sources over an LlamaIndex graph data structure. Parameters graph (Any) – query_configs (List[Dict]) – Return type None attribute graph: Any = None attribute query_configs: List[Dict] [Optional] async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] get_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – Return type List[langchain.schema.Document] class langchain.retrievers.LlamaIndexRetriever(*, index=None, query_kwargs=None)[source] Bases: langchain.schema.BaseRetriever, pydantic.main.BaseModel Question-answering with sources over an LlamaIndex data structure. Parameters index (Any) – query_kwargs (Dict) – Return type None attribute index: Any = None attribute query_kwargs: Dict [Optional] async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] get_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – Return type List[langchain.schema.Document] class langchain.retrievers.MergerRetriever(retrievers)[source] Bases: langchain.schema.BaseRetriever This class merges the results of multiple retrievers. Parameters retrievers (List[langchain.schema.BaseRetriever]) – A list of retrievers to merge. get_relevant_documents(query)[source] Get the relevant documents for a given query. Parameters query (str) – The query to search for. Returns A list of relevant documents. Return type List[langchain.schema.Document] async aget_relevant_documents(query)[source] Asynchronously get the relevant documents for a given query. Parameters query (str) – The query to search for. Returns A list of relevant documents. Return type List[langchain.schema.Document] merge_documents(query)[source] Merge the results of the retrievers. Parameters query (str) – The query to search for. Returns A list of merged documents. Return type List[langchain.schema.Document] async amerge_documents(query)[source] Asynchronously merge the results of the retrievers. Parameters query (str) – The query to search for. Returns A list of merged documents. Return type List[langchain.schema.Document] class langchain.retrievers.MetalRetriever(client, params=None)[source] Bases: langchain.schema.BaseRetriever Retriever that uses the Metal API. Parameters client (Any) – params (Optional[dict]) – get_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] class langchain.retrievers.MilvusRetriever(embedding_function, collection_name='LangChainCollection', connection_args=None, consistency_level='Session', search_params=None)[source] Bases: langchain.schema.BaseRetriever Retriever that uses the Milvus API. Parameters embedding_function (langchain.embeddings.base.Embeddings) – collection_name (str) – connection_args (Optional[Dict[str, Any]]) – consistency_level (str) – search_params (Optional[dict]) – add_texts(texts, metadatas=None)[source] Add text to the Milvus store Parameters texts (List[str]) – The text metadatas (List[dict]) – Metadata dicts, must line up with existing store Return type None get_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] class langchain.retrievers.PineconeHybridSearchRetriever(*, embeddings, sparse_encoder=None, index=None, top_k=4, alpha=0.5)[source] Bases: langchain.schema.BaseRetriever, pydantic.main.BaseModel Parameters embeddings (langchain.embeddings.base.Embeddings) – sparse_encoder (Any) – index (Any) – top_k (int) – alpha (float) – Return type None attribute alpha: float = 0.5 attribute embeddings: langchain.embeddings.base.Embeddings [Required] attribute index: Any = None attribute sparse_encoder: Any = None attribute top_k: int = 4 add_texts(texts, ids=None, metadatas=None)[source] Parameters texts (List[str]) – ids (Optional[List[str]]) – metadatas (Optional[List[dict]]) – Return type None async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] get_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] class langchain.retrievers.PubMedRetriever(*, top_k_results=3, load_max_docs=25, doc_content_chars_max=2000, load_all_available_meta=False, email='your_email@example.com', base_url_esearch='https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?', base_url_efetch='https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?', max_retry=5, sleep_time=0.2, ARXIV_MAX_QUERY_LENGTH=300)[source] Bases: langchain.schema.BaseRetriever, langchain.utilities.pupmed.PubMedAPIWrapper It is effectively a wrapper for PubMedAPIWrapper. It wraps load() to get_relevant_documents(). It uses all PubMedAPIWrapper arguments without any change. Parameters top_k_results (int) – load_max_docs (int) – doc_content_chars_max (int) – load_all_available_meta (bool) – email (str) – base_url_esearch (str) – base_url_efetch (str) – max_retry (int) – sleep_time (float) – ARXIV_MAX_QUERY_LENGTH (int) – Return type None async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] get_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] class langchain.retrievers.RemoteLangChainRetriever(*, url, headers=None, input_key='message', response_key='response', page_content_key='page_content', metadata_key='metadata')[source] Bases: langchain.schema.BaseRetriever, pydantic.main.BaseModel Parameters url (str) – headers (Optional[dict]) – input_key (str) – response_key (str) – page_content_key (str) – metadata_key (str) – Return type None attribute headers: Optional[dict] = None attribute input_key: str = 'message' attribute metadata_key: str = 'metadata' attribute page_content_key: str = 'page_content' attribute response_key: str = 'response' attribute url: str [Required] async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] get_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] class langchain.retrievers.SVMRetriever(*, embeddings, index=None, texts, k=4, relevancy_threshold=None)[source] Bases: langchain.schema.BaseRetriever, pydantic.main.BaseModel SVM Retriever. Parameters embeddings (langchain.embeddings.base.Embeddings) – index (Any) – texts (List[str]) – k (int) – relevancy_threshold (Optional[float]) – Return type None attribute embeddings: langchain.embeddings.base.Embeddings [Required] attribute index: Any = None attribute k: int = 4 attribute relevancy_threshold: Optional[float] = None attribute texts: List[str] [Required] async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] classmethod from_texts(texts, embeddings, **kwargs)[source] Parameters texts (List[str]) – embeddings (langchain.embeddings.base.Embeddings) – kwargs (Any) – Return type langchain.retrievers.svm.SVMRetriever get_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] class langchain.retrievers.SelfQueryRetriever(*, vectorstore, llm_chain, search_type='similarity', search_kwargs=None, structured_query_translator, verbose=False, use_original_query=False)[source] Bases: langchain.schema.BaseRetriever, pydantic.main.BaseModel Retriever that wraps around a vector store and uses an LLM to generate the vector store queries. Parameters vectorstore (langchain.vectorstores.base.VectorStore) – llm_chain (langchain.chains.llm.LLMChain) – search_type (str) – search_kwargs (dict) – structured_query_translator (langchain.chains.query_constructor.ir.Visitor) – verbose (bool) – use_original_query (bool) – Return type None attribute llm_chain: langchain.chains.llm.LLMChain [Required] The LLMChain for generating the vector store queries. attribute search_kwargs: dict [Optional] Keyword arguments to pass in to the vector store search. attribute search_type: str = 'similarity' The search type to perform on the vector store. attribute structured_query_translator: langchain.chains.query_constructor.ir.Visitor [Required] Translator for turning internal query language into vectorstore search params. attribute use_original_query: bool = False attribute vectorstore: langchain.vectorstores.base.VectorStore [Required] The underlying vector store from which documents will be retrieved. attribute verbose: bool = False Use original query instead of the revised new query from LLM async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] classmethod from_llm(llm, vectorstore, document_contents, metadata_field_info, structured_query_translator=None, chain_kwargs=None, enable_limit=False, use_original_query=False, **kwargs)[source] Parameters llm (langchain.base_language.BaseLanguageModel) – vectorstore (langchain.vectorstores.base.VectorStore) – document_contents (str) – metadata_field_info (List[langchain.chains.query_constructor.schema.AttributeInfo]) – structured_query_translator (Optional[langchain.chains.query_constructor.ir.Visitor]) – chain_kwargs (Optional[Dict]) – enable_limit (bool) – use_original_query (bool) – kwargs (Any) – Return type langchain.retrievers.self_query.base.SelfQueryRetriever get_relevant_documents(query, callbacks=None)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Returns List of relevant documents Return type List[langchain.schema.Document] class langchain.retrievers.TFIDFRetriever(*, vectorizer=None, docs, tfidf_array=None, k=4)[source] Bases: langchain.schema.BaseRetriever, pydantic.main.BaseModel Parameters vectorizer (Any) – docs (List[langchain.schema.Document]) – tfidf_array (Any) – k (int) – Return type None attribute docs: List[langchain.schema.Document] [Required] attribute k: int = 4 attribute tfidf_array: Any = None attribute vectorizer: Any = None async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] classmethod from_documents(documents, *, tfidf_params=None, **kwargs)[source] Parameters documents (Iterable[langchain.schema.Document]) – tfidf_params (Optional[Dict[str, Any]]) – kwargs (Any) – Return type langchain.retrievers.tfidf.TFIDFRetriever classmethod from_texts(texts, metadatas=None, tfidf_params=None, **kwargs)[source] Parameters texts (Iterable[str]) – metadatas (Optional[Iterable[dict]]) – tfidf_params (Optional[Dict[str, Any]]) – kwargs (Any) – Return type langchain.retrievers.tfidf.TFIDFRetriever get_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] class langchain.retrievers.TimeWeightedVectorStoreRetriever(*, vectorstore, search_kwargs=None, memory_stream=None, decay_rate=0.01, k=4, other_score_keys=[], default_salience=None)[source] Bases: langchain.schema.BaseRetriever, pydantic.main.BaseModel Retriever combining embedding similarity with recency. Parameters vectorstore (langchain.vectorstores.base.VectorStore) – search_kwargs (dict) – memory_stream (List[langchain.schema.Document]) – decay_rate (float) – k (int) – other_score_keys (List[str]) – default_salience (Optional[float]) – Return type None attribute decay_rate: float = 0.01 The exponential decay factor used as (1.0-decay_rate)**(hrs_passed). attribute default_salience: Optional[float] = None The salience to assign memories not retrieved from the vector store. None assigns no salience to documents not fetched from the vector store. attribute k: int = 4 The maximum number of documents to retrieve in a given call. attribute memory_stream: List[langchain.schema.Document] [Optional] The memory_stream of documents to search through. attribute other_score_keys: List[str] = [] Other keys in the metadata to factor into the score, e.g. β€˜importance’. attribute search_kwargs: dict [Optional] Keyword arguments to pass to the vectorstore similarity search. attribute vectorstore: langchain.vectorstores.base.VectorStore [Required] The vectorstore to store documents and determine salience. async aadd_documents(documents, **kwargs)[source] Add documents to vectorstore. Parameters documents (List[langchain.schema.Document]) – kwargs (Any) – Return type List[str] add_documents(documents, **kwargs)[source] Add documents to vectorstore. Parameters documents (List[langchain.schema.Document]) – kwargs (Any) – Return type List[str] async aget_relevant_documents(query)[source] Return documents that are relevant to the query. Parameters query (str) – Return type List[langchain.schema.Document] get_relevant_documents(query)[source] Return documents that are relevant to the query. Parameters query (str) – Return type List[langchain.schema.Document] get_salient_docs(query)[source] Return documents that are salient to the query. Parameters query (str) – Return type Dict[int, Tuple[langchain.schema.Document, float]] class langchain.retrievers.VespaRetriever(app, body, content_field, metadata_fields=None)[source] Bases: langchain.schema.BaseRetriever Retriever that uses the Vespa. Parameters app (Vespa) – body (Dict) – content_field (str) – metadata_fields (Optional[Sequence[str]]) – get_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] get_relevant_documents_with_filter(query, *, _filter=None)[source] Parameters query (str) – _filter (Optional[str]) – Return type List[langchain.schema.Document] classmethod from_params(url, content_field, *, k=None, metadata_fields=(), sources=None, _filter=None, yql=None, **kwargs)[source] Instantiate retriever from params. Parameters url (str) – Vespa app URL. content_field (str) – Field in results to return as Document page_content. k (Optional[int]) – Number of Documents to return. Defaults to None. metadata_fields (Sequence[str] or "*") – Fields in results to include in document metadata. Defaults to empty tuple (). sources (Sequence[str] or "*" or None) – Sources to retrieve from. Defaults to None. _filter (Optional[str]) – Document filter condition expressed in YQL. Defaults to None. yql (Optional[str]) – Full YQL query to be used. Should not be specified if _filter or sources are specified. Defaults to None. kwargs (Any) – Keyword arguments added to query body. Return type langchain.retrievers.vespa_retriever.VespaRetriever class langchain.retrievers.WeaviateHybridSearchRetriever(client, index_name, text_key, alpha=0.5, k=4, attributes=None, create_schema_if_missing=True)[source] Bases: langchain.schema.BaseRetriever Parameters client (Any) – index_name (str) – text_key (str) – alpha (float) – k (int) – attributes (Optional[List[str]]) – create_schema_if_missing (bool) – class Config[source] Bases: object Configuration for this pydantic object. extra = 'forbid' arbitrary_types_allowed = True add_documents(docs, **kwargs)[source] Upload documents to Weaviate. Parameters docs (List[langchain.schema.Document]) – kwargs (Any) – Return type List[str] get_relevant_documents(query, where_filter=None)[source] Look up similar documents in Weaviate. Parameters query (str) – where_filter (Optional[Dict[str, object]]) – Return type List[langchain.schema.Document] async aget_relevant_documents(query, where_filter=None)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for where_filter (Optional[Dict[str, object]]) – Returns List of relevant documents Return type List[langchain.schema.Document] class langchain.retrievers.WikipediaRetriever(*, wiki_client=None, top_k_results=3, lang='en', load_all_available_meta=False, doc_content_chars_max=4000)[source] Bases: langchain.schema.BaseRetriever, langchain.utilities.wikipedia.WikipediaAPIWrapper It is effectively a wrapper for WikipediaAPIWrapper. It wraps load() to get_relevant_documents(). It uses all WikipediaAPIWrapper arguments without any change. Parameters wiki_client (Any) – top_k_results (int) – lang (str) – load_all_available_meta (bool) – doc_content_chars_max (int) – Return type None async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] get_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] class langchain.retrievers.ZepRetriever(session_id, url, top_k=None)[source] Bases: langchain.schema.BaseRetriever A Retriever implementation for the Zep long-term memory store. Search your user’s long-term chat history with Zep. Note: You will need to provide the user’s session_id to use this retriever. More on Zep: Zep provides long-term conversation storage for LLM apps. The server stores, summarizes, embeds, indexes, and enriches conversational AI chat histories, and exposes them via simple, low-latency APIs. For server installation instructions, see: https://getzep.github.io/deployment/quickstart/ Parameters session_id (str) – url (str) – top_k (Optional[int]) – get_relevant_documents(query, metadata=None)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for metadata (Optional[Dict]) – Returns List of relevant documents Return type List[langchain.schema.Document] async aget_relevant_documents(query, metadata=None)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for metadata (Optional[Dict]) – Returns List of relevant documents Return type List[langchain.schema.Document] class langchain.retrievers.ZillizRetriever(embedding_function, collection_name='LangChainCollection', connection_args=None, consistency_level='Session', search_params=None)[source] Bases: langchain.schema.BaseRetriever Retriever that uses the Zilliz API. Parameters embedding_function (langchain.embeddings.base.Embeddings) – collection_name (str) – connection_args (Optional[Dict[str, Any]]) – consistency_level (str) – search_params (Optional[dict]) – add_texts(texts, metadatas=None)[source] Add text to the Zilliz store Parameters texts (List[str]) – The text metadatas (List[dict]) – Metadata dicts, must line up with existing store Return type None get_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] class langchain.retrievers.DocArrayRetriever(*, index=None, embeddings, search_field, content_field, search_type=SearchType.similarity, top_k=1, filters=None)[source] Bases: langchain.schema.BaseRetriever, pydantic.main.BaseModel Retriever class for DocArray Document Indices. Currently, supports 5 backends: InMemoryExactNNIndex, HnswDocumentIndex, QdrantDocumentIndex, ElasticDocIndex, and WeaviateDocumentIndex. Parameters index (Any) – embeddings (langchain.embeddings.base.Embeddings) – search_field (str) – content_field (str) – search_type (langchain.retrievers.docarray.SearchType) – top_k (int) – filters (Optional[Any]) – Return type None index One of the above-mentioned index instances embeddings Embedding model to represent text as vectors search_field Field to consider for searching in the documents. Should be an embedding/vector/tensor. content_field Field that represents the main content in your document schema. Will be used as a page_content. Everything else will go into metadata. search_type Type of search to perform (similarity / mmr) filters Filters applied for document retrieval. top_k Number of documents to return attribute content_field: str [Required] attribute embeddings: langchain.embeddings.base.Embeddings [Required] attribute filters: Optional[Any] = None attribute index: Any = None attribute search_field: str [Required] attribute search_type: langchain.retrievers.docarray.SearchType = SearchType.similarity attribute top_k: int = 1 async aget_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] get_relevant_documents(query)[source] Get documents relevant for a query. Parameters query (str) – string to find relevant documents for Returns List of relevant documents Return type List[langchain.schema.Document] Document compressors class langchain.retrievers.document_compressors.DocumentCompressorPipeline(*, transformers)[source] Bases: langchain.retrievers.document_compressors.base.BaseDocumentCompressor Document compressor that uses a pipeline of transformers. Parameters transformers (List[Union[langchain.schema.BaseDocumentTransformer, langchain.retrievers.document_compressors.base.BaseDocumentCompressor]]) – Return type None attribute transformers: List[Union[langchain.schema.BaseDocumentTransformer, langchain.retrievers.document_compressors.base.BaseDocumentCompressor]] [Required] List of document filters that are chained together and run in sequence. async acompress_documents(documents, query)[source] Compress retrieved documents given the query context. Parameters documents (Sequence[langchain.schema.Document]) – query (str) – Return type Sequence[langchain.schema.Document] compress_documents(documents, query)[source] Transform a list of documents. Parameters documents (Sequence[langchain.schema.Document]) – query (str) – Return type Sequence[langchain.schema.Document] class langchain.retrievers.document_compressors.EmbeddingsFilter(*, embeddings, similarity_fn=<function cosine_similarity>, k=20, similarity_threshold=None)[source] Bases: langchain.retrievers.document_compressors.base.BaseDocumentCompressor Parameters embeddings (langchain.embeddings.base.Embeddings) – similarity_fn (Callable) – k (Optional[int]) – similarity_threshold (Optional[float]) – Return type None attribute embeddings: langchain.embeddings.base.Embeddings [Required] Embeddings to use for embedding document contents and queries. attribute k: Optional[int] = 20 The number of relevant documents to return. Can be set to None, in which case similarity_threshold must be specified. Defaults to 20. attribute similarity_fn: Callable = <function cosine_similarity> Similarity function for comparing documents. Function expected to take as input two matrices (List[List[float]]) and return a matrix of scores where higher values indicate greater similarity. attribute similarity_threshold: Optional[float] = None Threshold for determining when two documents are similar enough to be considered redundant. Defaults to None, must be specified if k is set to None. async acompress_documents(documents, query)[source] Filter down documents. Parameters documents (Sequence[langchain.schema.Document]) – query (str) – Return type Sequence[langchain.schema.Document] compress_documents(documents, query)[source] Filter documents based on similarity of their embeddings to the query. Parameters documents (Sequence[langchain.schema.Document]) – query (str) – Return type Sequence[langchain.schema.Document] class langchain.retrievers.document_compressors.LLMChainExtractor(*, llm_chain, get_input=<function default_get_input>)[source] Bases: langchain.retrievers.document_compressors.base.BaseDocumentCompressor Parameters llm_chain (langchain.chains.llm.LLMChain) – get_input (Callable[[str, langchain.schema.Document], dict]) – Return type None attribute get_input: Callable[[str, langchain.schema.Document], dict] = <function default_get_input> Callable for constructing the chain input from the query and a Document. attribute llm_chain: langchain.chains.llm.LLMChain [Required] LLM wrapper to use for compressing documents. async acompress_documents(documents, query)[source] Compress page content of raw documents asynchronously. Parameters documents (Sequence[langchain.schema.Document]) – query (str) – Return type Sequence[langchain.schema.Document] compress_documents(documents, query)[source] Compress page content of raw documents. Parameters documents (Sequence[langchain.schema.Document]) – query (str) – Return type Sequence[langchain.schema.Document] classmethod from_llm(llm, prompt=None, get_input=None, llm_chain_kwargs=None)[source] Initialize from LLM. Parameters llm (langchain.base_language.BaseLanguageModel) – prompt (Optional[langchain.prompts.prompt.PromptTemplate]) – get_input (Optional[Callable[[str, langchain.schema.Document], str]]) – llm_chain_kwargs (Optional[dict]) – Return type langchain.retrievers.document_compressors.chain_extract.LLMChainExtractor class langchain.retrievers.document_compressors.LLMChainFilter(*, llm_chain, get_input=<function default_get_input>)[source] Bases: langchain.retrievers.document_compressors.base.BaseDocumentCompressor Filter that drops documents that aren’t relevant to the query. Parameters llm_chain (langchain.chains.llm.LLMChain) – get_input (Callable[[str, langchain.schema.Document], dict]) – Return type None attribute get_input: Callable[[str, langchain.schema.Document], dict] = <function default_get_input> Callable for constructing the chain input from the query and a Document. attribute llm_chain: langchain.chains.llm.LLMChain [Required] LLM wrapper to use for filtering documents. The chain prompt is expected to have a BooleanOutputParser. async acompress_documents(documents, query)[source] Filter down documents. Parameters documents (Sequence[langchain.schema.Document]) – query (str) – Return type Sequence[langchain.schema.Document] compress_documents(documents, query)[source] Filter down documents based on their relevance to the query. Parameters documents (Sequence[langchain.schema.Document]) – query (str) – Return type Sequence[langchain.schema.Document] classmethod from_llm(llm, prompt=None, **kwargs)[source] Parameters llm (langchain.base_language.BaseLanguageModel) – prompt (Optional[langchain.prompts.base.BasePromptTemplate]) – kwargs (Any) – Return type langchain.retrievers.document_compressors.chain_filter.LLMChainFilter class langchain.retrievers.document_compressors.CohereRerank(*, client, top_n=3, model='rerank-english-v2.0')[source] Bases: langchain.retrievers.document_compressors.base.BaseDocumentCompressor Parameters client (Client) – top_n (int) – model (str) – Return type None attribute client: Client [Required] attribute model: str = 'rerank-english-v2.0' attribute top_n: int = 3 async acompress_documents(documents, query)[source] Compress retrieved documents given the query context. Parameters documents (Sequence[langchain.schema.Document]) – query (str) – Return type Sequence[langchain.schema.Document] compress_documents(documents, query)[source] Compress retrieved documents given the query context. Parameters documents (Sequence[langchain.schema.Document]) – query (str) – Return type Sequence[langchain.schema.Document]
https://api.python.langchain.com/en/latest/modules/retrievers.html
c8bf6033-faa2-4a5c-8dcf-ab5de43b6b09
Chat Models class langchain.chat_models.ChatOpenAI(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, model='gpt-3.5-turbo', temperature=0.7, model_kwargs=None, openai_api_key=None, openai_api_base=None, openai_organization=None, openai_proxy=None, request_timeout=None, max_retries=6, streaming=False, n=1, max_tokens=None, tiktoken_model_name=None)[source] Bases: langchain.chat_models.base.BaseChatModel Wrapper around OpenAI Chat large language models. To use, you should have the openai python package installed, and the environment variable OPENAI_API_KEY set with your API key. Any parameters that are valid to be passed to the openai.create call can be passed in, even if not explicitly saved on this class. Example from langchain.chat_models import ChatOpenAI openai = ChatOpenAI(model_name="gpt-3.5-turbo") Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – model (str) – temperature (float) – model_kwargs (Dict[str, Any]) – openai_api_key (Optional[str]) – openai_api_base (Optional[str]) – openai_organization (Optional[str]) – openai_proxy (Optional[str]) – request_timeout (Optional[Union[float, Tuple[float, float]]]) – max_retries (int) – streaming (bool) – n (int) – max_tokens (Optional[int]) – tiktoken_model_name (Optional[str]) – Return type None attribute max_retries: int = 6 Maximum number of retries to make when generating. attribute max_tokens: Optional[int] = None Maximum number of tokens to generate. attribute model_kwargs: Dict[str, Any] [Optional] Holds any model parameters valid for create call not explicitly specified. attribute model_name: str = 'gpt-3.5-turbo' (alias 'model') Model name to use. attribute n: int = 1 Number of chat completions to generate for each prompt. attribute openai_api_base: Optional[str] = None attribute openai_api_key: Optional[str] = None Base URL path for API requests, leave blank if not using a proxy or service emulator. attribute openai_organization: Optional[str] = None attribute openai_proxy: Optional[str] = None attribute request_timeout: Optional[Union[float, Tuple[float, float]]] = None Timeout for requests to OpenAI completion API. Default is 600 seconds. attribute streaming: bool = False Whether to stream the results or not. attribute temperature: float = 0.7 What sampling temperature to use. attribute tiktoken_model_name: Optional[str] = None The model name to pass to tiktoken when using this class. Tiktoken is used to count the number of tokens in documents to constrain them to be under a certain limit. By default, when set to None, this will be the same as the embedding model name. However, there are some cases where you may want to use this Embedding class with a model name not supported by tiktoken. This can include when using Azure embeddings or when using one of the many model providers that expose an OpenAI-like API but with different models. In those cases, in order to avoid erroring when tiktoken is called, you can specify a model name to use here. completion_with_retry(**kwargs)[source] Use tenacity to retry the completion call. Parameters kwargs (Any) – Return type Any get_num_tokens_from_messages(messages)[source] Calculate num tokens for gpt-3.5-turbo and gpt-4 with tiktoken package. Official documentation: https://github.com/openai/openai-cookbook/blob/ main/examples/How_to_format_inputs_to_ChatGPT_models.ipynb Parameters messages (List[langchain.schema.BaseMessage]) – Return type int get_token_ids(text)[source] Get the tokens present in the text with tiktoken package. Parameters text (str) – Return type List[int] property lc_secrets: Dict[str, str] Return a map of constructor argument names to secret ids. eg. {β€œopenai_api_key”: β€œOPENAI_API_KEY”} property lc_serializable: bool Return whether or not the class is serializable. class langchain.chat_models.AzureChatOpenAI(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, model='gpt-3.5-turbo', temperature=0.7, model_kwargs=None, openai_api_key='', openai_api_base='', openai_organization='', openai_proxy='', request_timeout=None, max_retries=6, streaming=False, n=1, max_tokens=None, tiktoken_model_name=None, deployment_name='', openai_api_type='azure', openai_api_version='')[source] Bases: langchain.chat_models.openai.ChatOpenAI Wrapper around Azure OpenAI Chat Completion API. To use this class you must have a deployed model on Azure OpenAI. Use deployment_name in the constructor to refer to the β€œModel deployment name” in the Azure portal. In addition, you should have the openai python package installed, and the following environment variables set or passed in constructor in lower case: - OPENAI_API_TYPE (default: azure) - OPENAI_API_KEY - OPENAI_API_BASE - OPENAI_API_VERSION - OPENAI_PROXY For exmaple, if you have gpt-35-turbo deployed, with the deployment name 35-turbo-dev, the constructor should look like: AzureChatOpenAI( deployment_name="35-turbo-dev", openai_api_version="2023-03-15-preview", ) Be aware the API version may change. Any parameters that are valid to be passed to the openai.create call can be passed in, even if not explicitly saved on this class. Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – model (str) – temperature (float) – model_kwargs (Dict[str, Any]) – openai_api_key (str) – openai_api_base (str) – openai_organization (str) – openai_proxy (str) – request_timeout (Optional[Union[float, Tuple[float, float]]]) – max_retries (int) – streaming (bool) – n (int) – max_tokens (Optional[int]) – tiktoken_model_name (Optional[str]) – deployment_name (str) – openai_api_type (str) – openai_api_version (str) – Return type None attribute deployment_name: str = '' attribute openai_api_base: str = '' attribute openai_api_key: str = '' Base URL path for API requests, leave blank if not using a proxy or service emulator. attribute openai_api_type: str = 'azure' attribute openai_api_version: str = '' attribute openai_organization: str = '' attribute openai_proxy: str = '' class langchain.chat_models.FakeListChatModel(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, responses, i=0)[source] Bases: langchain.chat_models.base.SimpleChatModel Fake ChatModel for testing purposes. Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – responses (List) – i (int) – Return type None attribute i: int = 0 attribute responses: List [Required] class langchain.chat_models.PromptLayerChatOpenAI(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, model='gpt-3.5-turbo', temperature=0.7, model_kwargs=None, openai_api_key=None, openai_api_base=None, openai_organization=None, openai_proxy=None, request_timeout=None, max_retries=6, streaming=False, n=1, max_tokens=None, tiktoken_model_name=None, pl_tags=None, return_pl_id=False)[source] Bases: langchain.chat_models.openai.ChatOpenAI Wrapper around OpenAI Chat large language models and PromptLayer. To use, you should have the openai and promptlayer python package installed, and the environment variable OPENAI_API_KEY and PROMPTLAYER_API_KEY set with your openAI API key and promptlayer key respectively. All parameters that can be passed to the OpenAI LLM can also be passed here. The PromptLayerChatOpenAI adds to optional Parameters pl_tags (Optional[List[str]]) – List of strings to tag the request with. return_pl_id (Optional[bool]) – If True, the PromptLayer request ID will be returned in the generation_info field of the Generation object. cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – model (str) – temperature (float) – model_kwargs (Dict[str, Any]) – openai_api_key (Optional[str]) – openai_api_base (Optional[str]) – openai_organization (Optional[str]) – openai_proxy (Optional[str]) – request_timeout (Optional[Union[float, Tuple[float, float]]]) – max_retries (int) – streaming (bool) – n (int) – max_tokens (Optional[int]) – tiktoken_model_name (Optional[str]) – Return type None Example from langchain.chat_models import PromptLayerChatOpenAI openai = PromptLayerChatOpenAI(model_name="gpt-3.5-turbo") attribute pl_tags: Optional[List[str]] = None attribute return_pl_id: Optional[bool] = False class langchain.chat_models.ChatAnthropic(*, client=None, model='claude-v1', max_tokens_to_sample=256, temperature=None, top_k=None, top_p=None, streaming=False, default_request_timeout=None, anthropic_api_url=None, anthropic_api_key=None, HUMAN_PROMPT=None, AI_PROMPT=None, count_tokens=None, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None)[source] Bases: langchain.chat_models.base.BaseChatModel, langchain.llms.anthropic._AnthropicCommon Wrapper around Anthropic’s large language model. To use, you should have the anthropic python package installed, and the environment variable ANTHROPIC_API_KEY set with your API key, or pass it as a named parameter to the constructor. Example import anthropic from langchain.llms import Anthropic model = ChatAnthropic(model="<model_name>", anthropic_api_key="my-api-key") Parameters client (Any) – model (str) – max_tokens_to_sample (int) – temperature (Optional[float]) – top_k (Optional[int]) – top_p (Optional[float]) – streaming (bool) – default_request_timeout (Optional[Union[float, Tuple[float, float]]]) – anthropic_api_url (Optional[str]) – anthropic_api_key (Optional[str]) – HUMAN_PROMPT (Optional[str]) – AI_PROMPT (Optional[str]) – count_tokens (Optional[Callable[[str], int]]) – cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – Return type None get_num_tokens(text)[source] Calculate number of tokens. Parameters text (str) – Return type int property lc_serializable: bool Return whether or not the class is serializable. class langchain.chat_models.ChatGooglePalm(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, model_name='models/chat-bison-001', google_api_key=None, temperature=None, top_p=None, top_k=None, n=1)[source] Bases: langchain.chat_models.base.BaseChatModel, pydantic.main.BaseModel Wrapper around Google’s PaLM Chat API. To use you must have the google.generativeai Python package installed and either: The GOOGLE_API_KEY` environment varaible set with your API key, or Pass your API key using the google_api_key kwarg to the ChatGoogle constructor. Example from langchain.chat_models import ChatGooglePalm chat = ChatGooglePalm() Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (Any) – model_name (str) – google_api_key (Optional[str]) – temperature (Optional[float]) – top_p (Optional[float]) – top_k (Optional[int]) – n (int) – Return type None attribute google_api_key: Optional[str] = None attribute model_name: str = 'models/chat-bison-001' Model name to use. attribute n: int = 1 Number of chat completions to generate for each prompt. Note that the API may not return the full n completions if duplicates are generated. attribute temperature: Optional[float] = None Run inference with this temperature. Must by in the closed interval [0.0, 1.0]. attribute top_k: Optional[int] = None Decode using top-k sampling: consider the set of top_k most probable tokens. Must be positive. attribute top_p: Optional[float] = None Decode using nucleus sampling: consider the smallest set of tokens whose probability sum is at least top_p. Must be in the closed interval [0.0, 1.0]. class langchain.chat_models.ChatVertexAI(*, cache=None, verbose=None, callbacks=None, callback_manager=None, tags=None, client=None, model_name='chat-bison', temperature=0.0, max_output_tokens=128, top_p=0.95, top_k=40, stop=None, project=None, location='us-central1', credentials=None)[source] Bases: langchain.llms.vertexai._VertexAICommon, langchain.chat_models.base.BaseChatModel Wrapper around Vertex AI large language models. Parameters cache (Optional[bool]) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – tags (Optional[List[str]]) – client (_LanguageModel) – model_name (str) – temperature (float) – max_output_tokens (int) – top_p (float) – top_k (int) – stop (Optional[List[str]]) – project (Optional[str]) – location (str) – credentials (Any) – Return type None attribute model_name: str = 'chat-bison' Model name to use.
https://api.python.langchain.com/en/latest/modules/chat_models.html
30ea5497-b36c-4b13-9417-1b31e3763e0e
Prompt Templates Prompt template classes. class langchain.prompts.AIMessagePromptTemplate(*, prompt, additional_kwargs=None)[source] Bases: langchain.prompts.chat.BaseStringMessagePromptTemplate Parameters prompt (langchain.prompts.base.StringPromptTemplate) – additional_kwargs (dict) – Return type None format(**kwargs)[source] To a BaseMessage. Parameters kwargs (Any) – Return type langchain.schema.BaseMessage class langchain.prompts.BaseChatPromptTemplate(*, input_variables, output_parser=None, partial_variables=None)[source] Bases: langchain.prompts.base.BasePromptTemplate, abc.ABC Parameters input_variables (List[str]) – output_parser (Optional[langchain.schema.BaseOutputParser]) – partial_variables (Mapping[str, Union[str, Callable[[], str]]]) – Return type None format(**kwargs)[source] Format the prompt with the inputs. Parameters kwargs (Any) – Any arguments to be passed to the prompt template. Returns A formatted string. Return type str Example: prompt.format(variable1="foo") abstract format_messages(**kwargs)[source] Format kwargs into a list of messages. Parameters kwargs (Any) – Return type List[langchain.schema.BaseMessage] format_prompt(**kwargs)[source] Create Chat Messages. Parameters kwargs (Any) – Return type langchain.schema.PromptValue class langchain.prompts.BasePromptTemplate(*, input_variables, output_parser=None, partial_variables=None)[source] Bases: langchain.load.serializable.Serializable, abc.ABC Base class for all prompt templates, returning a prompt. Parameters input_variables (List[str]) – output_parser (Optional[langchain.schema.BaseOutputParser]) – partial_variables (Mapping[str, Union[str, Callable[[], str]]]) – Return type None attribute input_variables: List[str] [Required] A list of the names of the variables the prompt template expects. attribute output_parser: Optional[langchain.schema.BaseOutputParser] = None How to parse the output of calling an LLM on this formatted prompt. attribute partial_variables: Mapping[str, Union[str, Callable[[], str]]] [Optional] dict(**kwargs)[source] Return dictionary representation of prompt. Parameters kwargs (Any) – Return type Dict abstract format(**kwargs)[source] Format the prompt with the inputs. Parameters kwargs (Any) – Any arguments to be passed to the prompt template. Returns A formatted string. Return type str Example: prompt.format(variable1="foo") abstract format_prompt(**kwargs)[source] Create Chat Messages. Parameters kwargs (Any) – Return type langchain.schema.PromptValue partial(**kwargs)[source] Return a partial of the prompt template. Parameters kwargs (Union[str, Callable[[], str]]) – Return type langchain.prompts.base.BasePromptTemplate save(file_path)[source] Save the prompt. Parameters file_path (Union[pathlib.Path, str]) – Path to directory to save prompt to. Return type None Example: .. code-block:: python prompt.save(file_path=”path/prompt.yaml”) property lc_serializable: bool Return whether or not the class is serializable. class langchain.prompts.ChatMessagePromptTemplate(*, prompt, additional_kwargs=None, role)[source] Bases: langchain.prompts.chat.BaseStringMessagePromptTemplate Parameters prompt (langchain.prompts.base.StringPromptTemplate) – additional_kwargs (dict) – role (str) – Return type None attribute role: str [Required] format(**kwargs)[source] To a BaseMessage. Parameters kwargs (Any) – Return type langchain.schema.BaseMessage class langchain.prompts.ChatPromptTemplate(*, input_variables, output_parser=None, partial_variables=None, messages)[source] Bases: langchain.prompts.chat.BaseChatPromptTemplate, abc.ABC Parameters input_variables (List[str]) – output_parser (Optional[langchain.schema.BaseOutputParser]) – partial_variables (Mapping[str, Union[str, Callable[[], str]]]) – messages (List[Union[langchain.prompts.chat.BaseMessagePromptTemplate, langchain.schema.BaseMessage]]) – Return type None attribute input_variables: List[str] [Required] A list of the names of the variables the prompt template expects. attribute messages: List[Union[BaseMessagePromptTemplate, BaseMessage]] [Required] format(**kwargs)[source] Format the prompt with the inputs. Parameters kwargs (Any) – Any arguments to be passed to the prompt template. Returns A formatted string. Return type str Example: prompt.format(variable1="foo") format_messages(**kwargs)[source] Format kwargs into a list of messages. Parameters kwargs (Any) – Return type List[langchain.schema.BaseMessage] classmethod from_messages(messages)[source] Parameters messages (Sequence[Union[langchain.prompts.chat.BaseMessagePromptTemplate, langchain.schema.BaseMessage]]) – Return type langchain.prompts.chat.ChatPromptTemplate classmethod from_role_strings(string_messages)[source] Parameters string_messages (List[Tuple[str, str]]) – Return type langchain.prompts.chat.ChatPromptTemplate classmethod from_strings(string_messages)[source] Parameters string_messages (List[Tuple[Type[langchain.prompts.chat.BaseMessagePromptTemplate], str]]) – Return type langchain.prompts.chat.ChatPromptTemplate classmethod from_template(template, **kwargs)[source] Parameters template (str) – kwargs (Any) – Return type langchain.prompts.chat.ChatPromptTemplate partial(**kwargs)[source] Return a partial of the prompt template. Parameters kwargs (Union[str, Callable[[], str]]) – Return type langchain.prompts.base.BasePromptTemplate save(file_path)[source] Save the prompt. Parameters file_path (Union[pathlib.Path, str]) – Path to directory to save prompt to. Return type None Example: .. code-block:: python prompt.save(file_path=”path/prompt.yaml”) class langchain.prompts.FewShotPromptTemplate(*, input_variables, output_parser=None, partial_variables=None, examples=None, example_selector=None, example_prompt, suffix, example_separator='\n\n', prefix='', template_format='f-string', validate_template=True)[source] Bases: langchain.prompts.base.StringPromptTemplate Prompt template that contains few shot examples. Parameters input_variables (List[str]) – output_parser (Optional[langchain.schema.BaseOutputParser]) – partial_variables (Mapping[str, Union[str, Callable[[], str]]]) – examples (Optional[List[dict]]) – example_selector (Optional[langchain.prompts.example_selector.base.BaseExampleSelector]) – example_prompt (langchain.prompts.prompt.PromptTemplate) – suffix (str) – example_separator (str) – prefix (str) – template_format (str) – validate_template (bool) – Return type None attribute example_prompt: langchain.prompts.prompt.PromptTemplate [Required] PromptTemplate used to format an individual example. attribute example_selector: Optional[langchain.prompts.example_selector.base.BaseExampleSelector] = None ExampleSelector to choose the examples to format into the prompt. Either this or examples should be provided. attribute example_separator: str = '\n\n' String separator used to join the prefix, the examples, and suffix. attribute examples: Optional[List[dict]] = None Examples to format into the prompt. Either this or example_selector should be provided. attribute input_variables: List[str] [Required] A list of the names of the variables the prompt template expects. attribute prefix: str = '' A prompt template string to put before the examples. attribute suffix: str [Required] A prompt template string to put after the examples. attribute template_format: str = 'f-string' The format of the prompt template. Options are: β€˜f-string’, β€˜jinja2’. attribute validate_template: bool = True Whether or not to try validating the template. dict(**kwargs)[source] Return a dictionary of the prompt. Parameters kwargs (Any) – Return type Dict format(**kwargs)[source] Format the prompt with the inputs. Parameters kwargs (Any) – Any arguments to be passed to the prompt template. Returns A formatted string. Return type str Example: prompt.format(variable1="foo") property lc_serializable: bool Return whether or not the class is serializable. class langchain.prompts.FewShotPromptWithTemplates(*, input_variables, output_parser=None, partial_variables=None, examples=None, example_selector=None, example_prompt, suffix, example_separator='\n\n', prefix=None, template_format='f-string', validate_template=True)[source] Bases: langchain.prompts.base.StringPromptTemplate Prompt template that contains few shot examples. Parameters input_variables (List[str]) – output_parser (Optional[langchain.schema.BaseOutputParser]) – partial_variables (Mapping[str, Union[str, Callable[[], str]]]) – examples (Optional[List[dict]]) – example_selector (Optional[langchain.prompts.example_selector.base.BaseExampleSelector]) – example_prompt (langchain.prompts.prompt.PromptTemplate) – suffix (langchain.prompts.base.StringPromptTemplate) – example_separator (str) – prefix (Optional[langchain.prompts.base.StringPromptTemplate]) – template_format (str) – validate_template (bool) – Return type None attribute example_prompt: langchain.prompts.prompt.PromptTemplate [Required] PromptTemplate used to format an individual example. attribute example_selector: Optional[langchain.prompts.example_selector.base.BaseExampleSelector] = None ExampleSelector to choose the examples to format into the prompt. Either this or examples should be provided. attribute example_separator: str = '\n\n' String separator used to join the prefix, the examples, and suffix. attribute examples: Optional[List[dict]] = None Examples to format into the prompt. Either this or example_selector should be provided. attribute input_variables: List[str] [Required] A list of the names of the variables the prompt template expects. attribute prefix: Optional[langchain.prompts.base.StringPromptTemplate] = None A PromptTemplate to put before the examples. attribute suffix: langchain.prompts.base.StringPromptTemplate [Required] A PromptTemplate to put after the examples. attribute template_format: str = 'f-string' The format of the prompt template. Options are: β€˜f-string’, β€˜jinja2’. attribute validate_template: bool = True Whether or not to try validating the template. dict(**kwargs)[source] Return a dictionary of the prompt. Parameters kwargs (Any) – Return type Dict format(**kwargs)[source] Format the prompt with the inputs. Parameters kwargs (Any) – Any arguments to be passed to the prompt template. Returns A formatted string. Return type str Example: prompt.format(variable1="foo") class langchain.prompts.HumanMessagePromptTemplate(*, prompt, additional_kwargs=None)[source] Bases: langchain.prompts.chat.BaseStringMessagePromptTemplate Parameters prompt (langchain.prompts.base.StringPromptTemplate) – additional_kwargs (dict) – Return type None format(**kwargs)[source] To a BaseMessage. Parameters kwargs (Any) – Return type langchain.schema.BaseMessage class langchain.prompts.LengthBasedExampleSelector(*, examples, example_prompt, get_text_length=<function _get_length_based>, max_length=2048, example_text_lengths=[])[source] Bases: langchain.prompts.example_selector.base.BaseExampleSelector, pydantic.main.BaseModel Select examples based on length. Parameters examples (List[dict]) – example_prompt (langchain.prompts.prompt.PromptTemplate) – get_text_length (Callable[[str], int]) – max_length (int) – example_text_lengths (List[int]) – Return type None attribute example_prompt: langchain.prompts.prompt.PromptTemplate [Required] Prompt template used to format the examples. attribute examples: List[dict] [Required] A list of the examples that the prompt template expects. attribute get_text_length: Callable[[str], int] = <function _get_length_based> Function to measure prompt length. Defaults to word count. attribute max_length: int = 2048 Max length for the prompt, beyond which examples are cut. add_example(example)[source] Add new example to list. Parameters example (Dict[str, str]) – Return type None select_examples(input_variables)[source] Select which examples to use based on the input lengths. Parameters input_variables (Dict[str, str]) – Return type List[dict] class langchain.prompts.MaxMarginalRelevanceExampleSelector(*, vectorstore, k=4, example_keys=None, input_keys=None, fetch_k=20)[source] Bases: langchain.prompts.example_selector.semantic_similarity.SemanticSimilarityExampleSelector ExampleSelector that selects examples based on Max Marginal Relevance. This was shown to improve performance in this paper: https://arxiv.org/pdf/2211.13892.pdf Parameters vectorstore (langchain.vectorstores.base.VectorStore) – k (int) – example_keys (Optional[List[str]]) – input_keys (Optional[List[str]]) – fetch_k (int) – Return type None attribute example_keys: Optional[List[str]] = None Optional keys to filter examples to. attribute fetch_k: int = 20 Number of examples to fetch to rerank. attribute input_keys: Optional[List[str]] = None Optional keys to filter input to. If provided, the search is based on the input variables instead of all variables. attribute k: int = 4 Number of examples to select. attribute vectorstore: langchain.vectorstores.base.VectorStore [Required] VectorStore than contains information about examples. classmethod from_examples(examples, embeddings, vectorstore_cls, k=4, input_keys=None, fetch_k=20, **vectorstore_cls_kwargs)[source] Create k-shot example selector using example list and embeddings. Reshuffles examples dynamically based on query similarity. Parameters examples (List[dict]) – List of examples to use in the prompt. embeddings (langchain.embeddings.base.Embeddings) – An iniialized embedding API interface, e.g. OpenAIEmbeddings(). vectorstore_cls (Type[langchain.vectorstores.base.VectorStore]) – A vector store DB interface class, e.g. FAISS. k (int) – Number of examples to select input_keys (Optional[List[str]]) – If provided, the search is based on the input variables instead of all variables. vectorstore_cls_kwargs (Any) – optional kwargs containing url for vector store fetch_k (int) – Returns The ExampleSelector instantiated, backed by a vector store. Return type langchain.prompts.example_selector.semantic_similarity.MaxMarginalRelevanceExampleSelector select_examples(input_variables)[source] Select which examples to use based on semantic similarity. Parameters input_variables (Dict[str, str]) – Return type List[dict] class langchain.prompts.MessagesPlaceholder(*, variable_name)[source] Bases: langchain.prompts.chat.BaseMessagePromptTemplate Prompt template that assumes variable is already list of messages. Parameters variable_name (str) – Return type None attribute variable_name: str [Required] format_messages(**kwargs)[source] To a BaseMessage. Parameters kwargs (Any) – Return type List[langchain.schema.BaseMessage] property input_variables: List[str] Input variables for this prompt template. class langchain.prompts.NGramOverlapExampleSelector(*, examples, example_prompt, threshold=- 1.0)[source] Bases: langchain.prompts.example_selector.base.BaseExampleSelector, pydantic.main.BaseModel Select and order examples based on ngram overlap score (sentence_bleu score). https://www.nltk.org/_modules/nltk/translate/bleu_score.html https://aclanthology.org/P02-1040.pdf Parameters examples (List[dict]) – example_prompt (langchain.prompts.prompt.PromptTemplate) – threshold (float) – Return type None attribute example_prompt: langchain.prompts.prompt.PromptTemplate [Required] Prompt template used to format the examples. attribute examples: List[dict] [Required] A list of the examples that the prompt template expects. attribute threshold: float = -1.0 Threshold at which algorithm stops. Set to -1.0 by default. For negative threshold: select_examples sorts examples by ngram_overlap_score, but excludes none. For threshold greater than 1.0: select_examples excludes all examples, and returns an empty list. For threshold equal to 0.0: select_examples sorts examples by ngram_overlap_score, and excludes examples with no ngram overlap with input. add_example(example)[source] Add new example to list. Parameters example (Dict[str, str]) – Return type None select_examples(input_variables)[source] Return list of examples sorted by ngram_overlap_score with input. Descending order. Excludes any examples with ngram_overlap_score less than or equal to threshold. Parameters input_variables (Dict[str, str]) – Return type List[dict] class langchain.prompts.PipelinePromptTemplate(*, input_variables, output_parser=None, partial_variables=None, final_prompt, pipeline_prompts)[source] Bases: langchain.prompts.base.BasePromptTemplate A prompt template for composing multiple prompts together. This can be useful when you want to reuse parts of prompts. A PipelinePrompt consists of two main parts: final_prompt: This is the final prompt that is returned pipeline_prompts: This is a list of tuples, consistingof a string (name) and a Prompt Template. Each PromptTemplate will be formatted and then passed to future prompt templates as a variable with the same name as name Parameters input_variables (List[str]) – output_parser (Optional[langchain.schema.BaseOutputParser]) – partial_variables (Mapping[str, Union[str, Callable[[], str]]]) – final_prompt (langchain.prompts.base.BasePromptTemplate) – pipeline_prompts (List[Tuple[str, langchain.prompts.base.BasePromptTemplate]]) – Return type None attribute final_prompt: langchain.prompts.base.BasePromptTemplate [Required] attribute pipeline_prompts: List[Tuple[str, langchain.prompts.base.BasePromptTemplate]] [Required] format(**kwargs)[source] Format the prompt with the inputs. Parameters kwargs (Any) – Any arguments to be passed to the prompt template. Returns A formatted string. Return type str Example: prompt.format(variable1="foo") format_prompt(**kwargs)[source] Create Chat Messages. Parameters kwargs (Any) – Return type langchain.schema.PromptValue langchain.prompts.Prompt alias of langchain.prompts.prompt.PromptTemplate class langchain.prompts.PromptTemplate(*, input_variables, output_parser=None, partial_variables=None, template, template_format='f-string', validate_template=True)[source] Bases: langchain.prompts.base.StringPromptTemplate Schema to represent a prompt for an LLM. Example from langchain import PromptTemplate prompt = PromptTemplate(input_variables=["foo"], template="Say {foo}") Parameters input_variables (List[str]) – output_parser (Optional[langchain.schema.BaseOutputParser]) – partial_variables (Mapping[str, Union[str, Callable[[], str]]]) – template (str) – template_format (str) – validate_template (bool) – Return type None attribute input_variables: List[str] [Required] A list of the names of the variables the prompt template expects. attribute template: str [Required] The prompt template. attribute template_format: str = 'f-string' The format of the prompt template. Options are: β€˜f-string’, β€˜jinja2’. attribute validate_template: bool = True Whether or not to try validating the template. format(**kwargs)[source] Format the prompt with the inputs. Parameters kwargs (Any) – Any arguments to be passed to the prompt template. Returns A formatted string. Return type str Example: prompt.format(variable1="foo") classmethod from_examples(examples, suffix, input_variables, example_separator='\n\n', prefix='', **kwargs)[source] Take examples in list format with prefix and suffix to create a prompt. Intended to be used as a way to dynamically create a prompt from examples. Parameters examples (List[str]) – List of examples to use in the prompt. suffix (str) – String to go after the list of examples. Should generally set up the user’s input. input_variables (List[str]) – A list of variable names the final prompt template will expect. example_separator (str) – The separator to use in between examples. Defaults to two new line characters. prefix (str) – String that should go before any examples. Generally includes examples. Default to an empty string. kwargs (Any) – Returns The final prompt generated. Return type langchain.prompts.prompt.PromptTemplate classmethod from_file(template_file, input_variables, **kwargs)[source] Load a prompt from a file. Parameters template_file (Union[str, pathlib.Path]) – The path to the file containing the prompt template. input_variables (List[str]) – A list of variable names the final prompt template will expect. kwargs (Any) – Returns The prompt loaded from the file. Return type langchain.prompts.prompt.PromptTemplate classmethod from_template(template, **kwargs)[source] Load a prompt template from a template. Parameters template (str) – kwargs (Any) – Return type langchain.prompts.prompt.PromptTemplate property lc_attributes: Dict[str, Any] Return a list of attribute names that should be included in the serialized kwargs. These attributes must be accepted by the constructor. class langchain.prompts.SemanticSimilarityExampleSelector(*, vectorstore, k=4, example_keys=None, input_keys=None)[source] Bases: langchain.prompts.example_selector.base.BaseExampleSelector, pydantic.main.BaseModel Example selector that selects examples based on SemanticSimilarity. Parameters vectorstore (langchain.vectorstores.base.VectorStore) – k (int) – example_keys (Optional[List[str]]) – input_keys (Optional[List[str]]) – Return type None attribute example_keys: Optional[List[str]] = None Optional keys to filter examples to. attribute input_keys: Optional[List[str]] = None Optional keys to filter input to. If provided, the search is based on the input variables instead of all variables. attribute k: int = 4 Number of examples to select. attribute vectorstore: langchain.vectorstores.base.VectorStore [Required] VectorStore than contains information about examples. add_example(example)[source] Add new example to vectorstore. Parameters example (Dict[str, str]) – Return type str classmethod from_examples(examples, embeddings, vectorstore_cls, k=4, input_keys=None, **vectorstore_cls_kwargs)[source] Create k-shot example selector using example list and embeddings. Reshuffles examples dynamically based on query similarity. Parameters examples (List[dict]) – List of examples to use in the prompt. embeddings (langchain.embeddings.base.Embeddings) – An initialized embedding API interface, e.g. OpenAIEmbeddings(). vectorstore_cls (Type[langchain.vectorstores.base.VectorStore]) – A vector store DB interface class, e.g. FAISS. k (int) – Number of examples to select input_keys (Optional[List[str]]) – If provided, the search is based on the input variables instead of all variables. vectorstore_cls_kwargs (Any) – optional kwargs containing url for vector store Returns The ExampleSelector instantiated, backed by a vector store. Return type langchain.prompts.example_selector.semantic_similarity.SemanticSimilarityExampleSelector select_examples(input_variables)[source] Select which examples to use based on semantic similarity. Parameters input_variables (Dict[str, str]) – Return type List[dict] class langchain.prompts.StringPromptTemplate(*, input_variables, output_parser=None, partial_variables=None)[source] Bases: langchain.prompts.base.BasePromptTemplate, abc.ABC String prompt should expose the format method, returning a prompt. Parameters input_variables (List[str]) – output_parser (Optional[langchain.schema.BaseOutputParser]) – partial_variables (Mapping[str, Union[str, Callable[[], str]]]) – Return type None format_prompt(**kwargs)[source] Create Chat Messages. Parameters kwargs (Any) – Return type langchain.schema.PromptValue class langchain.prompts.SystemMessagePromptTemplate(*, prompt, additional_kwargs=None)[source] Bases: langchain.prompts.chat.BaseStringMessagePromptTemplate Parameters prompt (langchain.prompts.base.StringPromptTemplate) – additional_kwargs (dict) – Return type None format(**kwargs)[source] To a BaseMessage. Parameters kwargs (Any) – Return type langchain.schema.BaseMessage langchain.prompts.load_prompt(path)[source] Unified method for loading a prompt from LangChainHub or local fs. Parameters path (Union[str, pathlib.Path]) – Return type langchain.prompts.base.BasePromptTemplate
https://api.python.langchain.com/en/latest/modules/prompts.html
f631b90a-9e53-4c8e-9640-d7fea38e324c
Example Selector Logic for selecting examples to include in prompts. class langchain.prompts.example_selector.LengthBasedExampleSelector(*, examples, example_prompt, get_text_length=<function _get_length_based>, max_length=2048, example_text_lengths=[])[source] Bases: langchain.prompts.example_selector.base.BaseExampleSelector, pydantic.main.BaseModel Select examples based on length. Parameters examples (List[dict]) – example_prompt (langchain.prompts.prompt.PromptTemplate) – get_text_length (Callable[[str], int]) – max_length (int) – example_text_lengths (List[int]) – Return type None attribute example_prompt: langchain.prompts.prompt.PromptTemplate [Required] Prompt template used to format the examples. attribute examples: List[dict] [Required] A list of the examples that the prompt template expects. attribute get_text_length: Callable[[str], int] = <function _get_length_based> Function to measure prompt length. Defaults to word count. attribute max_length: int = 2048 Max length for the prompt, beyond which examples are cut. add_example(example)[source] Add new example to list. Parameters example (Dict[str, str]) – Return type None select_examples(input_variables)[source] Select which examples to use based on the input lengths. Parameters input_variables (Dict[str, str]) – Return type List[dict] class langchain.prompts.example_selector.MaxMarginalRelevanceExampleSelector(*, vectorstore, k=4, example_keys=None, input_keys=None, fetch_k=20)[source] Bases: langchain.prompts.example_selector.semantic_similarity.SemanticSimilarityExampleSelector ExampleSelector that selects examples based on Max Marginal Relevance. This was shown to improve performance in this paper: https://arxiv.org/pdf/2211.13892.pdf Parameters vectorstore (langchain.vectorstores.base.VectorStore) – k (int) – example_keys (Optional[List[str]]) – input_keys (Optional[List[str]]) – fetch_k (int) – Return type None attribute fetch_k: int = 20 Number of examples to fetch to rerank. classmethod from_examples(examples, embeddings, vectorstore_cls, k=4, input_keys=None, fetch_k=20, **vectorstore_cls_kwargs)[source] Create k-shot example selector using example list and embeddings. Reshuffles examples dynamically based on query similarity. Parameters examples (List[dict]) – List of examples to use in the prompt. embeddings (langchain.embeddings.base.Embeddings) – An iniialized embedding API interface, e.g. OpenAIEmbeddings(). vectorstore_cls (Type[langchain.vectorstores.base.VectorStore]) – A vector store DB interface class, e.g. FAISS. k (int) – Number of examples to select input_keys (Optional[List[str]]) – If provided, the search is based on the input variables instead of all variables. vectorstore_cls_kwargs (Any) – optional kwargs containing url for vector store fetch_k (int) – Returns The ExampleSelector instantiated, backed by a vector store. Return type langchain.prompts.example_selector.semantic_similarity.MaxMarginalRelevanceExampleSelector select_examples(input_variables)[source] Select which examples to use based on semantic similarity. Parameters input_variables (Dict[str, str]) – Return type List[dict] class langchain.prompts.example_selector.NGramOverlapExampleSelector(*, examples, example_prompt, threshold=- 1.0)[source] Bases: langchain.prompts.example_selector.base.BaseExampleSelector, pydantic.main.BaseModel Select and order examples based on ngram overlap score (sentence_bleu score). https://www.nltk.org/_modules/nltk/translate/bleu_score.html https://aclanthology.org/P02-1040.pdf Parameters examples (List[dict]) – example_prompt (langchain.prompts.prompt.PromptTemplate) – threshold (float) – Return type None attribute example_prompt: langchain.prompts.prompt.PromptTemplate [Required] Prompt template used to format the examples. attribute examples: List[dict] [Required] A list of the examples that the prompt template expects. attribute threshold: float = -1.0 Threshold at which algorithm stops. Set to -1.0 by default. For negative threshold: select_examples sorts examples by ngram_overlap_score, but excludes none. For threshold greater than 1.0: select_examples excludes all examples, and returns an empty list. For threshold equal to 0.0: select_examples sorts examples by ngram_overlap_score, and excludes examples with no ngram overlap with input. add_example(example)[source] Add new example to list. Parameters example (Dict[str, str]) – Return type None select_examples(input_variables)[source] Return list of examples sorted by ngram_overlap_score with input. Descending order. Excludes any examples with ngram_overlap_score less than or equal to threshold. Parameters input_variables (Dict[str, str]) – Return type List[dict] class langchain.prompts.example_selector.SemanticSimilarityExampleSelector(*, vectorstore, k=4, example_keys=None, input_keys=None)[source] Bases: langchain.prompts.example_selector.base.BaseExampleSelector, pydantic.main.BaseModel Example selector that selects examples based on SemanticSimilarity. Parameters vectorstore (langchain.vectorstores.base.VectorStore) – k (int) – example_keys (Optional[List[str]]) – input_keys (Optional[List[str]]) – Return type None attribute example_keys: Optional[List[str]] = None Optional keys to filter examples to. attribute input_keys: Optional[List[str]] = None Optional keys to filter input to. If provided, the search is based on the input variables instead of all variables. attribute k: int = 4 Number of examples to select. attribute vectorstore: langchain.vectorstores.base.VectorStore [Required] VectorStore than contains information about examples. add_example(example)[source] Add new example to vectorstore. Parameters example (Dict[str, str]) – Return type str classmethod from_examples(examples, embeddings, vectorstore_cls, k=4, input_keys=None, **vectorstore_cls_kwargs)[source] Create k-shot example selector using example list and embeddings. Reshuffles examples dynamically based on query similarity. Parameters examples (List[dict]) – List of examples to use in the prompt. embeddings (langchain.embeddings.base.Embeddings) – An initialized embedding API interface, e.g. OpenAIEmbeddings(). vectorstore_cls (Type[langchain.vectorstores.base.VectorStore]) – A vector store DB interface class, e.g. FAISS. k (int) – Number of examples to select input_keys (Optional[List[str]]) – If provided, the search is based on the input variables instead of all variables. vectorstore_cls_kwargs (Any) – optional kwargs containing url for vector store Returns The ExampleSelector instantiated, backed by a vector store. Return type langchain.prompts.example_selector.semantic_similarity.SemanticSimilarityExampleSelector select_examples(input_variables)[source] Select which examples to use based on semantic similarity. Parameters input_variables (Dict[str, str]) – Return type List[dict]
https://api.python.langchain.com/en/latest/modules/example_selector.html
cfa50aec-8749-4b8e-af15-9d7641c9caa9
Document Transformers Transform documents langchain.document_transformers.get_stateful_documents(documents)[source] Convert a list of documents to a list of documents with state. Parameters documents (Sequence[langchain.schema.Document]) – The documents to convert. Returns A list of documents with state. Return type Sequence[langchain.document_transformers._DocumentWithState] class langchain.document_transformers.EmbeddingsRedundantFilter(*, embeddings, similarity_fn=<function cosine_similarity>, similarity_threshold=0.95)[source] Bases: langchain.schema.BaseDocumentTransformer, pydantic.main.BaseModel Filter that drops redundant documents by comparing their embeddings. Parameters embeddings (langchain.embeddings.base.Embeddings) – similarity_fn (Callable) – similarity_threshold (float) – Return type None attribute embeddings: langchain.embeddings.base.Embeddings [Required] Embeddings to use for embedding document contents. attribute similarity_fn: Callable = <function cosine_similarity> Similarity function for comparing documents. Function expected to take as input two matrices (List[List[float]]) and return a matrix of scores where higher values indicate greater similarity. attribute similarity_threshold: float = 0.95 Threshold for determining when two documents are similar enough to be considered redundant. async atransform_documents(documents, **kwargs)[source] Asynchronously transform a list of documents. Parameters documents (Sequence[langchain.schema.Document]) – kwargs (Any) – Return type Sequence[langchain.schema.Document] transform_documents(documents, **kwargs)[source] Filter down documents. Parameters documents (Sequence[langchain.schema.Document]) – kwargs (Any) – Return type Sequence[langchain.schema.Document] Text Splitters Functionality for splitting text. class langchain.text_splitter.TextSplitter(chunk_size=4000, chunk_overlap=200, length_function=<built-in function len>, keep_separator=False, add_start_index=False)[source] Bases: langchain.schema.BaseDocumentTransformer, abc.ABC Interface for splitting text into chunks. Parameters chunk_size (int) – chunk_overlap (int) – length_function (Callable[[str], int]) – keep_separator (bool) – add_start_index (bool) – Return type None abstract split_text(text)[source] Split text into multiple components. Parameters text (str) – Return type List[str] create_documents(texts, metadatas=None)[source] Create documents from a list of texts. Parameters texts (List[str]) – metadatas (Optional[List[dict]]) – Return type List[langchain.schema.Document] split_documents(documents)[source] Split documents. Parameters documents (Iterable[langchain.schema.Document]) – Return type List[langchain.schema.Document] classmethod from_huggingface_tokenizer(tokenizer, **kwargs)[source] Text splitter that uses HuggingFace tokenizer to count length. Parameters tokenizer (Any) – kwargs (Any) – Return type langchain.text_splitter.TextSplitter classmethod from_tiktoken_encoder(encoding_name='gpt2', model_name=None, allowed_special={}, disallowed_special='all', **kwargs)[source] Text splitter that uses tiktoken encoder to count length. Parameters encoding_name (str) – model_name (Optional[str]) – allowed_special (Union[Literal['all'], typing.AbstractSet[str]]) – disallowed_special (Union[Literal['all'], typing.Collection[str]]) – kwargs (Any) – Return type langchain.text_splitter.TS transform_documents(documents, **kwargs)[source] Transform sequence of documents by splitting them. Parameters documents (Sequence[langchain.schema.Document]) – kwargs (Any) – Return type Sequence[langchain.schema.Document] async atransform_documents(documents, **kwargs)[source] Asynchronously transform a sequence of documents by splitting them. Parameters documents (Sequence[langchain.schema.Document]) – kwargs (Any) – Return type Sequence[langchain.schema.Document] class langchain.text_splitter.CharacterTextSplitter(separator='\n\n', **kwargs)[source] Bases: langchain.text_splitter.TextSplitter Implementation of splitting text that looks at characters. Parameters separator (str) – kwargs (Any) – Return type None split_text(text)[source] Split incoming text and return chunks. Parameters text (str) – Return type List[str] class langchain.text_splitter.LineType[source] Bases: TypedDict Line type as typed dict. metadata: Dict[str, str] content: str class langchain.text_splitter.HeaderType[source] Bases: TypedDict Header type as typed dict. level: int name: str data: str class langchain.text_splitter.MarkdownHeaderTextSplitter(headers_to_split_on, return_each_line=False)[source] Bases: object Implementation of splitting markdown files based on specified headers. Parameters headers_to_split_on (List[Tuple[str, str]]) – return_each_line (bool) – aggregate_lines_to_chunks(lines)[source] Combine lines with common metadata into chunks :param lines: Line of text / associated header metadata Parameters lines (List[langchain.text_splitter.LineType]) – Return type List[langchain.schema.Document] split_text(text)[source] Split markdown file :param text: Markdown file Parameters text (str) – Return type List[langchain.schema.Document] class langchain.text_splitter.Tokenizer(chunk_overlap: 'int', tokens_per_chunk: 'int', decode: 'Callable[[list[int]], str]', encode: 'Callable[[str], List[int]]')[source] Bases: object Parameters chunk_overlap (int) – tokens_per_chunk (int) – decode (Callable[[list[int]], str]) – encode (Callable[[str], List[int]]) – Return type None chunk_overlap: int tokens_per_chunk: int decode: Callable[[list[int]], str] encode: Callable[[str], List[int]] langchain.text_splitter.split_text_on_tokens(*, text, tokenizer)[source] Split incoming text and return chunks. Parameters text (str) – tokenizer (langchain.text_splitter.Tokenizer) – Return type List[str] class langchain.text_splitter.TokenTextSplitter(encoding_name='gpt2', model_name=None, allowed_special={}, disallowed_special='all', **kwargs)[source] Bases: langchain.text_splitter.TextSplitter Implementation of splitting text that looks at tokens. Parameters encoding_name (str) – model_name (Optional[str]) – allowed_special (Union[Literal['all'], AbstractSet[str]]) – disallowed_special (Union[Literal['all'], Collection[str]]) – kwargs (Any) – Return type None split_text(text)[source] Split text into multiple components. Parameters text (str) – Return type List[str] class langchain.text_splitter.SentenceTransformersTokenTextSplitter(chunk_overlap=50, model_name='sentence-transformers/all-mpnet-base-v2', tokens_per_chunk=None, **kwargs)[source] Bases: langchain.text_splitter.TextSplitter Implementation of splitting text that looks at tokens. Parameters chunk_overlap (int) – model_name (str) – tokens_per_chunk (Optional[int]) – kwargs (Any) – Return type None split_text(text)[source] Split text into multiple components. Parameters text (str) – Return type List[str] count_tokens(*, text)[source] Parameters text (str) – Return type int class langchain.text_splitter.Language(value, names=None, *, module=None, qualname=None, type=None, start=1, boundary=None)[source] Bases: str, enum.Enum CPP = 'cpp' GO = 'go' JAVA = 'java' JS = 'js' PHP = 'php' PROTO = 'proto' PYTHON = 'python' RST = 'rst' RUBY = 'ruby' RUST = 'rust' SCALA = 'scala' SWIFT = 'swift' MARKDOWN = 'markdown' LATEX = 'latex' HTML = 'html' SOL = 'sol' class langchain.text_splitter.RecursiveCharacterTextSplitter(separators=None, keep_separator=True, **kwargs)[source] Bases: langchain.text_splitter.TextSplitter Implementation of splitting text that looks at characters. Recursively tries to split by different characters to find one that works. Parameters separators (Optional[List[str]]) – keep_separator (bool) – kwargs (Any) – Return type None split_text(text)[source] Split text into multiple components. Parameters text (str) – Return type List[str] classmethod from_language(language, **kwargs)[source] Parameters language (langchain.text_splitter.Language) – kwargs (Any) – Return type langchain.text_splitter.RecursiveCharacterTextSplitter static get_separators_for_language(language)[source] Parameters language (langchain.text_splitter.Language) – Return type List[str] class langchain.text_splitter.NLTKTextSplitter(separator='\n\n', **kwargs)[source] Bases: langchain.text_splitter.TextSplitter Implementation of splitting text that looks at sentences using NLTK. Parameters separator (str) – kwargs (Any) – Return type None split_text(text)[source] Split incoming text and return chunks. Parameters text (str) – Return type List[str] class langchain.text_splitter.SpacyTextSplitter(separator='\n\n', pipeline='en_core_web_sm', **kwargs)[source] Bases: langchain.text_splitter.TextSplitter Implementation of splitting text that looks at sentences using Spacy. Parameters separator (str) – pipeline (str) – kwargs (Any) – Return type None split_text(text)[source] Split incoming text and return chunks. Parameters text (str) – Return type List[str] class langchain.text_splitter.PythonCodeTextSplitter(**kwargs)[source] Bases: langchain.text_splitter.RecursiveCharacterTextSplitter Attempts to split the text along Python syntax. Parameters kwargs (Any) – Return type None class langchain.text_splitter.MarkdownTextSplitter(**kwargs)[source] Bases: langchain.text_splitter.RecursiveCharacterTextSplitter Attempts to split the text along Markdown-formatted headings. Parameters kwargs (Any) – Return type None class langchain.text_splitter.LatexTextSplitter(**kwargs)[source] Bases: langchain.text_splitter.RecursiveCharacterTextSplitter Attempts to split the text along Latex-formatted layout elements. Parameters kwargs (Any) – Return type None
https://api.python.langchain.com/en/latest/modules/document_transformers.html
a697e986-4657-4695-97fa-2c921a92a446
Agents Interface for agents. class langchain.agents.Agent(*, llm_chain, output_parser, allowed_tools=None)[source] Bases: langchain.agents.agent.BaseSingleActionAgent Class responsible for calling the language model and deciding the action. This is driven by an LLMChain. The prompt in the LLMChain MUST include a variable called β€œagent_scratchpad” where the agent can put its intermediary work. Parameters llm_chain (langchain.chains.llm.LLMChain) – output_parser (langchain.agents.agent.AgentOutputParser) – allowed_tools (Optional[List[str]]) – Return type None attribute allowed_tools: Optional[List[str]] = None attribute llm_chain: langchain.chains.llm.LLMChain [Required] attribute output_parser: langchain.agents.agent.AgentOutputParser [Required] async aplan(intermediate_steps, callbacks=None, **kwargs)[source] Given input, decided what to do. Parameters intermediate_steps (List[Tuple[langchain.schema.AgentAction, str]]) – Steps the LLM has taken to date, along with observations callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to run. **kwargs – User inputs. kwargs (Any) – Returns Action specifying what tool to use. Return type Union[langchain.schema.AgentAction, langchain.schema.AgentFinish] abstract classmethod create_prompt(tools)[source] Create a prompt for this class. Parameters tools (Sequence[langchain.tools.base.BaseTool]) – Return type langchain.prompts.base.BasePromptTemplate dict(**kwargs)[source] Return dictionary representation of agent. Parameters kwargs (Any) – Return type Dict classmethod from_llm_and_tools(llm, tools, callback_manager=None, output_parser=None, **kwargs)[source] Construct an agent from an LLM and tools. Parameters llm (langchain.base_language.BaseLanguageModel) – tools (Sequence[langchain.tools.base.BaseTool]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – output_parser (Optional[langchain.agents.agent.AgentOutputParser]) – kwargs (Any) – Return type langchain.agents.agent.Agent get_allowed_tools()[source] Return type Optional[List[str]] get_full_inputs(intermediate_steps, **kwargs)[source] Create the full inputs for the LLMChain from intermediate steps. Parameters intermediate_steps (List[Tuple[langchain.schema.AgentAction, str]]) – kwargs (Any) – Return type Dict[str, Any] plan(intermediate_steps, callbacks=None, **kwargs)[source] Given input, decided what to do. Parameters intermediate_steps (List[Tuple[langchain.schema.AgentAction, str]]) – Steps the LLM has taken to date, along with observations callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to run. **kwargs – User inputs. kwargs (Any) – Returns Action specifying what tool to use. Return type Union[langchain.schema.AgentAction, langchain.schema.AgentFinish] return_stopped_response(early_stopping_method, intermediate_steps, **kwargs)[source] Return response when agent has been stopped due to max iterations. Parameters early_stopping_method (str) – intermediate_steps (List[Tuple[langchain.schema.AgentAction, str]]) – kwargs (Any) – Return type langchain.schema.AgentFinish tool_run_logging_kwargs()[source] Return type Dict abstract property llm_prefix: str Prefix to append the LLM call with. abstract property observation_prefix: str Prefix to append the observation with. property return_values: List[str] Return values of the agent. class langchain.agents.AgentExecutor(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, agent, tools, return_intermediate_steps=False, max_iterations=15, max_execution_time=None, early_stopping_method='force', handle_parsing_errors=False)[source] Bases: langchain.chains.base.Chain Consists of an agent using tools. Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – agent (Union[langchain.agents.agent.BaseSingleActionAgent, langchain.agents.agent.BaseMultiActionAgent]) – tools (Sequence[langchain.tools.base.BaseTool]) – return_intermediate_steps (bool) – max_iterations (Optional[int]) – max_execution_time (Optional[float]) – early_stopping_method (str) – handle_parsing_errors (Union[bool, str, Callable[[langchain.schema.OutputParserException], str]]) – Return type None attribute agent: Union[BaseSingleActionAgent, BaseMultiActionAgent] [Required] The agent to run for creating a plan and determining actions to take at each step of the execution loop. attribute early_stopping_method: str = 'force' The method to use for early stopping if the agent never returns AgentFinish. Either β€˜force’ or β€˜generate’. β€œforce” returns a string saying that it stopped because it met atime or iteration limit. β€œgenerate” calls the agent’s LLM Chain one final time to generatea final answer based on the previous steps. attribute handle_parsing_errors: Union[bool, str, Callable[[OutputParserException], str]] = False How to handle errors raised by the agent’s output parser.Defaults to False, which raises the error. sIf true, the error will be sent back to the LLM as an observation. If a string, the string itself will be sent to the LLM as an observation. If a callable function, the function will be called with the exception as an argument, and the result of that function will be passed to the agentas an observation. attribute max_execution_time: Optional[float] = None The maximum amount of wall clock time to spend in the execution loop. attribute max_iterations: Optional[int] = 15 The maximum number of steps to take before ending the execution loop. Setting to β€˜None’ could lead to an infinite loop. attribute return_intermediate_steps: bool = False Whether to return the agent’s trajectory of intermediate steps at the end in addition to the final output. attribute tools: Sequence[BaseTool] [Required] The valid tools the agent can call. classmethod from_agent_and_tools(agent, tools, callback_manager=None, **kwargs)[source] Create from agent and tools. Parameters agent (Union[langchain.agents.agent.BaseSingleActionAgent, langchain.agents.agent.BaseMultiActionAgent]) – tools (Sequence[langchain.tools.base.BaseTool]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – kwargs (Any) – Return type langchain.agents.agent.AgentExecutor lookup_tool(name)[source] Lookup tool by name. Parameters name (str) – Return type langchain.tools.base.BaseTool save(file_path)[source] Raise error - saving not supported for Agent Executors. Parameters file_path (Union[pathlib.Path, str]) – Return type None save_agent(file_path)[source] Save the underlying agent. Parameters file_path (Union[pathlib.Path, str]) – Return type None class langchain.agents.AgentOutputParser[source] Bases: langchain.schema.BaseOutputParser Return type None abstract parse(text)[source] Parse text into agent action/finish. Parameters text (str) – Return type Union[langchain.schema.AgentAction, langchain.schema.AgentFinish] class langchain.agents.AgentType(value, names=None, *, module=None, qualname=None, type=None, start=1, boundary=None)[source] Bases: str, enum.Enum Enumerator with the Agent types. ZERO_SHOT_REACT_DESCRIPTION = 'zero-shot-react-description' REACT_DOCSTORE = 'react-docstore' SELF_ASK_WITH_SEARCH = 'self-ask-with-search' CONVERSATIONAL_REACT_DESCRIPTION = 'conversational-react-description' CHAT_ZERO_SHOT_REACT_DESCRIPTION = 'chat-zero-shot-react-description' CHAT_CONVERSATIONAL_REACT_DESCRIPTION = 'chat-conversational-react-description' STRUCTURED_CHAT_ZERO_SHOT_REACT_DESCRIPTION = 'structured-chat-zero-shot-react-description' OPENAI_FUNCTIONS = 'openai-functions' OPENAI_MULTI_FUNCTIONS = 'openai-multi-functions' class langchain.agents.BaseMultiActionAgent[source] Bases: pydantic.main.BaseModel Base Agent class. Return type None abstract async aplan(intermediate_steps, callbacks=None, **kwargs)[source] Given input, decided what to do. Parameters intermediate_steps (List[Tuple[langchain.schema.AgentAction, str]]) – Steps the LLM has taken to date, along with observations callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to run. **kwargs – User inputs. kwargs (Any) – Returns Actions specifying what tool to use. Return type Union[List[langchain.schema.AgentAction], langchain.schema.AgentFinish] dict(**kwargs)[source] Return dictionary representation of agent. Parameters kwargs (Any) – Return type Dict get_allowed_tools()[source] Return type Optional[List[str]] abstract plan(intermediate_steps, callbacks=None, **kwargs)[source] Given input, decided what to do. Parameters intermediate_steps (List[Tuple[langchain.schema.AgentAction, str]]) – Steps the LLM has taken to date, along with observations callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to run. **kwargs – User inputs. kwargs (Any) – Returns Actions specifying what tool to use. Return type Union[List[langchain.schema.AgentAction], langchain.schema.AgentFinish] return_stopped_response(early_stopping_method, intermediate_steps, **kwargs)[source] Return response when agent has been stopped due to max iterations. Parameters early_stopping_method (str) – intermediate_steps (List[Tuple[langchain.schema.AgentAction, str]]) – kwargs (Any) – Return type langchain.schema.AgentFinish save(file_path)[source] Save the agent. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the agent to. Return type None Example: .. code-block:: python # If working with agent executor agent.agent.save(file_path=”path/agent.yaml”) tool_run_logging_kwargs()[source] Return type Dict property return_values: List[str] Return values of the agent. class langchain.agents.BaseSingleActionAgent[source] Bases: pydantic.main.BaseModel Base Agent class. Return type None abstract async aplan(intermediate_steps, callbacks=None, **kwargs)[source] Given input, decided what to do. Parameters intermediate_steps (List[Tuple[langchain.schema.AgentAction, str]]) – Steps the LLM has taken to date, along with observations callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to run. **kwargs – User inputs. kwargs (Any) – Returns Action specifying what tool to use. Return type Union[langchain.schema.AgentAction, langchain.schema.AgentFinish] dict(**kwargs)[source] Return dictionary representation of agent. Parameters kwargs (Any) – Return type Dict classmethod from_llm_and_tools(llm, tools, callback_manager=None, **kwargs)[source] Parameters llm (langchain.base_language.BaseLanguageModel) – tools (Sequence[langchain.tools.base.BaseTool]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – kwargs (Any) – Return type langchain.agents.agent.BaseSingleActionAgent get_allowed_tools()[source] Return type Optional[List[str]] abstract plan(intermediate_steps, callbacks=None, **kwargs)[source] Given input, decided what to do. Parameters intermediate_steps (List[Tuple[langchain.schema.AgentAction, str]]) – Steps the LLM has taken to date, along with observations callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to run. **kwargs – User inputs. kwargs (Any) – Returns Action specifying what tool to use. Return type Union[langchain.schema.AgentAction, langchain.schema.AgentFinish] return_stopped_response(early_stopping_method, intermediate_steps, **kwargs)[source] Return response when agent has been stopped due to max iterations. Parameters early_stopping_method (str) – intermediate_steps (List[Tuple[langchain.schema.AgentAction, str]]) – kwargs (Any) – Return type langchain.schema.AgentFinish save(file_path)[source] Save the agent. Parameters file_path (Union[pathlib.Path, str]) – Path to file to save the agent to. Return type None Example: .. code-block:: python # If working with agent executor agent.agent.save(file_path=”path/agent.yaml”) tool_run_logging_kwargs()[source] Return type Dict property return_values: List[str] Return values of the agent. class langchain.agents.ConversationalAgent(*, llm_chain, output_parser=None, allowed_tools=None, ai_prefix='AI')[source] Bases: langchain.agents.agent.Agent An agent designed to hold a conversation in addition to using tools. Parameters llm_chain (langchain.chains.llm.LLMChain) – output_parser (langchain.agents.agent.AgentOutputParser) – allowed_tools (Optional[List[str]]) – ai_prefix (str) – Return type None attribute ai_prefix: str = 'AI' attribute output_parser: langchain.agents.agent.AgentOutputParser [Optional] classmethod create_prompt(tools, prefix='Assistant is a large language model trained by OpenAI.\n\nAssistant is designed to be able to assist with a wide range of tasks, from answering simple questions to providing in-depth explanations and discussions on a wide range of topics. As a language model, Assistant is able to generate human-like text based on the input it receives, allowing it to engage in natural-sounding conversations and provide responses that are coherent and relevant to the topic at hand.\n\nAssistant is constantly learning and improving, and its capabilities are constantly evolving. It is able to process and understand large amounts of text, and can use this knowledge to provide accurate and informative responses to a wide range of questions. Additionally, Assistant is able to generate its own text based on the input it receives, allowing it to engage in discussions and provide explanations and descriptions on a wide range of topics.\n\nOverall, Assistant is a powerful tool that can help with a wide range of tasks and provide valuable insights and information on a wide range of topics. Whether you need help with a specific question or just want to have a conversation about a particular topic, Assistant is here to assist.\n\nTOOLS:\n------\n\nAssistant has access to the following tools:', suffix='Begin!\n\nPrevious conversation history:\n{chat_history}\n\nNew input: {input}\n{agent_scratchpad}', format_instructions='To use a tool, please use the following format:\n\n```\nThought: Do I need to use a tool? Yes\nAction: the action to take, should be one of [{tool_names}]\nAction Input: the input to the action\nObservation: the result of the action\n```\n\nWhen you have a response to say to the Human, or if you do not need to use a tool, you MUST use the format:\n\n```\nThought: Do I need to use a tool? No\n{ai_prefix}: [your response here]\n```', ai_prefix='AI', human_prefix='Human', input_variables=None)[source] Create prompt in the style of the zero shot agent. Parameters tools (Sequence[langchain.tools.base.BaseTool]) – List of tools the agent will have access to, used to format the prompt. prefix (str) – String to put before the list of tools. suffix (str) – String to put after the list of tools. ai_prefix (str) – String to use before AI output. human_prefix (str) – String to use before human output. input_variables (Optional[List[str]]) – List of input variables the final prompt will expect. format_instructions (str) – Returns A PromptTemplate with the template assembled from the pieces here. Return type langchain.prompts.prompt.PromptTemplate classmethod from_llm_and_tools(llm, tools, callback_manager=None, output_parser=None, prefix='Assistant is a large language model trained by OpenAI.\n\nAssistant is designed to be able to assist with a wide range of tasks, from answering simple questions to providing in-depth explanations and discussions on a wide range of topics. As a language model, Assistant is able to generate human-like text based on the input it receives, allowing it to engage in natural-sounding conversations and provide responses that are coherent and relevant to the topic at hand.\n\nAssistant is constantly learning and improving, and its capabilities are constantly evolving. It is able to process and understand large amounts of text, and can use this knowledge to provide accurate and informative responses to a wide range of questions. Additionally, Assistant is able to generate its own text based on the input it receives, allowing it to engage in discussions and provide explanations and descriptions on a wide range of topics.\n\nOverall, Assistant is a powerful tool that can help with a wide range of tasks and provide valuable insights and information on a wide range of topics. Whether you need help with a specific question or just want to have a conversation about a particular topic, Assistant is here to assist.\n\nTOOLS:\n------\n\nAssistant has access to the following tools:', suffix='Begin!\n\nPrevious conversation history:\n{chat_history}\n\nNew input: {input}\n{agent_scratchpad}', format_instructions='To use a tool, please use the following format:\n\n```\nThought: Do I need to use a tool? Yes\nAction: the action to take, should be one of [{tool_names}]\nAction Input: the input to the action\nObservation: the result of the action\n```\n\nWhen you have a response to say to the Human, or if you do not need to use a tool, you MUST use the format:\n\n```\nThought: Do I need to use a tool? No\n{ai_prefix}: [your response here]\n```', ai_prefix='AI', human_prefix='Human', input_variables=None, **kwargs)[source] Construct an agent from an LLM and tools. Parameters llm (langchain.base_language.BaseLanguageModel) – tools (Sequence[langchain.tools.base.BaseTool]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – output_parser (Optional[langchain.agents.agent.AgentOutputParser]) – prefix (str) – suffix (str) – format_instructions (str) – ai_prefix (str) – human_prefix (str) – input_variables (Optional[List[str]]) – kwargs (Any) – Return type langchain.agents.agent.Agent property llm_prefix: str Prefix to append the llm call with. property observation_prefix: str Prefix to append the observation with. class langchain.agents.ConversationalChatAgent(*, llm_chain, output_parser=None, allowed_tools=None, template_tool_response="TOOL RESPONSE: \n---------------------\n{observation}\n\nUSER'S INPUT\n--------------------\n\nOkay, so what is the response to my last comment? If using information obtained from the tools you must mention it explicitly without mentioning the tool names - I have forgotten all TOOL RESPONSES! Remember to respond with a markdown code snippet of a json blob with a single action, and NOTHING else.")[source] Bases: langchain.agents.agent.Agent An agent designed to hold a conversation in addition to using tools. Parameters llm_chain (langchain.chains.llm.LLMChain) – output_parser (langchain.agents.agent.AgentOutputParser) – allowed_tools (Optional[List[str]]) – template_tool_response (str) – Return type None attribute output_parser: langchain.agents.agent.AgentOutputParser [Optional] attribute template_tool_response: str = "TOOL RESPONSE: \n---------------------\n{observation}\n\nUSER'S INPUT\n--------------------\n\nOkay, so what is the response to my last comment? If using information obtained from the tools you must mention it explicitly without mentioning the tool names - I have forgotten all TOOL RESPONSES! Remember to respond with a markdown code snippet of a json blob with a single action, and NOTHING else." classmethod create_prompt(tools, system_message='Assistant is a large language model trained by OpenAI.\n\nAssistant is designed to be able to assist with a wide range of tasks, from answering simple questions to providing in-depth explanations and discussions on a wide range of topics. As a language model, Assistant is able to generate human-like text based on the input it receives, allowing it to engage in natural-sounding conversations and provide responses that are coherent and relevant to the topic at hand.\n\nAssistant is constantly learning and improving, and its capabilities are constantly evolving. It is able to process and understand large amounts of text, and can use this knowledge to provide accurate and informative responses to a wide range of questions. Additionally, Assistant is able to generate its own text based on the input it receives, allowing it to engage in discussions and provide explanations and descriptions on a wide range of topics.\n\nOverall, Assistant is a powerful system that can help with a wide range of tasks and provide valuable insights and information on a wide range of topics. Whether you need help with a specific question or just want to have a conversation about a particular topic, Assistant is here to assist.', human_message="TOOLS\n------\nAssistant can ask the user to use tools to look up information that may be helpful in answering the users original question. The tools the human can use are:\n\n{{tools}}\n\n{format_instructions}\n\nUSER'S INPUT\n--------------------\nHere is the user's input (remember to respond with a markdown code snippet of a json blob with a single action, and NOTHING else):\n\n{{{{input}}}}", input_variables=None, output_parser=None)[source] Create a prompt for this class. Parameters tools (Sequence[langchain.tools.base.BaseTool]) – system_message (str) – human_message (str) – input_variables (Optional[List[str]]) – output_parser (Optional[langchain.schema.BaseOutputParser]) – Return type langchain.prompts.base.BasePromptTemplate classmethod from_llm_and_tools(llm, tools, callback_manager=None, output_parser=None, system_message='Assistant is a large language model trained by OpenAI.\n\nAssistant is designed to be able to assist with a wide range of tasks, from answering simple questions to providing in-depth explanations and discussions on a wide range of topics. As a language model, Assistant is able to generate human-like text based on the input it receives, allowing it to engage in natural-sounding conversations and provide responses that are coherent and relevant to the topic at hand.\n\nAssistant is constantly learning and improving, and its capabilities are constantly evolving. It is able to process and understand large amounts of text, and can use this knowledge to provide accurate and informative responses to a wide range of questions. Additionally, Assistant is able to generate its own text based on the input it receives, allowing it to engage in discussions and provide explanations and descriptions on a wide range of topics.\n\nOverall, Assistant is a powerful system that can help with a wide range of tasks and provide valuable insights and information on a wide range of topics. Whether you need help with a specific question or just want to have a conversation about a particular topic, Assistant is here to assist.', human_message="TOOLS\n------\nAssistant can ask the user to use tools to look up information that may be helpful in answering the users original question. The tools the human can use are:\n\n{{tools}}\n\n{format_instructions}\n\nUSER'S INPUT\n--------------------\nHere is the user's input (remember to respond with a markdown code snippet of a json blob with a single action, and NOTHING else):\n\n{{{{input}}}}", input_variables=None, **kwargs)[source] Construct an agent from an LLM and tools. Parameters llm (langchain.base_language.BaseLanguageModel) – tools (Sequence[langchain.tools.base.BaseTool]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – output_parser (Optional[langchain.agents.agent.AgentOutputParser]) – system_message (str) – human_message (str) – input_variables (Optional[List[str]]) – kwargs (Any) – Return type langchain.agents.agent.Agent property llm_prefix: str Prefix to append the llm call with. property observation_prefix: str Prefix to append the observation with. class langchain.agents.LLMSingleActionAgent(*, llm_chain, output_parser, stop)[source] Bases: langchain.agents.agent.BaseSingleActionAgent Parameters llm_chain (langchain.chains.llm.LLMChain) – output_parser (langchain.agents.agent.AgentOutputParser) – stop (List[str]) – Return type None attribute llm_chain: langchain.chains.llm.LLMChain [Required] attribute output_parser: langchain.agents.agent.AgentOutputParser [Required] attribute stop: List[str] [Required] async aplan(intermediate_steps, callbacks=None, **kwargs)[source] Given input, decided what to do. Parameters intermediate_steps (List[Tuple[langchain.schema.AgentAction, str]]) – Steps the LLM has taken to date, along with observations callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to run. **kwargs – User inputs. kwargs (Any) – Returns Action specifying what tool to use. Return type Union[langchain.schema.AgentAction, langchain.schema.AgentFinish] dict(**kwargs)[source] Return dictionary representation of agent. Parameters kwargs (Any) – Return type Dict plan(intermediate_steps, callbacks=None, **kwargs)[source] Given input, decided what to do. Parameters intermediate_steps (List[Tuple[langchain.schema.AgentAction, str]]) – Steps the LLM has taken to date, along with observations callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Callbacks to run. **kwargs – User inputs. kwargs (Any) – Returns Action specifying what tool to use. Return type Union[langchain.schema.AgentAction, langchain.schema.AgentFinish] tool_run_logging_kwargs()[source] Return type Dict class langchain.agents.MRKLChain(*, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, agent, tools, return_intermediate_steps=False, max_iterations=15, max_execution_time=None, early_stopping_method='force', handle_parsing_errors=False)[source] Bases: langchain.agents.agent.AgentExecutor Chain that implements the MRKL system. Example from langchain import OpenAI, MRKLChain from langchain.chains.mrkl.base import ChainConfig llm = OpenAI(temperature=0) prompt = PromptTemplate(...) chains = [...] mrkl = MRKLChain.from_chains(llm=llm, prompt=prompt) Parameters memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – agent (Union[langchain.agents.agent.BaseSingleActionAgent, langchain.agents.agent.BaseMultiActionAgent]) – tools (Sequence[langchain.tools.base.BaseTool]) – return_intermediate_steps (bool) – max_iterations (Optional[int]) – max_execution_time (Optional[float]) – early_stopping_method (str) – handle_parsing_errors (Union[bool, str, Callable[[langchain.schema.OutputParserException], str]]) – Return type None classmethod from_chains(llm, chains, **kwargs)[source] User friendly way to initialize the MRKL chain. This is intended to be an easy way to get up and running with the MRKL chain. Parameters llm (langchain.base_language.BaseLanguageModel) – The LLM to use as the agent LLM. chains (List[langchain.agents.mrkl.base.ChainConfig]) – The chains the MRKL system has access to. **kwargs – parameters to be passed to initialization. kwargs (Any) – Returns An initialized MRKL chain. Return type langchain.agents.agent.AgentExecutor Example from langchain import LLMMathChain, OpenAI, SerpAPIWrapper, MRKLChain from langchain.chains.mrkl.base import ChainConfig llm = OpenAI(temperature=0) search = SerpAPIWrapper() llm_math_chain = LLMMathChain(llm=llm) chains = [ ChainConfig( action_name = "Search", action=search.search, action_description="useful for searching" ), ChainConfig( action_name="Calculator", action=llm_math_chain.run, action_description="useful for doing math" ) ] mrkl = MRKLChain.from_chains(llm, chains) class langchain.agents.OpenAIFunctionsAgent(*, llm, tools, prompt)[source] Bases: langchain.agents.agent.BaseSingleActionAgent An Agent driven by OpenAIs function powered API. Parameters llm (langchain.base_language.BaseLanguageModel) – This should be an instance of ChatOpenAI, specifically a model that supports using functions. tools (Sequence[langchain.tools.base.BaseTool]) – The tools this agent has access to. prompt (langchain.prompts.base.BasePromptTemplate) – The prompt for this agent, should support agent_scratchpad as one of the variables. For an easy way to construct this prompt, use OpenAIFunctionsAgent.create_prompt(…) Return type None attribute llm: langchain.base_language.BaseLanguageModel [Required] attribute prompt: langchain.prompts.base.BasePromptTemplate [Required] attribute tools: Sequence[langchain.tools.base.BaseTool] [Required] async aplan(intermediate_steps, callbacks=None, **kwargs)[source] Given input, decided what to do. Parameters intermediate_steps (List[Tuple[langchain.schema.AgentAction, str]]) – Steps the LLM has taken to date, along with observations **kwargs – User inputs. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Returns Action specifying what tool to use. Return type Union[langchain.schema.AgentAction, langchain.schema.AgentFinish] classmethod create_prompt(system_message=SystemMessage(content='You are a helpful AI assistant.', additional_kwargs={}), extra_prompt_messages=None)[source] Create prompt for this agent. Parameters system_message (Optional[langchain.schema.SystemMessage]) – Message to use as the system message that will be the first in the prompt. extra_prompt_messages (Optional[List[langchain.prompts.chat.BaseMessagePromptTemplate]]) – Prompt messages that will be placed between the system message and the new human input. Returns A prompt template to pass into this agent. Return type langchain.prompts.base.BasePromptTemplate classmethod from_llm_and_tools(llm, tools, callback_manager=None, extra_prompt_messages=None, system_message=SystemMessage(content='You are a helpful AI assistant.', additional_kwargs={}), **kwargs)[source] Construct an agent from an LLM and tools. Parameters llm (langchain.base_language.BaseLanguageModel) – tools (Sequence[langchain.tools.base.BaseTool]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – extra_prompt_messages (Optional[List[langchain.prompts.chat.BaseMessagePromptTemplate]]) – system_message (Optional[langchain.schema.SystemMessage]) – kwargs (Any) – Return type langchain.agents.agent.BaseSingleActionAgent get_allowed_tools()[source] Get allowed tools. Return type List[str] plan(intermediate_steps, callbacks=None, **kwargs)[source] Given input, decided what to do. Parameters intermediate_steps (List[Tuple[langchain.schema.AgentAction, str]]) – Steps the LLM has taken to date, along with observations **kwargs – User inputs. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – kwargs (Any) – Returns Action specifying what tool to use. Return type Union[langchain.schema.AgentAction, langchain.schema.AgentFinish] property functions: List[dict] property input_keys: List[str] Get input keys. Input refers to user input here. class langchain.agents.ReActChain(llm, docstore, *, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, agent, tools, return_intermediate_steps=False, max_iterations=15, max_execution_time=None, early_stopping_method='force', handle_parsing_errors=False)[source] Bases: langchain.agents.agent.AgentExecutor Chain that implements the ReAct paper. Example from langchain import ReActChain, OpenAI react = ReAct(llm=OpenAI()) Parameters llm (langchain.base_language.BaseLanguageModel) – docstore (langchain.docstore.base.Docstore) – memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – agent (Union[langchain.agents.agent.BaseSingleActionAgent, langchain.agents.agent.BaseMultiActionAgent]) – tools (Sequence[langchain.tools.base.BaseTool]) – return_intermediate_steps (bool) – max_iterations (Optional[int]) – max_execution_time (Optional[float]) – early_stopping_method (str) – handle_parsing_errors (Union[bool, str, Callable[[langchain.schema.OutputParserException], str]]) – Return type None class langchain.agents.ReActTextWorldAgent(*, llm_chain, output_parser=None, allowed_tools=None)[source] Bases: langchain.agents.react.base.ReActDocstoreAgent Agent for the ReAct TextWorld chain. Parameters llm_chain (langchain.chains.llm.LLMChain) – output_parser (langchain.agents.agent.AgentOutputParser) – allowed_tools (Optional[List[str]]) – Return type None classmethod create_prompt(tools)[source] Return default prompt. Parameters tools (Sequence[langchain.tools.base.BaseTool]) – Return type langchain.prompts.base.BasePromptTemplate class langchain.agents.SelfAskWithSearchChain(llm, search_chain, *, memory=None, callbacks=None, callback_manager=None, verbose=None, tags=None, agent, tools, return_intermediate_steps=False, max_iterations=15, max_execution_time=None, early_stopping_method='force', handle_parsing_errors=False)[source] Bases: langchain.agents.agent.AgentExecutor Chain that does self ask with search. Example from langchain import SelfAskWithSearchChain, OpenAI, GoogleSerperAPIWrapper search_chain = GoogleSerperAPIWrapper() self_ask = SelfAskWithSearchChain(llm=OpenAI(), search_chain=search_chain) Parameters llm (langchain.base_language.BaseLanguageModel) – search_chain (Union[langchain.utilities.google_serper.GoogleSerperAPIWrapper, langchain.utilities.serpapi.SerpAPIWrapper]) – memory (Optional[langchain.schema.BaseMemory]) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – verbose (bool) – tags (Optional[List[str]]) – agent (Union[langchain.agents.agent.BaseSingleActionAgent, langchain.agents.agent.BaseMultiActionAgent]) – tools (Sequence[langchain.tools.base.BaseTool]) – return_intermediate_steps (bool) – max_iterations (Optional[int]) – max_execution_time (Optional[float]) – early_stopping_method (str) – handle_parsing_errors (Union[bool, str, Callable[[langchain.schema.OutputParserException], str]]) – Return type None class langchain.agents.StructuredChatAgent(*, llm_chain, output_parser=None, allowed_tools=None)[source] Bases: langchain.agents.agent.Agent Parameters llm_chain (langchain.chains.llm.LLMChain) – output_parser (langchain.agents.agent.AgentOutputParser) – allowed_tools (Optional[List[str]]) – Return type None attribute output_parser: langchain.agents.agent.AgentOutputParser [Optional] classmethod create_prompt(tools, prefix='Respond to the human as helpfully and accurately as possible. You have access to the following tools:', suffix='Begin! Reminder to ALWAYS respond with a valid json blob of a single action. Use tools if necessary. Respond directly if appropriate. Format is Action:```$JSON_BLOB```then Observation:.\nThought:', human_message_template='{input}\n\n{agent_scratchpad}', format_instructions='Use a json blob to specify a tool by providing an action key (tool name) and an action_input key (tool input).\n\nValid "action" values: "Final Answer" or {tool_names}\n\nProvide only ONE action per $JSON_BLOB, as shown:\n\n```\n{{{{\nΒ  "action": $TOOL_NAME,\nΒ  "action_input": $INPUT\n}}}}\n```\n\nFollow this format:\n\nQuestion: input question to answer\nThought: consider previous and subsequent steps\nAction:\n```\n$JSON_BLOB\n```\nObservation: action result\n... (repeat Thought/Action/Observation N times)\nThought: I know what to respond\nAction:\n```\n{{{{\nΒ  "action": "Final Answer",\nΒ  "action_input": "Final response to human"\n}}}}\n```', input_variables=None, memory_prompts=None)[source] Create a prompt for this class. Parameters tools (Sequence[langchain.tools.base.BaseTool]) – prefix (str) – suffix (str) – human_message_template (str) – format_instructions (str) – input_variables (Optional[List[str]]) – memory_prompts (Optional[List[langchain.prompts.base.BasePromptTemplate]]) – Return type langchain.prompts.base.BasePromptTemplate classmethod from_llm_and_tools(llm, tools, callback_manager=None, output_parser=None, prefix='Respond to the human as helpfully and accurately as possible. You have access to the following tools:', suffix='Begin! Reminder to ALWAYS respond with a valid json blob of a single action. Use tools if necessary. Respond directly if appropriate. Format is Action:```$JSON_BLOB```then Observation:.\nThought:', human_message_template='{input}\n\n{agent_scratchpad}', format_instructions='Use a json blob to specify a tool by providing an action key (tool name) and an action_input key (tool input).\n\nValid "action" values: "Final Answer" or {tool_names}\n\nProvide only ONE action per $JSON_BLOB, as shown:\n\n```\n{{{{\nΒ  "action": $TOOL_NAME,\nΒ  "action_input": $INPUT\n}}}}\n```\n\nFollow this format:\n\nQuestion: input question to answer\nThought: consider previous and subsequent steps\nAction:\n```\n$JSON_BLOB\n```\nObservation: action result\n... (repeat Thought/Action/Observation N times)\nThought: I know what to respond\nAction:\n```\n{{{{\nΒ  "action": "Final Answer",\nΒ  "action_input": "Final response to human"\n}}}}\n```', input_variables=None, memory_prompts=None, **kwargs)[source] Construct an agent from an LLM and tools. Parameters llm (langchain.base_language.BaseLanguageModel) – tools (Sequence[langchain.tools.base.BaseTool]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – output_parser (Optional[langchain.agents.agent.AgentOutputParser]) – prefix (str) – suffix (str) – human_message_template (str) – format_instructions (str) – input_variables (Optional[List[str]]) – memory_prompts (Optional[List[langchain.prompts.base.BasePromptTemplate]]) – kwargs (Any) – Return type langchain.agents.agent.Agent property llm_prefix: str Prefix to append the llm call with. property observation_prefix: str Prefix to append the observation with. class langchain.agents.Tool(name, func, description, *, args_schema=None, return_direct=False, verbose=False, callbacks=None, callback_manager=None, handle_tool_error=False, coroutine=None)[source] Bases: langchain.tools.base.BaseTool Tool that takes in function or coroutine directly. Parameters name (str) – func (Callable[[...], str]) – description (str) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – return_direct (bool) – verbose (bool) – callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – handle_tool_error (Optional[Union[bool, str, Callable[[langchain.tools.base.ToolException], str]]]) – coroutine (Optional[Callable[[...], Awaitable[str]]]) – Return type None attribute coroutine: Optional[Callable[[...], Awaitable[str]]] = None The asynchronous version of the function. attribute description: str = '' Used to tell the model how/when/why to use the tool. You can provide few-shot examples as a part of the description. attribute func: Callable[[...], str] [Required] The function to run when the tool is called. classmethod from_function(func, name, description, return_direct=False, args_schema=None, **kwargs)[source] Initialize tool from a function. Parameters func (Callable) – name (str) – description (str) – return_direct (bool) – args_schema (Optional[Type[pydantic.main.BaseModel]]) – kwargs (Any) – Return type langchain.tools.base.Tool property args: dict The tool’s input arguments. class langchain.agents.ZeroShotAgent(*, llm_chain, output_parser=None, allowed_tools=None)[source] Bases: langchain.agents.agent.Agent Agent for the MRKL chain. Parameters llm_chain (langchain.chains.llm.LLMChain) – output_parser (langchain.agents.agent.AgentOutputParser) – allowed_tools (Optional[List[str]]) – Return type None attribute output_parser: langchain.agents.agent.AgentOutputParser [Optional] classmethod create_prompt(tools, prefix='Answer the following questions as best you can. You have access to the following tools:', suffix='Begin!\n\nQuestion: {input}\nThought:{agent_scratchpad}', format_instructions='Use the following format:\n\nQuestion: the input question you must answer\nThought: you should always think about what to do\nAction: the action to take, should be one of [{tool_names}]\nAction Input: the input to the action\nObservation: the result of the action\n... (this Thought/Action/Action Input/Observation can repeat N times)\nThought: I now know the final answer\nFinal Answer: the final answer to the original input question', input_variables=None)[source] Create prompt in the style of the zero shot agent. Parameters tools (Sequence[langchain.tools.base.BaseTool]) – List of tools the agent will have access to, used to format the prompt. prefix (str) – String to put before the list of tools. suffix (str) – String to put after the list of tools. input_variables (Optional[List[str]]) – List of input variables the final prompt will expect. format_instructions (str) – Returns A PromptTemplate with the template assembled from the pieces here. Return type langchain.prompts.prompt.PromptTemplate classmethod from_llm_and_tools(llm, tools, callback_manager=None, output_parser=None, prefix='Answer the following questions as best you can. You have access to the following tools:', suffix='Begin!\n\nQuestion: {input}\nThought:{agent_scratchpad}', format_instructions='Use the following format:\n\nQuestion: the input question you must answer\nThought: you should always think about what to do\nAction: the action to take, should be one of [{tool_names}]\nAction Input: the input to the action\nObservation: the result of the action\n... (this Thought/Action/Action Input/Observation can repeat N times)\nThought: I now know the final answer\nFinal Answer: the final answer to the original input question', input_variables=None, **kwargs)[source] Construct an agent from an LLM and tools. Parameters llm (langchain.base_language.BaseLanguageModel) – tools (Sequence[langchain.tools.base.BaseTool]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – output_parser (Optional[langchain.agents.agent.AgentOutputParser]) – prefix (str) – suffix (str) – format_instructions (str) – input_variables (Optional[List[str]]) – kwargs (Any) – Return type langchain.agents.agent.Agent property llm_prefix: str Prefix to append the llm call with. property observation_prefix: str Prefix to append the observation with. langchain.agents.create_csv_agent(llm, path, pandas_kwargs=None, **kwargs)[source] Create csv agent by loading to a dataframe and using pandas agent. Parameters llm (langchain.base_language.BaseLanguageModel) – path (Union[str, List[str]]) – pandas_kwargs (Optional[dict]) – kwargs (Any) – Return type langchain.agents.agent.AgentExecutor langchain.agents.create_json_agent(llm, toolkit, callback_manager=None, prefix='You are an agent designed to interact with JSON.\nYour goal is to return a final answer by interacting with the JSON.\nYou have access to the following tools which help you learn more about the JSON you are interacting with.\nOnly use the below tools. Only use the information returned by the below tools to construct your final answer.\nDo not make up any information that is not contained in the JSON.\nYour input to the tools should be in the form of `data["key"][0]` where `data` is the JSON blob you are interacting with, and the syntax used is Python. \nYou should only use keys that you know for a fact exist. You must validate that a key exists by seeing it previously when calling `json_spec_list_keys`. \nIf you have not seen a key in one of those responses, you cannot use it.\nYou should only add one key at a time to the path. You cannot add multiple keys at once.\nIf you encounter a "KeyError", go back to the previous key, look at the available keys, and try again.\n\nIf the question does not seem to be related to the JSON, just return "I don\'t know" as the answer.\nAlways begin your interaction with the `json_spec_list_keys` tool with input "data" to see what keys exist in the JSON.\n\nNote that sometimes the value at a given path is large. In this case, you will get an error "Value is a large dictionary, should explore its keys directly".\nIn this case, you should ALWAYS follow up by using the `json_spec_list_keys` tool to see what keys exist at that path.\nDo not simply refer the user to the JSON or a section of the JSON, as this is not a valid answer. Keep digging until you find the answer and explicitly return it.\n', suffix='Begin!"\n\nQuestion: {input}\nThought: I should look at the keys that exist in data to see what I have access to\n{agent_scratchpad}', format_instructions='Use the following format:\n\nQuestion: the input question you must answer\nThought: you should always think about what to do\nAction: the action to take, should be one of [{tool_names}]\nAction Input: the input to the action\nObservation: the result of the action\n... (this Thought/Action/Action Input/Observation can repeat N times)\nThought: I now know the final answer\nFinal Answer: the final answer to the original input question', input_variables=None, verbose=False, agent_executor_kwargs=None, **kwargs)[source] Construct a json agent from an LLM and tools. Parameters llm (langchain.base_language.BaseLanguageModel) – toolkit (langchain.agents.agent_toolkits.json.toolkit.JsonToolkit) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – prefix (str) – suffix (str) – format_instructions (str) – input_variables (Optional[List[str]]) – verbose (bool) – agent_executor_kwargs (Optional[Dict[str, Any]]) – kwargs (Dict[str, Any]) – Return type langchain.agents.agent.AgentExecutor langchain.agents.create_openapi_agent(llm, toolkit, callback_manager=None, prefix="You are an agent designed to answer questions by making web requests to an API given the openapi spec.\n\nIf the question does not seem related to the API, return I don't know. Do not make up an answer.\nOnly use information provided by the tools to construct your response.\n\nFirst, find the base URL needed to make the request.\n\nSecond, find the relevant paths needed to answer the question. Take note that, sometimes, you might need to make more than one request to more than one path to answer the question.\n\nThird, find the required parameters needed to make the request. For GET requests, these are usually URL parameters and for POST requests, these are request body parameters.\n\nFourth, make the requests needed to answer the question. Ensure that you are sending the correct parameters to the request by checking which parameters are required. For parameters with a fixed set of values, please use the spec to look at which values are allowed.\n\nUse the exact parameter names as listed in the spec, do not make up any names or abbreviate the names of parameters.\nIf you get a not found error, ensure that you are using a path that actually exists in the spec.\n", suffix='Begin!\n\nQuestion: {input}\nThought: I should explore the spec to find the base url for the API.\n{agent_scratchpad}', format_instructions='Use the following format:\n\nQuestion: the input question you must answer\nThought: you should always think about what to do\nAction: the action to take, should be one of [{tool_names}]\nAction Input: the input to the action\nObservation: the result of the action\n... (this Thought/Action/Action Input/Observation can repeat N times)\nThought: I now know the final answer\nFinal Answer: the final answer to the original input question', input_variables=None, max_iterations=15, max_execution_time=None, early_stopping_method='force', verbose=False, return_intermediate_steps=False, agent_executor_kwargs=None, **kwargs)[source] Construct a json agent from an LLM and tools. Parameters llm (langchain.base_language.BaseLanguageModel) – toolkit (langchain.agents.agent_toolkits.openapi.toolkit.OpenAPIToolkit) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – prefix (str) – suffix (str) – format_instructions (str) – input_variables (Optional[List[str]]) – max_iterations (Optional[int]) – max_execution_time (Optional[float]) – early_stopping_method (str) – verbose (bool) – return_intermediate_steps (bool) – agent_executor_kwargs (Optional[Dict[str, Any]]) – kwargs (Dict[str, Any]) – Return type langchain.agents.agent.AgentExecutor langchain.agents.create_pandas_dataframe_agent(llm, df, agent_type=AgentType.ZERO_SHOT_REACT_DESCRIPTION, callback_manager=None, prefix=None, suffix=None, input_variables=None, verbose=False, return_intermediate_steps=False, max_iterations=15, max_execution_time=None, early_stopping_method='force', agent_executor_kwargs=None, include_df_in_prompt=True, **kwargs)[source] Construct a pandas agent from an LLM and dataframe. Parameters llm (langchain.base_language.BaseLanguageModel) – df (Any) – agent_type (langchain.agents.agent_types.AgentType) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – prefix (Optional[str]) – suffix (Optional[str]) – input_variables (Optional[List[str]]) – verbose (bool) – return_intermediate_steps (bool) – max_iterations (Optional[int]) – max_execution_time (Optional[float]) – early_stopping_method (str) – agent_executor_kwargs (Optional[Dict[str, Any]]) – include_df_in_prompt (Optional[bool]) – kwargs (Dict[str, Any]) – Return type langchain.agents.agent.AgentExecutor langchain.agents.create_pbi_agent(llm, toolkit, powerbi=None, callback_manager=None, prefix='You are an agent designed to help users interact with a PowerBI Dataset.\n\nAgent has access to a tool that can write a query based on the question and then run those against PowerBI, Microsofts business intelligence tool. The questions from the users should be interpreted as related to the dataset that is available and not general questions about the world. If the question does not seem related to the dataset, just return "This does not appear to be part of this dataset." as the answer.\n\nGiven an input question, ask to run the questions against the dataset, then look at the results and return the answer, the answer should be a complete sentence that answers the question, if multiple rows are asked find a way to write that in a easily readable format for a human, also make sure to represent numbers in readable ways, like 1M instead of 1000000. Unless the user specifies a specific number of examples they wish to obtain, always limit your query to at most {top_k} results.\n', suffix='Begin!\n\nQuestion: {input}\nThought: I can first ask which tables I have, then how each table is defined and then ask the query tool the question I need, and finally create a nice sentence that answers the question.\n{agent_scratchpad}', format_instructions='Use the following format:\n\nQuestion: the input question you must answer\nThought: you should always think about what to do\nAction: the action to take, should be one of [{tool_names}]\nAction Input: the input to the action\nObservation: the result of the action\n... (this Thought/Action/Action Input/Observation can repeat N times)\nThought: I now know the final answer\nFinal Answer: the final answer to the original input question', examples=None, input_variables=None, top_k=10, verbose=False, agent_executor_kwargs=None, **kwargs)[source] Construct a pbi agent from an LLM and tools. Parameters llm (langchain.base_language.BaseLanguageModel) – toolkit (Optional[langchain.agents.agent_toolkits.powerbi.toolkit.PowerBIToolkit]) – powerbi (Optional[langchain.utilities.powerbi.PowerBIDataset]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – prefix (str) – suffix (str) – format_instructions (str) – examples (Optional[str]) – input_variables (Optional[List[str]]) – top_k (int) – verbose (bool) – agent_executor_kwargs (Optional[Dict[str, Any]]) – kwargs (Dict[str, Any]) – Return type langchain.agents.agent.AgentExecutor langchain.agents.create_pbi_chat_agent(llm, toolkit, powerbi=None, callback_manager=None, output_parser=None, prefix='Assistant is a large language model built to help users interact with a PowerBI Dataset.\n\nAssistant has access to a tool that can write a query based on the question and then run those against PowerBI, Microsofts business intelligence tool. The questions from the users should be interpreted as related to the dataset that is available and not general questions about the world. If the question does not seem related to the dataset, just return "This does not appear to be part of this dataset." as the answer.\n\nGiven an input question, ask to run the questions against the dataset, then look at the results and return the answer, the answer should be a complete sentence that answers the question, if multiple rows are asked find a way to write that in a easily readable format for a human, also make sure to represent numbers in readable ways, like 1M instead of 1000000. Unless the user specifies a specific number of examples they wish to obtain, always limit your query to at most {top_k} results.\n', suffix="TOOLS\n------\nAssistant can ask the user to use tools to look up information that may be helpful in answering the users original question. The tools the human can use are:\n\n{{tools}}\n\n{format_instructions}\n\nUSER'S INPUT\n--------------------\nHere is the user's input (remember to respond with a markdown code snippet of a json blob with a single action, and NOTHING else):\n\n{{{{input}}}}\n", examples=None, input_variables=None, memory=None, top_k=10, verbose=False, agent_executor_kwargs=None, **kwargs)[source] Construct a pbi agent from an Chat LLM and tools. If you supply only a toolkit and no powerbi dataset, the same LLM is used for both. Parameters llm (langchain.chat_models.base.BaseChatModel) – toolkit (Optional[langchain.agents.agent_toolkits.powerbi.toolkit.PowerBIToolkit]) – powerbi (Optional[langchain.utilities.powerbi.PowerBIDataset]) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – output_parser (Optional[langchain.agents.agent.AgentOutputParser]) – prefix (str) – suffix (str) – examples (Optional[str]) – input_variables (Optional[List[str]]) – memory (Optional[langchain.memory.chat_memory.BaseChatMemory]) – top_k (int) – verbose (bool) – agent_executor_kwargs (Optional[Dict[str, Any]]) – kwargs (Dict[str, Any]) – Return type langchain.agents.agent.AgentExecutor langchain.agents.create_spark_dataframe_agent(llm, df, callback_manager=None, prefix='\nYou are working with a spark dataframe in Python. The name of the dataframe is `df`.\nYou should use the tools below to answer the question posed of you:', suffix='\nThis is the result of `print(df.first())`:\n{df}\n\nBegin!\nQuestion: {input}\n{agent_scratchpad}', input_variables=None, verbose=False, return_intermediate_steps=False, max_iterations=15, max_execution_time=None, early_stopping_method='force', agent_executor_kwargs=None, **kwargs)[source] Construct a spark agent from an LLM and dataframe. Parameters llm (langchain.llms.base.BaseLLM) – df (Any) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – prefix (str) – suffix (str) – input_variables (Optional[List[str]]) – verbose (bool) – return_intermediate_steps (bool) – max_iterations (Optional[int]) – max_execution_time (Optional[float]) – early_stopping_method (str) – agent_executor_kwargs (Optional[Dict[str, Any]]) – kwargs (Dict[str, Any]) – Return type langchain.agents.agent.AgentExecutor langchain.agents.create_spark_sql_agent(llm, toolkit, callback_manager=None, prefix='You are an agent designed to interact with Spark SQL.\nGiven an input question, create a syntactically correct Spark SQL query to run, then look at the results of the query and return the answer.\nUnless the user specifies a specific number of examples they wish to obtain, always limit your query to at most {top_k} results.\nYou can order the results by a relevant column to return the most interesting examples in the database.\nNever query for all the columns from a specific table, only ask for the relevant columns given the question.\nYou have access to tools for interacting with the database.\nOnly use the below tools. Only use the information returned by the below tools to construct your final answer.\nYou MUST double check your query before executing it. If you get an error while executing a query, rewrite the query and try again.\n\nDO NOT make any DML statements (INSERT, UPDATE, DELETE, DROP etc.) to the database.\n\nIf the question does not seem related to the database, just return "I don\'t know" as the answer.\n', suffix='Begin!\n\nQuestion: {input}\nThought: I should look at the tables in the database to see what I can query.\n{agent_scratchpad}', format_instructions='Use the following format:\n\nQuestion: the input question you must answer\nThought: you should always think about what to do\nAction: the action to take, should be one of [{tool_names}]\nAction Input: the input to the action\nObservation: the result of the action\n... (this Thought/Action/Action Input/Observation can repeat N times)\nThought: I now know the final answer\nFinal Answer: the final answer to the original input question', input_variables=None, top_k=10, max_iterations=15, max_execution_time=None, early_stopping_method='force', verbose=False, agent_executor_kwargs=None, **kwargs)[source] Construct a sql agent from an LLM and tools. Parameters llm (langchain.base_language.BaseLanguageModel) – toolkit (langchain.agents.agent_toolkits.spark_sql.toolkit.SparkSQLToolkit) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – prefix (str) – suffix (str) – format_instructions (str) – input_variables (Optional[List[str]]) – top_k (int) – max_iterations (Optional[int]) – max_execution_time (Optional[float]) – early_stopping_method (str) – verbose (bool) – agent_executor_kwargs (Optional[Dict[str, Any]]) – kwargs (Dict[str, Any]) – Return type langchain.agents.agent.AgentExecutor langchain.agents.create_sql_agent(llm, toolkit, agent_type=AgentType.ZERO_SHOT_REACT_DESCRIPTION, callback_manager=None, prefix='You are an agent designed to interact with a SQL database.\nGiven an input question, create a syntactically correct {dialect} query to run, then look at the results of the query and return the answer.\nUnless the user specifies a specific number of examples they wish to obtain, always limit your query to at most {top_k} results.\nYou can order the results by a relevant column to return the most interesting examples in the database.\nNever query for all the columns from a specific table, only ask for the relevant columns given the question.\nYou have access to tools for interacting with the database.\nOnly use the below tools. Only use the information returned by the below tools to construct your final answer.\nYou MUST double check your query before executing it. If you get an error while executing a query, rewrite the query and try again.\n\nDO NOT make any DML statements (INSERT, UPDATE, DELETE, DROP etc.) to the database.\n\nIf the question does not seem related to the database, just return "I don\'t know" as the answer.\n', suffix=None, format_instructions='Use the following format:\n\nQuestion: the input question you must answer\nThought: you should always think about what to do\nAction: the action to take, should be one of [{tool_names}]\nAction Input: the input to the action\nObservation: the result of the action\n... (this Thought/Action/Action Input/Observation can repeat N times)\nThought: I now know the final answer\nFinal Answer: the final answer to the original input question', input_variables=None, top_k=10, max_iterations=15, max_execution_time=None, early_stopping_method='force', verbose=False, agent_executor_kwargs=None, **kwargs)[source] Construct a sql agent from an LLM and tools. Parameters llm (langchain.base_language.BaseLanguageModel) – toolkit (langchain.agents.agent_toolkits.sql.toolkit.SQLDatabaseToolkit) – agent_type (langchain.agents.agent_types.AgentType) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – prefix (str) – suffix (Optional[str]) – format_instructions (str) – input_variables (Optional[List[str]]) – top_k (int) – max_iterations (Optional[int]) – max_execution_time (Optional[float]) – early_stopping_method (str) – verbose (bool) – agent_executor_kwargs (Optional[Dict[str, Any]]) – kwargs (Dict[str, Any]) – Return type langchain.agents.agent.AgentExecutor langchain.agents.create_vectorstore_agent(llm, toolkit, callback_manager=None, prefix='You are an agent designed to answer questions about sets of documents.\nYou have access to tools for interacting with the documents, and the inputs to the tools are questions.\nSometimes, you will be asked to provide sources for your questions, in which case you should use the appropriate tool to do so.\nIf the question does not seem relevant to any of the tools provided, just return "I don\'t know" as the answer.\n', verbose=False, agent_executor_kwargs=None, **kwargs)[source] Construct a vectorstore agent from an LLM and tools. Parameters llm (langchain.base_language.BaseLanguageModel) – toolkit (langchain.agents.agent_toolkits.vectorstore.toolkit.VectorStoreToolkit) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – prefix (str) – verbose (bool) – agent_executor_kwargs (Optional[Dict[str, Any]]) – kwargs (Dict[str, Any]) – Return type langchain.agents.agent.AgentExecutor langchain.agents.create_vectorstore_router_agent(llm, toolkit, callback_manager=None, prefix='You are an agent designed to answer questions.\nYou have access to tools for interacting with different sources, and the inputs to the tools are questions.\nYour main task is to decide which of the tools is relevant for answering question at hand.\nFor complex questions, you can break the question down into sub questions and use tools to answers the sub questions.\n', verbose=False, agent_executor_kwargs=None, **kwargs)[source] Construct a vectorstore router agent from an LLM and tools. Parameters llm (langchain.base_language.BaseLanguageModel) – toolkit (langchain.agents.agent_toolkits.vectorstore.toolkit.VectorStoreRouterToolkit) – callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – prefix (str) – verbose (bool) – agent_executor_kwargs (Optional[Dict[str, Any]]) – kwargs (Dict[str, Any]) – Return type langchain.agents.agent.AgentExecutor langchain.agents.get_all_tool_names()[source] Get a list of all possible tool names. Return type List[str] langchain.agents.initialize_agent(tools, llm, agent=None, callback_manager=None, agent_path=None, agent_kwargs=None, *, tags=None, **kwargs)[source] Load an agent executor given tools and LLM. Parameters tools (Sequence[langchain.tools.base.BaseTool]) – List of tools this agent has access to. llm (langchain.base_language.BaseLanguageModel) – Language model to use as the agent. agent (Optional[langchain.agents.agent_types.AgentType]) – Agent type to use. If None and agent_path is also None, will default to AgentType.ZERO_SHOT_REACT_DESCRIPTION. callback_manager (Optional[langchain.callbacks.base.BaseCallbackManager]) – CallbackManager to use. Global callback manager is used if not provided. Defaults to None. agent_path (Optional[str]) – Path to serialized agent to use. agent_kwargs (Optional[dict]) – Additional key word arguments to pass to the underlying agent tags (Optional[Sequence[str]]) – Tags to apply to the traced runs. **kwargs – Additional key word arguments passed to the agent executor kwargs (Any) – Returns An agent executor Return type langchain.agents.agent.AgentExecutor langchain.agents.load_agent(path, **kwargs)[source] Unified method for loading a agent from LangChainHub or local fs. Parameters path (Union[str, pathlib.Path]) – kwargs (Any) – Return type Union[langchain.agents.agent.BaseSingleActionAgent, langchain.agents.agent.BaseMultiActionAgent] langchain.agents.load_huggingface_tool(task_or_repo_id, model_repo_id=None, token=None, remote=False, **kwargs)[source] Loads a tool from the HuggingFace Hub. Parameters task_or_repo_id (str) – Task or model repo id. model_repo_id (Optional[str]) – Optional model repo id. token (Optional[str]) – Optional token. remote (bool) – Optional remote. Defaults to False. **kwargs – kwargs (Any) – Returns A tool. Return type langchain.tools.base.BaseTool langchain.agents.load_tools(tool_names, llm=None, callbacks=None, **kwargs)[source] Load tools based on their name. Parameters tool_names (List[str]) – name of tools to load. llm (Optional[langchain.base_language.BaseLanguageModel]) – Optional language model, may be needed to initialize certain tools. callbacks (Optional[Union[List[langchain.callbacks.base.BaseCallbackHandler], langchain.callbacks.base.BaseCallbackManager]]) – Optional callback manager or list of callback handlers. If not provided, default global callback manager will be used. kwargs (Any) – Returns List of tools. Return type List[langchain.tools.base.BaseTool] langchain.agents.tool(*args, return_direct=False, args_schema=None, infer_schema=True)[source] Make tools out of functions, can be used with or without arguments. Parameters *args – The arguments to the tool. return_direct (bool) – Whether to return directly from the tool rather than continuing the agent loop. args_schema (Optional[Type[pydantic.main.BaseModel]]) – optional argument schema for user to specify infer_schema (bool) – Whether to infer the schema of the arguments from the function’s signature. This also makes the resultant tool accept a dictionary input to its run() function. args (Union[str, Callable]) – Return type Callable Requires: Function must be of type (str) -> str Function must have a docstring Examples @tool def search_api(query: str) -> str: # Searches the API for the query. return @tool("search", return_direct=True) def search_api(query: str) -> str: # Searches the API for the query. return
https://api.python.langchain.com/en/latest/modules/agents.html
00a73b87-368a-4d4b-b6fe-acf310c654c4
Source code for langchain.requests """Lightweight wrapper around requests library, with async support.""" from contextlib import asynccontextmanager from typing import Any, AsyncGenerator, Dict, Optional import aiohttp import requests from pydantic import BaseModel, Extra class Requests(BaseModel): """Wrapper around requests to handle auth and async. The main purpose of this wrapper is to handle authentication (by saving headers) and enable easy async methods on the same base object. """ headers: Optional[Dict[str, str]] = None aiosession: Optional[aiohttp.ClientSession] = None class Config: """Configuration for this pydantic object.""" extra = Extra.forbid arbitrary_types_allowed = True def get(self, url: str, **kwargs: Any) -> requests.Response: """GET the URL and return the text.""" return requests.get(url, headers=self.headers, **kwargs) def post(self, url: str, data: Dict[str, Any], **kwargs: Any) -> requests.Response: """POST to the URL and return the text.""" return requests.post(url, json=data, headers=self.headers, **kwargs) def patch(self, url: str, data: Dict[str, Any], **kwargs: Any) -> requests.Response: """PATCH the URL and return the text.""" return requests.patch(url, json=data, headers=self.headers, **kwargs) def put(self, url: str, data: Dict[str, Any], **kwargs: Any) -> requests.Response: """PUT the URL and return the text.""" return requests.put(url, json=data, headers=self.headers, **kwargs) def delete(self, url: str, **kwargs: Any) -> requests.Response: """DELETE the URL and return the text.""" return requests.delete(url, headers=self.headers, **kwargs) @asynccontextmanager async def _arequest( self, method: str, url: str, **kwargs: Any ) -> AsyncGenerator[aiohttp.ClientResponse, None]: """Make an async request.""" if not self.aiosession: async with aiohttp.ClientSession() as session: async with session.request( method, url, headers=self.headers, **kwargs ) as response: yield response else: async with self.aiosession.request( method, url, headers=self.headers, **kwargs ) as response: yield response @asynccontextmanager async def aget( self, url: str, **kwargs: Any ) -> AsyncGenerator[aiohttp.ClientResponse, None]: """GET the URL and return the text asynchronously.""" async with self._arequest("GET", url, **kwargs) as response: yield response @asynccontextmanager async def apost( self, url: str, data: Dict[str, Any], **kwargs: Any ) -> AsyncGenerator[aiohttp.ClientResponse, None]: """POST to the URL and return the text asynchronously.""" async with self._arequest("POST", url, **kwargs) as response: yield response @asynccontextmanager async def apatch( self, url: str, data: Dict[str, Any], **kwargs: Any ) -> AsyncGenerator[aiohttp.ClientResponse, None]: """PATCH the URL and return the text asynchronously.""" async with self._arequest("PATCH", url, **kwargs) as response: yield response @asynccontextmanager async def aput( self, url: str, data: Dict[str, Any], **kwargs: Any ) -> AsyncGenerator[aiohttp.ClientResponse, None]: """PUT the URL and return the text asynchronously.""" async with self._arequest("PUT", url, **kwargs) as response: yield response @asynccontextmanager async def adelete( self, url: str, **kwargs: Any ) -> AsyncGenerator[aiohttp.ClientResponse, None]: """DELETE the URL and return the text asynchronously.""" async with self._arequest("DELETE", url, **kwargs) as response: yield response [docs]class TextRequestsWrapper(BaseModel): """Lightweight wrapper around requests library. The main purpose of this wrapper is to always return a text output. """ headers: Optional[Dict[str, str]] = None aiosession: Optional[aiohttp.ClientSession] = None class Config: """Configuration for this pydantic object.""" extra = Extra.forbid arbitrary_types_allowed = True @property def requests(self) -> Requests: return Requests(headers=self.headers, aiosession=self.aiosession) [docs] def get(self, url: str, **kwargs: Any) -> str: """GET the URL and return the text.""" return self.requests.get(url, **kwargs).text [docs] def post(self, url: str, data: Dict[str, Any], **kwargs: Any) -> str: """POST to the URL and return the text.""" return self.requests.post(url, data, **kwargs).text [docs] def patch(self, url: str, data: Dict[str, Any], **kwargs: Any) -> str: """PATCH the URL and return the text.""" return self.requests.patch(url, data, **kwargs).text [docs] def put(self, url: str, data: Dict[str, Any], **kwargs: Any) -> str: """PUT the URL and return the text.""" return self.requests.put(url, data, **kwargs).text [docs] def delete(self, url: str, **kwargs: Any) -> str: """DELETE the URL and return the text.""" return self.requests.delete(url, **kwargs).text [docs] async def aget(self, url: str, **kwargs: Any) -> str: """GET the URL and return the text asynchronously.""" async with self.requests.aget(url, **kwargs) as response: return await response.text() [docs] async def apost(self, url: str, data: Dict[str, Any], **kwargs: Any) -> str: """POST to the URL and return the text asynchronously.""" async with self.requests.apost(url, **kwargs) as response: return await response.text() [docs] async def apatch(self, url: str, data: Dict[str, Any], **kwargs: Any) -> str: """PATCH the URL and return the text asynchronously.""" async with self.requests.apatch(url, **kwargs) as response: return await response.text() [docs] async def aput(self, url: str, data: Dict[str, Any], **kwargs: Any) -> str: """PUT the URL and return the text asynchronously.""" async with self.requests.aput(url, **kwargs) as response: return await response.text() [docs] async def adelete(self, url: str, **kwargs: Any) -> str: """DELETE the URL and return the text asynchronously.""" async with self.requests.adelete(url, **kwargs) as response: return await response.text() # For backwards compatibility RequestsWrapper = TextRequestsWrapper
https://api.python.langchain.com/en/latest/_modules/langchain/requests.html
64279c2e-c500-4306-9936-fc32d5ed5c43
Source code for langchain.text_splitter """Functionality for splitting text.""" from __future__ import annotations import copy import logging import re from abc import ABC, abstractmethod from dataclasses import dataclass from enum import Enum from typing import ( AbstractSet, Any, Callable, Collection, Dict, Iterable, List, Literal, Optional, Sequence, Tuple, Type, TypedDict, TypeVar, Union, cast, ) from langchain.docstore.document import Document from langchain.schema import BaseDocumentTransformer logger = logging.getLogger(__name__) TS = TypeVar("TS", bound="TextSplitter") def _split_text_with_regex( text: str, separator: str, keep_separator: bool ) -> List[str]: # Now that we have the separator, split the text if separator: if keep_separator: # The parentheses in the pattern keep the delimiters in the result. _splits = re.split(f"({separator})", text) splits = [_splits[i] + _splits[i + 1] for i in range(1, len(_splits), 2)] if len(_splits) % 2 == 0: splits += _splits[-1:] splits = [_splits[0]] + splits else: splits = text.split(separator) else: splits = list(text) return [s for s in splits if s != ""] [docs]class TextSplitter(BaseDocumentTransformer, ABC): """Interface for splitting text into chunks.""" def __init__( self, chunk_size: int = 4000, chunk_overlap: int = 200, length_function: Callable[[str], int] = len, keep_separator: bool = False, add_start_index: bool = False, ) -> None: """Create a new TextSplitter. Args: chunk_size: Maximum size of chunks to return chunk_overlap: Overlap in characters between chunks length_function: Function that measures the length of given chunks keep_separator: Whether or not to keep the separator in the chunks add_start_index: If `True`, includes chunk's start index in metadata """ if chunk_overlap > chunk_size: raise ValueError( f"Got a larger chunk overlap ({chunk_overlap}) than chunk size " f"({chunk_size}), should be smaller." ) self._chunk_size = chunk_size self._chunk_overlap = chunk_overlap self._length_function = length_function self._keep_separator = keep_separator self._add_start_index = add_start_index [docs] @abstractmethod def split_text(self, text: str) -> List[str]: """Split text into multiple components.""" [docs] def create_documents( self, texts: List[str], metadatas: Optional[List[dict]] = None ) -> List[Document]: """Create documents from a list of texts.""" _metadatas = metadatas or [{}] * len(texts) documents = [] for i, text in enumerate(texts): index = -1 for chunk in self.split_text(text): metadata = copy.deepcopy(_metadatas[i]) if self._add_start_index: index = text.find(chunk, index + 1) metadata["start_index"] = index new_doc = Document(page_content=chunk, metadata=metadata) documents.append(new_doc) return documents [docs] def split_documents(self, documents: Iterable[Document]) -> List[Document]: """Split documents.""" texts, metadatas = [], [] for doc in documents: texts.append(doc.page_content) metadatas.append(doc.metadata) return self.create_documents(texts, metadatas=metadatas) def _join_docs(self, docs: List[str], separator: str) -> Optional[str]: text = separator.join(docs) text = text.strip() if text == "": return None else: return text def _merge_splits(self, splits: Iterable[str], separator: str) -> List[str]: # We now want to combine these smaller pieces into medium size # chunks to send to the LLM. separator_len = self._length_function(separator) docs = [] current_doc: List[str] = [] total = 0 for d in splits: _len = self._length_function(d) if ( total + _len + (separator_len if len(current_doc) > 0 else 0) > self._chunk_size ): if total > self._chunk_size: logger.warning( f"Created a chunk of size {total}, " f"which is longer than the specified {self._chunk_size}" ) if len(current_doc) > 0: doc = self._join_docs(current_doc, separator) if doc is not None: docs.append(doc) # Keep on popping if: # - we have a larger chunk than in the chunk overlap # - or if we still have any chunks and the length is long while total > self._chunk_overlap or ( total + _len + (separator_len if len(current_doc) > 0 else 0) > self._chunk_size and total > 0 ): total -= self._length_function(current_doc[0]) + ( separator_len if len(current_doc) > 1 else 0 ) current_doc = current_doc[1:] current_doc.append(d) total += _len + (separator_len if len(current_doc) > 1 else 0) doc = self._join_docs(current_doc, separator) if doc is not None: docs.append(doc) return docs [docs] @classmethod def from_huggingface_tokenizer(cls, tokenizer: Any, **kwargs: Any) -> TextSplitter: """Text splitter that uses HuggingFace tokenizer to count length.""" try: from transformers import PreTrainedTokenizerBase if not isinstance(tokenizer, PreTrainedTokenizerBase): raise ValueError( "Tokenizer received was not an instance of PreTrainedTokenizerBase" ) def _huggingface_tokenizer_length(text: str) -> int: return len(tokenizer.encode(text)) except ImportError: raise ValueError( "Could not import transformers python package. " "Please install it with `pip install transformers`." ) return cls(length_function=_huggingface_tokenizer_length, **kwargs) [docs] @classmethod def from_tiktoken_encoder( cls: Type[TS], encoding_name: str = "gpt2", model_name: Optional[str] = None, allowed_special: Union[Literal["all"], AbstractSet[str]] = set(), disallowed_special: Union[Literal["all"], Collection[str]] = "all", **kwargs: Any, ) -> TS: """Text splitter that uses tiktoken encoder to count length.""" try: import tiktoken except ImportError: raise ImportError( "Could not import tiktoken python package. " "This is needed in order to calculate max_tokens_for_prompt. " "Please install it with `pip install tiktoken`." ) if model_name is not None: enc = tiktoken.encoding_for_model(model_name) else: enc = tiktoken.get_encoding(encoding_name) def _tiktoken_encoder(text: str) -> int: return len( enc.encode( text, allowed_special=allowed_special, disallowed_special=disallowed_special, ) ) if issubclass(cls, TokenTextSplitter): extra_kwargs = { "encoding_name": encoding_name, "model_name": model_name, "allowed_special": allowed_special, "disallowed_special": disallowed_special, } kwargs = {**kwargs, **extra_kwargs} return cls(length_function=_tiktoken_encoder, **kwargs) [docs] def transform_documents( self, documents: Sequence[Document], **kwargs: Any ) -> Sequence[Document]: """Transform sequence of documents by splitting them.""" return self.split_documents(list(documents)) [docs] async def atransform_documents( self, documents: Sequence[Document], **kwargs: Any ) -> Sequence[Document]: """Asynchronously transform a sequence of documents by splitting them.""" raise NotImplementedError [docs]class CharacterTextSplitter(TextSplitter): """Implementation of splitting text that looks at characters.""" def __init__(self, separator: str = "\n\n", **kwargs: Any) -> None: """Create a new TextSplitter.""" super().__init__(**kwargs) self._separator = separator [docs] def split_text(self, text: str) -> List[str]: """Split incoming text and return chunks.""" # First we naively split the large input into a bunch of smaller ones. splits = _split_text_with_regex(text, self._separator, self._keep_separator) _separator = "" if self._keep_separator else self._separator return self._merge_splits(splits, _separator) [docs]class LineType(TypedDict): """Line type as typed dict.""" metadata: Dict[str, str] content: str [docs]class HeaderType(TypedDict): """Header type as typed dict.""" level: int name: str data: str [docs]class MarkdownHeaderTextSplitter: """Implementation of splitting markdown files based on specified headers.""" def __init__( self, headers_to_split_on: List[Tuple[str, str]], return_each_line: bool = False ): """Create a new MarkdownHeaderTextSplitter. Args: headers_to_split_on: Headers we want to track return_each_line: Return each line w/ associated headers """ # Output line-by-line or aggregated into chunks w/ common headers self.return_each_line = return_each_line # Given the headers we want to split on, # (e.g., "#, ##, etc") order by length self.headers_to_split_on = sorted( headers_to_split_on, key=lambda split: len(split[0]), reverse=True ) [docs] def aggregate_lines_to_chunks(self, lines: List[LineType]) -> List[Document]: """Combine lines with common metadata into chunks Args: lines: Line of text / associated header metadata """ aggregated_chunks: List[LineType] = [] for line in lines: if ( aggregated_chunks and aggregated_chunks[-1]["metadata"] == line["metadata"] ): # If the last line in the aggregated list # has the same metadata as the current line, # append the current content to the last lines's content aggregated_chunks[-1]["content"] += " \n" + line["content"] else: # Otherwise, append the current line to the aggregated list aggregated_chunks.append(line) return [ Document(page_content=chunk["content"], metadata=chunk["metadata"]) for chunk in aggregated_chunks ] [docs] def split_text(self, text: str) -> List[Document]: """Split markdown file Args: text: Markdown file""" # Split the input text by newline character ("\n"). lines = text.split("\n") # Final output lines_with_metadata: List[LineType] = [] # Content and metadata of the chunk currently being processed current_content: List[str] = [] current_metadata: Dict[str, str] = {} # Keep track of the nested header structure # header_stack: List[Dict[str, Union[int, str]]] = [] header_stack: List[HeaderType] = [] initial_metadata: Dict[str, str] = {} for line in lines: stripped_line = line.strip() # Check each line against each of the header types (e.g., #, ##) for sep, name in self.headers_to_split_on: # Check if line starts with a header that we intend to split on if stripped_line.startswith(sep) and ( # Header with no text OR header is followed by space # Both are valid conditions that sep is being used a header len(stripped_line) == len(sep) or stripped_line[len(sep)] == " " ): # Ensure we are tracking the header as metadata if name is not None: # Get the current header level current_header_level = sep.count("#") # Pop out headers of lower or same level from the stack while ( header_stack and header_stack[-1]["level"] >= current_header_level ): # We have encountered a new header # at the same or higher level popped_header = header_stack.pop() # Clear the metadata for the # popped header in initial_metadata if popped_header["name"] in initial_metadata: initial_metadata.pop(popped_header["name"]) # Push the current header to the stack header: HeaderType = { "level": current_header_level, "name": name, "data": stripped_line[len(sep) :].strip(), } header_stack.append(header) # Update initial_metadata with the current header initial_metadata[name] = header["data"] # Add the previous line to the lines_with_metadata # only if current_content is not empty if current_content: lines_with_metadata.append( { "content": "\n".join(current_content), "metadata": current_metadata.copy(), } ) current_content.clear() break else: if stripped_line: current_content.append(stripped_line) elif current_content: lines_with_metadata.append( { "content": "\n".join(current_content), "metadata": current_metadata.copy(), } ) current_content.clear() current_metadata = initial_metadata.copy() if current_content: lines_with_metadata.append( {"content": "\n".join(current_content), "metadata": current_metadata} ) # lines_with_metadata has each line with associated header metadata # aggregate these into chunks based on common metadata if not self.return_each_line: return self.aggregate_lines_to_chunks(lines_with_metadata) else: return [ Document(page_content=chunk["content"], metadata=chunk["metadata"]) for chunk in lines_with_metadata ] # should be in newer Python versions (3.10+) # @dataclass(frozen=True, kw_only=True, slots=True) [docs]@dataclass(frozen=True) class Tokenizer: chunk_overlap: int tokens_per_chunk: int decode: Callable[[list[int]], str] encode: Callable[[str], List[int]] [docs]def split_text_on_tokens(*, text: str, tokenizer: Tokenizer) -> List[str]: """Split incoming text and return chunks.""" splits: List[str] = [] input_ids = tokenizer.encode(text) start_idx = 0 cur_idx = min(start_idx + tokenizer.tokens_per_chunk, len(input_ids)) chunk_ids = input_ids[start_idx:cur_idx] while start_idx < len(input_ids): splits.append(tokenizer.decode(chunk_ids)) start_idx += tokenizer.tokens_per_chunk - tokenizer.chunk_overlap cur_idx = min(start_idx + tokenizer.tokens_per_chunk, len(input_ids)) chunk_ids = input_ids[start_idx:cur_idx] return splits [docs]class TokenTextSplitter(TextSplitter): """Implementation of splitting text that looks at tokens.""" def __init__( self, encoding_name: str = "gpt2", model_name: Optional[str] = None, allowed_special: Union[Literal["all"], AbstractSet[str]] = set(), disallowed_special: Union[Literal["all"], Collection[str]] = "all", **kwargs: Any, ) -> None: """Create a new TextSplitter.""" super().__init__(**kwargs) try: import tiktoken except ImportError: raise ImportError( "Could not import tiktoken python package. " "This is needed in order to for TokenTextSplitter. " "Please install it with `pip install tiktoken`." ) if model_name is not None: enc = tiktoken.encoding_for_model(model_name) else: enc = tiktoken.get_encoding(encoding_name) self._tokenizer = enc self._allowed_special = allowed_special self._disallowed_special = disallowed_special [docs] def split_text(self, text: str) -> List[str]: def _encode(_text: str) -> List[int]: return self._tokenizer.encode( _text, allowed_special=self._allowed_special, disallowed_special=self._disallowed_special, ) tokenizer = Tokenizer( chunk_overlap=self._chunk_overlap, tokens_per_chunk=self._chunk_size, decode=self._tokenizer.decode, encode=_encode, ) return split_text_on_tokens(text=text, tokenizer=tokenizer) [docs]class SentenceTransformersTokenTextSplitter(TextSplitter): """Implementation of splitting text that looks at tokens.""" def __init__( self, chunk_overlap: int = 50, model_name: str = "sentence-transformers/all-mpnet-base-v2", tokens_per_chunk: Optional[int] = None, **kwargs: Any, ) -> None: """Create a new TextSplitter.""" super().__init__(**kwargs, chunk_overlap=chunk_overlap) try: from sentence_transformers import SentenceTransformer except ImportError: raise ImportError( "Could not import sentence_transformer python package. " "This is needed in order to for SentenceTransformersTokenTextSplitter. " "Please install it with `pip install sentence-transformers`." ) self.model_name = model_name self._model = SentenceTransformer(self.model_name) self.tokenizer = self._model.tokenizer self._initialize_chunk_configuration(tokens_per_chunk=tokens_per_chunk) def _initialize_chunk_configuration( self, *, tokens_per_chunk: Optional[int] ) -> None: self.maximum_tokens_per_chunk = cast(int, self._model.max_seq_length) if tokens_per_chunk is None: self.tokens_per_chunk = self.maximum_tokens_per_chunk else: self.tokens_per_chunk = tokens_per_chunk if self.tokens_per_chunk > self.maximum_tokens_per_chunk: raise ValueError( f"The token limit of the models '{self.model_name}'" f" is: {self.maximum_tokens_per_chunk}." f" Argument tokens_per_chunk={self.tokens_per_chunk}" f" > maximum token limit." ) [docs] def split_text(self, text: str) -> List[str]: def encode_strip_start_and_stop_token_ids(text: str) -> List[int]: return self._encode(text)[1:-1] tokenizer = Tokenizer( chunk_overlap=self._chunk_overlap, tokens_per_chunk=self.tokens_per_chunk, decode=self.tokenizer.decode, encode=encode_strip_start_and_stop_token_ids, ) return split_text_on_tokens(text=text, tokenizer=tokenizer) [docs] def count_tokens(self, *, text: str) -> int: return len(self._encode(text)) _max_length_equal_32_bit_integer = 2**32 def _encode(self, text: str) -> List[int]: token_ids_with_start_and_end_token_ids = self.tokenizer.encode( text, max_length=self._max_length_equal_32_bit_integer, truncation="do_not_truncate", ) return token_ids_with_start_and_end_token_ids [docs]class Language(str, Enum): CPP = "cpp" GO = "go" JAVA = "java" JS = "js" PHP = "php" PROTO = "proto" PYTHON = "python" RST = "rst" RUBY = "ruby" RUST = "rust" SCALA = "scala" SWIFT = "swift" MARKDOWN = "markdown" LATEX = "latex" HTML = "html" SOL = "sol" [docs]class RecursiveCharacterTextSplitter(TextSplitter): """Implementation of splitting text that looks at characters. Recursively tries to split by different characters to find one that works. """ def __init__( self, separators: Optional[List[str]] = None, keep_separator: bool = True, **kwargs: Any, ) -> None: """Create a new TextSplitter.""" super().__init__(keep_separator=keep_separator, **kwargs) self._separators = separators or ["\n\n", "\n", " ", ""] def _split_text(self, text: str, separators: List[str]) -> List[str]: """Split incoming text and return chunks.""" final_chunks = [] # Get appropriate separator to use separator = separators[-1] new_separators = [] for i, _s in enumerate(separators): if _s == "": separator = _s break if re.search(_s, text): separator = _s new_separators = separators[i + 1 :] break splits = _split_text_with_regex(text, separator, self._keep_separator) # Now go merging things, recursively splitting longer texts. _good_splits = [] _separator = "" if self._keep_separator else separator for s in splits: if self._length_function(s) < self._chunk_size: _good_splits.append(s) else: if _good_splits: merged_text = self._merge_splits(_good_splits, _separator) final_chunks.extend(merged_text) _good_splits = [] if not new_separators: final_chunks.append(s) else: other_info = self._split_text(s, new_separators) final_chunks.extend(other_info) if _good_splits: merged_text = self._merge_splits(_good_splits, _separator) final_chunks.extend(merged_text) return final_chunks [docs] def split_text(self, text: str) -> List[str]: return self._split_text(text, self._separators) [docs] @classmethod def from_language( cls, language: Language, **kwargs: Any ) -> RecursiveCharacterTextSplitter: separators = cls.get_separators_for_language(language) return cls(separators=separators, **kwargs) [docs] @staticmethod def get_separators_for_language(language: Language) -> List[str]: if language == Language.CPP: return [ # Split along class definitions "\nclass ", # Split along function definitions "\nvoid ", "\nint ", "\nfloat ", "\ndouble ", # Split along control flow statements "\nif ", "\nfor ", "\nwhile ", "\nswitch ", "\ncase ", # Split by the normal type of lines "\n\n", "\n", " ", "", ] elif language == Language.GO: return [ # Split along function definitions "\nfunc ", "\nvar ", "\nconst ", "\ntype ", # Split along control flow statements "\nif ", "\nfor ", "\nswitch ", "\ncase ", # Split by the normal type of lines "\n\n", "\n", " ", "", ] elif language == Language.JAVA: return [ # Split along class definitions "\nclass ", # Split along method definitions "\npublic ", "\nprotected ", "\nprivate ", "\nstatic ", # Split along control flow statements "\nif ", "\nfor ", "\nwhile ", "\nswitch ", "\ncase ", # Split by the normal type of lines "\n\n", "\n", " ", "", ] elif language == Language.JS: return [ # Split along function definitions "\nfunction ", "\nconst ", "\nlet ", "\nvar ", "\nclass ", # Split along control flow statements "\nif ", "\nfor ", "\nwhile ", "\nswitch ", "\ncase ", "\ndefault ", # Split by the normal type of lines "\n\n", "\n", " ", "", ] elif language == Language.PHP: return [ # Split along function definitions "\nfunction ", # Split along class definitions "\nclass ", # Split along control flow statements "\nif ", "\nforeach ", "\nwhile ", "\ndo ", "\nswitch ", "\ncase ", # Split by the normal type of lines "\n\n", "\n", " ", "", ] elif language == Language.PROTO: return [ # Split along message definitions "\nmessage ", # Split along service definitions "\nservice ", # Split along enum definitions "\nenum ", # Split along option definitions "\noption ", # Split along import statements "\nimport ", # Split along syntax declarations "\nsyntax ", # Split by the normal type of lines "\n\n", "\n", " ", "", ] elif language == Language.PYTHON: return [ # First, try to split along class definitions "\nclass ", "\ndef ", "\n\tdef ", # Now split by the normal type of lines "\n\n", "\n", " ", "", ] elif language == Language.RST: return [ # Split along section titles "\n=+\n", "\n-+\n", "\n\*+\n", # Split along directive markers "\n\n.. *\n\n", # Split by the normal type of lines "\n\n", "\n", " ", "", ] elif language == Language.RUBY: return [ # Split along method definitions "\ndef ", "\nclass ", # Split along control flow statements "\nif ", "\nunless ", "\nwhile ", "\nfor ", "\ndo ", "\nbegin ", "\nrescue ", # Split by the normal type of lines "\n\n", "\n", " ", "", ] elif language == Language.RUST: return [ # Split along function definitions "\nfn ", "\nconst ", "\nlet ", # Split along control flow statements "\nif ", "\nwhile ", "\nfor ", "\nloop ", "\nmatch ", "\nconst ", # Split by the normal type of lines "\n\n", "\n", " ", "", ] elif language == Language.SCALA: return [ # Split along class definitions "\nclass ", "\nobject ", # Split along method definitions "\ndef ", "\nval ", "\nvar ", # Split along control flow statements "\nif ", "\nfor ", "\nwhile ", "\nmatch ", "\ncase ", # Split by the normal type of lines "\n\n", "\n", " ", "", ] elif language == Language.SWIFT: return [ # Split along function definitions "\nfunc ", # Split along class definitions "\nclass ", "\nstruct ", "\nenum ", # Split along control flow statements "\nif ", "\nfor ", "\nwhile ", "\ndo ", "\nswitch ", "\ncase ", # Split by the normal type of lines "\n\n", "\n", " ", "", ] elif language == Language.MARKDOWN: return [ # First, try to split along Markdown headings (starting with level 2) "\n#{1,6} ", # Note the alternative syntax for headings (below) is not handled here # Heading level 2 # --------------- # End of code block "```\n", # Horizontal lines "\n\*\*\*+\n", "\n---+\n", "\n___+\n", # Note that this splitter doesn't handle horizontal lines defined # by *three or more* of ***, ---, or ___, but this is not handled "\n\n", "\n", " ", "", ] elif language == Language.LATEX: return [ # First, try to split along Latex sections "\n\\\chapter{", "\n\\\section{", "\n\\\subsection{", "\n\\\subsubsection{", # Now split by environments "\n\\\begin{enumerate}", "\n\\\begin{itemize}", "\n\\\begin{description}", "\n\\\begin{list}", "\n\\\begin{quote}", "\n\\\begin{quotation}", "\n\\\begin{verse}", "\n\\\begin{verbatim}", # Now split by math environments "\n\\\begin{align}", "$$", "$", # Now split by the normal type of lines " ", "", ] elif language == Language.HTML: return [ # First, try to split along HTML tags "<body", "<div", "<p", "<br", "<li", "<h1", "<h2", "<h3", "<h4", "<h5", "<h6", "<span", "<table", "<tr", "<td", "<th", "<ul", "<ol", "<header", "<footer", "<nav", # Head "<head", "<style", "<script", "<meta", "<title", "", ] elif language == Language.SOL: return [ # Split along compiler informations definitions "\npragma ", "\nusing ", # Split along contract definitions "\ncontract ", "\ninterface ", "\nlibrary ", # Split along method definitions "\nconstructor ", "\ntype ", "\nfunction ", "\nevent ", "\nmodifier ", "\nerror ", "\nstruct ", "\nenum ", # Split along control flow statements "\nif ", "\nfor ", "\nwhile ", "\ndo while ", "\nassembly ", # Split by the normal type of lines "\n\n", "\n", " ", "", ] else: raise ValueError( f"Language {language} is not supported! " f"Please choose from {list(Language)}" ) [docs]class NLTKTextSplitter(TextSplitter): """Implementation of splitting text that looks at sentences using NLTK.""" def __init__(self, separator: str = "\n\n", **kwargs: Any) -> None: """Initialize the NLTK splitter.""" super().__init__(**kwargs) try: from nltk.tokenize import sent_tokenize self._tokenizer = sent_tokenize except ImportError: raise ImportError( "NLTK is not installed, please install it with `pip install nltk`." ) self._separator = separator [docs] def split_text(self, text: str) -> List[str]: """Split incoming text and return chunks.""" # First we naively split the large input into a bunch of smaller ones. splits = self._tokenizer(text) return self._merge_splits(splits, self._separator) [docs]class SpacyTextSplitter(TextSplitter): """Implementation of splitting text that looks at sentences using Spacy.""" def __init__( self, separator: str = "\n\n", pipeline: str = "en_core_web_sm", **kwargs: Any ) -> None: """Initialize the spacy text splitter.""" super().__init__(**kwargs) try: import spacy except ImportError: raise ImportError( "Spacy is not installed, please install it with `pip install spacy`." ) self._tokenizer = spacy.load(pipeline) self._separator = separator [docs] def split_text(self, text: str) -> List[str]: """Split incoming text and return chunks.""" splits = (str(s) for s in self._tokenizer(text).sents) return self._merge_splits(splits, self._separator) # For backwards compatibility [docs]class PythonCodeTextSplitter(RecursiveCharacterTextSplitter): """Attempts to split the text along Python syntax.""" def __init__(self, **kwargs: Any) -> None: """Initialize a PythonCodeTextSplitter.""" separators = self.get_separators_for_language(Language.PYTHON) super().__init__(separators=separators, **kwargs) [docs]class MarkdownTextSplitter(RecursiveCharacterTextSplitter): """Attempts to split the text along Markdown-formatted headings.""" def __init__(self, **kwargs: Any) -> None: """Initialize a MarkdownTextSplitter.""" separators = self.get_separators_for_language(Language.MARKDOWN) super().__init__(separators=separators, **kwargs) [docs]class LatexTextSplitter(RecursiveCharacterTextSplitter): """Attempts to split the text along Latex-formatted layout elements.""" def __init__(self, **kwargs: Any) -> None: """Initialize a LatexTextSplitter.""" separators = self.get_separators_for_language(Language.LATEX) super().__init__(separators=separators, **kwargs)
https://api.python.langchain.com/en/latest/_modules/langchain/text_splitter.html
11bb2cef-40ed-4a73-ab1d-b2444d0d686c
Source code for langchain.schema """Common schema objects.""" from __future__ import annotations from abc import ABC, abstractmethod from dataclasses import dataclass from typing import ( Any, Dict, Generic, List, NamedTuple, Optional, Sequence, TypeVar, Union, ) from uuid import UUID from pydantic import BaseModel, Field, root_validator from langchain.load.serializable import Serializable RUN_KEY = "__run" [docs]def get_buffer_string( messages: List[BaseMessage], human_prefix: str = "Human", ai_prefix: str = "AI" ) -> str: """Get buffer string of messages.""" string_messages = [] for m in messages: if isinstance(m, HumanMessage): role = human_prefix elif isinstance(m, AIMessage): role = ai_prefix elif isinstance(m, SystemMessage): role = "System" elif isinstance(m, FunctionMessage): role = "Function" elif isinstance(m, ChatMessage): role = m.role else: raise ValueError(f"Got unsupported message type: {m}") message = f"{role}: {m.content}" if isinstance(m, AIMessage) and "function_call" in m.additional_kwargs: message += f"{m.additional_kwargs['function_call']}" string_messages.append(message) return "\n".join(string_messages) [docs]@dataclass class AgentAction: """Agent's action to take.""" tool: str tool_input: Union[str, dict] log: str [docs]class AgentFinish(NamedTuple): """Agent's return value.""" return_values: dict log: str [docs]class Generation(Serializable): """Output of a single generation.""" text: str """Generated text output.""" generation_info: Optional[Dict[str, Any]] = None """Raw generation info response from the provider""" """May include things like reason for finishing (e.g. in OpenAI)""" # TODO: add log probs @property def lc_serializable(self) -> bool: """This class is LangChain serializable.""" return True [docs]class BaseMessage(Serializable): """Message object.""" content: str additional_kwargs: dict = Field(default_factory=dict) @property @abstractmethod def type(self) -> str: """Type of the message, used for serialization.""" @property def lc_serializable(self) -> bool: """This class is LangChain serializable.""" return True [docs]class HumanMessage(BaseMessage): """Type of message that is spoken by the human.""" example: bool = False @property def type(self) -> str: """Type of the message, used for serialization.""" return "human" [docs]class AIMessage(BaseMessage): """Type of message that is spoken by the AI.""" example: bool = False @property def type(self) -> str: """Type of the message, used for serialization.""" return "ai" [docs]class SystemMessage(BaseMessage): """Type of message that is a system message.""" @property def type(self) -> str: """Type of the message, used for serialization.""" return "system" [docs]class FunctionMessage(BaseMessage): name: str @property def type(self) -> str: """Type of the message, used for serialization.""" return "function" [docs]class ChatMessage(BaseMessage): """Type of message with arbitrary speaker.""" role: str @property def type(self) -> str: """Type of the message, used for serialization.""" return "chat" def _message_to_dict(message: BaseMessage) -> dict: return {"type": message.type, "data": message.dict()} [docs]def messages_to_dict(messages: List[BaseMessage]) -> List[dict]: """Convert messages to dict. Args: messages: List of messages to convert. Returns: List of dicts. """ return [_message_to_dict(m) for m in messages] def _message_from_dict(message: dict) -> BaseMessage: _type = message["type"] if _type == "human": return HumanMessage(**message["data"]) elif _type == "ai": return AIMessage(**message["data"]) elif _type == "system": return SystemMessage(**message["data"]) elif _type == "chat": return ChatMessage(**message["data"]) else: raise ValueError(f"Got unexpected type: {_type}") [docs]def messages_from_dict(messages: List[dict]) -> List[BaseMessage]: """Convert messages from dict. Args: messages: List of messages (dicts) to convert. Returns: List of messages (BaseMessages). """ return [_message_from_dict(m) for m in messages] [docs]class ChatGeneration(Generation): """Output of a single generation.""" text = "" message: BaseMessage @root_validator def set_text(cls, values: Dict[str, Any]) -> Dict[str, Any]: values["text"] = values["message"].content return values [docs]class RunInfo(BaseModel): """Class that contains all relevant metadata for a Run.""" run_id: UUID [docs]class ChatResult(BaseModel): """Class that contains all relevant information for a Chat Result.""" generations: List[ChatGeneration] """List of the things generated.""" llm_output: Optional[dict] = None """For arbitrary LLM provider specific output.""" [docs]class LLMResult(BaseModel): """Class that contains all relevant information for an LLM Result.""" generations: List[List[Generation]] """List of the things generated. This is List[List[]] because each input could have multiple generations.""" llm_output: Optional[dict] = None """For arbitrary LLM provider specific output.""" run: Optional[List[RunInfo]] = None """Run metadata.""" [docs] def flatten(self) -> List[LLMResult]: """Flatten generations into a single list.""" llm_results = [] for i, gen_list in enumerate(self.generations): # Avoid double counting tokens in OpenAICallback if i == 0: llm_results.append( LLMResult( generations=[gen_list], llm_output=self.llm_output, ) ) else: if self.llm_output is not None: llm_output = self.llm_output.copy() llm_output["token_usage"] = dict() else: llm_output = None llm_results.append( LLMResult( generations=[gen_list], llm_output=llm_output, ) ) return llm_results def __eq__(self, other: object) -> bool: if not isinstance(other, LLMResult): return NotImplemented return ( self.generations == other.generations and self.llm_output == other.llm_output ) [docs]class PromptValue(Serializable, ABC): [docs] @abstractmethod def to_string(self) -> str: """Return prompt as string.""" [docs] @abstractmethod def to_messages(self) -> List[BaseMessage]: """Return prompt as messages.""" [docs]class BaseMemory(Serializable, ABC): """Base interface for memory in chains.""" class Config: """Configuration for this pydantic object.""" arbitrary_types_allowed = True @property @abstractmethod def memory_variables(self) -> List[str]: """Input keys this memory class will load dynamically.""" [docs] @abstractmethod def load_memory_variables(self, inputs: Dict[str, Any]) -> Dict[str, Any]: """Return key-value pairs given the text input to the chain. If None, return all memories """ [docs] @abstractmethod def save_context(self, inputs: Dict[str, Any], outputs: Dict[str, str]) -> None: """Save the context of this model run to memory.""" [docs] @abstractmethod def clear(self) -> None: """Clear memory contents.""" [docs]class BaseChatMessageHistory(ABC): """Base interface for chat message history See `ChatMessageHistory` for default implementation. """ """ Example: .. code-block:: python class FileChatMessageHistory(BaseChatMessageHistory): storage_path: str session_id: str @property def messages(self): with open(os.path.join(storage_path, session_id), 'r:utf-8') as f: messages = json.loads(f.read()) return messages_from_dict(messages) def add_message(self, message: BaseMessage) -> None: messages = self.messages.append(_message_to_dict(message)) with open(os.path.join(storage_path, session_id), 'w') as f: json.dump(f, messages) def clear(self): with open(os.path.join(storage_path, session_id), 'w') as f: f.write("[]") """ messages: List[BaseMessage] [docs] def add_user_message(self, message: str) -> None: """Add a user message to the store""" self.add_message(HumanMessage(content=message)) [docs] def add_ai_message(self, message: str) -> None: """Add an AI message to the store""" self.add_message(AIMessage(content=message)) [docs] def add_message(self, message: BaseMessage) -> None: """Add a self-created message to the store""" raise NotImplementedError [docs] @abstractmethod def clear(self) -> None: """Remove all messages from the store""" [docs]class Document(Serializable): """Interface for interacting with a document.""" page_content: str metadata: dict = Field(default_factory=dict) [docs]class BaseRetriever(ABC): """Base interface for retrievers.""" [docs] @abstractmethod def get_relevant_documents(self, query: str) -> List[Document]: """Get documents relevant for a query. Args: query: string to find relevant documents for Returns: List of relevant documents """ [docs] @abstractmethod async def aget_relevant_documents(self, query: str) -> List[Document]: """Get documents relevant for a query. Args: query: string to find relevant documents for Returns: List of relevant documents """ # For backwards compatibility Memory = BaseMemory T = TypeVar("T") [docs]class BaseLLMOutputParser(Serializable, ABC, Generic[T]): [docs] @abstractmethod def parse_result(self, result: List[Generation]) -> T: """Parse LLM Result.""" [docs]class BaseOutputParser(BaseLLMOutputParser, ABC, Generic[T]): """Class to parse the output of an LLM call. Output parsers help structure language model responses. """ [docs] def parse_result(self, result: List[Generation]) -> T: return self.parse(result[0].text) [docs] @abstractmethod def parse(self, text: str) -> T: """Parse the output of an LLM call. A method which takes in a string (assumed output of a language model ) and parses it into some structure. Args: text: output of language model Returns: structured output """ [docs] def parse_with_prompt(self, completion: str, prompt: PromptValue) -> Any: """Optional method to parse the output of an LLM call with a prompt. The prompt is largely provided in the event the OutputParser wants to retry or fix the output in some way, and needs information from the prompt to do so. Args: completion: output of language model prompt: prompt value Returns: structured output """ return self.parse(completion) [docs] def get_format_instructions(self) -> str: """Instructions on how the LLM output should be formatted.""" raise NotImplementedError @property def _type(self) -> str: """Return the type key.""" raise NotImplementedError( f"_type property is not implemented in class {self.__class__.__name__}." " This is required for serialization." ) [docs] def dict(self, **kwargs: Any) -> Dict: """Return dictionary representation of output parser.""" output_parser_dict = super().dict() output_parser_dict["_type"] = self._type return output_parser_dict [docs]class NoOpOutputParser(BaseOutputParser[str]): """Output parser that just returns the text as is.""" @property def lc_serializable(self) -> bool: return True @property def _type(self) -> str: return "default" [docs] def parse(self, text: str) -> str: return text [docs]class OutputParserException(ValueError): """Exception that output parsers should raise to signify a parsing error. This exists to differentiate parsing errors from other code or execution errors that also may arise inside the output parser. OutputParserExceptions will be available to catch and handle in ways to fix the parsing error, while other errors will be raised. """ def __init__( self, error: Any, observation: str | None = None, llm_output: str | None = None, send_to_llm: bool = False, ): super(OutputParserException, self).__init__(error) if send_to_llm: if observation is None or llm_output is None: raise ValueError( "Arguments 'observation' & 'llm_output'" " are required if 'send_to_llm' is True" ) self.observation = observation self.llm_output = llm_output self.send_to_llm = send_to_llm [docs]class BaseDocumentTransformer(ABC): """Base interface for transforming documents.""" [docs] @abstractmethod def transform_documents( self, documents: Sequence[Document], **kwargs: Any ) -> Sequence[Document]: """Transform a list of documents.""" [docs] @abstractmethod async def atransform_documents( self, documents: Sequence[Document], **kwargs: Any ) -> Sequence[Document]: """Asynchronously transform a list of documents."""
https://api.python.langchain.com/en/latest/_modules/langchain/schema.html
f90ea9cb-01b6-4419-8455-44ec9c6f15c3
Source code for langchain.document_transformers """Transform documents""" from typing import Any, Callable, List, Sequence import numpy as np from pydantic import BaseModel, Field from langchain.embeddings.base import Embeddings from langchain.math_utils import cosine_similarity from langchain.schema import BaseDocumentTransformer, Document class _DocumentWithState(Document): """Wrapper for a document that includes arbitrary state.""" state: dict = Field(default_factory=dict) """State associated with the document.""" def to_document(self) -> Document: """Convert the DocumentWithState to a Document.""" return Document(page_content=self.page_content, metadata=self.metadata) @classmethod def from_document(cls, doc: Document) -> "_DocumentWithState": """Create a DocumentWithState from a Document.""" if isinstance(doc, cls): return doc return cls(page_content=doc.page_content, metadata=doc.metadata) [docs]def get_stateful_documents( documents: Sequence[Document], ) -> Sequence[_DocumentWithState]: """Convert a list of documents to a list of documents with state. Args: documents: The documents to convert. Returns: A list of documents with state. """ return [_DocumentWithState.from_document(doc) for doc in documents] def _filter_similar_embeddings( embedded_documents: List[List[float]], similarity_fn: Callable, threshold: float ) -> List[int]: """Filter redundant documents based on the similarity of their embeddings.""" similarity = np.tril(similarity_fn(embedded_documents, embedded_documents), k=-1) redundant = np.where(similarity > threshold) redundant_stacked = np.column_stack(redundant) redundant_sorted = np.argsort(similarity[redundant])[::-1] included_idxs = set(range(len(embedded_documents))) for first_idx, second_idx in redundant_stacked[redundant_sorted]: if first_idx in included_idxs and second_idx in included_idxs: # Default to dropping the second document of any highly similar pair. included_idxs.remove(second_idx) return list(sorted(included_idxs)) def _get_embeddings_from_stateful_docs( embeddings: Embeddings, documents: Sequence[_DocumentWithState] ) -> List[List[float]]: if len(documents) and "embedded_doc" in documents[0].state: embedded_documents = [doc.state["embedded_doc"] for doc in documents] else: embedded_documents = embeddings.embed_documents( [d.page_content for d in documents] ) for doc, embedding in zip(documents, embedded_documents): doc.state["embedded_doc"] = embedding return embedded_documents [docs]class EmbeddingsRedundantFilter(BaseDocumentTransformer, BaseModel): """Filter that drops redundant documents by comparing their embeddings.""" embeddings: Embeddings """Embeddings to use for embedding document contents.""" similarity_fn: Callable = cosine_similarity """Similarity function for comparing documents. Function expected to take as input two matrices (List[List[float]]) and return a matrix of scores where higher values indicate greater similarity.""" similarity_threshold: float = 0.95 """Threshold for determining when two documents are similar enough to be considered redundant.""" class Config: """Configuration for this pydantic object.""" arbitrary_types_allowed = True [docs] def transform_documents( self, documents: Sequence[Document], **kwargs: Any ) -> Sequence[Document]: """Filter down documents.""" stateful_documents = get_stateful_documents(documents) embedded_documents = _get_embeddings_from_stateful_docs( self.embeddings, stateful_documents ) included_idxs = _filter_similar_embeddings( embedded_documents, self.similarity_fn, self.similarity_threshold ) return [stateful_documents[i] for i in sorted(included_idxs)] [docs] async def atransform_documents( self, documents: Sequence[Document], **kwargs: Any ) -> Sequence[Document]: raise NotImplementedError
https://api.python.langchain.com/en/latest/_modules/langchain/document_transformers.html
4899e0e4-1602-471e-8bd2-9c58e78677eb
Source code for langchain.vectorstores.clickhouse """Wrapper around open source ClickHouse VectorSearch capability.""" from __future__ import annotations import json import logging from hashlib import sha1 from threading import Thread from typing import Any, Dict, Iterable, List, Optional, Tuple, Union from pydantic import BaseSettings from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.vectorstores.base import VectorStore logger = logging.getLogger() def has_mul_sub_str(s: str, *args: Any) -> bool: """ Check if a string contains multiple substrings. Args: s: string to check. *args: substrings to check. Returns: True if all substrings are in the string, False otherwise. """ for a in args: if a not in s: return False return True [docs]class ClickhouseSettings(BaseSettings): """ClickHouse Client Configuration Attribute: clickhouse_host (str) : An URL to connect to MyScale backend. Defaults to 'localhost'. clickhouse_port (int) : URL port to connect with HTTP. Defaults to 8443. username (str) : Username to login. Defaults to None. password (str) : Password to login. Defaults to None. index_type (str): index type string. index_param (list): index build parameter. index_query_params(dict): index query parameters. database (str) : Database name to find the table. Defaults to 'default'. table (str) : Table name to operate on. Defaults to 'vector_table'. metric (str) : Metric to compute distance, supported are ('angular', 'euclidean', 'manhattan', 'hamming', 'dot'). Defaults to 'angular'. https://github.com/spotify/annoy/blob/main/src/annoymodule.cc#L149-L169 column_map (Dict) : Column type map to project column name onto langchain semantics. Must have keys: `text`, `id`, `vector`, must be same size to number of columns. For example: .. code-block:: python { 'id': 'text_id', 'uuid': 'global_unique_id' 'embedding': 'text_embedding', 'document': 'text_plain', 'metadata': 'metadata_dictionary_in_json', } Defaults to identity map. """ host: str = "localhost" port: int = 8123 username: Optional[str] = None password: Optional[str] = None index_type: str = "annoy" # Annoy supports L2Distance and cosineDistance. index_param: Optional[Union[List, Dict]] = ["'L2Distance'", 100] index_query_params: Dict[str, str] = {} column_map: Dict[str, str] = { "id": "id", "uuid": "uuid", "document": "document", "embedding": "embedding", "metadata": "metadata", } database: str = "default" table: str = "langchain" metric: str = "angular" def __getitem__(self, item: str) -> Any: return getattr(self, item) class Config: env_file = ".env" env_prefix = "clickhouse_" env_file_encoding = "utf-8" [docs]class Clickhouse(VectorStore): """Wrapper around ClickHouse vector database You need a `clickhouse-connect` python package, and a valid account to connect to ClickHouse. ClickHouse can not only search with simple vector indexes, it also supports complex query with multiple conditions, constraints and even sub-queries. For more information, please visit [ClickHouse official site](https://clickhouse.com/clickhouse) """ def __init__( self, embedding: Embeddings, config: Optional[ClickhouseSettings] = None, **kwargs: Any, ) -> None: """ClickHouse Wrapper to LangChain embedding_function (Embeddings): config (ClickHouseSettings): Configuration to ClickHouse Client Other keyword arguments will pass into [clickhouse-connect](https://docs.clickhouse.com/) """ try: from clickhouse_connect import get_client except ImportError: raise ValueError( "Could not import clickhouse connect python package. " "Please install it with `pip install clickhouse-connect`." ) try: from tqdm import tqdm self.pgbar = tqdm except ImportError: # Just in case if tqdm is not installed self.pgbar = lambda x, **kwargs: x super().__init__() if config is not None: self.config = config else: self.config = ClickhouseSettings() assert self.config assert self.config.host and self.config.port assert ( self.config.column_map and self.config.database and self.config.table and self.config.metric ) for k in ["id", "embedding", "document", "metadata", "uuid"]: assert k in self.config.column_map assert self.config.metric in [ "angular", "euclidean", "manhattan", "hamming", "dot", ] # initialize the schema dim = len(embedding.embed_query("test")) index_params = ( ( ",".join([f"'{k}={v}'" for k, v in self.config.index_param.items()]) if self.config.index_param else "" ) if isinstance(self.config.index_param, Dict) else ",".join([str(p) for p in self.config.index_param]) if isinstance(self.config.index_param, List) else self.config.index_param ) self.schema = f"""\ CREATE TABLE IF NOT EXISTS {self.config.database}.{self.config.table}( {self.config.column_map['id']} Nullable(String), {self.config.column_map['document']} Nullable(String), {self.config.column_map['embedding']} Array(Float32), {self.config.column_map['metadata']} JSON, {self.config.column_map['uuid']} UUID DEFAULT generateUUIDv4(), CONSTRAINT cons_vec_len CHECK length({self.config.column_map['embedding']}) = {dim}, INDEX vec_idx {self.config.column_map['embedding']} TYPE \ {self.config.index_type}({index_params}) GRANULARITY 1000 ) ENGINE = MergeTree ORDER BY uuid SETTINGS index_granularity = 8192\ """ self.dim = dim self.BS = "\\" self.must_escape = ("\\", "'") self.embedding_function = embedding self.dist_order = "ASC" # Only support ConsingDistance and L2Distance # Create a connection to clickhouse self.client = get_client( host=self.config.host, port=self.config.port, username=self.config.username, password=self.config.password, **kwargs, ) # Enable JSON type self.client.command("SET allow_experimental_object_type=1") # Enable Annoy index self.client.command("SET allow_experimental_annoy_index=1") self.client.command(self.schema) [docs] def escape_str(self, value: str) -> str: return "".join(f"{self.BS}{c}" if c in self.must_escape else c for c in value) def _build_insert_sql(self, transac: Iterable, column_names: Iterable[str]) -> str: ks = ",".join(column_names) _data = [] for n in transac: n = ",".join([f"'{self.escape_str(str(_n))}'" for _n in n]) _data.append(f"({n})") i_str = f""" INSERT INTO TABLE {self.config.database}.{self.config.table}({ks}) VALUES {','.join(_data)} """ return i_str def _insert(self, transac: Iterable, column_names: Iterable[str]) -> None: _insert_query = self._build_insert_sql(transac, column_names) self.client.command(_insert_query) [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, batch_size: int = 32, ids: Optional[Iterable[str]] = None, **kwargs: Any, ) -> List[str]: """Insert more texts through the embeddings and add to the VectorStore. Args: texts: Iterable of strings to add to the VectorStore. ids: Optional list of ids to associate with the texts. batch_size: Batch size of insertion metadata: Optional column data to be inserted Returns: List of ids from adding the texts into the VectorStore. """ # Embed and create the documents ids = ids or [sha1(t.encode("utf-8")).hexdigest() for t in texts] colmap_ = self.config.column_map transac = [] column_names = { colmap_["id"]: ids, colmap_["document"]: texts, colmap_["embedding"]: self.embedding_function.embed_documents(list(texts)), } metadatas = metadatas or [{} for _ in texts] column_names[colmap_["metadata"]] = map(json.dumps, metadatas) assert len(set(colmap_) - set(column_names)) >= 0 keys, values = zip(*column_names.items()) try: t = None for v in self.pgbar( zip(*values), desc="Inserting data...", total=len(metadatas) ): assert ( len(v[keys.index(self.config.column_map["embedding"])]) == self.dim ) transac.append(v) if len(transac) == batch_size: if t: t.join() t = Thread(target=self._insert, args=[transac, keys]) t.start() transac = [] if len(transac) > 0: if t: t.join() self._insert(transac, keys) return [i for i in ids] except Exception as e: logger.error(f"\033[91m\033[1m{type(e)}\033[0m \033[95m{str(e)}\033[0m") return [] [docs] @classmethod def from_texts( cls, texts: List[str], embedding: Embeddings, metadatas: Optional[List[Dict[Any, Any]]] = None, config: Optional[ClickhouseSettings] = None, text_ids: Optional[Iterable[str]] = None, batch_size: int = 32, **kwargs: Any, ) -> Clickhouse: """Create ClickHouse wrapper with existing texts Args: embedding_function (Embeddings): Function to extract text embedding texts (Iterable[str]): List or tuple of strings to be added config (ClickHouseSettings, Optional): ClickHouse configuration text_ids (Optional[Iterable], optional): IDs for the texts. Defaults to None. batch_size (int, optional): Batchsize when transmitting data to ClickHouse. Defaults to 32. metadata (List[dict], optional): metadata to texts. Defaults to None. Other keyword arguments will pass into [clickhouse-connect](https://clickhouse.com/docs/en/integrations/python#clickhouse-connect-driver-api) Returns: ClickHouse Index """ ctx = cls(embedding, config, **kwargs) ctx.add_texts(texts, ids=text_ids, batch_size=batch_size, metadatas=metadatas) return ctx def __repr__(self) -> str: """Text representation for ClickHouse Vector Store, prints backends, username and schemas. Easy to use with `str(ClickHouse())` Returns: repr: string to show connection info and data schema """ _repr = f"\033[92m\033[1m{self.config.database}.{self.config.table} @ " _repr += f"{self.config.host}:{self.config.port}\033[0m\n\n" _repr += f"\033[1musername: {self.config.username}\033[0m\n\nTable Schema:\n" _repr += "-" * 51 + "\n" for r in self.client.query( f"DESC {self.config.database}.{self.config.table}" ).named_results(): _repr += ( f"|\033[94m{r['name']:24s}\033[0m|\033[96m{r['type']:24s}\033[0m|\n" ) _repr += "-" * 51 + "\n" return _repr def _build_query_sql( self, q_emb: List[float], topk: int, where_str: Optional[str] = None ) -> str: q_emb_str = ",".join(map(str, q_emb)) if where_str: where_str = f"PREWHERE {where_str}" else: where_str = "" settings_strs = [] if self.config.index_query_params: for k in self.config.index_query_params: settings_strs.append(f"SETTING {k}={self.config.index_query_params[k]}") q_str = f""" SELECT {self.config.column_map['document']}, {self.config.column_map['metadata']}, dist FROM {self.config.database}.{self.config.table} {where_str} ORDER BY L2Distance({self.config.column_map['embedding']}, [{q_emb_str}]) AS dist {self.dist_order} LIMIT {topk} {' '.join(settings_strs)} """ return q_str [docs] def similarity_search( self, query: str, k: int = 4, where_str: Optional[str] = None, **kwargs: Any ) -> List[Document]: """Perform a similarity search with ClickHouse Args: query (str): query string k (int, optional): Top K neighbors to retrieve. Defaults to 4. where_str (Optional[str], optional): where condition string. Defaults to None. NOTE: Please do not let end-user to fill this and always be aware of SQL injection. When dealing with metadatas, remember to use `{self.metadata_column}.attribute` instead of `attribute` alone. The default name for it is `metadata`. Returns: List[Document]: List of Documents """ return self.similarity_search_by_vector( self.embedding_function.embed_query(query), k, where_str, **kwargs ) [docs] def similarity_search_by_vector( self, embedding: List[float], k: int = 4, where_str: Optional[str] = None, **kwargs: Any, ) -> List[Document]: """Perform a similarity search with ClickHouse by vectors Args: query (str): query string k (int, optional): Top K neighbors to retrieve. Defaults to 4. where_str (Optional[str], optional): where condition string. Defaults to None. NOTE: Please do not let end-user to fill this and always be aware of SQL injection. When dealing with metadatas, remember to use `{self.metadata_column}.attribute` instead of `attribute` alone. The default name for it is `metadata`. Returns: List[Document]: List of (Document, similarity) """ q_str = self._build_query_sql(embedding, k, where_str) try: return [ Document( page_content=r[self.config.column_map["document"]], metadata=r[self.config.column_map["metadata"]], ) for r in self.client.query(q_str).named_results() ] except Exception as e: logger.error(f"\033[91m\033[1m{type(e)}\033[0m \033[95m{str(e)}\033[0m") return [] [docs] def similarity_search_with_relevance_scores( self, query: str, k: int = 4, where_str: Optional[str] = None, **kwargs: Any ) -> List[Tuple[Document, float]]: """Perform a similarity search with ClickHouse Args: query (str): query string k (int, optional): Top K neighbors to retrieve. Defaults to 4. where_str (Optional[str], optional): where condition string. Defaults to None. NOTE: Please do not let end-user to fill this and always be aware of SQL injection. When dealing with metadatas, remember to use `{self.metadata_column}.attribute` instead of `attribute` alone. The default name for it is `metadata`. Returns: List[Document]: List of documents """ q_str = self._build_query_sql( self.embedding_function.embed_query(query), k, where_str ) try: return [ ( Document( page_content=r[self.config.column_map["document"]], metadata=r[self.config.column_map["metadata"]], ), r["dist"], ) for r in self.client.query(q_str).named_results() ] except Exception as e: logger.error(f"\033[91m\033[1m{type(e)}\033[0m \033[95m{str(e)}\033[0m") return [] [docs] def drop(self) -> None: """ Helper function: Drop data """ self.client.command( f"DROP TABLE IF EXISTS {self.config.database}.{self.config.table}" ) @property def metadata_column(self) -> str: return self.config.column_map["metadata"]
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/clickhouse.html
211b670b-f857-449e-9721-b44bc7f2bae3
Source code for langchain.vectorstores.alibabacloud_opensearch import json import logging import numbers from hashlib import sha1 from typing import Any, Dict, Iterable, List, Optional, Tuple from langchain.embeddings.base import Embeddings from langchain.schema import Document from langchain.vectorstores.base import VectorStore logger = logging.getLogger() [docs]class AlibabaCloudOpenSearchSettings: """Opensearch Client Configuration Attribute: endpoint (str) : The endpoint of opensearch instance, You can find it from the console of Alibaba Cloud OpenSearch. instance_id (str) : The identify of opensearch instance, You can find it from the console of Alibaba Cloud OpenSearch. datasource_name (str): The name of the data source specified when creating it. username (str) : The username specified when purchasing the instance. password (str) : The password specified when purchasing the instance. embedding_index_name (str) : The name of the vector attribute specified when configuring the instance attributes. field_name_mapping (Dict) : Using field name mapping between opensearch vector store and opensearch instance configuration table field names: { 'id': 'The id field name map of index document.', 'document': 'The text field name map of index document.', 'embedding': 'In the embedding field of the opensearch instance, the values must be in float16 multivalue type and separated by commas.', 'metadata_field_x': 'Metadata field mapping includes the mapped field name and operator in the mapping value, separated by a comma between the mapped field name and the operator.', } """ endpoint: str instance_id: str username: str password: str datasource_name: str embedding_index_name: str field_name_mapping: Dict[str, str] = { "id": "id", "document": "document", "embedding": "embedding", "metadata_field_x": "metadata_field_x,operator", } def __init__( self, endpoint: str, instance_id: str, username: str, password: str, datasource_name: str, embedding_index_name: str, field_name_mapping: Dict[str, str], ) -> None: self.endpoint = endpoint self.instance_id = instance_id self.username = username self.password = password self.datasource_name = datasource_name self.embedding_index_name = embedding_index_name self.field_name_mapping = field_name_mapping def __getitem__(self, item: str) -> Any: return getattr(self, item) def create_metadata(fields: Dict[str, Any]) -> Dict[str, Any]: """Create metadata from fields. Args: fields: The fields of the document. The fields must be a dict. Returns: metadata: The metadata of the document. The metadata must be a dict. """ metadata: Dict[str, Any] = {} for key, value in fields.items(): if key == "id" or key == "document" or key == "embedding": continue metadata[key] = value return metadata [docs]class AlibabaCloudOpenSearch(VectorStore): """Alibaba Cloud OpenSearch Vector Store""" def __init__( self, embedding: Embeddings, config: AlibabaCloudOpenSearchSettings, **kwargs: Any, ) -> None: try: from alibabacloud_ha3engine import client, models from alibabacloud_tea_util import models as util_models except ImportError: raise ValueError( "Could not import alibaba cloud opensearch python package. " "Please install it with `pip install alibabacloud-ha3engine`." ) self.config = config self.embedding = embedding self.runtime = util_models.RuntimeOptions( connect_timeout=5000, read_timeout=10000, autoretry=False, ignore_ssl=False, max_idle_conns=50, ) self.ha3EngineClient = client.Client( models.Config( endpoint=config.endpoint, instance_id=config.instance_id, protocol="http", access_user_name=config.username, access_pass_word=config.password, ) ) self.options_headers: Dict[str, str] = {} [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, **kwargs: Any, ) -> List[str]: def _upsert(push_doc_list: List[Dict]) -> List[str]: if push_doc_list is None or len(push_doc_list) == 0: return [] try: push_request = models.PushDocumentsRequestModel( self.options_headers, push_doc_list ) push_response = self.ha3EngineClient.push_documents( self.config.datasource_name, field_name_map["id"], push_request ) json_response = json.loads(push_response.body) if json_response["status"] == "OK": return [ push_doc["fields"][field_name_map["id"]] for push_doc in push_doc_list ] return [] except Exception as e: logger.error( f"add doc to endpoint:{self.config.endpoint} " f"instance_id:{self.config.instance_id} failed.", e, ) raise e from alibabacloud_ha3engine import models ids = [sha1(t.encode("utf-8")).hexdigest() for t in texts] embeddings = self.embedding.embed_documents(list(texts)) metadatas = metadatas or [{} for _ in texts] field_name_map = self.config.field_name_mapping add_doc_list = [] text_list = list(texts) for idx, doc_id in enumerate(ids): embedding = embeddings[idx] if idx < len(embeddings) else None metadata = metadatas[idx] if idx < len(metadatas) else None text = text_list[idx] if idx < len(text_list) else None add_doc: Dict[str, Any] = dict() add_doc_fields: Dict[str, Any] = dict() add_doc_fields.__setitem__(field_name_map["id"], doc_id) add_doc_fields.__setitem__(field_name_map["document"], text) if embedding is not None: add_doc_fields.__setitem__( field_name_map["embedding"], ",".join(str(unit) for unit in embedding), ) if metadata is not None: for md_key, md_value in metadata.items(): add_doc_fields.__setitem__( field_name_map[md_key].split(",")[0], md_value ) add_doc.__setitem__("fields", add_doc_fields) add_doc.__setitem__("cmd", "add") add_doc_list.append(add_doc) return _upsert(add_doc_list) [docs] def similarity_search( self, query: str, k: int = 4, search_filter: Optional[Dict[str, Any]] = None, **kwargs: Any, ) -> List[Document]: embedding = self.embedding.embed_query(query) return self.create_results( self.inner_embedding_query( embedding=embedding, search_filter=search_filter, k=k ) ) [docs] def similarity_search_with_relevance_scores( self, query: str, k: int = 4, search_filter: Optional[dict] = None, **kwargs: Any, ) -> List[Tuple[Document, float]]: embedding: List[float] = self.embedding.embed_query(query) return self.create_results_with_score( self.inner_embedding_query( embedding=embedding, search_filter=search_filter, k=k ) ) [docs] def similarity_search_by_vector( self, embedding: List[float], k: int = 4, search_filter: Optional[dict] = None, **kwargs: Any, ) -> List[Document]: return self.create_results( self.inner_embedding_query( embedding=embedding, search_filter=search_filter, k=k ) ) [docs] def inner_embedding_query( self, embedding: List[float], search_filter: Optional[Dict[str, Any]] = None, k: int = 4, ) -> Dict[str, Any]: def generate_embedding_query() -> str: tmp_search_config_str = ( f"config=start:0,hit:{k},format:json&&cluster=general&&kvpairs=" f"first_formula:proxima_score({self.config.embedding_index_name})&&sort=+RANK" ) tmp_query_str = ( f"&&query={self.config.embedding_index_name}:" + "'" + ",".join(str(x) for x in embedding) + "'" ) if search_filter is not None: filter_clause = "&&filter=" + " AND ".join( [ create_filter(md_key, md_value) for md_key, md_value in search_filter.items() ] ) tmp_query_str += filter_clause return tmp_search_config_str + tmp_query_str def create_filter(md_key: str, md_value: Any) -> str: md_filter_expr = self.config.field_name_mapping[md_key] if md_filter_expr is None: return "" expr = md_filter_expr.split(",") if len(expr) != 2: logger.error( f"filter {md_filter_expr} express is not correct, " f"must contain mapping field and operator." ) return "" md_filter_key = expr[0].strip() md_filter_operator = expr[1].strip() if isinstance(md_value, numbers.Number): return f"{md_filter_key} {md_filter_operator} {md_value}" return f'{md_filter_key}{md_filter_operator}"{md_value}"' def search_data(single_query_str: str) -> Dict[str, Any]: search_query = models.SearchQuery(query=single_query_str) search_request = models.SearchRequestModel( self.options_headers, search_query ) return json.loads(self.ha3EngineClient.search(search_request).body) from alibabacloud_ha3engine import models try: query_str = generate_embedding_query() json_response = search_data(query_str) if len(json_response["errors"]) != 0: logger.error( f"query {self.config.endpoint} {self.config.instance_id} " f"errors:{json_response['errors']} failed." ) else: return json_response except Exception as e: logger.error( f"query instance endpoint:{self.config.endpoint} " f"instance_id:{self.config.instance_id} failed.", e, ) return {} [docs] def create_results(self, json_result: Dict[str, Any]) -> List[Document]: items = json_result["result"]["items"] query_result_list: List[Document] = [] for item in items: fields = item["fields"] query_result_list.append( Document( page_content=fields[self.config.field_name_mapping["document"]], metadata=create_metadata(fields), ) ) return query_result_list [docs] def create_results_with_score( self, json_result: Dict[str, Any] ) -> List[Tuple[Document, float]]: items = json_result["result"]["items"] query_result_list: List[Tuple[Document, float]] = [] for item in items: fields = item["fields"] query_result_list.append( ( Document( page_content=fields[self.config.field_name_mapping["document"]], metadata=create_metadata(fields), ), float(item["sortExprValues"][0]), ) ) return query_result_list [docs] @classmethod def from_texts( cls, texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, config: Optional[AlibabaCloudOpenSearchSettings] = None, **kwargs: Any, ) -> "AlibabaCloudOpenSearch": if config is None: raise Exception("config can't be none") ctx = cls(embedding, config, **kwargs) ctx.add_texts(texts=texts, metadatas=metadatas) return ctx [docs] @classmethod def from_documents( cls, documents: List[Document], embedding: Embeddings, ids: Optional[List[str]] = None, config: Optional[AlibabaCloudOpenSearchSettings] = None, **kwargs: Any, ) -> "AlibabaCloudOpenSearch": if config is None: raise Exception("config can't be none") texts = [d.page_content for d in documents] metadatas = [d.metadata for d in documents] return cls.from_texts( texts=texts, embedding=embedding, metadatas=metadatas, config=config, **kwargs, )
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/alibabacloud_opensearch.html
3d3563a0-755a-41b7-8c5d-c7617ff1d62f
Source code for langchain.vectorstores.rocksetdb """Wrapper around Rockset vector database.""" from __future__ import annotations import logging from enum import Enum from typing import Any, Iterable, List, Optional, Tuple from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.vectorstores.base import VectorStore logger = logging.getLogger(__name__) [docs]class Rockset(VectorStore): """Wrapper arpund Rockset vector database. To use, you should have the `rockset` python package installed. Note that to use this, the collection being used must already exist in your Rockset instance. You must also ensure you use a Rockset ingest transformation to apply `VECTOR_ENFORCE` on the column being used to store `embedding_key` in the collection. See: https://rockset.com/blog/introducing-vector-search-on-rockset/ for more details Everything below assumes `commons` Rockset workspace. TODO: Add support for workspace args. Example: .. code-block:: python from langchain.vectorstores import Rockset from langchain.embeddings.openai import OpenAIEmbeddings import rockset # Make sure you use the right host (region) for your Rockset instance # and APIKEY has both read-write access to your collection. rs = rockset.RocksetClient(host=rockset.Regions.use1a1, api_key="***") collection_name = "langchain_demo" embeddings = OpenAIEmbeddings() vectorstore = Rockset(rs, collection_name, embeddings, "description", "description_embedding") """ def __init__( self, client: Any, embeddings: Embeddings, collection_name: str, text_key: str, embedding_key: str, ): """Initialize with Rockset client. Args: client: Rockset client object collection: Rockset collection to insert docs / query embeddings: Langchain Embeddings object to use to generate embedding for given text. text_key: column in Rockset collection to use to store the text embedding_key: column in Rockset collection to use to store the embedding. Note: We must apply `VECTOR_ENFORCE()` on this column via Rockset ingest transformation. """ try: from rockset import RocksetClient except ImportError: raise ImportError( "Could not import rockset client python package. " "Please install it with `pip install rockset`." ) if not isinstance(client, RocksetClient): raise ValueError( f"client should be an instance of rockset.RocksetClient, " f"got {type(client)}" ) # TODO: check that `collection_name` exists in rockset. Create if not. self._client = client self._collection_name = collection_name self._embeddings = embeddings self._text_key = text_key self._embedding_key = embedding_key [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, batch_size: int = 32, **kwargs: Any, ) -> List[str]: """Run more texts through the embeddings and add to the vectorstore Args: texts: Iterable of strings to add to the vectorstore. metadatas: Optional list of metadatas associated with the texts. ids: Optional list of ids to associate with the texts. batch_size: Send documents in batches to rockset. Returns: List of ids from adding the texts into the vectorstore. """ batch: list[dict] = [] stored_ids = [] for i, text in enumerate(texts): if len(batch) == batch_size: stored_ids += self._write_documents_to_rockset(batch) batch = [] doc = {} if metadatas and len(metadatas) > i: doc = metadatas[i] if ids and len(ids) > i: doc["_id"] = ids[i] doc[self._text_key] = text doc[self._embedding_key] = self._embeddings.embed_query(text) batch.append(doc) if len(batch) > 0: stored_ids += self._write_documents_to_rockset(batch) batch = [] return stored_ids [docs] @classmethod def from_texts( cls, texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, client: Any = None, collection_name: str = "", text_key: str = "", embedding_key: str = "", ids: Optional[List[str]] = None, batch_size: int = 32, **kwargs: Any, ) -> Rockset: """Create Rockset wrapper with existing texts. This is intended as a quicker way to get started. """ # Sanitize imputs assert client is not None, "Rockset Client cannot be None" assert collection_name, "Collection name cannot be empty" assert text_key, "Text key name cannot be empty" assert embedding_key, "Embedding key cannot be empty" rockset = cls(client, embedding, collection_name, text_key, embedding_key) rockset.add_texts(texts, metadatas, ids, batch_size) return rockset # Rockset supports these vector distance functions. [docs] class DistanceFunction(Enum): COSINE_SIM = "COSINE_SIM" EUCLIDEAN_DIST = "EUCLIDEAN_DIST" DOT_PRODUCT = "DOT_PRODUCT" # how to sort results for "similarity" [docs] def order_by(self) -> str: if self.value == "EUCLIDEAN_DIST": return "ASC" return "DESC" [docs] def similarity_search_with_relevance_scores( self, query: str, k: int = 4, distance_func: DistanceFunction = DistanceFunction.COSINE_SIM, where_str: Optional[str] = None, **kwargs: Any, ) -> List[Tuple[Document, float]]: """Perform a similarity search with Rockset Args: query (str): Text to look up documents similar to. distance_func (DistanceFunction): how to compute distance between two vectors in Rockset. k (int, optional): Top K neighbors to retrieve. Defaults to 4. where_str (Optional[str], optional): Metadata filters supplied as a SQL `where` condition string. Defaults to None. eg. "price<=70.0 AND brand='Nintendo'" NOTE: Please do not let end-user to fill this and always be aware of SQL injection. Returns: List[Tuple[Document, float]]: List of documents with their relevance score """ return self.similarity_search_by_vector_with_relevance_scores( self._embeddings.embed_query(query), k, distance_func, where_str, **kwargs, ) [docs] def similarity_search( self, query: str, k: int = 4, distance_func: DistanceFunction = DistanceFunction.COSINE_SIM, where_str: Optional[str] = None, **kwargs: Any, ) -> List[Document]: """Same as `similarity_search_with_relevance_scores` but doesn't return the scores. """ return self.similarity_search_by_vector( self._embeddings.embed_query(query), k, distance_func, where_str, **kwargs, ) [docs] def similarity_search_by_vector( self, embedding: List[float], k: int = 4, distance_func: DistanceFunction = DistanceFunction.COSINE_SIM, where_str: Optional[str] = None, **kwargs: Any, ) -> List[Document]: """Accepts a query_embedding (vector), and returns documents with similar embeddings.""" docs_and_scores = self.similarity_search_by_vector_with_relevance_scores( embedding, k, distance_func, where_str, **kwargs ) return [doc for doc, _ in docs_and_scores] [docs] def similarity_search_by_vector_with_relevance_scores( self, embedding: List[float], k: int = 4, distance_func: DistanceFunction = DistanceFunction.COSINE_SIM, where_str: Optional[str] = None, **kwargs: Any, ) -> List[Tuple[Document, float]]: """Accepts a query_embedding (vector), and returns documents with similar embeddings along with their relevance scores.""" q_str = self._build_query_sql(embedding, distance_func, k, where_str) try: query_response = self._client.Queries.query(sql={"query": q_str}) except Exception as e: logger.error("Exception when querying Rockset: %s\n", e) return [] finalResult: list[Tuple[Document, float]] = [] for document in query_response.results: metadata = {} assert isinstance( document, dict ), "document should be of type `dict[str,Any]`. But found: `{}`".format( type(document) ) for k, v in document.items(): if k == self._text_key: assert isinstance( v, str ), "page content stored in column `{}` must be of type `str`. \ But found: `{}`".format( self._text_key, type(v) ) page_content = v elif k == "dist": assert isinstance( v, float ), "Computed distance between vectors must of type `float`. \ But found {}".format( type(v) ) score = v elif k not in ["_id", "_event_time", "_meta"]: # These columns are populated by Rockset when documents are # inserted. No need to return them in metadata dict. metadata[k] = v finalResult.append( (Document(page_content=page_content, metadata=metadata), score) ) return finalResult # Helper functions def _build_query_sql( self, query_embedding: List[float], distance_func: DistanceFunction, k: int = 4, where_str: Optional[str] = None, ) -> str: """Builds Rockset SQL query to query similar vectors to query_vector""" q_embedding_str = ",".join(map(str, query_embedding)) distance_str = f"""{distance_func.value}({self._embedding_key}, \ [{q_embedding_str}]) as dist""" where_str = f"WHERE {where_str}\n" if where_str else "" return f"""\ SELECT * EXCEPT({self._embedding_key}), {distance_str} FROM {self._collection_name} {where_str}\ ORDER BY dist {distance_func.order_by()} LIMIT {str(k)} """ def _write_documents_to_rockset(self, batch: List[dict]) -> List[str]: add_doc_res = self._client.Documents.add_documents( collection=self._collection_name, data=batch ) return [doc_status._id for doc_status in add_doc_res.data] [docs] def delete_texts(self, ids: List[str]) -> None: """Delete a list of docs from the Rockset collection""" try: from rockset.models import DeleteDocumentsRequestData except ImportError: raise ImportError( "Could not import rockset client python package. " "Please install it with `pip install rockset`." ) self._client.Documents.delete_documents( collection=self._collection_name, data=[DeleteDocumentsRequestData(id=i) for i in ids], )
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/rocksetdb.html
46fd788a-4565-4e92-80dc-b71052deb56f
Source code for langchain.vectorstores.base """Interface for vector stores.""" from __future__ import annotations import asyncio import warnings from abc import ABC, abstractmethod from functools import partial from typing import ( Any, ClassVar, Collection, Dict, Iterable, List, Optional, Tuple, Type, TypeVar, ) from pydantic import BaseModel, Field, root_validator from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.schema import BaseRetriever VST = TypeVar("VST", bound="VectorStore") [docs]class VectorStore(ABC): """Interface for vector stores.""" [docs] @abstractmethod def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, **kwargs: Any, ) -> List[str]: """Run more texts through the embeddings and add to the vectorstore. Args: texts: Iterable of strings to add to the vectorstore. metadatas: Optional list of metadatas associated with the texts. kwargs: vectorstore specific parameters Returns: List of ids from adding the texts into the vectorstore. """ [docs] def delete(self, ids: List[str]) -> Optional[bool]: """Delete by vector ID. Args: ids: List of ids to delete. Returns: Optional[bool]: True if deletion is successful, False otherwise, None if not implemented. """ raise NotImplementedError( "delete_by_id method must be implemented by subclass." ) [docs] async def aadd_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, **kwargs: Any, ) -> List[str]: """Run more texts through the embeddings and add to the vectorstore.""" raise NotImplementedError [docs] def add_documents(self, documents: List[Document], **kwargs: Any) -> List[str]: """Run more documents through the embeddings and add to the vectorstore. Args: documents (List[Document]: Documents to add to the vectorstore. Returns: List[str]: List of IDs of the added texts. """ # TODO: Handle the case where the user doesn't provide ids on the Collection texts = [doc.page_content for doc in documents] metadatas = [doc.metadata for doc in documents] return self.add_texts(texts, metadatas, **kwargs) [docs] async def aadd_documents( self, documents: List[Document], **kwargs: Any ) -> List[str]: """Run more documents through the embeddings and add to the vectorstore. Args: documents (List[Document]: Documents to add to the vectorstore. Returns: List[str]: List of IDs of the added texts. """ texts = [doc.page_content for doc in documents] metadatas = [doc.metadata for doc in documents] return await self.aadd_texts(texts, metadatas, **kwargs) [docs] def search(self, query: str, search_type: str, **kwargs: Any) -> List[Document]: """Return docs most similar to query using specified search type.""" if search_type == "similarity": return self.similarity_search(query, **kwargs) elif search_type == "mmr": return self.max_marginal_relevance_search(query, **kwargs) else: raise ValueError( f"search_type of {search_type} not allowed. Expected " "search_type to be 'similarity' or 'mmr'." ) [docs] async def asearch( self, query: str, search_type: str, **kwargs: Any ) -> List[Document]: """Return docs most similar to query using specified search type.""" if search_type == "similarity": return await self.asimilarity_search(query, **kwargs) elif search_type == "mmr": return await self.amax_marginal_relevance_search(query, **kwargs) else: raise ValueError( f"search_type of {search_type} not allowed. Expected " "search_type to be 'similarity' or 'mmr'." ) [docs] @abstractmethod def similarity_search( self, query: str, k: int = 4, **kwargs: Any ) -> List[Document]: """Return docs most similar to query.""" [docs] def similarity_search_with_relevance_scores( self, query: str, k: int = 4, **kwargs: Any, ) -> List[Tuple[Document, float]]: """Return docs and relevance scores in the range [0, 1]. 0 is dissimilar, 1 is most similar. Args: query: input text k: Number of Documents to return. Defaults to 4. **kwargs: kwargs to be passed to similarity search. Should include: score_threshold: Optional, a floating point value between 0 to 1 to filter the resulting set of retrieved docs Returns: List of Tuples of (doc, similarity_score) """ docs_and_similarities = self._similarity_search_with_relevance_scores( query, k=k, **kwargs ) if any( similarity < 0.0 or similarity > 1.0 for _, similarity in docs_and_similarities ): warnings.warn( "Relevance scores must be between" f" 0 and 1, got {docs_and_similarities}" ) score_threshold = kwargs.get("score_threshold") if score_threshold is not None: docs_and_similarities = [ (doc, similarity) for doc, similarity in docs_and_similarities if similarity >= score_threshold ] if len(docs_and_similarities) == 0: warnings.warn( "No relevant docs were retrieved using the relevance score" f" threshold {score_threshold}" ) return docs_and_similarities def _similarity_search_with_relevance_scores( self, query: str, k: int = 4, **kwargs: Any, ) -> List[Tuple[Document, float]]: """Return docs and relevance scores, normalized on a scale from 0 to 1. 0 is dissimilar, 1 is most similar. """ raise NotImplementedError [docs] async def asimilarity_search_with_relevance_scores( self, query: str, k: int = 4, **kwargs: Any ) -> List[Tuple[Document, float]]: """Return docs most similar to query.""" # This is a temporary workaround to make the similarity search # asynchronous. The proper solution is to make the similarity search # asynchronous in the vector store implementations. func = partial(self.similarity_search_with_relevance_scores, query, k, **kwargs) return await asyncio.get_event_loop().run_in_executor(None, func) [docs] async def asimilarity_search( self, query: str, k: int = 4, **kwargs: Any ) -> List[Document]: """Return docs most similar to query.""" # This is a temporary workaround to make the similarity search # asynchronous. The proper solution is to make the similarity search # asynchronous in the vector store implementations. func = partial(self.similarity_search, query, k, **kwargs) return await asyncio.get_event_loop().run_in_executor(None, func) [docs] def similarity_search_by_vector( self, embedding: List[float], k: int = 4, **kwargs: Any ) -> List[Document]: """Return docs most similar to embedding vector. Args: embedding: Embedding to look up documents similar to. k: Number of Documents to return. Defaults to 4. Returns: List of Documents most similar to the query vector. """ raise NotImplementedError [docs] async def asimilarity_search_by_vector( self, embedding: List[float], k: int = 4, **kwargs: Any ) -> List[Document]: """Return docs most similar to embedding vector.""" # This is a temporary workaround to make the similarity search # asynchronous. The proper solution is to make the similarity search # asynchronous in the vector store implementations. func = partial(self.similarity_search_by_vector, embedding, k, **kwargs) return await asyncio.get_event_loop().run_in_executor(None, func) [docs] def max_marginal_relevance_search( self, query: str, k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. fetch_k: Number of Documents to fetch to pass to MMR algorithm. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. Returns: List of Documents selected by maximal marginal relevance. """ raise NotImplementedError [docs] async def amax_marginal_relevance_search( self, query: str, k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance.""" # This is a temporary workaround to make the similarity search # asynchronous. The proper solution is to make the similarity search # asynchronous in the vector store implementations. func = partial( self.max_marginal_relevance_search, query, k, fetch_k, lambda_mult, **kwargs ) return await asyncio.get_event_loop().run_in_executor(None, func) [docs] def max_marginal_relevance_search_by_vector( self, embedding: List[float], k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: embedding: Embedding to look up documents similar to. k: Number of Documents to return. Defaults to 4. fetch_k: Number of Documents to fetch to pass to MMR algorithm. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. Returns: List of Documents selected by maximal marginal relevance. """ raise NotImplementedError [docs] async def amax_marginal_relevance_search_by_vector( self, embedding: List[float], k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance.""" raise NotImplementedError [docs] @classmethod def from_documents( cls: Type[VST], documents: List[Document], embedding: Embeddings, **kwargs: Any, ) -> VST: """Return VectorStore initialized from documents and embeddings.""" texts = [d.page_content for d in documents] metadatas = [d.metadata for d in documents] return cls.from_texts(texts, embedding, metadatas=metadatas, **kwargs) [docs] @classmethod async def afrom_documents( cls: Type[VST], documents: List[Document], embedding: Embeddings, **kwargs: Any, ) -> VST: """Return VectorStore initialized from documents and embeddings.""" texts = [d.page_content for d in documents] metadatas = [d.metadata for d in documents] return await cls.afrom_texts(texts, embedding, metadatas=metadatas, **kwargs) [docs] @classmethod @abstractmethod def from_texts( cls: Type[VST], texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, **kwargs: Any, ) -> VST: """Return VectorStore initialized from texts and embeddings.""" [docs] @classmethod async def afrom_texts( cls: Type[VST], texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, **kwargs: Any, ) -> VST: """Return VectorStore initialized from texts and embeddings.""" raise NotImplementedError [docs] def as_retriever(self, **kwargs: Any) -> VectorStoreRetriever: return VectorStoreRetriever(vectorstore=self, **kwargs) class VectorStoreRetriever(BaseRetriever, BaseModel): vectorstore: VectorStore search_type: str = "similarity" search_kwargs: dict = Field(default_factory=dict) allowed_search_types: ClassVar[Collection[str]] = ( "similarity", "similarity_score_threshold", "mmr", ) class Config: """Configuration for this pydantic object.""" arbitrary_types_allowed = True @root_validator() def validate_search_type(cls, values: Dict) -> Dict: """Validate search type.""" search_type = values["search_type"] if search_type not in cls.allowed_search_types: raise ValueError( f"search_type of {search_type} not allowed. Valid values are: " f"{cls.allowed_search_types}" ) if search_type == "similarity_score_threshold": score_threshold = values["search_kwargs"].get("score_threshold") if (score_threshold is None) or (not isinstance(score_threshold, float)): raise ValueError( "`score_threshold` is not specified with a float value(0~1) " "in `search_kwargs`." ) return values def get_relevant_documents(self, query: str) -> List[Document]: if self.search_type == "similarity": docs = self.vectorstore.similarity_search(query, **self.search_kwargs) elif self.search_type == "similarity_score_threshold": docs_and_similarities = ( self.vectorstore.similarity_search_with_relevance_scores( query, **self.search_kwargs ) ) docs = [doc for doc, _ in docs_and_similarities] elif self.search_type == "mmr": docs = self.vectorstore.max_marginal_relevance_search( query, **self.search_kwargs ) else: raise ValueError(f"search_type of {self.search_type} not allowed.") return docs async def aget_relevant_documents(self, query: str) -> List[Document]: if self.search_type == "similarity": docs = await self.vectorstore.asimilarity_search( query, **self.search_kwargs ) elif self.search_type == "similarity_score_threshold": docs_and_similarities = ( await self.vectorstore.asimilarity_search_with_relevance_scores( query, **self.search_kwargs ) ) docs = [doc for doc, _ in docs_and_similarities] elif self.search_type == "mmr": docs = await self.vectorstore.amax_marginal_relevance_search( query, **self.search_kwargs ) else: raise ValueError(f"search_type of {self.search_type} not allowed.") return docs def add_documents(self, documents: List[Document], **kwargs: Any) -> List[str]: """Add documents to vectorstore.""" return self.vectorstore.add_documents(documents, **kwargs) async def aadd_documents( self, documents: List[Document], **kwargs: Any ) -> List[str]: """Add documents to vectorstore.""" return await self.vectorstore.aadd_documents(documents, **kwargs)
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/base.html
e854f09b-19f2-4b3b-9c98-43e5a46926b0
Source code for langchain.vectorstores.awadb """Wrapper around AwaDB for embedding vectors""" from __future__ import annotations import logging import uuid from typing import TYPE_CHECKING, Any, Iterable, List, Optional, Tuple, Type from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.vectorstores.base import VectorStore # from pydantic import BaseModel, Field, root_validator if TYPE_CHECKING: import awadb logger = logging.getLogger() DEFAULT_TOPN = 4 [docs]class AwaDB(VectorStore): """Interface implemented by AwaDB vector stores.""" _DEFAULT_TABLE_NAME = "langchain_awadb" def __init__( self, table_name: str = _DEFAULT_TABLE_NAME, embedding_model: Optional[Embeddings] = None, log_and_data_dir: Optional[str] = None, client: Optional[awadb.Client] = None, ) -> None: """Initialize with AwaDB client.""" try: import awadb except ImportError: raise ValueError( "Could not import awadb python package. " "Please install it with `pip install awadb`." ) if client is not None: self.awadb_client = client else: if log_and_data_dir is not None: self.awadb_client = awadb.Client(log_and_data_dir) else: self.awadb_client = awadb.Client() if table_name == self._DEFAULT_TABLE_NAME: table_name += "_" table_name += str(uuid.uuid4()).split("-")[-1] self.awadb_client.Create(table_name) self.table2embeddings: dict[str, Embeddings] = {} if embedding_model is not None: self.table2embeddings[table_name] = embedding_model self.using_table_name = table_name [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, is_duplicate_texts: Optional[bool] = None, **kwargs: Any, ) -> List[str]: """Run more texts through the embeddings and add to the vectorstore. Args: texts: Iterable of strings to add to the vectorstore. metadatas: Optional list of metadatas associated with the texts. is_duplicate_texts: Optional whether to duplicate texts. kwargs: vectorstore specific parameters. Returns: List of ids from adding the texts into the vectorstore. """ if self.awadb_client is None: raise ValueError("AwaDB client is None!!!") embeddings = None if self.using_table_name in self.table2embeddings: embeddings = self.table2embeddings[self.using_table_name].embed_documents( list(texts) ) return self.awadb_client.AddTexts( "embedding_text", "text_embedding", texts, embeddings, metadatas, is_duplicate_texts, ) [docs] def load_local( self, table_name: str, **kwargs: Any, ) -> bool: if self.awadb_client is None: raise ValueError("AwaDB client is None!!!") return self.awadb_client.Load(table_name) [docs] def similarity_search( self, query: str, k: int = DEFAULT_TOPN, **kwargs: Any, ) -> List[Document]: """Return docs most similar to query.""" if self.awadb_client is None: raise ValueError("AwaDB client is None!!!") embedding = None if self.using_table_name in self.table2embeddings: embedding = self.table2embeddings[self.using_table_name].embed_query(query) else: from awadb import llm_embedding llm = llm_embedding.LLMEmbedding() embedding = llm.Embedding(query) return self.similarity_search_by_vector(embedding, k) [docs] def similarity_search_with_score( self, query: str, k: int = DEFAULT_TOPN, **kwargs: Any, ) -> List[Tuple[Document, float]]: """Return docs and relevance scores, normalized on a scale from 0 to 1. 0 is dissimilar, 1 is most similar. """ if self.awadb_client is None: raise ValueError("AwaDB client is None!!!") embedding = None if self.using_table_name in self.table2embeddings: embedding = self.table2embeddings[self.using_table_name].embed_query(query) else: from awadb import llm_embedding llm = llm_embedding.LLMEmbedding() embedding = llm.Embedding(query) results: List[Tuple[Document, float]] = [] scores: List[float] = [] retrieval_docs = self.similarity_search_by_vector(embedding, k, scores) L2_Norm = 0.0 for score in scores: L2_Norm = L2_Norm + score * score L2_Norm = pow(L2_Norm, 0.5) doc_no = 0 for doc in retrieval_docs: doc_tuple = (doc, 1 - (scores[doc_no] / L2_Norm)) results.append(doc_tuple) doc_no = doc_no + 1 return results [docs] def similarity_search_with_relevance_scores( self, query: str, k: int = DEFAULT_TOPN, **kwargs: Any, ) -> List[Tuple[Document, float]]: """Return docs and relevance scores, normalized on a scale from 0 to 1. 0 is dissimilar, 1 is most similar. """ if self.awadb_client is None: raise ValueError("AwaDB client is None!!!") embedding = None if self.using_table_name in self.table2embeddings: embedding = self.table2embeddings[self.using_table_name].embed_query(query) show_results = self.awadb_client.Search(embedding, k) results: List[Tuple[Document, float]] = [] if show_results.__len__() == 0: return results scores: List[float] = [] retrieval_docs = self.similarity_search_by_vector(embedding, k, scores) L2_Norm = 0.0 for score in scores: L2_Norm = L2_Norm + score * score L2_Norm = pow(L2_Norm, 0.5) doc_no = 0 for doc in retrieval_docs: doc_tuple = (doc, 1 - scores[doc_no] / L2_Norm) results.append(doc_tuple) doc_no = doc_no + 1 return results [docs] def similarity_search_by_vector( self, embedding: Optional[List[float]] = None, k: int = DEFAULT_TOPN, scores: Optional[list] = None, **kwargs: Any, ) -> List[Document]: """Return docs most similar to embedding vector. Args: embedding: Embedding to look up documents similar to. k: Number of Documents to return. Defaults to 4. Returns: List of Documents most similar to the query vector. """ if self.awadb_client is None: raise ValueError("AwaDB client is None!!!") results: List[Document] = [] if embedding is None: return results show_results = self.awadb_client.Search(embedding, k) if show_results.__len__() == 0: return results for item_detail in show_results[0]["ResultItems"]: content = "" meta_data = {} for item_key in item_detail: if ( item_key == "Field@0" and self.using_table_name in self.table2embeddings ): # text for the document content = item_detail[item_key] elif item_key == "embedding_text": content = item_detail[item_key] elif ( item_key == "Field@1" or item_key == "text_embedding" ): # embedding field for the document continue elif item_key == "score": # L2 distance if scores is not None: score = item_detail[item_key] scores.append(score) else: meta_data[item_key] = item_detail[item_key] results.append(Document(page_content=content, metadata=meta_data)) return results [docs] def create_table( self, table_name: str, **kwargs: Any, ) -> bool: """Create a new table.""" if self.awadb_client is None: return False ret = self.awadb_client.Create(table_name) if ret: self.using_table_name = table_name return ret [docs] def use( self, table_name: str, **kwargs: Any, ) -> bool: """Use the specified table. Don't know the tables, please invoke list_tables.""" if self.awadb_client is None: return False ret = self.awadb_client.Use(table_name) if ret: self.using_table_name = table_name return ret [docs] def list_tables( self, **kwargs: Any, ) -> List[str]: """List all the tables created by the client.""" if self.awadb_client is None: return [] return self.awadb_client.ListAllTables() [docs] def get_current_table( self, **kwargs: Any, ) -> str: """Get the current table.""" return self.using_table_name [docs] @classmethod def from_texts( cls: Type[AwaDB], texts: List[str], embedding: Optional[Embeddings] = None, metadatas: Optional[List[dict]] = None, table_name: str = _DEFAULT_TABLE_NAME, logging_and_data_dir: Optional[str] = None, client: Optional[awadb.Client] = None, **kwargs: Any, ) -> AwaDB: """Create an AwaDB vectorstore from a raw documents. Args: texts (List[str]): List of texts to add to the table. embedding (Optional[Embeddings]): Embedding function. Defaults to None. metadatas (Optional[List[dict]]): List of metadatas. Defaults to None. table_name (str): Name of the table to create. logging_and_data_dir (Optional[str]): Directory of logging and persistence. client (Optional[awadb.Client]): AwaDB client Returns: AwaDB: AwaDB vectorstore. """ awadb_client = cls( table_name=table_name, embedding_model=embedding, log_and_data_dir=logging_and_data_dir, client=client, ) awadb_client.add_texts(texts=texts, metadatas=metadatas) return awadb_client [docs] @classmethod def from_documents( cls: Type[AwaDB], documents: List[Document], embedding: Optional[Embeddings] = None, table_name: str = _DEFAULT_TABLE_NAME, logging_and_data_dir: Optional[str] = None, client: Optional[awadb.Client] = None, **kwargs: Any, ) -> AwaDB: """Create an AwaDB vectorstore from a list of documents. If a logging_and_data_dir specified, the table will be persisted there. Args: documents (List[Document]): List of documents to add to the vectorstore. embedding (Optional[Embeddings]): Embedding function. Defaults to None. table_name (str): Name of the table to create. logging_and_data_dir (Optional[str]): Directory to persist the table. client (Optional[awadb.Client]): AwaDB client Returns: AwaDB: AwaDB vectorstore. """ texts = [doc.page_content for doc in documents] metadatas = [doc.metadata for doc in documents] return cls.from_texts( texts=texts, embedding=embedding, metadatas=metadatas, table_name=table_name, logging_and_data_dir=logging_and_data_dir, client=client, )
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/awadb.html
c3b8d61c-2db6-4497-9c84-32f596ffd713
Source code for langchain.vectorstores.milvus """Wrapper around the Milvus vector database.""" from __future__ import annotations import logging from typing import Any, Iterable, List, Optional, Tuple, Union from uuid import uuid4 import numpy as np from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.vectorstores.base import VectorStore from langchain.vectorstores.utils import maximal_marginal_relevance logger = logging.getLogger(__name__) DEFAULT_MILVUS_CONNECTION = { "host": "localhost", "port": "19530", "user": "", "password": "", "secure": False, } [docs]class Milvus(VectorStore): """Wrapper around the Milvus vector database.""" def __init__( self, embedding_function: Embeddings, collection_name: str = "LangChainCollection", connection_args: Optional[dict[str, Any]] = None, consistency_level: str = "Session", index_params: Optional[dict] = None, search_params: Optional[dict] = None, drop_old: Optional[bool] = False, ): """Initialize wrapper around the milvus vector database. In order to use this you need to have `pymilvus` installed and a running Milvus/Zilliz Cloud instance. See the following documentation for how to run a Milvus instance: https://milvus.io/docs/install_standalone-docker.md If looking for a hosted Milvus, take a looka this documentation: https://zilliz.com/cloud IF USING L2/IP metric IT IS HIGHLY SUGGESTED TO NORMALIZE YOUR DATA. The connection args used for this class comes in the form of a dict, here are a few of the options: address (str): The actual address of Milvus instance. Example address: "localhost:19530" uri (str): The uri of Milvus instance. Example uri: "http://randomwebsite:19530", "tcp:foobarsite:19530", "https://ok.s3.south.com:19530". host (str): The host of Milvus instance. Default at "localhost", PyMilvus will fill in the default host if only port is provided. port (str/int): The port of Milvus instance. Default at 19530, PyMilvus will fill in the default port if only host is provided. user (str): Use which user to connect to Milvus instance. If user and password are provided, we will add related header in every RPC call. password (str): Required when user is provided. The password corresponding to the user. secure (bool): Default is false. If set to true, tls will be enabled. client_key_path (str): If use tls two-way authentication, need to write the client.key path. client_pem_path (str): If use tls two-way authentication, need to write the client.pem path. ca_pem_path (str): If use tls two-way authentication, need to write the ca.pem path. server_pem_path (str): If use tls one-way authentication, need to write the server.pem path. server_name (str): If use tls, need to write the common name. Args: embedding_function (Embeddings): Function used to embed the text. collection_name (str): Which Milvus collection to use. Defaults to "LangChainCollection". connection_args (Optional[dict[str, any]]): The arguments for connection to Milvus/Zilliz instance. Defaults to DEFAULT_MILVUS_CONNECTION. consistency_level (str): The consistency level to use for a collection. Defaults to "Session". index_params (Optional[dict]): Which index params to use. Defaults to HNSW/AUTOINDEX depending on service. search_params (Optional[dict]): Which search params to use. Defaults to default of index. drop_old (Optional[bool]): Whether to drop the current collection. Defaults to False. """ try: from pymilvus import Collection, utility except ImportError: raise ValueError( "Could not import pymilvus python package. " "Please install it with `pip install pymilvus`." ) # Default search params when one is not provided. self.default_search_params = { "IVF_FLAT": {"metric_type": "L2", "params": {"nprobe": 10}}, "IVF_SQ8": {"metric_type": "L2", "params": {"nprobe": 10}}, "IVF_PQ": {"metric_type": "L2", "params": {"nprobe": 10}}, "HNSW": {"metric_type": "L2", "params": {"ef": 10}}, "RHNSW_FLAT": {"metric_type": "L2", "params": {"ef": 10}}, "RHNSW_SQ": {"metric_type": "L2", "params": {"ef": 10}}, "RHNSW_PQ": {"metric_type": "L2", "params": {"ef": 10}}, "IVF_HNSW": {"metric_type": "L2", "params": {"nprobe": 10, "ef": 10}}, "ANNOY": {"metric_type": "L2", "params": {"search_k": 10}}, "AUTOINDEX": {"metric_type": "L2", "params": {}}, } self.embedding_func = embedding_function self.collection_name = collection_name self.index_params = index_params self.search_params = search_params self.consistency_level = consistency_level # In order for a collection to be compatible, pk needs to be auto'id and int self._primary_field = "pk" # In order for compatiblility, the text field will need to be called "text" self._text_field = "text" # In order for compatbility, the vector field needs to be called "vector" self._vector_field = "vector" self.fields: list[str] = [] # Create the connection to the server if connection_args is None: connection_args = DEFAULT_MILVUS_CONNECTION self.alias = self._create_connection_alias(connection_args) self.col: Optional[Collection] = None # Grab the existing colection if it exists if utility.has_collection(self.collection_name, using=self.alias): self.col = Collection( self.collection_name, using=self.alias, ) # If need to drop old, drop it if drop_old and isinstance(self.col, Collection): self.col.drop() self.col = None # Initialize the vector store self._init() def _create_connection_alias(self, connection_args: dict) -> str: """Create the connection to the Milvus server.""" from pymilvus import MilvusException, connections # Grab the connection arguments that are used for checking existing connection host: str = connection_args.get("host", None) port: Union[str, int] = connection_args.get("port", None) address: str = connection_args.get("address", None) uri: str = connection_args.get("uri", None) user = connection_args.get("user", None) # Order of use is host/port, uri, address if host is not None and port is not None: given_address = str(host) + ":" + str(port) elif uri is not None: given_address = uri.split("https://")[1] elif address is not None: given_address = address else: given_address = None logger.debug("Missing standard address type for reuse atttempt") # User defaults to empty string when getting connection info if user is not None: tmp_user = user else: tmp_user = "" # If a valid address was given, then check if a connection exists if given_address is not None: for con in connections.list_connections(): addr = connections.get_connection_addr(con[0]) if ( con[1] and ("address" in addr) and (addr["address"] == given_address) and ("user" in addr) and (addr["user"] == tmp_user) ): logger.debug("Using previous connection: %s", con[0]) return con[0] # Generate a new connection if one doesnt exist alias = uuid4().hex try: connections.connect(alias=alias, **connection_args) logger.debug("Created new connection using: %s", alias) return alias except MilvusException as e: logger.error("Failed to create new connection using: %s", alias) raise e def _init( self, embeddings: Optional[list] = None, metadatas: Optional[list[dict]] = None ) -> None: if embeddings is not None: self._create_collection(embeddings, metadatas) self._extract_fields() self._create_index() self._create_search_params() self._load() def _create_collection( self, embeddings: list, metadatas: Optional[list[dict]] = None ) -> None: from pymilvus import ( Collection, CollectionSchema, DataType, FieldSchema, MilvusException, ) from pymilvus.orm.types import infer_dtype_bydata # Determine embedding dim dim = len(embeddings[0]) fields = [] # Determine metadata schema if metadatas: # Create FieldSchema for each entry in metadata. for key, value in metadatas[0].items(): # Infer the corresponding datatype of the metadata dtype = infer_dtype_bydata(value) # Datatype isnt compatible if dtype == DataType.UNKNOWN or dtype == DataType.NONE: logger.error( "Failure to create collection, unrecognized dtype for key: %s", key, ) raise ValueError(f"Unrecognized datatype for {key}.") # Dataype is a string/varchar equivalent elif dtype == DataType.VARCHAR: fields.append(FieldSchema(key, DataType.VARCHAR, max_length=65_535)) else: fields.append(FieldSchema(key, dtype)) # Create the text field fields.append( FieldSchema(self._text_field, DataType.VARCHAR, max_length=65_535) ) # Create the primary key field fields.append( FieldSchema( self._primary_field, DataType.INT64, is_primary=True, auto_id=True ) ) # Create the vector field, supports binary or float vectors fields.append( FieldSchema(self._vector_field, infer_dtype_bydata(embeddings[0]), dim=dim) ) # Create the schema for the collection schema = CollectionSchema(fields) # Create the collection try: self.col = Collection( name=self.collection_name, schema=schema, consistency_level=self.consistency_level, using=self.alias, ) except MilvusException as e: logger.error( "Failed to create collection: %s error: %s", self.collection_name, e ) raise e def _extract_fields(self) -> None: """Grab the existing fields from the Collection""" from pymilvus import Collection if isinstance(self.col, Collection): schema = self.col.schema for x in schema.fields: self.fields.append(x.name) # Since primary field is auto-id, no need to track it self.fields.remove(self._primary_field) def _get_index(self) -> Optional[dict[str, Any]]: """Return the vector index information if it exists""" from pymilvus import Collection if isinstance(self.col, Collection): for x in self.col.indexes: if x.field_name == self._vector_field: return x.to_dict() return None def _create_index(self) -> None: """Create a index on the collection""" from pymilvus import Collection, MilvusException if isinstance(self.col, Collection) and self._get_index() is None: try: # If no index params, use a default HNSW based one if self.index_params is None: self.index_params = { "metric_type": "L2", "index_type": "HNSW", "params": {"M": 8, "efConstruction": 64}, } try: self.col.create_index( self._vector_field, index_params=self.index_params, using=self.alias, ) # If default did not work, most likely on Zilliz Cloud except MilvusException: # Use AUTOINDEX based index self.index_params = { "metric_type": "L2", "index_type": "AUTOINDEX", "params": {}, } self.col.create_index( self._vector_field, index_params=self.index_params, using=self.alias, ) logger.debug( "Successfully created an index on collection: %s", self.collection_name, ) except MilvusException as e: logger.error( "Failed to create an index on collection: %s", self.collection_name ) raise e def _create_search_params(self) -> None: """Generate search params based on the current index type""" from pymilvus import Collection if isinstance(self.col, Collection) and self.search_params is None: index = self._get_index() if index is not None: index_type: str = index["index_param"]["index_type"] metric_type: str = index["index_param"]["metric_type"] self.search_params = self.default_search_params[index_type] self.search_params["metric_type"] = metric_type def _load(self) -> None: """Load the collection if available.""" from pymilvus import Collection if isinstance(self.col, Collection) and self._get_index() is not None: self.col.load() [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, timeout: Optional[int] = None, batch_size: int = 1000, **kwargs: Any, ) -> List[str]: """Insert text data into Milvus. Inserting data when the collection has not be made yet will result in creating a new Collection. The data of the first entity decides the schema of the new collection, the dim is extracted from the first embedding and the columns are decided by the first metadata dict. Metada keys will need to be present for all inserted values. At the moment there is no None equivalent in Milvus. Args: texts (Iterable[str]): The texts to embed, it is assumed that they all fit in memory. metadatas (Optional[List[dict]]): Metadata dicts attached to each of the texts. Defaults to None. timeout (Optional[int]): Timeout for each batch insert. Defaults to None. batch_size (int, optional): Batch size to use for insertion. Defaults to 1000. Raises: MilvusException: Failure to add texts Returns: List[str]: The resulting keys for each inserted element. """ from pymilvus import Collection, MilvusException texts = list(texts) try: embeddings = self.embedding_func.embed_documents(texts) except NotImplementedError: embeddings = [self.embedding_func.embed_query(x) for x in texts] if len(embeddings) == 0: logger.debug("Nothing to insert, skipping.") return [] # If the collection hasnt been initialized yet, perform all steps to do so if not isinstance(self.col, Collection): self._init(embeddings, metadatas) # Dict to hold all insert columns insert_dict: dict[str, list] = { self._text_field: texts, self._vector_field: embeddings, } # Collect the metadata into the insert dict. if metadatas is not None: for d in metadatas: for key, value in d.items(): if key in self.fields: insert_dict.setdefault(key, []).append(value) # Total insert count vectors: list = insert_dict[self._vector_field] total_count = len(vectors) pks: list[str] = [] assert isinstance(self.col, Collection) for i in range(0, total_count, batch_size): # Grab end index end = min(i + batch_size, total_count) # Convert dict to list of lists batch for insertion insert_list = [insert_dict[x][i:end] for x in self.fields] # Insert into the collection. try: res: Collection res = self.col.insert(insert_list, timeout=timeout, **kwargs) pks.extend(res.primary_keys) except MilvusException as e: logger.error( "Failed to insert batch starting at entity: %s/%s", i, total_count ) raise e return pks [docs] def similarity_search( self, query: str, k: int = 4, param: Optional[dict] = None, expr: Optional[str] = None, timeout: Optional[int] = None, **kwargs: Any, ) -> List[Document]: """Perform a similarity search against the query string. Args: query (str): The text to search. k (int, optional): How many results to return. Defaults to 4. param (dict, optional): The search params for the index type. Defaults to None. expr (str, optional): Filtering expression. Defaults to None. timeout (int, optional): How long to wait before timeout error. Defaults to None. kwargs: Collection.search() keyword arguments. Returns: List[Document]: Document results for search. """ if self.col is None: logger.debug("No existing collection to search.") return [] res = self.similarity_search_with_score( query=query, k=k, param=param, expr=expr, timeout=timeout, **kwargs ) return [doc for doc, _ in res] [docs] def similarity_search_by_vector( self, embedding: List[float], k: int = 4, param: Optional[dict] = None, expr: Optional[str] = None, timeout: Optional[int] = None, **kwargs: Any, ) -> List[Document]: """Perform a similarity search against the query string. Args: embedding (List[float]): The embedding vector to search. k (int, optional): How many results to return. Defaults to 4. param (dict, optional): The search params for the index type. Defaults to None. expr (str, optional): Filtering expression. Defaults to None. timeout (int, optional): How long to wait before timeout error. Defaults to None. kwargs: Collection.search() keyword arguments. Returns: List[Document]: Document results for search. """ if self.col is None: logger.debug("No existing collection to search.") return [] res = self.similarity_search_with_score_by_vector( embedding=embedding, k=k, param=param, expr=expr, timeout=timeout, **kwargs ) return [doc for doc, _ in res] [docs] def similarity_search_with_score( self, query: str, k: int = 4, param: Optional[dict] = None, expr: Optional[str] = None, timeout: Optional[int] = None, **kwargs: Any, ) -> List[Tuple[Document, float]]: """Perform a search on a query string and return results with score. For more information about the search parameters, take a look at the pymilvus documentation found here: https://milvus.io/api-reference/pymilvus/v2.2.6/Collection/search().md Args: query (str): The text being searched. k (int, optional): The amount of results ot return. Defaults to 4. param (dict): The search params for the specified index. Defaults to None. expr (str, optional): Filtering expression. Defaults to None. timeout (int, optional): How long to wait before timeout error. Defaults to None. kwargs: Collection.search() keyword arguments. Returns: List[float], List[Tuple[Document, any, any]]: """ if self.col is None: logger.debug("No existing collection to search.") return [] # Embed the query text. embedding = self.embedding_func.embed_query(query) res = self.similarity_search_with_score_by_vector( embedding=embedding, k=k, param=param, expr=expr, timeout=timeout, **kwargs ) return res [docs] def similarity_search_with_score_by_vector( self, embedding: List[float], k: int = 4, param: Optional[dict] = None, expr: Optional[str] = None, timeout: Optional[int] = None, **kwargs: Any, ) -> List[Tuple[Document, float]]: """Perform a search on a query string and return results with score. For more information about the search parameters, take a look at the pymilvus documentation found here: https://milvus.io/api-reference/pymilvus/v2.2.6/Collection/search().md Args: embedding (List[float]): The embedding vector being searched. k (int, optional): The amount of results ot return. Defaults to 4. param (dict): The search params for the specified index. Defaults to None. expr (str, optional): Filtering expression. Defaults to None. timeout (int, optional): How long to wait before timeout error. Defaults to None. kwargs: Collection.search() keyword arguments. Returns: List[Tuple[Document, float]]: Result doc and score. """ if self.col is None: logger.debug("No existing collection to search.") return [] if param is None: param = self.search_params # Determine result metadata fields. output_fields = self.fields[:] output_fields.remove(self._vector_field) # Perform the search. res = self.col.search( data=[embedding], anns_field=self._vector_field, param=param, limit=k, expr=expr, output_fields=output_fields, timeout=timeout, **kwargs, ) # Organize results. ret = [] for result in res[0]: meta = {x: result.entity.get(x) for x in output_fields} doc = Document(page_content=meta.pop(self._text_field), metadata=meta) pair = (doc, result.score) ret.append(pair) return ret [docs] def max_marginal_relevance_search( self, query: str, k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, param: Optional[dict] = None, expr: Optional[str] = None, timeout: Optional[int] = None, **kwargs: Any, ) -> List[Document]: """Perform a search and return results that are reordered by MMR. Args: query (str): The text being searched. k (int, optional): How many results to give. Defaults to 4. fetch_k (int, optional): Total results to select k from. Defaults to 20. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5 param (dict, optional): The search params for the specified index. Defaults to None. expr (str, optional): Filtering expression. Defaults to None. timeout (int, optional): How long to wait before timeout error. Defaults to None. kwargs: Collection.search() keyword arguments. Returns: List[Document]: Document results for search. """ if self.col is None: logger.debug("No existing collection to search.") return [] embedding = self.embedding_func.embed_query(query) return self.max_marginal_relevance_search_by_vector( embedding=embedding, k=k, fetch_k=fetch_k, lambda_mult=lambda_mult, param=param, expr=expr, timeout=timeout, **kwargs, ) [docs] def max_marginal_relevance_search_by_vector( self, embedding: list[float], k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, param: Optional[dict] = None, expr: Optional[str] = None, timeout: Optional[int] = None, **kwargs: Any, ) -> List[Document]: """Perform a search and return results that are reordered by MMR. Args: embedding (str): The embedding vector being searched. k (int, optional): How many results to give. Defaults to 4. fetch_k (int, optional): Total results to select k from. Defaults to 20. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5 param (dict, optional): The search params for the specified index. Defaults to None. expr (str, optional): Filtering expression. Defaults to None. timeout (int, optional): How long to wait before timeout error. Defaults to None. kwargs: Collection.search() keyword arguments. Returns: List[Document]: Document results for search. """ if self.col is None: logger.debug("No existing collection to search.") return [] if param is None: param = self.search_params # Determine result metadata fields. output_fields = self.fields[:] output_fields.remove(self._vector_field) # Perform the search. res = self.col.search( data=[embedding], anns_field=self._vector_field, param=param, limit=fetch_k, expr=expr, output_fields=output_fields, timeout=timeout, **kwargs, ) # Organize results. ids = [] documents = [] scores = [] for result in res[0]: meta = {x: result.entity.get(x) for x in output_fields} doc = Document(page_content=meta.pop(self._text_field), metadata=meta) documents.append(doc) scores.append(result.score) ids.append(result.id) vectors = self.col.query( expr=f"{self._primary_field} in {ids}", output_fields=[self._primary_field, self._vector_field], timeout=timeout, ) # Reorganize the results from query to match search order. vectors = {x[self._primary_field]: x[self._vector_field] for x in vectors} ordered_result_embeddings = [vectors[x] for x in ids] # Get the new order of results. new_ordering = maximal_marginal_relevance( np.array(embedding), ordered_result_embeddings, k=k, lambda_mult=lambda_mult ) # Reorder the values and return. ret = [] for x in new_ordering: # Function can return -1 index if x == -1: break else: ret.append(documents[x]) return ret [docs] @classmethod def from_texts( cls, texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, collection_name: str = "LangChainCollection", connection_args: dict[str, Any] = DEFAULT_MILVUS_CONNECTION, consistency_level: str = "Session", index_params: Optional[dict] = None, search_params: Optional[dict] = None, drop_old: bool = False, **kwargs: Any, ) -> Milvus: """Create a Milvus collection, indexes it with HNSW, and insert data. Args: texts (List[str]): Text data. embedding (Embeddings): Embedding function. metadatas (Optional[List[dict]]): Metadata for each text if it exists. Defaults to None. collection_name (str, optional): Collection name to use. Defaults to "LangChainCollection". connection_args (dict[str, Any], optional): Connection args to use. Defaults to DEFAULT_MILVUS_CONNECTION. consistency_level (str, optional): Which consistency level to use. Defaults to "Session". index_params (Optional[dict], optional): Which index_params to use. Defaults to None. search_params (Optional[dict], optional): Which search params to use. Defaults to None. drop_old (Optional[bool], optional): Whether to drop the collection with that name if it exists. Defaults to False. Returns: Milvus: Milvus Vector Store """ vector_db = cls( embedding_function=embedding, collection_name=collection_name, connection_args=connection_args, consistency_level=consistency_level, index_params=index_params, search_params=search_params, drop_old=drop_old, **kwargs, ) vector_db.add_texts(texts=texts, metadatas=metadatas) return vector_db
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/milvus.html
8fb7cb62-979e-4302-9a82-5bbcf885a61a
Source code for langchain.vectorstores.elastic_vector_search """Wrapper around Elasticsearch vector database.""" from __future__ import annotations import uuid from abc import ABC from typing import ( TYPE_CHECKING, Any, Dict, Iterable, List, Mapping, Optional, Tuple, Union, ) from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.utils import get_from_env from langchain.vectorstores.base import VectorStore if TYPE_CHECKING: from elasticsearch import Elasticsearch def _default_text_mapping(dim: int) -> Dict: return { "properties": { "text": {"type": "text"}, "vector": {"type": "dense_vector", "dims": dim}, } } def _default_script_query(query_vector: List[float], filter: Optional[dict]) -> Dict: if filter: ((key, value),) = filter.items() filter = {"match": {f"metadata.{key}.keyword": f"{value}"}} else: filter = {"match_all": {}} return { "script_score": { "query": filter, "script": { "source": "cosineSimilarity(params.query_vector, 'vector') + 1.0", "params": {"query_vector": query_vector}, }, } } # ElasticVectorSearch is a concrete implementation of the abstract base class # VectorStore, which defines a common interface for all vector database # implementations. By inheriting from the ABC class, ElasticVectorSearch can be # defined as an abstract base class itself, allowing the creation of subclasses with # their own specific implementations. If you plan to subclass ElasticVectorSearch, # you can inherit from it and define your own implementation of the necessary methods # and attributes. [docs]class ElasticVectorSearch(VectorStore, ABC): """Wrapper around Elasticsearch as a vector database. To connect to an Elasticsearch instance that does not require login credentials, pass the Elasticsearch URL and index name along with the embedding object to the constructor. Example: .. code-block:: python from langchain import ElasticVectorSearch from langchain.embeddings import OpenAIEmbeddings embedding = OpenAIEmbeddings() elastic_vector_search = ElasticVectorSearch( elasticsearch_url="http://localhost:9200", index_name="test_index", embedding=embedding ) To connect to an Elasticsearch instance that requires login credentials, including Elastic Cloud, use the Elasticsearch URL format https://username:password@es_host:9243. For example, to connect to Elastic Cloud, create the Elasticsearch URL with the required authentication details and pass it to the ElasticVectorSearch constructor as the named parameter elasticsearch_url. You can obtain your Elastic Cloud URL and login credentials by logging in to the Elastic Cloud console at https://cloud.elastic.co, selecting your deployment, and navigating to the "Deployments" page. To obtain your Elastic Cloud password for the default "elastic" user: 1. Log in to the Elastic Cloud console at https://cloud.elastic.co 2. Go to "Security" > "Users" 3. Locate the "elastic" user and click "Edit" 4. Click "Reset password" 5. Follow the prompts to reset the password The format for Elastic Cloud URLs is https://username:password@cluster_id.region_id.gcp.cloud.es.io:9243. Example: .. code-block:: python from langchain import ElasticVectorSearch from langchain.embeddings import OpenAIEmbeddings embedding = OpenAIEmbeddings() elastic_host = "cluster_id.region_id.gcp.cloud.es.io" elasticsearch_url = f"https://username:password@{elastic_host}:9243" elastic_vector_search = ElasticVectorSearch( elasticsearch_url=elasticsearch_url, index_name="test_index", embedding=embedding ) Args: elasticsearch_url (str): The URL for the Elasticsearch instance. index_name (str): The name of the Elasticsearch index for the embeddings. embedding (Embeddings): An object that provides the ability to embed text. It should be an instance of a class that subclasses the Embeddings abstract base class, such as OpenAIEmbeddings() Raises: ValueError: If the elasticsearch python package is not installed. """ def __init__( self, elasticsearch_url: str, index_name: str, embedding: Embeddings, *, ssl_verify: Optional[Dict[str, Any]] = None, ): """Initialize with necessary components.""" try: import elasticsearch except ImportError: raise ImportError( "Could not import elasticsearch python package. " "Please install it with `pip install elasticsearch`." ) self.embedding = embedding self.index_name = index_name _ssl_verify = ssl_verify or {} try: self.client = elasticsearch.Elasticsearch(elasticsearch_url, **_ssl_verify) except ValueError as e: raise ValueError( f"Your elasticsearch client string is mis-formatted. Got error: {e} " ) [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, refresh_indices: bool = True, ids: Optional[List[str]] = None, **kwargs: Any, ) -> List[str]: """Run more texts through the embeddings and add to the vectorstore. Args: texts: Iterable of strings to add to the vectorstore. metadatas: Optional list of metadatas associated with the texts. refresh_indices: bool to refresh ElasticSearch indices Returns: List of ids from adding the texts into the vectorstore. """ try: from elasticsearch.exceptions import NotFoundError from elasticsearch.helpers import bulk except ImportError: raise ImportError( "Could not import elasticsearch python package. " "Please install it with `pip install elasticsearch`." ) requests = [] ids = ids or [str(uuid.uuid4()) for _ in texts] embeddings = self.embedding.embed_documents(list(texts)) dim = len(embeddings[0]) mapping = _default_text_mapping(dim) # check to see if the index already exists try: self.client.indices.get(index=self.index_name) except NotFoundError: # TODO would be nice to create index before embedding, # just to save expensive steps for last self.create_index(self.client, self.index_name, mapping) for i, text in enumerate(texts): metadata = metadatas[i] if metadatas else {} request = { "_op_type": "index", "_index": self.index_name, "vector": embeddings[i], "text": text, "metadata": metadata, "_id": ids[i], } requests.append(request) bulk(self.client, requests) if refresh_indices: self.client.indices.refresh(index=self.index_name) return ids [docs] def similarity_search( self, query: str, k: int = 4, filter: Optional[dict] = None, **kwargs: Any ) -> List[Document]: """Return docs most similar to query. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. Returns: List of Documents most similar to the query. """ docs_and_scores = self.similarity_search_with_score(query, k, filter=filter) documents = [d[0] for d in docs_and_scores] return documents [docs] def similarity_search_with_score( self, query: str, k: int = 4, filter: Optional[dict] = None, **kwargs: Any ) -> List[Tuple[Document, float]]: """Return docs most similar to query. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. Returns: List of Documents most similar to the query. """ embedding = self.embedding.embed_query(query) script_query = _default_script_query(embedding, filter) response = self.client_search( self.client, self.index_name, script_query, size=k ) hits = [hit for hit in response["hits"]["hits"]] docs_and_scores = [ ( Document( page_content=hit["_source"]["text"], metadata=hit["_source"]["metadata"], ), hit["_score"], ) for hit in hits ] return docs_and_scores [docs] @classmethod def from_texts( cls, texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, elasticsearch_url: Optional[str] = None, index_name: Optional[str] = None, refresh_indices: bool = True, **kwargs: Any, ) -> ElasticVectorSearch: """Construct ElasticVectorSearch wrapper from raw documents. This is a user-friendly interface that: 1. Embeds documents. 2. Creates a new index for the embeddings in the Elasticsearch instance. 3. Adds the documents to the newly created Elasticsearch index. This is intended to be a quick way to get started. Example: .. code-block:: python from langchain import ElasticVectorSearch from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() elastic_vector_search = ElasticVectorSearch.from_texts( texts, embeddings, elasticsearch_url="http://localhost:9200" ) """ elasticsearch_url = elasticsearch_url or get_from_env( "elasticsearch_url", "ELASTICSEARCH_URL" ) index_name = index_name or uuid.uuid4().hex vectorsearch = cls(elasticsearch_url, index_name, embedding, **kwargs) vectorsearch.add_texts( texts, metadatas=metadatas, refresh_indices=refresh_indices ) return vectorsearch [docs] def create_index(self, client: Any, index_name: str, mapping: Dict) -> None: version_num = client.info()["version"]["number"][0] version_num = int(version_num) if version_num >= 8: client.indices.create(index=index_name, mappings=mapping) else: client.indices.create(index=index_name, body={"mappings": mapping}) [docs] def client_search( self, client: Any, index_name: str, script_query: Dict, size: int ) -> Any: version_num = client.info()["version"]["number"][0] version_num = int(version_num) if version_num >= 8: response = client.search(index=index_name, query=script_query, size=size) else: response = client.search( index=index_name, body={"query": script_query, "size": size} ) return response [docs] def delete(self, ids: List[str]) -> None: """Delete by vector IDs. Args: ids: List of ids to delete. """ # TODO: Check if this can be done in bulk for id in ids: self.client.delete(index=self.index_name, id=id) class ElasticKnnSearch(ElasticVectorSearch): """ A class for performing k-Nearest Neighbors (k-NN) search on an Elasticsearch index. The class is designed for a text search scenario where documents are text strings and their embeddings are vector representations of those strings. """ def __init__( self, index_name: str, embedding: Embeddings, es_connection: Optional["Elasticsearch"] = None, es_cloud_id: Optional[str] = None, es_user: Optional[str] = None, es_password: Optional[str] = None, vector_query_field: Optional[str] = "vector", query_field: Optional[str] = "text", ): """ Initializes an instance of the ElasticKnnSearch class and sets up the Elasticsearch client. Args: index_name: The name of the Elasticsearch index. embedding: An instance of the Embeddings class, used to generate vector representations of text strings. es_connection: An existing Elasticsearch connection. es_cloud_id: The Cloud ID of the Elasticsearch instance. Required if creating a new connection. es_user: The username for the Elasticsearch instance. Required if creating a new connection. es_password: The password for the Elasticsearch instance. Required if creating a new connection. """ try: import elasticsearch except ImportError: raise ImportError( "Could not import elasticsearch python package. " "Please install it with `pip install elasticsearch`." ) self.embedding = embedding self.index_name = index_name self.query_field = query_field self.vector_query_field = vector_query_field # If a pre-existing Elasticsearch connection is provided, use it. if es_connection is not None: self.client = es_connection else: # If credentials for a new Elasticsearch connection are provided, # create a new connection. if es_cloud_id and es_user and es_password: self.client = elasticsearch.Elasticsearch( cloud_id=es_cloud_id, basic_auth=(es_user, es_password) ) else: raise ValueError( """Either provide a pre-existing Elasticsearch connection, \ or valid credentials for creating a new connection.""" ) @staticmethod def _default_knn_mapping(dims: int) -> Dict: """Generates a default index mapping for kNN search.""" return { "properties": { "text": {"type": "text"}, "vector": { "type": "dense_vector", "dims": dims, "index": True, "similarity": "dot_product", }, } } def _default_knn_query( self, query_vector: Optional[List[float]] = None, query: Optional[str] = None, model_id: Optional[str] = None, k: Optional[int] = 10, num_candidates: Optional[int] = 10, ) -> Dict: knn: Dict = { "field": self.vector_query_field, "k": k, "num_candidates": num_candidates, } # Case 1: `query_vector` is provided, but not `model_id` -> use query_vector if query_vector and not model_id: knn["query_vector"] = query_vector # Case 2: `query` and `model_id` are provided, -> use query_vector_builder elif query and model_id: knn["query_vector_builder"] = { "text_embedding": { "model_id": model_id, # use 'model_id' argument "model_text": query, # use 'query' argument } } else: raise ValueError( "Either `query_vector` or `model_id` must be provided, but not both." ) return knn def knn_search( self, query: Optional[str] = None, k: Optional[int] = 10, query_vector: Optional[List[float]] = None, model_id: Optional[str] = None, size: Optional[int] = 10, source: Optional[bool] = True, fields: Optional[ Union[List[Mapping[str, Any]], Tuple[Mapping[str, Any], ...], None] ] = None, ) -> Dict: """ Performs a k-nearest neighbor (k-NN) search on the Elasticsearch index. The search can be conducted using either a raw query vector or a model ID. The method first generates the body of the search query, which can be interpreted by Elasticsearch. It then performs the k-NN search on the Elasticsearch index and returns the results. Args: query: The query or queries to be used for the search. Required if `query_vector` is not provided. k: The number of nearest neighbors to return. Defaults to 10. query_vector: The query vector to be used for the search. Required if `query` is not provided. model_id: The ID of the model to use for generating the query vector, if `query` is provided. size: The number of search hits to return. Defaults to 10. source: Whether to include the source of each hit in the results. fields: The fields to include in the source of each hit. If None, all fields are included. vector_query_field: Field name to use in knn search if not default 'vector' Returns: The search results. Raises: ValueError: If neither `query_vector` nor `model_id` is provided, or if both are provided. """ knn_query_body = self._default_knn_query( query_vector=query_vector, query=query, model_id=model_id, k=k ) # Perform the kNN search on the Elasticsearch index and return the results. res = self.client.search( index=self.index_name, knn=knn_query_body, size=size, source=source, fields=fields, ) return dict(res) def knn_hybrid_search( self, query: Optional[str] = None, k: Optional[int] = 10, query_vector: Optional[List[float]] = None, model_id: Optional[str] = None, size: Optional[int] = 10, source: Optional[bool] = True, knn_boost: Optional[float] = 0.9, query_boost: Optional[float] = 0.1, fields: Optional[ Union[List[Mapping[str, Any]], Tuple[Mapping[str, Any], ...], None] ] = None, ) -> Dict[Any, Any]: """Performs a hybrid k-nearest neighbor (k-NN) and text-based search on the Elasticsearch index. The search can be conducted using either a raw query vector or a model ID. The method first generates the body of the k-NN search query and the text-based query, which can be interpreted by Elasticsearch. It then performs the hybrid search on the Elasticsearch index and returns the results. Args: query: The query or queries to be used for the search. Required if `query_vector` is not provided. k: The number of nearest neighbors to return. Defaults to 10. query_vector: The query vector to be used for the search. Required if `query` is not provided. model_id: The ID of the model to use for generating the query vector, if `query` is provided. size: The number of search hits to return. Defaults to 10. source: Whether to include the source of each hit in the results. knn_boost: The boost factor for the k-NN part of the search. query_boost: The boost factor for the text-based part of the search. fields The fields to include in the source of each hit. If None, all fields are included. Defaults to None. vector_query_field: Field name to use in knn search if not default 'vector' query_field: Field name to use in search if not default 'text' Returns: The search results. Raises: ValueError: If neither `query_vector` nor `model_id` is provided, or if both are provided. """ knn_query_body = self._default_knn_query( query_vector=query_vector, query=query, model_id=model_id, k=k ) # Modify the knn_query_body to add a "boost" parameter knn_query_body["boost"] = knn_boost # Generate the body of the standard Elasticsearch query match_query_body = { "match": {self.query_field: {"query": query, "boost": query_boost}} } # Perform the hybrid search on the Elasticsearch index and return the results. res = self.client.search( index=self.index_name, query=match_query_body, knn=knn_query_body, fields=fields, size=size, source=source, ) return dict(res)
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/elastic_vector_search.html
bc831b53-5810-4ade-a742-d8dc2cc1dba9
Source code for langchain.vectorstores.mongodb_atlas from __future__ import annotations import logging from typing import ( TYPE_CHECKING, Any, Dict, Generator, Iterable, List, Optional, Tuple, TypeVar, Union, ) from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.vectorstores.base import VectorStore if TYPE_CHECKING: from pymongo.collection import Collection MongoDBDocumentType = TypeVar("MongoDBDocumentType", bound=Dict[str, Any]) logger = logging.getLogger(__name__) DEFAULT_INSERT_BATCH_SIZE = 100 [docs]class MongoDBAtlasVectorSearch(VectorStore): """Wrapper around MongoDB Atlas Vector Search. To use, you should have both: - the ``pymongo`` python package installed - a connection string associated with a MongoDB Atlas Cluster having deployed an Atlas Search index Example: .. code-block:: python from langchain.vectorstores import MongoDBAtlasVectorSearch from langchain.embeddings.openai import OpenAIEmbeddings from pymongo import MongoClient mongo_client = MongoClient("<YOUR-CONNECTION-STRING>") collection = mongo_client["<db_name>"]["<collection_name>"] embeddings = OpenAIEmbeddings() vectorstore = MongoDBAtlasVectorSearch(collection, embeddings) """ def __init__( self, collection: Collection[MongoDBDocumentType], embedding: Embeddings, *, index_name: str = "default", text_key: str = "text", embedding_key: str = "embedding", ): """ Args: collection: MongoDB collection to add the texts to. embedding: Text embedding model to use. text_key: MongoDB field that will contain the text for each document. embedding_key: MongoDB field that will contain the embedding for each document. """ self._collection = collection self._embedding = embedding self._index_name = index_name self._text_key = text_key self._embedding_key = embedding_key [docs] @classmethod def from_connection_string( cls, connection_string: str, namespace: str, embedding: Embeddings, **kwargs: Any, ) -> MongoDBAtlasVectorSearch: try: from pymongo import MongoClient except ImportError: raise ImportError( "Could not import pymongo, please install it with " "`pip install pymongo`." ) client: MongoClient = MongoClient(connection_string) db_name, collection_name = namespace.split(".") collection = client[db_name][collection_name] return cls(collection, embedding, **kwargs) [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[Dict[str, Any]]] = None, **kwargs: Any, ) -> List: """Run more texts through the embeddings and add to the vectorstore. Args: texts: Iterable of strings to add to the vectorstore. metadatas: Optional list of metadatas associated with the texts. Returns: List of ids from adding the texts into the vectorstore. """ batch_size = kwargs.get("batch_size", DEFAULT_INSERT_BATCH_SIZE) _metadatas: Union[List, Generator] = metadatas or ({} for _ in texts) texts_batch = [] metadatas_batch = [] result_ids = [] for i, (text, metadata) in enumerate(zip(texts, _metadatas)): texts_batch.append(text) metadatas_batch.append(metadata) if (i + 1) % batch_size == 0: result_ids.extend(self._insert_texts(texts_batch, metadatas_batch)) texts_batch = [] metadatas_batch = [] if texts_batch: result_ids.extend(self._insert_texts(texts_batch, metadatas_batch)) return result_ids def _insert_texts(self, texts: List[str], metadatas: List[Dict[str, Any]]) -> List: if not texts: return [] # Embed and create the documents embeddings = self._embedding.embed_documents(texts) to_insert = [ {self._text_key: t, self._embedding_key: embedding, **m} for t, m, embedding in zip(texts, metadatas, embeddings) ] # insert the documents in MongoDB Atlas insert_result = self._collection.insert_many(to_insert) return insert_result.inserted_ids [docs] def similarity_search_with_score( self, query: str, *, k: int = 4, pre_filter: Optional[dict] = None, post_filter_pipeline: Optional[List[Dict]] = None, ) -> List[Tuple[Document, float]]: """Return MongoDB documents most similar to query, along with scores. Use the knnBeta Operator available in MongoDB Atlas Search This feature is in early access and available only for evaluation purposes, to validate functionality, and to gather feedback from a small closed group of early access users. It is not recommended for production deployments as we may introduce breaking changes. For more: https://www.mongodb.com/docs/atlas/atlas-search/knn-beta Args: query: Text to look up documents similar to. k: Optional Number of Documents to return. Defaults to 4. pre_filter: Optional Dictionary of argument(s) to prefilter on document fields. post_filter_pipeline: Optional Pipeline of MongoDB aggregation stages following the knnBeta search. Returns: List of Documents most similar to the query and score for each """ knn_beta = { "vector": self._embedding.embed_query(query), "path": self._embedding_key, "k": k, } if pre_filter: knn_beta["filter"] = pre_filter pipeline = [ { "$search": { "index": self._index_name, "knnBeta": knn_beta, } }, {"$project": {"score": {"$meta": "searchScore"}, self._embedding_key: 0}}, ] if post_filter_pipeline is not None: pipeline.extend(post_filter_pipeline) cursor = self._collection.aggregate(pipeline) docs = [] for res in cursor: text = res.pop(self._text_key) score = res.pop("score") docs.append((Document(page_content=text, metadata=res), score)) return docs [docs] def similarity_search( self, query: str, k: int = 4, pre_filter: Optional[dict] = None, post_filter_pipeline: Optional[List[Dict]] = None, **kwargs: Any, ) -> List[Document]: """Return MongoDB documents most similar to query. Use the knnBeta Operator available in MongoDB Atlas Search This feature is in early access and available only for evaluation purposes, to validate functionality, and to gather feedback from a small closed group of early access users. It is not recommended for production deployments as we may introduce breaking changes. For more: https://www.mongodb.com/docs/atlas/atlas-search/knn-beta Args: query: Text to look up documents similar to. k: Optional Number of Documents to return. Defaults to 4. pre_filter: Optional Dictionary of argument(s) to prefilter on document fields. post_filter_pipeline: Optional Pipeline of MongoDB aggregation stages following the knnBeta search. Returns: List of Documents most similar to the query and score for each """ docs_and_scores = self.similarity_search_with_score( query, k=k, pre_filter=pre_filter, post_filter_pipeline=post_filter_pipeline, ) return [doc for doc, _ in docs_and_scores] [docs] @classmethod def from_texts( cls, texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, collection: Optional[Collection[MongoDBDocumentType]] = None, **kwargs: Any, ) -> MongoDBAtlasVectorSearch: """Construct MongoDBAtlasVectorSearch wrapper from raw documents. This is a user-friendly interface that: 1. Embeds documents. 2. Adds the documents to a provided MongoDB Atlas Vector Search index (Lucene) This is intended to be a quick way to get started. Example: .. code-block:: python from pymongo import MongoClient from langchain.vectorstores import MongoDBAtlasVectorSearch from langchain.embeddings import OpenAIEmbeddings client = MongoClient("<YOUR-CONNECTION-STRING>") collection = mongo_client["<db_name>"]["<collection_name>"] embeddings = OpenAIEmbeddings() vectorstore = MongoDBAtlasVectorSearch.from_texts( texts, embeddings, metadatas=metadatas, collection=collection ) """ if collection is None: raise ValueError("Must provide 'collection' named parameter.") vecstore = cls(collection, embedding, **kwargs) vecstore.add_texts(texts, metadatas=metadatas) return vecstore
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/mongodb_atlas.html
0239964e-13d8-4344-9872-2623ed923446
Source code for langchain.vectorstores.clarifai from __future__ import annotations import logging import os import traceback from typing import Any, Iterable, List, Optional, Tuple import requests from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.vectorstores.base import VectorStore logger = logging.getLogger(__name__) [docs]class Clarifai(VectorStore): """Wrapper around Clarifai AI platform's vector store. To use, you should have the ``clarifai`` python package installed. Example: .. code-block:: python from langchain.vectorstores import Clarifai from langchain.embeddings.openai import OpenAIEmbeddings embeddings = OpenAIEmbeddings() vectorstore = Clarifai("langchain_store", embeddings.embed_query) """ def __init__( self, user_id: Optional[str] = None, app_id: Optional[str] = None, pat: Optional[str] = None, number_of_docs: Optional[int] = None, api_base: Optional[str] = None, ) -> None: """Initialize with Clarifai client. Args: user_id (Optional[str], optional): User ID. Defaults to None. app_id (Optional[str], optional): App ID. Defaults to None. pat (Optional[str], optional): Personal access token. Defaults to None. number_of_docs (Optional[int], optional): Number of documents to return during vector search. Defaults to None. api_base (Optional[str], optional): API base. Defaults to None. Raises: ValueError: If user ID, app ID or personal access token is not provided. """ try: from clarifai.auth.helper import DEFAULT_BASE, ClarifaiAuthHelper from clarifai.client import create_stub except ImportError: raise ValueError( "Could not import clarifai python package. " "Please install it with `pip install clarifai`." ) if api_base is None: self._api_base = DEFAULT_BASE self._user_id = user_id or os.environ.get("CLARIFAI_USER_ID") self._app_id = app_id or os.environ.get("CLARIFAI_APP_ID") self._pat = pat or os.environ.get("CLARIFAI_PAT_KEY") if self._user_id is None or self._app_id is None or self._pat is None: raise ValueError( "Could not find CLARIFAI_USER_ID, CLARIFAI_APP_ID or\ CLARIFAI_PAT in your environment. " "Please set those env variables with a valid user ID, \ app ID and personal access token \ from https://clarifai.com/settings/security." ) self._auth = ClarifaiAuthHelper( user_id=self._user_id, app_id=self._app_id, pat=self._pat, base=self._api_base, ) self._stub = create_stub(self._auth) self._userDataObject = self._auth.get_user_app_id_proto() self._number_of_docs = number_of_docs def _post_text_input(self, text: str, metadata: dict) -> str: """Post text to Clarifai and return the ID of the input. Args: text (str): Text to post. metadata (dict): Metadata to post. Returns: str: ID of the input. """ try: from clarifai_grpc.grpc.api import resources_pb2, service_pb2 from clarifai_grpc.grpc.api.status import status_code_pb2 from google.protobuf.struct_pb2 import Struct # type: ignore except ImportError as e: raise ImportError( "Could not import clarifai python package. " "Please install it with `pip install clarifai`." ) from e input_metadata = Struct() input_metadata.update(metadata) post_inputs_response = self._stub.PostInputs( service_pb2.PostInputsRequest( user_app_id=self._userDataObject, inputs=[ resources_pb2.Input( data=resources_pb2.Data( text=resources_pb2.Text(raw=text), metadata=input_metadata, ) ) ], ) ) if post_inputs_response.status.code != status_code_pb2.SUCCESS: logger.error(post_inputs_response.status) raise Exception( "Post inputs failed, status: " + post_inputs_response.status.description ) input_id = post_inputs_response.inputs[0].id return input_id [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, **kwargs: Any, ) -> List[str]: """Add texts to the Clarifai vectorstore. This will push the text to a Clarifai application. Application use base workflow that create and store embedding for each text. Make sure you are using a base workflow that is compatible with text (such as Language Understanding). Args: texts (Iterable[str]): Texts to add to the vectorstore. metadatas (Optional[List[dict]], optional): Optional list of metadatas. ids (Optional[List[str]], optional): Optional list of IDs. Returns: List[str]: List of IDs of the added texts. """ assert len(list(texts)) > 0, "No texts provided to add to the vectorstore." if metadatas is not None: assert len(list(texts)) == len( metadatas ), "Number of texts and metadatas should be the same." input_ids = [] for idx, text in enumerate(texts): try: metadata = metadatas[idx] if metadatas else {} input_id = self._post_text_input(text, metadata) input_ids.append(input_id) logger.debug(f"Input {input_id} posted successfully.") except Exception as error: logger.warning(f"Post inputs failed: {error}") traceback.print_exc() return input_ids [docs] def similarity_search_with_score( self, query: str, k: int = 4, filter: Optional[dict] = None, namespace: Optional[str] = None, **kwargs: Any, ) -> List[Tuple[Document, float]]: """Run similarity search with score using Clarifai. Args: query (str): Query text to search for. k (int): Number of results to return. Defaults to 4. filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. Returns: List[Document]: List of documents most simmilar to the query text. """ try: from clarifai_grpc.grpc.api import resources_pb2, service_pb2 from clarifai_grpc.grpc.api.status import status_code_pb2 from google.protobuf import json_format # type: ignore except ImportError as e: raise ImportError( "Could not import clarifai python package. " "Please install it with `pip install clarifai`." ) from e # Get number of docs to return if self._number_of_docs is not None: k = self._number_of_docs post_annotations_searches_response = self._stub.PostAnnotationsSearches( service_pb2.PostAnnotationsSearchesRequest( user_app_id=self._userDataObject, searches=[ resources_pb2.Search( query=resources_pb2.Query( ranks=[ resources_pb2.Rank( annotation=resources_pb2.Annotation( data=resources_pb2.Data( text=resources_pb2.Text(raw=query), ) ) ) ] ) ) ], pagination=service_pb2.Pagination(page=1, per_page=k), ) ) # Check if search was successful if post_annotations_searches_response.status.code != status_code_pb2.SUCCESS: raise Exception( "Post searches failed, status: " + post_annotations_searches_response.status.description ) # Retrieve hits hits = post_annotations_searches_response.hits docs_and_scores = [] # Iterate over hits and retrieve metadata and text for hit in hits: metadata = json_format.MessageToDict(hit.input.data.metadata) request = requests.get(hit.input.data.text.url) # override encoding by real educated guess as provided by chardet request.encoding = request.apparent_encoding requested_text = request.text logger.debug( f"\tScore {hit.score:.2f} for annotation: {hit.annotation.id}\ off input: {hit.input.id}, text: {requested_text[:125]}" ) docs_and_scores.append( (Document(page_content=requested_text, metadata=metadata), hit.score) ) return docs_and_scores [docs] def similarity_search( self, query: str, k: int = 4, **kwargs: Any, ) -> List[Document]: """Run similarity search using Clarifai. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. Returns: List of Documents most similar to the query and score for each """ docs_and_scores = self.similarity_search_with_score(query, **kwargs) return [doc for doc, _ in docs_and_scores] [docs] @classmethod def from_texts( cls, texts: List[str], embedding: Optional[Embeddings] = None, metadatas: Optional[List[dict]] = None, user_id: Optional[str] = None, app_id: Optional[str] = None, pat: Optional[str] = None, number_of_docs: Optional[int] = None, api_base: Optional[str] = None, **kwargs: Any, ) -> Clarifai: """Create a Clarifai vectorstore from a list of texts. Args: user_id (str): User ID. app_id (str): App ID. texts (List[str]): List of texts to add. pat (Optional[str]): Personal access token. Defaults to None. number_of_docs (Optional[int]): Number of documents to return during vector search. Defaults to None. api_base (Optional[str]): API base. Defaults to None. metadatas (Optional[List[dict]]): Optional list of metadatas. Defaults to None. Returns: Clarifai: Clarifai vectorstore. """ clarifai_vector_db = cls( user_id=user_id, app_id=app_id, pat=pat, number_of_docs=number_of_docs, api_base=api_base, ) clarifai_vector_db.add_texts(texts=texts, metadatas=metadatas) return clarifai_vector_db [docs] @classmethod def from_documents( cls, documents: List[Document], embedding: Optional[Embeddings] = None, user_id: Optional[str] = None, app_id: Optional[str] = None, pat: Optional[str] = None, number_of_docs: Optional[int] = None, api_base: Optional[str] = None, **kwargs: Any, ) -> Clarifai: """Create a Clarifai vectorstore from a list of documents. Args: user_id (str): User ID. app_id (str): App ID. documents (List[Document]): List of documents to add. pat (Optional[str]): Personal access token. Defaults to None. number_of_docs (Optional[int]): Number of documents to return during vector search. Defaults to None. api_base (Optional[str]): API base. Defaults to None. Returns: Clarifai: Clarifai vectorstore. """ texts = [doc.page_content for doc in documents] metadatas = [doc.metadata for doc in documents] return cls.from_texts( user_id=user_id, app_id=app_id, texts=texts, pat=pat, number_of_docs=number_of_docs, api_base=api_base, metadatas=metadatas, )
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/clarifai.html
076f8df6-bb46-4200-b23f-a215208be56b
Source code for langchain.vectorstores.chroma """Wrapper around ChromaDB embeddings platform.""" from __future__ import annotations import logging import uuid from typing import TYPE_CHECKING, Any, Dict, Iterable, List, Optional, Tuple, Type import numpy as np from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.utils import xor_args from langchain.vectorstores.base import VectorStore from langchain.vectorstores.utils import maximal_marginal_relevance if TYPE_CHECKING: import chromadb import chromadb.config from chromadb.api.types import ID, OneOrMany, Where, WhereDocument logger = logging.getLogger() DEFAULT_K = 4 # Number of Documents to return. def _results_to_docs(results: Any) -> List[Document]: return [doc for doc, _ in _results_to_docs_and_scores(results)] def _results_to_docs_and_scores(results: Any) -> List[Tuple[Document, float]]: return [ # TODO: Chroma can do batch querying, # we shouldn't hard code to the 1st result (Document(page_content=result[0], metadata=result[1] or {}), result[2]) for result in zip( results["documents"][0], results["metadatas"][0], results["distances"][0], ) ] [docs]class Chroma(VectorStore): """Wrapper around ChromaDB embeddings platform. To use, you should have the ``chromadb`` python package installed. Example: .. code-block:: python from langchain.vectorstores import Chroma from langchain.embeddings.openai import OpenAIEmbeddings embeddings = OpenAIEmbeddings() vectorstore = Chroma("langchain_store", embeddings) """ _LANGCHAIN_DEFAULT_COLLECTION_NAME = "langchain" def __init__( self, collection_name: str = _LANGCHAIN_DEFAULT_COLLECTION_NAME, embedding_function: Optional[Embeddings] = None, persist_directory: Optional[str] = None, client_settings: Optional[chromadb.config.Settings] = None, collection_metadata: Optional[Dict] = None, client: Optional[chromadb.Client] = None, ) -> None: """Initialize with Chroma client.""" try: import chromadb import chromadb.config except ImportError: raise ValueError( "Could not import chromadb python package. " "Please install it with `pip install chromadb`." ) if client is not None: self._client = client else: if client_settings: self._client_settings = client_settings else: self._client_settings = chromadb.config.Settings() if persist_directory is not None: self._client_settings = chromadb.config.Settings( chroma_db_impl="duckdb+parquet", persist_directory=persist_directory, ) self._client = chromadb.Client(self._client_settings) self._embedding_function = embedding_function self._persist_directory = persist_directory self._collection = self._client.get_or_create_collection( name=collection_name, embedding_function=self._embedding_function.embed_documents if self._embedding_function is not None else None, metadata=collection_metadata, ) @xor_args(("query_texts", "query_embeddings")) def __query_collection( self, query_texts: Optional[List[str]] = None, query_embeddings: Optional[List[List[float]]] = None, n_results: int = 4, where: Optional[Dict[str, str]] = None, **kwargs: Any, ) -> List[Document]: """Query the chroma collection.""" try: import chromadb # noqa: F401 except ImportError: raise ValueError( "Could not import chromadb python package. " "Please install it with `pip install chromadb`." ) return self._collection.query( query_texts=query_texts, query_embeddings=query_embeddings, n_results=n_results, where=where, **kwargs, ) [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, **kwargs: Any, ) -> List[str]: """Run more texts through the embeddings and add to the vectorstore. Args: texts (Iterable[str]): Texts to add to the vectorstore. metadatas (Optional[List[dict]], optional): Optional list of metadatas. ids (Optional[List[str]], optional): Optional list of IDs. Returns: List[str]: List of IDs of the added texts. """ # TODO: Handle the case where the user doesn't provide ids on the Collection if ids is None: ids = [str(uuid.uuid1()) for _ in texts] embeddings = None if self._embedding_function is not None: embeddings = self._embedding_function.embed_documents(list(texts)) self._collection.upsert( metadatas=metadatas, embeddings=embeddings, documents=texts, ids=ids ) return ids [docs] def similarity_search( self, query: str, k: int = DEFAULT_K, filter: Optional[Dict[str, str]] = None, **kwargs: Any, ) -> List[Document]: """Run similarity search with Chroma. Args: query (str): Query text to search for. k (int): Number of results to return. Defaults to 4. filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. Returns: List[Document]: List of documents most similar to the query text. """ docs_and_scores = self.similarity_search_with_score(query, k, filter=filter) return [doc for doc, _ in docs_and_scores] [docs] def similarity_search_by_vector( self, embedding: List[float], k: int = DEFAULT_K, filter: Optional[Dict[str, str]] = None, **kwargs: Any, ) -> List[Document]: """Return docs most similar to embedding vector. Args: embedding (str): Embedding to look up documents similar to. k (int): Number of Documents to return. Defaults to 4. filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. Returns: List of Documents most similar to the query vector. """ results = self.__query_collection( query_embeddings=embedding, n_results=k, where=filter ) return _results_to_docs(results) [docs] def similarity_search_with_score( self, query: str, k: int = DEFAULT_K, filter: Optional[Dict[str, str]] = None, **kwargs: Any, ) -> List[Tuple[Document, float]]: """Run similarity search with Chroma with distance. Args: query (str): Query text to search for. k (int): Number of results to return. Defaults to 4. filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. Returns: List[Tuple[Document, float]]: List of documents most similar to the query text and cosine distance in float for each. Lower score represents more similarity. """ if self._embedding_function is None: results = self.__query_collection( query_texts=[query], n_results=k, where=filter ) else: query_embedding = self._embedding_function.embed_query(query) results = self.__query_collection( query_embeddings=[query_embedding], n_results=k, where=filter ) return _results_to_docs_and_scores(results) def _similarity_search_with_relevance_scores( self, query: str, k: int = 4, **kwargs: Any, ) -> List[Tuple[Document, float]]: return self.similarity_search_with_score(query, k, **kwargs) [docs] def max_marginal_relevance_search_by_vector( self, embedding: List[float], k: int = DEFAULT_K, fetch_k: int = 20, lambda_mult: float = 0.5, filter: Optional[Dict[str, str]] = None, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: embedding: Embedding to look up documents similar to. k: Number of Documents to return. Defaults to 4. fetch_k: Number of Documents to fetch to pass to MMR algorithm. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. Returns: List of Documents selected by maximal marginal relevance. """ results = self.__query_collection( query_embeddings=embedding, n_results=fetch_k, where=filter, include=["metadatas", "documents", "distances", "embeddings"], ) mmr_selected = maximal_marginal_relevance( np.array(embedding, dtype=np.float32), results["embeddings"][0], k=k, lambda_mult=lambda_mult, ) candidates = _results_to_docs(results) selected_results = [r for i, r in enumerate(candidates) if i in mmr_selected] return selected_results [docs] def max_marginal_relevance_search( self, query: str, k: int = DEFAULT_K, fetch_k: int = 20, lambda_mult: float = 0.5, filter: Optional[Dict[str, str]] = None, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. fetch_k: Number of Documents to fetch to pass to MMR algorithm. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. Returns: List of Documents selected by maximal marginal relevance. """ if self._embedding_function is None: raise ValueError( "For MMR search, you must specify an embedding function on" "creation." ) embedding = self._embedding_function.embed_query(query) docs = self.max_marginal_relevance_search_by_vector( embedding, k, fetch_k, lambda_mul=lambda_mult, filter=filter ) return docs [docs] def delete_collection(self) -> None: """Delete the collection.""" self._client.delete_collection(self._collection.name) [docs] def get( self, ids: Optional[OneOrMany[ID]] = None, where: Optional[Where] = None, limit: Optional[int] = None, offset: Optional[int] = None, where_document: Optional[WhereDocument] = None, include: Optional[List[str]] = None, ) -> Dict[str, Any]: """Gets the collection. Args: ids: The ids of the embeddings to get. Optional. where: A Where type dict used to filter results by. E.g. `{"color" : "red", "price": 4.20}`. Optional. limit: The number of documents to return. Optional. offset: The offset to start returning results from. Useful for paging results with limit. Optional. where_document: A WhereDocument type dict used to filter by the documents. E.g. `{$contains: {"text": "hello"}}`. Optional. include: A list of what to include in the results. Can contain `"embeddings"`, `"metadatas"`, `"documents"`. Ids are always included. Defaults to `["metadatas", "documents"]`. Optional. """ kwargs = { "ids": ids, "where": where, "limit": limit, "offset": offset, "where_document": where_document, } if include is not None: kwargs["include"] = include return self._collection.get(**kwargs) [docs] def persist(self) -> None: """Persist the collection. This can be used to explicitly persist the data to disk. It will also be called automatically when the object is destroyed. """ if self._persist_directory is None: raise ValueError( "You must specify a persist_directory on" "creation to persist the collection." ) self._client.persist() [docs] def update_document(self, document_id: str, document: Document) -> None: """Update a document in the collection. Args: document_id (str): ID of the document to update. document (Document): Document to update. """ text = document.page_content metadata = document.metadata if self._embedding_function is None: raise ValueError( "For update, you must specify an embedding function on creation." ) embeddings = self._embedding_function.embed_documents([text]) self._collection.update( ids=[document_id], embeddings=embeddings, documents=[text], metadatas=[metadata], ) [docs] @classmethod def from_texts( cls: Type[Chroma], texts: List[str], embedding: Optional[Embeddings] = None, metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, collection_name: str = _LANGCHAIN_DEFAULT_COLLECTION_NAME, persist_directory: Optional[str] = None, client_settings: Optional[chromadb.config.Settings] = None, client: Optional[chromadb.Client] = None, **kwargs: Any, ) -> Chroma: """Create a Chroma vectorstore from a raw documents. If a persist_directory is specified, the collection will be persisted there. Otherwise, the data will be ephemeral in-memory. Args: texts (List[str]): List of texts to add to the collection. collection_name (str): Name of the collection to create. persist_directory (Optional[str]): Directory to persist the collection. embedding (Optional[Embeddings]): Embedding function. Defaults to None. metadatas (Optional[List[dict]]): List of metadatas. Defaults to None. ids (Optional[List[str]]): List of document IDs. Defaults to None. client_settings (Optional[chromadb.config.Settings]): Chroma client settings Returns: Chroma: Chroma vectorstore. """ chroma_collection = cls( collection_name=collection_name, embedding_function=embedding, persist_directory=persist_directory, client_settings=client_settings, client=client, ) chroma_collection.add_texts(texts=texts, metadatas=metadatas, ids=ids) return chroma_collection [docs] @classmethod def from_documents( cls: Type[Chroma], documents: List[Document], embedding: Optional[Embeddings] = None, ids: Optional[List[str]] = None, collection_name: str = _LANGCHAIN_DEFAULT_COLLECTION_NAME, persist_directory: Optional[str] = None, client_settings: Optional[chromadb.config.Settings] = None, client: Optional[chromadb.Client] = None, # Add this line **kwargs: Any, ) -> Chroma: """Create a Chroma vectorstore from a list of documents. If a persist_directory is specified, the collection will be persisted there. Otherwise, the data will be ephemeral in-memory. Args: collection_name (str): Name of the collection to create. persist_directory (Optional[str]): Directory to persist the collection. ids (Optional[List[str]]): List of document IDs. Defaults to None. documents (List[Document]): List of documents to add to the vectorstore. embedding (Optional[Embeddings]): Embedding function. Defaults to None. client_settings (Optional[chromadb.config.Settings]): Chroma client settings Returns: Chroma: Chroma vectorstore. """ texts = [doc.page_content for doc in documents] metadatas = [doc.metadata for doc in documents] return cls.from_texts( texts=texts, embedding=embedding, metadatas=metadatas, ids=ids, collection_name=collection_name, persist_directory=persist_directory, client_settings=client_settings, client=client, ) [docs] def delete(self, ids: List[str]) -> None: """Delete by vector IDs. Args: ids: List of ids to delete. """ self._collection.delete(ids=ids)
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/chroma.html
501f6fba-6bcb-410a-bd6f-dabf4ba0ef6a
Source code for langchain.vectorstores.qdrant """Wrapper around Qdrant vector database.""" from __future__ import annotations import uuid import warnings from itertools import islice from operator import itemgetter from typing import ( TYPE_CHECKING, Any, Callable, Dict, Iterable, List, Optional, Sequence, Tuple, Type, Union, ) import numpy as np from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.vectorstores import VectorStore from langchain.vectorstores.utils import maximal_marginal_relevance if TYPE_CHECKING: from qdrant_client.conversions import common_types from qdrant_client.http import models as rest DictFilter = Dict[str, Union[str, int, bool, dict, list]] MetadataFilter = Union[DictFilter, common_types.Filter] [docs]class Qdrant(VectorStore): """Wrapper around Qdrant vector database. To use you should have the ``qdrant-client`` package installed. Example: .. code-block:: python from qdrant_client import QdrantClient from langchain import Qdrant client = QdrantClient() collection_name = "MyCollection" qdrant = Qdrant(client, collection_name, embedding_function) """ CONTENT_KEY = "page_content" METADATA_KEY = "metadata" def __init__( self, client: Any, collection_name: str, embeddings: Optional[Embeddings] = None, content_payload_key: str = CONTENT_KEY, metadata_payload_key: str = METADATA_KEY, embedding_function: Optional[Callable] = None, # deprecated ): """Initialize with necessary components.""" try: import qdrant_client except ImportError: raise ValueError( "Could not import qdrant-client python package. " "Please install it with `pip install qdrant-client`." ) if not isinstance(client, qdrant_client.QdrantClient): raise ValueError( f"client should be an instance of qdrant_client.QdrantClient, " f"got {type(client)}" ) if embeddings is None and embedding_function is None: raise ValueError( "`embeddings` value can't be None. Pass `Embeddings` instance." ) if embeddings is not None and embedding_function is not None: raise ValueError( "Both `embeddings` and `embedding_function` are passed. " "Use `embeddings` only." ) self.embeddings = embeddings self._embeddings_function = embedding_function self.client: qdrant_client.QdrantClient = client self.collection_name = collection_name self.content_payload_key = content_payload_key or self.CONTENT_KEY self.metadata_payload_key = metadata_payload_key or self.METADATA_KEY if embedding_function is not None: warnings.warn( "Using `embedding_function` is deprecated. " "Pass `Embeddings` instance to `embeddings` instead." ) if not isinstance(embeddings, Embeddings): warnings.warn( "`embeddings` should be an instance of `Embeddings`." "Using `embeddings` as `embedding_function` which is deprecated" ) self._embeddings_function = embeddings self.embeddings = None [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, ids: Optional[Sequence[str]] = None, batch_size: int = 64, **kwargs: Any, ) -> List[str]: """Run more texts through the embeddings and add to the vectorstore. Args: texts: Iterable of strings to add to the vectorstore. metadatas: Optional list of metadatas associated with the texts. ids: Optional list of ids to associate with the texts. Ids have to be uuid-like strings. batch_size: How many vectors upload per-request. Default: 64 Returns: List of ids from adding the texts into the vectorstore. """ from qdrant_client.http import models as rest added_ids = [] texts_iterator = iter(texts) metadatas_iterator = iter(metadatas or []) ids_iterator = iter(ids or [uuid.uuid4().hex for _ in iter(texts)]) while batch_texts := list(islice(texts_iterator, batch_size)): # Take the corresponding metadata and id for each text in a batch batch_metadatas = list(islice(metadatas_iterator, batch_size)) or None batch_ids = list(islice(ids_iterator, batch_size)) self.client.upsert( collection_name=self.collection_name, points=rest.Batch.construct( ids=batch_ids, vectors=self._embed_texts(batch_texts), payloads=self._build_payloads( batch_texts, batch_metadatas, self.content_payload_key, self.metadata_payload_key, ), ), ) added_ids.extend(batch_ids) return added_ids [docs] def similarity_search( self, query: str, k: int = 4, filter: Optional[MetadataFilter] = None, search_params: Optional[common_types.SearchParams] = None, offset: int = 0, score_threshold: Optional[float] = None, consistency: Optional[common_types.ReadConsistency] = None, **kwargs: Any, ) -> List[Document]: """Return docs most similar to query. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. filter: Filter by metadata. Defaults to None. search_params: Additional search params offset: Offset of the first result to return. May be used to paginate results. Note: large offset values may cause performance issues. score_threshold: Define a minimal score threshold for the result. If defined, less similar results will not be returned. Score of the returned result might be higher or smaller than the threshold depending on the Distance function used. E.g. for cosine similarity only higher scores will be returned. consistency: Read consistency of the search. Defines how many replicas should be queried before returning the result. Values: - int - number of replicas to query, values should present in all queried replicas - 'majority' - query all replicas, but return values present in the majority of replicas - 'quorum' - query the majority of replicas, return values present in all of them - 'all' - query all replicas, and return values present in all replicas Returns: List of Documents most similar to the query. """ results = self.similarity_search_with_score( query, k, filter=filter, search_params=search_params, offset=offset, score_threshold=score_threshold, consistency=consistency, **kwargs, ) return list(map(itemgetter(0), results)) [docs] def similarity_search_with_score( self, query: str, k: int = 4, filter: Optional[MetadataFilter] = None, search_params: Optional[common_types.SearchParams] = None, offset: int = 0, score_threshold: Optional[float] = None, consistency: Optional[common_types.ReadConsistency] = None, **kwargs: Any, ) -> List[Tuple[Document, float]]: """Return docs most similar to query. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. filter: Filter by metadata. Defaults to None. search_params: Additional search params offset: Offset of the first result to return. May be used to paginate results. Note: large offset values may cause performance issues. score_threshold: Define a minimal score threshold for the result. If defined, less similar results will not be returned. Score of the returned result might be higher or smaller than the threshold depending on the Distance function used. E.g. for cosine similarity only higher scores will be returned. consistency: Read consistency of the search. Defines how many replicas should be queried before returning the result. Values: - int - number of replicas to query, values should present in all queried replicas - 'majority' - query all replicas, but return values present in the majority of replicas - 'quorum' - query the majority of replicas, return values present in all of them - 'all' - query all replicas, and return values present in all replicas Returns: List of documents most similar to the query text and cosine distance in float for each. Lower score represents more similarity. """ return self.similarity_search_with_score_by_vector( self._embed_query(query), k, filter=filter, search_params=search_params, offset=offset, score_threshold=score_threshold, consistency=consistency, **kwargs, ) [docs] def similarity_search_by_vector( self, embedding: List[float], k: int = 4, filter: Optional[MetadataFilter] = None, search_params: Optional[common_types.SearchParams] = None, offset: int = 0, score_threshold: Optional[float] = None, consistency: Optional[common_types.ReadConsistency] = None, **kwargs: Any, ) -> List[Document]: """Return docs most similar to embedding vector. Args: embedding: Embedding vector to look up documents similar to. k: Number of Documents to return. Defaults to 4. filter: Filter by metadata. Defaults to None. search_params: Additional search params offset: Offset of the first result to return. May be used to paginate results. Note: large offset values may cause performance issues. score_threshold: Define a minimal score threshold for the result. If defined, less similar results will not be returned. Score of the returned result might be higher or smaller than the threshold depending on the Distance function used. E.g. for cosine similarity only higher scores will be returned. consistency: Read consistency of the search. Defines how many replicas should be queried before returning the result. Values: - int - number of replicas to query, values should present in all queried replicas - 'majority' - query all replicas, but return values present in the majority of replicas - 'quorum' - query the majority of replicas, return values present in all of them - 'all' - query all replicas, and return values present in all replicas Returns: List of Documents most similar to the query. """ results = self.similarity_search_with_score_by_vector( embedding, k, filter=filter, search_params=search_params, offset=offset, score_threshold=score_threshold, consistency=consistency, **kwargs, ) return list(map(itemgetter(0), results)) [docs] def similarity_search_with_score_by_vector( self, embedding: List[float], k: int = 4, filter: Optional[MetadataFilter] = None, search_params: Optional[common_types.SearchParams] = None, offset: int = 0, score_threshold: Optional[float] = None, consistency: Optional[common_types.ReadConsistency] = None, **kwargs: Any, ) -> List[Tuple[Document, float]]: """Return docs most similar to embedding vector. Args: embedding: Embedding vector to look up documents similar to. k: Number of Documents to return. Defaults to 4. filter: Filter by metadata. Defaults to None. search_params: Additional search params offset: Offset of the first result to return. May be used to paginate results. Note: large offset values may cause performance issues. score_threshold: Define a minimal score threshold for the result. If defined, less similar results will not be returned. Score of the returned result might be higher or smaller than the threshold depending on the Distance function used. E.g. for cosine similarity only higher scores will be returned. consistency: Read consistency of the search. Defines how many replicas should be queried before returning the result. Values: - int - number of replicas to query, values should present in all queried replicas - 'majority' - query all replicas, but return values present in the majority of replicas - 'quorum' - query the majority of replicas, return values present in all of them - 'all' - query all replicas, and return values present in all replicas Returns: List of documents most similar to the query text and cosine distance in float for each. Lower score represents more similarity. """ if filter is not None and isinstance(filter, dict): warnings.warn( "Using dict as a `filter` is deprecated. Please use qdrant-client " "filters directly: " "https://qdrant.tech/documentation/concepts/filtering/", DeprecationWarning, ) qdrant_filter = self._qdrant_filter_from_dict(filter) else: qdrant_filter = filter results = self.client.search( collection_name=self.collection_name, query_vector=embedding, query_filter=qdrant_filter, search_params=search_params, limit=k, offset=offset, with_payload=True, with_vectors=False, # Langchain does not expect vectors to be returned score_threshold=score_threshold, consistency=consistency, **kwargs, ) return [ ( self._document_from_scored_point( result, self.content_payload_key, self.metadata_payload_key ), result.score, ) for result in results ] def _similarity_search_with_relevance_scores( self, query: str, k: int = 4, **kwargs: Any, ) -> List[Tuple[Document, float]]: """Return docs and relevance scores in the range [0, 1]. 0 is dissimilar, 1 is most similar. Args: query: input text k: Number of Documents to return. Defaults to 4. **kwargs: kwargs to be passed to similarity search. Should include: score_threshold: Optional, a floating point value between 0 to 1 to filter the resulting set of retrieved docs Returns: List of Tuples of (doc, similarity_score) """ return self.similarity_search_with_score(query, k, **kwargs) [docs] def max_marginal_relevance_search( self, query: str, k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. fetch_k: Number of Documents to fetch to pass to MMR algorithm. Defaults to 20. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. Returns: List of Documents selected by maximal marginal relevance. """ embedding = self._embed_query(query) results = self.client.search( collection_name=self.collection_name, query_vector=embedding, with_payload=True, with_vectors=True, limit=fetch_k, ) embeddings = [result.vector for result in results] mmr_selected = maximal_marginal_relevance( np.array(embedding), embeddings, k=k, lambda_mult=lambda_mult ) return [ self._document_from_scored_point( results[i], self.content_payload_key, self.metadata_payload_key ) for i in mmr_selected ] [docs] @classmethod def from_texts( cls: Type[Qdrant], texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, ids: Optional[Sequence[str]] = None, location: Optional[str] = None, url: Optional[str] = None, port: Optional[int] = 6333, grpc_port: int = 6334, prefer_grpc: bool = False, https: Optional[bool] = None, api_key: Optional[str] = None, prefix: Optional[str] = None, timeout: Optional[float] = None, host: Optional[str] = None, path: Optional[str] = None, collection_name: Optional[str] = None, distance_func: str = "Cosine", content_payload_key: str = CONTENT_KEY, metadata_payload_key: str = METADATA_KEY, batch_size: int = 64, shard_number: Optional[int] = None, replication_factor: Optional[int] = None, write_consistency_factor: Optional[int] = None, on_disk_payload: Optional[bool] = None, hnsw_config: Optional[common_types.HnswConfigDiff] = None, optimizers_config: Optional[common_types.OptimizersConfigDiff] = None, wal_config: Optional[common_types.WalConfigDiff] = None, quantization_config: Optional[common_types.QuantizationConfig] = None, init_from: Optional[common_types.InitFrom] = None, **kwargs: Any, ) -> Qdrant: """Construct Qdrant wrapper from a list of texts. Args: texts: A list of texts to be indexed in Qdrant. embedding: A subclass of `Embeddings`, responsible for text vectorization. metadatas: An optional list of metadata. If provided it has to be of the same length as a list of texts. ids: Optional list of ids to associate with the texts. Ids have to be uuid-like strings. location: If `:memory:` - use in-memory Qdrant instance. If `str` - use it as a `url` parameter. If `None` - fallback to relying on `host` and `port` parameters. url: either host or str of "Optional[scheme], host, Optional[port], Optional[prefix]". Default: `None` port: Port of the REST API interface. Default: 6333 grpc_port: Port of the gRPC interface. Default: 6334 prefer_grpc: If true - use gPRC interface whenever possible in custom methods. Default: False https: If true - use HTTPS(SSL) protocol. Default: None api_key: API key for authentication in Qdrant Cloud. Default: None prefix: If not None - add prefix to the REST URL path. Example: service/v1 will result in http://localhost:6333/service/v1/{qdrant-endpoint} for REST API. Default: None timeout: Timeout for REST and gRPC API requests. Default: 5.0 seconds for REST and unlimited for gRPC host: Host name of Qdrant service. If url and host are None, set to 'localhost'. Default: None path: Path in which the vectors will be stored while using local mode. Default: None collection_name: Name of the Qdrant collection to be used. If not provided, it will be created randomly. Default: None distance_func: Distance function. One of: "Cosine" / "Euclid" / "Dot". Default: "Cosine" content_payload_key: A payload key used to store the content of the document. Default: "page_content" metadata_payload_key: A payload key used to store the metadata of the document. Default: "metadata" batch_size: How many vectors upload per-request. Default: 64 shard_number: Number of shards in collection. Default is 1, minimum is 1. replication_factor: Replication factor for collection. Default is 1, minimum is 1. Defines how many copies of each shard will be created. Have effect only in distributed mode. write_consistency_factor: Write consistency factor for collection. Default is 1, minimum is 1. Defines how many replicas should apply the operation for us to consider it successful. Increasing this number will make the collection more resilient to inconsistencies, but will also make it fail if not enough replicas are available. Does not have any performance impact. Have effect only in distributed mode. on_disk_payload: If true - point`s payload will not be stored in memory. It will be read from the disk every time it is requested. This setting saves RAM by (slightly) increasing the response time. Note: those payload values that are involved in filtering and are indexed - remain in RAM. hnsw_config: Params for HNSW index optimizers_config: Params for optimizer wal_config: Params for Write-Ahead-Log quantization_config: Params for quantization, if None - quantization will be disabled init_from: Use data stored in another collection to initialize this collection **kwargs: Additional arguments passed directly into REST client initialization This is a user-friendly interface that: 1. Creates embeddings, one for each text 2. Initializes the Qdrant database as an in-memory docstore by default (and overridable to a remote docstore) 3. Adds the text embeddings to the Qdrant database This is intended to be a quick way to get started. Example: .. code-block:: python from langchain import Qdrant from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() qdrant = Qdrant.from_texts(texts, embeddings, "localhost") """ try: import qdrant_client except ImportError: raise ValueError( "Could not import qdrant-client python package. " "Please install it with `pip install qdrant-client`." ) from qdrant_client.http import models as rest # Just do a single quick embedding to get vector size partial_embeddings = embedding.embed_documents(texts[:1]) vector_size = len(partial_embeddings[0]) collection_name = collection_name or uuid.uuid4().hex distance_func = distance_func.upper() client = qdrant_client.QdrantClient( location=location, url=url, port=port, grpc_port=grpc_port, prefer_grpc=prefer_grpc, https=https, api_key=api_key, prefix=prefix, timeout=timeout, host=host, path=path, **kwargs, ) client.recreate_collection( collection_name=collection_name, vectors_config=rest.VectorParams( size=vector_size, distance=rest.Distance[distance_func], ), shard_number=shard_number, replication_factor=replication_factor, write_consistency_factor=write_consistency_factor, on_disk_payload=on_disk_payload, hnsw_config=hnsw_config, optimizers_config=optimizers_config, wal_config=wal_config, quantization_config=quantization_config, init_from=init_from, timeout=timeout, # type: ignore[arg-type] ) texts_iterator = iter(texts) metadatas_iterator = iter(metadatas or []) ids_iterator = iter(ids or [uuid.uuid4().hex for _ in iter(texts)]) while batch_texts := list(islice(texts_iterator, batch_size)): # Take the corresponding metadata and id for each text in a batch batch_metadatas = list(islice(metadatas_iterator, batch_size)) or None batch_ids = list(islice(ids_iterator, batch_size)) # Generate the embeddings for all the texts in a batch batch_embeddings = embedding.embed_documents(batch_texts) client.upsert( collection_name=collection_name, points=rest.Batch.construct( ids=batch_ids, vectors=batch_embeddings, payloads=cls._build_payloads( batch_texts, batch_metadatas, content_payload_key, metadata_payload_key, ), ), ) return cls( client=client, collection_name=collection_name, embeddings=embedding, content_payload_key=content_payload_key, metadata_payload_key=metadata_payload_key, ) @classmethod def _build_payloads( cls, texts: Iterable[str], metadatas: Optional[List[dict]], content_payload_key: str, metadata_payload_key: str, ) -> List[dict]: payloads = [] for i, text in enumerate(texts): if text is None: raise ValueError( "At least one of the texts is None. Please remove it before " "calling .from_texts or .add_texts on Qdrant instance." ) metadata = metadatas[i] if metadatas is not None else None payloads.append( { content_payload_key: text, metadata_payload_key: metadata, } ) return payloads @classmethod def _document_from_scored_point( cls, scored_point: Any, content_payload_key: str, metadata_payload_key: str, ) -> Document: return Document( page_content=scored_point.payload.get(content_payload_key), metadata=scored_point.payload.get(metadata_payload_key) or {}, ) def _build_condition(self, key: str, value: Any) -> List[rest.FieldCondition]: from qdrant_client.http import models as rest out = [] if isinstance(value, dict): for _key, value in value.items(): out.extend(self._build_condition(f"{key}.{_key}", value)) elif isinstance(value, list): for _value in value: if isinstance(_value, dict): out.extend(self._build_condition(f"{key}[]", _value)) else: out.extend(self._build_condition(f"{key}", _value)) else: out.append( rest.FieldCondition( key=f"{self.metadata_payload_key}.{key}", match=rest.MatchValue(value=value), ) ) return out def _qdrant_filter_from_dict( self, filter: Optional[DictFilter] ) -> Optional[rest.Filter]: from qdrant_client.http import models as rest if not filter: return None return rest.Filter( must=[ condition for key, value in filter.items() for condition in self._build_condition(key, value) ] ) def _embed_query(self, query: str) -> List[float]: """Embed query text. Used to provide backward compatibility with `embedding_function` argument. Args: query: Query text. Returns: List of floats representing the query embedding. """ if self.embeddings is not None: embedding = self.embeddings.embed_query(query) else: if self._embeddings_function is not None: embedding = self._embeddings_function(query) else: raise ValueError("Neither of embeddings or embedding_function is set") return embedding.tolist() if hasattr(embedding, "tolist") else embedding def _embed_texts(self, texts: Iterable[str]) -> List[List[float]]: """Embed search texts. Used to provide backward compatibility with `embedding_function` argument. Args: texts: Iterable of texts to embed. Returns: List of floats representing the texts embedding. """ if self.embeddings is not None: embeddings = self.embeddings.embed_documents(list(texts)) if hasattr(embeddings, "tolist"): embeddings = embeddings.tolist() elif self._embeddings_function is not None: embeddings = [] for text in texts: embedding = self._embeddings_function(text) if hasattr(embeddings, "tolist"): embedding = embedding.tolist() embeddings.append(embedding) else: raise ValueError("Neither of embeddings or embedding_function is set") return embeddings
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/qdrant.html
406b321a-f513-49f7-90b6-c76f3a498f72
Source code for langchain.vectorstores.azuresearch """Wrapper around Azure Cognitive Search.""" from __future__ import annotations import base64 import json import logging import uuid from typing import ( TYPE_CHECKING, Any, Callable, Dict, Iterable, List, Optional, Tuple, Type, ) import numpy as np from pydantic import BaseModel, root_validator from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.schema import BaseRetriever from langchain.utils import get_from_env from langchain.vectorstores.base import VectorStore logger = logging.getLogger() if TYPE_CHECKING: from azure.search.documents import SearchClient # Allow overriding field names for Azure Search FIELDS_ID = get_from_env( key="AZURESEARCH_FIELDS_ID", env_key="AZURESEARCH_FIELDS_ID", default="id" ) FIELDS_CONTENT = get_from_env( key="AZURESEARCH_FIELDS_CONTENT", env_key="AZURESEARCH_FIELDS_CONTENT", default="content", ) FIELDS_CONTENT_VECTOR = get_from_env( key="AZURESEARCH_FIELDS_CONTENT_VECTOR", env_key="AZURESEARCH_FIELDS_CONTENT_VECTOR", default="content_vector", ) FIELDS_METADATA = get_from_env( key="AZURESEARCH_FIELDS_TAG", env_key="AZURESEARCH_FIELDS_TAG", default="metadata" ) MAX_UPLOAD_BATCH_SIZE = 1000 def _get_search_client( endpoint: str, key: str, index_name: str, embedding_function: Callable, semantic_configuration_name: Optional[str] = None, ) -> SearchClient: from azure.core.credentials import AzureKeyCredential from azure.core.exceptions import ResourceNotFoundError from azure.identity import DefaultAzureCredential from azure.search.documents import SearchClient from azure.search.documents.indexes import SearchIndexClient from azure.search.documents.indexes.models import ( PrioritizedFields, SearchableField, SearchField, SearchFieldDataType, SearchIndex, SemanticConfiguration, SemanticField, SemanticSettings, SimpleField, VectorSearch, VectorSearchAlgorithmConfiguration, ) if key is None: credential = DefaultAzureCredential() else: credential = AzureKeyCredential(key) index_client: SearchIndexClient = SearchIndexClient( endpoint=endpoint, credential=credential ) try: index_client.get_index(name=index_name) except ResourceNotFoundError: # Fields configuration fields = [ SimpleField( name=FIELDS_ID, type=SearchFieldDataType.String, key=True, filterable=True, ), SearchableField( name=FIELDS_CONTENT, type=SearchFieldDataType.String, searchable=True, retrievable=True, ), SearchField( name=FIELDS_CONTENT_VECTOR, type=SearchFieldDataType.Collection(SearchFieldDataType.Single), searchable=True, dimensions=len(embedding_function("Text")), vector_search_configuration="default", ), SearchableField( name=FIELDS_METADATA, type=SearchFieldDataType.String, searchable=True, retrievable=True, ), ] # Vector search configuration vector_search = VectorSearch( algorithm_configurations=[ VectorSearchAlgorithmConfiguration( name="default", kind="hnsw", hnsw_parameters={ "m": 4, "efConstruction": 400, "efSearch": 500, "metric": "cosine", }, ) ] ) # Create the semantic settings with the configuration semantic_settings = ( None if semantic_configuration_name is None else SemanticSettings( configurations=[ SemanticConfiguration( name=semantic_configuration_name, prioritized_fields=PrioritizedFields( prioritized_content_fields=[ SemanticField(field_name=FIELDS_CONTENT) ], ), ) ] ) ) # Create the search index with the semantic settings and vector search index = SearchIndex( name=index_name, fields=fields, vector_search=vector_search, semantic_settings=semantic_settings, ) index_client.create_index(index) # Create the search client return SearchClient(endpoint=endpoint, index_name=index_name, credential=credential) [docs]class AzureSearch(VectorStore): def __init__( self, azure_search_endpoint: str, azure_search_key: str, index_name: str, embedding_function: Callable, search_type: str = "hybrid", semantic_configuration_name: Optional[str] = None, semantic_query_language: str = "en-us", **kwargs: Any, ): """Initialize with necessary components.""" # Initialize base class self.embedding_function = embedding_function self.client = _get_search_client( azure_search_endpoint, azure_search_key, index_name, embedding_function, semantic_configuration_name, ) self.search_type = search_type self.semantic_configuration_name = semantic_configuration_name self.semantic_query_language = semantic_query_language [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, **kwargs: Any, ) -> List[str]: """Add texts data to an existing index.""" keys = kwargs.get("keys") ids = [] # Write data to index data = [] for i, text in enumerate(texts): # Use provided key otherwise use default key key = keys[i] if keys else str(uuid.uuid4()) # Encoding key for Azure Search valid characters key = base64.urlsafe_b64encode(bytes(key, "utf-8")).decode("ascii") metadata = metadatas[i] if metadatas else {} # Add data to index data.append( { "@search.action": "upload", FIELDS_ID: key, FIELDS_CONTENT: text, FIELDS_CONTENT_VECTOR: np.array( self.embedding_function(text), dtype=np.float32 ).tolist(), FIELDS_METADATA: json.dumps(metadata), } ) ids.append(key) # Upload data in batches if len(data) == MAX_UPLOAD_BATCH_SIZE: response = self.client.upload_documents(documents=data) # Check if all documents were successfully uploaded if not all([r.succeeded for r in response]): raise Exception(response) # Reset data data = [] # Considering case where data is an exact multiple of batch-size entries if len(data) == 0: return ids # Upload data to index response = self.client.upload_documents(documents=data) # Check if all documents were successfully uploaded if all([r.succeeded for r in response]): return ids else: raise Exception(response) [docs] def similarity_search( self, query: str, k: int = 4, **kwargs: Any ) -> List[Document]: search_type = kwargs.get("search_type", self.search_type) if search_type == "similarity": docs = self.vector_search(query, k=k, **kwargs) elif search_type == "hybrid": docs = self.hybrid_search(query, k=k, **kwargs) elif search_type == "semantic_hybrid": docs = self.semantic_hybrid_search(query, k=k, **kwargs) else: raise ValueError(f"search_type of {search_type} not allowed.") return docs [docs] def vector_search(self, query: str, k: int = 4, **kwargs: Any) -> List[Document]: """ Returns the most similar indexed documents to the query text. Args: query (str): The query text for which to find similar documents. k (int): The number of documents to return. Default is 4. Returns: List[Document]: A list of documents that are most similar to the query text. """ docs_and_scores = self.vector_search_with_score( query, k=k, filters=kwargs.get("filters", None) ) return [doc for doc, _ in docs_and_scores] [docs] def vector_search_with_score( self, query: str, k: int = 4, filters: Optional[str] = None ) -> List[Tuple[Document, float]]: """Return docs most similar to query. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. Returns: List of Documents most similar to the query and score for each """ from azure.search.documents.models import Vector results = self.client.search( search_text="", vector=Vector( value=np.array( self.embedding_function(query), dtype=np.float32 ).tolist(), k=k, fields=FIELDS_CONTENT_VECTOR, ), select=[f"{FIELDS_ID},{FIELDS_CONTENT},{FIELDS_METADATA}"], filter=filters, ) # Convert results to Document objects docs = [ ( Document( page_content=result[FIELDS_CONTENT], metadata=json.loads(result[FIELDS_METADATA]), ), float(result["@search.score"]), ) for result in results ] return docs [docs] def hybrid_search(self, query: str, k: int = 4, **kwargs: Any) -> List[Document]: """ Returns the most similar indexed documents to the query text. Args: query (str): The query text for which to find similar documents. k (int): The number of documents to return. Default is 4. Returns: List[Document]: A list of documents that are most similar to the query text. """ docs_and_scores = self.hybrid_search_with_score( query, k=k, filters=kwargs.get("filters", None) ) return [doc for doc, _ in docs_and_scores] [docs] def hybrid_search_with_score( self, query: str, k: int = 4, filters: Optional[str] = None ) -> List[Tuple[Document, float]]: """Return docs most similar to query with an hybrid query. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. Returns: List of Documents most similar to the query and score for each """ from azure.search.documents.models import Vector results = self.client.search( search_text=query, vector=Vector( value=np.array( self.embedding_function(query), dtype=np.float32 ).tolist(), k=k, fields=FIELDS_CONTENT_VECTOR, ), select=[f"{FIELDS_ID},{FIELDS_CONTENT},{FIELDS_METADATA}"], filter=filters, top=k, ) # Convert results to Document objects docs = [ ( Document( page_content=result[FIELDS_CONTENT], metadata=json.loads(result[FIELDS_METADATA]), ), float(result["@search.score"]), ) for result in results ] return docs [docs] def semantic_hybrid_search( self, query: str, k: int = 4, **kwargs: Any ) -> List[Document]: """ Returns the most similar indexed documents to the query text. Args: query (str): The query text for which to find similar documents. k (int): The number of documents to return. Default is 4. Returns: List[Document]: A list of documents that are most similar to the query text. """ docs_and_scores = self.semantic_hybrid_search_with_score( query, k=k, filters=kwargs.get("filters", None) ) return [doc for doc, _ in docs_and_scores] [docs] def semantic_hybrid_search_with_score( self, query: str, k: int = 4, filters: Optional[str] = None ) -> List[Tuple[Document, float]]: """Return docs most similar to query with an hybrid query. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. Returns: List of Documents most similar to the query and score for each """ from azure.search.documents.models import Vector results = self.client.search( search_text=query, vector=Vector( value=np.array( self.embedding_function(query), dtype=np.float32 ).tolist(), k=50, # Hardcoded value to maximize L2 retrieval fields=FIELDS_CONTENT_VECTOR, ), select=[f"{FIELDS_ID},{FIELDS_CONTENT},{FIELDS_METADATA}"], filter=filters, query_type="semantic", query_language=self.semantic_query_language, semantic_configuration_name=self.semantic_configuration_name, query_caption="extractive", query_answer="extractive", top=k, ) # Get Semantic Answers semantic_answers = results.get_answers() semantic_answers_dict = {} for semantic_answer in semantic_answers: semantic_answers_dict[semantic_answer.key] = { "text": semantic_answer.text, "highlights": semantic_answer.highlights, } # Convert results to Document objects docs = [ ( Document( page_content=result["content"], metadata={ **json.loads(result["metadata"]), **{ "captions": { "text": result.get("@search.captions", [{}])[0].text, "highlights": result.get("@search.captions", [{}])[ 0 ].highlights, } if result.get("@search.captions") else {}, "answers": semantic_answers_dict.get( json.loads(result["metadata"]).get("key"), "" ), }, }, ), float(result["@search.score"]), ) for result in results ] return docs [docs] @classmethod def from_texts( cls: Type[AzureSearch], texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, azure_search_endpoint: str = "", azure_search_key: str = "", index_name: str = "langchain-index", **kwargs: Any, ) -> AzureSearch: # Creating a new Azure Search instance azure_search = cls( azure_search_endpoint, azure_search_key, index_name, embedding.embed_query, ) azure_search.add_texts(texts, metadatas, **kwargs) return azure_search class AzureSearchVectorStoreRetriever(BaseRetriever, BaseModel): vectorstore: AzureSearch search_type: str = "hybrid" k: int = 4 class Config: """Configuration for this pydantic object.""" arbitrary_types_allowed = True @root_validator() def validate_search_type(cls, values: Dict) -> Dict: """Validate search type.""" if "search_type" in values: search_type = values["search_type"] if search_type not in ("similarity", "hybrid", "semantic_hybrid"): raise ValueError(f"search_type of {search_type} not allowed.") return values def get_relevant_documents(self, query: str) -> List[Document]: if self.search_type == "similarity": docs = self.vectorstore.vector_search(query, k=self.k) elif self.search_type == "hybrid": docs = self.vectorstore.hybrid_search(query, k=self.k) elif self.search_type == "semantic_hybrid": docs = self.vectorstore.semantic_hybrid_search(query, k=self.k) else: raise ValueError(f"search_type of {self.search_type} not allowed.") return docs async def aget_relevant_documents(self, query: str) -> List[Document]: raise NotImplementedError( "AzureSearchVectorStoreRetriever does not support async" )
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/azuresearch.html
ccb07509-aa60-40dd-978e-b0177d7a7122
Source code for langchain.vectorstores.cassandra """Wrapper around Cassandra vector-store capabilities, based on cassIO.""" from __future__ import annotations import hashlib import typing from typing import Any, Iterable, List, Optional, Tuple, Type, TypeVar import numpy as np if typing.TYPE_CHECKING: from cassandra.cluster import Session from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.vectorstores.base import VectorStore from langchain.vectorstores.utils import maximal_marginal_relevance CVST = TypeVar("CVST", bound="Cassandra") # a positive number of seconds to expire entries, or None for no expiration. CASSANDRA_VECTORSTORE_DEFAULT_TTL_SECONDS = None def _hash(_input: str) -> str: """Use a deterministic hashing approach.""" return hashlib.md5(_input.encode()).hexdigest() [docs]class Cassandra(VectorStore): """Wrapper around Cassandra embeddings platform. There is no notion of a default table name, since each embedding function implies its own vector dimension, which is part of the schema. Example: .. code-block:: python from langchain.vectorstores import Cassandra from langchain.embeddings.openai import OpenAIEmbeddings embeddings = OpenAIEmbeddings() session = ... keyspace = 'my_keyspace' vectorstore = Cassandra(embeddings, session, keyspace, 'my_doc_archive') """ _embedding_dimension: int | None def _getEmbeddingDimension(self) -> int: if self._embedding_dimension is None: self._embedding_dimension = len( self.embedding.embed_query("This is a sample sentence.") ) return self._embedding_dimension def __init__( self, embedding: Embeddings, session: Session, keyspace: str, table_name: str, ttl_seconds: int | None = CASSANDRA_VECTORSTORE_DEFAULT_TTL_SECONDS, ) -> None: try: from cassio.vector import VectorTable except (ImportError, ModuleNotFoundError): raise ImportError( "Could not import cassio python package. " "Please install it with `pip install cassio`." ) """Create a vector table.""" self.embedding = embedding self.session = session self.keyspace = keyspace self.table_name = table_name self.ttl_seconds = ttl_seconds # self._embedding_dimension = None # self.table = VectorTable( session=session, keyspace=keyspace, table=table_name, embedding_dimension=self._getEmbeddingDimension(), auto_id=False, # the `add_texts` contract admits user-provided ids ) [docs] def delete_collection(self) -> None: """ Just an alias for `clear` (to better align with other VectorStore implementations). """ self.clear() [docs] def clear(self) -> None: """Empty the collection.""" self.table.clear() [docs] def delete_by_document_id(self, document_id: str) -> None: return self.table.delete(document_id) [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, **kwargs: Any, ) -> List[str]: """Run more texts through the embeddings and add to the vectorstore. Args: texts (Iterable[str]): Texts to add to the vectorstore. metadatas (Optional[List[dict]], optional): Optional list of metadatas. ids (Optional[List[str]], optional): Optional list of IDs. Returns: List[str]: List of IDs of the added texts. """ _texts = list(texts) # lest it be a generator or something if ids is None: # unless otherwise specified, we have deterministic IDs: # re-inserting an existing document will not create a duplicate. # (and effectively update the metadata) ids = [_hash(text) for text in _texts] if metadatas is None: metadatas = [{} for _ in _texts] # ttl_seconds = kwargs.get("ttl_seconds", self.ttl_seconds) # embedding_vectors = self.embedding.embed_documents(_texts) for text, embedding_vector, text_id, metadata in zip( _texts, embedding_vectors, ids, metadatas ): self.table.put( document=text, embedding_vector=embedding_vector, document_id=text_id, metadata=metadata, ttl_seconds=ttl_seconds, ) # return ids # id-returning search facilities [docs] def similarity_search_with_score_id_by_vector( self, embedding: List[float], k: int = 4, ) -> List[Tuple[Document, float, str]]: """Return docs most similar to embedding vector. No support for `filter` query (on metadata) along with vector search. Args: embedding (str): Embedding to look up documents similar to. k (int): Number of Documents to return. Defaults to 4. Returns: List of (Document, score, id), the most similar to the query vector. """ hits = self.table.search( embedding_vector=embedding, top_k=k, metric="cos", metric_threshold=None, ) # We stick to 'cos' distance as it can be normalized on a 0-1 axis # (1=most relevant), as required by this class' contract. return [ ( Document( page_content=hit["document"], metadata=hit["metadata"], ), 0.5 + 0.5 * hit["distance"], hit["document_id"], ) for hit in hits ] [docs] def similarity_search_with_score_id( self, query: str, k: int = 4, **kwargs: Any, ) -> List[Tuple[Document, float, str]]: embedding_vector = self.embedding.embed_query(query) return self.similarity_search_with_score_id_by_vector( embedding=embedding_vector, k=k, ) # id-unaware search facilities [docs] def similarity_search_with_score_by_vector( self, embedding: List[float], k: int = 4, ) -> List[Tuple[Document, float]]: """Return docs most similar to embedding vector. No support for `filter` query (on metadata) along with vector search. Args: embedding (str): Embedding to look up documents similar to. k (int): Number of Documents to return. Defaults to 4. Returns: List of (Document, score), the most similar to the query vector. """ return [ (doc, score) for (doc, score, docId) in self.similarity_search_with_score_id_by_vector( embedding=embedding, k=k, ) ] [docs] def similarity_search( self, query: str, k: int = 4, **kwargs: Any, ) -> List[Document]: # embedding_vector = self.embedding.embed_query(query) return self.similarity_search_by_vector( embedding_vector, k, **kwargs, ) [docs] def similarity_search_by_vector( self, embedding: List[float], k: int = 4, **kwargs: Any, ) -> List[Document]: return [ doc for doc, _ in self.similarity_search_with_score_by_vector( embedding, k, ) ] [docs] def similarity_search_with_score( self, query: str, k: int = 4, **kwargs: Any, ) -> List[Tuple[Document, float]]: embedding_vector = self.embedding.embed_query(query) return self.similarity_search_with_score_by_vector( embedding_vector, k, ) # Even though this is a `_`-method, # it is apparently used by VectorSearch parent class # in an exposed method (`similarity_search_with_relevance_scores`). # So we implement it (hmm). def _similarity_search_with_relevance_scores( self, query: str, k: int = 4, **kwargs: Any, ) -> List[Tuple[Document, float]]: return self.similarity_search_with_score( query, k, **kwargs, ) [docs] def max_marginal_relevance_search_by_vector( self, embedding: List[float], k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: embedding: Embedding to look up documents similar to. k: Number of Documents to return. fetch_k: Number of Documents to fetch to pass to MMR algorithm. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Returns: List of Documents selected by maximal marginal relevance. """ prefetchHits = self.table.search( embedding_vector=embedding, top_k=fetch_k, metric="cos", metric_threshold=None, ) # let the mmr utility pick the *indices* in the above array mmrChosenIndices = maximal_marginal_relevance( np.array(embedding, dtype=np.float32), [pfHit["embedding_vector"] for pfHit in prefetchHits], k=k, lambda_mult=lambda_mult, ) mmrHits = [ pfHit for pfIndex, pfHit in enumerate(prefetchHits) if pfIndex in mmrChosenIndices ] return [ Document( page_content=hit["document"], metadata=hit["metadata"], ) for hit in mmrHits ] [docs] def max_marginal_relevance_search( self, query: str, k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: query: Text to look up documents similar to. k: Number of Documents to return. fetch_k: Number of Documents to fetch to pass to MMR algorithm. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Optional. Returns: List of Documents selected by maximal marginal relevance. """ embedding_vector = self.embedding.embed_query(query) return self.max_marginal_relevance_search_by_vector( embedding_vector, k, fetch_k, lambda_mult=lambda_mult, ) [docs] @classmethod def from_texts( cls: Type[CVST], texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, **kwargs: Any, ) -> CVST: """Create a Cassandra vectorstore from raw texts. No support for specifying text IDs Returns: a Cassandra vectorstore. """ session: Session = kwargs["session"] keyspace: str = kwargs["keyspace"] table_name: str = kwargs["table_name"] cassandraStore = cls( embedding=embedding, session=session, keyspace=keyspace, table_name=table_name, ) cassandraStore.add_texts(texts=texts, metadatas=metadatas) return cassandraStore [docs] @classmethod def from_documents( cls: Type[CVST], documents: List[Document], embedding: Embeddings, **kwargs: Any, ) -> CVST: """Create a Cassandra vectorstore from a document list. No support for specifying text IDs Returns: a Cassandra vectorstore. """ texts = [doc.page_content for doc in documents] metadatas = [doc.metadata for doc in documents] session: Session = kwargs["session"] keyspace: str = kwargs["keyspace"] table_name: str = kwargs["table_name"] return cls.from_texts( texts=texts, metadatas=metadatas, embedding=embedding, session=session, keyspace=keyspace, table_name=table_name, )
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/cassandra.html
cd6529f4-b3f5-4c28-8a1a-488d935a11f6
Source code for langchain.vectorstores.lancedb """Wrapper around LanceDB vector database""" from __future__ import annotations import uuid from typing import Any, Iterable, List, Optional from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.vectorstores.base import VectorStore [docs]class LanceDB(VectorStore): """Wrapper around LanceDB vector database. To use, you should have ``lancedb`` python package installed. Example: .. code-block:: python db = lancedb.connect('./lancedb') table = db.open_table('my_table') vectorstore = LanceDB(table, embedding_function) vectorstore.add_texts(['text1', 'text2']) result = vectorstore.similarity_search('text1') """ def __init__( self, connection: Any, embedding: Embeddings, vector_key: Optional[str] = "vector", id_key: Optional[str] = "id", text_key: Optional[str] = "text", ): """Initialize with Lance DB connection""" try: import lancedb except ImportError: raise ValueError( "Could not import lancedb python package. " "Please install it with `pip install lancedb`." ) if not isinstance(connection, lancedb.db.LanceTable): raise ValueError( "connection should be an instance of lancedb.db.LanceTable, ", f"got {type(connection)}", ) self._connection = connection self._embedding = embedding self._vector_key = vector_key self._id_key = id_key self._text_key = text_key [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, **kwargs: Any, ) -> List[str]: """Turn texts into embedding and add it to the database Args: texts: Iterable of strings to add to the vectorstore. metadatas: Optional list of metadatas associated with the texts. ids: Optional list of ids to associate with the texts. Returns: List of ids of the added texts. """ # Embed texts and create documents docs = [] ids = ids or [str(uuid.uuid4()) for _ in texts] embeddings = self._embedding.embed_documents(list(texts)) for idx, text in enumerate(texts): embedding = embeddings[idx] metadata = metadatas[idx] if metadatas else {} docs.append( { self._vector_key: embedding, self._id_key: ids[idx], self._text_key: text, **metadata, } ) self._connection.add(docs) return ids [docs] def similarity_search( self, query: str, k: int = 4, **kwargs: Any ) -> List[Document]: """Return documents most similar to the query Args: query: String to query the vectorstore with. k: Number of documents to return. Returns: List of documents most similar to the query. """ embedding = self._embedding.embed_query(query) docs = self._connection.search(embedding).limit(k).to_df() return [ Document( page_content=row[self._text_key], metadata=row[docs.columns != self._text_key], ) for _, row in docs.iterrows() ] [docs] @classmethod def from_texts( cls, texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, connection: Any = None, vector_key: Optional[str] = "vector", id_key: Optional[str] = "id", text_key: Optional[str] = "text", **kwargs: Any, ) -> LanceDB: instance = LanceDB( connection, embedding, vector_key, id_key, text_key, ) instance.add_texts(texts, metadatas=metadatas, **kwargs) return instance
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/lancedb.html
36705cea-01dc-4cc4-94b5-df9781dee535
Source code for langchain.vectorstores.sklearn """ Wrapper around scikit-learn NearestNeighbors implementation. The vector store can be persisted in json, bson or parquet format. """ import json import math import os from abc import ABC, abstractmethod from typing import Any, Dict, Iterable, List, Literal, Optional, Tuple, Type from uuid import uuid4 from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.utils import guard_import from langchain.vectorstores.base import VectorStore from langchain.vectorstores.utils import maximal_marginal_relevance DEFAULT_K = 4 # Number of Documents to return. DEFAULT_FETCH_K = 20 # Number of Documents to initially fetch during MMR search. class BaseSerializer(ABC): """Abstract base class for saving and loading data.""" def __init__(self, persist_path: str) -> None: self.persist_path = persist_path @classmethod @abstractmethod def extension(cls) -> str: """The file extension suggested by this serializer (without dot).""" @abstractmethod def save(self, data: Any) -> None: """Saves the data to the persist_path""" @abstractmethod def load(self) -> Any: """Loads the data from the persist_path""" class JsonSerializer(BaseSerializer): """Serializes data in json using the json package from python standard library.""" @classmethod def extension(cls) -> str: return "json" def save(self, data: Any) -> None: with open(self.persist_path, "w") as fp: json.dump(data, fp) def load(self) -> Any: with open(self.persist_path, "r") as fp: return json.load(fp) class BsonSerializer(BaseSerializer): """Serializes data in binary json using the bson python package.""" def __init__(self, persist_path: str) -> None: super().__init__(persist_path) self.bson = guard_import("bson") @classmethod def extension(cls) -> str: return "bson" def save(self, data: Any) -> None: with open(self.persist_path, "wb") as fp: fp.write(self.bson.dumps(data)) def load(self) -> Any: with open(self.persist_path, "rb") as fp: return self.bson.loads(fp.read()) class ParquetSerializer(BaseSerializer): """Serializes data in Apache Parquet format using the pyarrow package.""" def __init__(self, persist_path: str) -> None: super().__init__(persist_path) self.pd = guard_import("pandas") self.pa = guard_import("pyarrow") self.pq = guard_import("pyarrow.parquet") @classmethod def extension(cls) -> str: return "parquet" def save(self, data: Any) -> None: df = self.pd.DataFrame(data) table = self.pa.Table.from_pandas(df) if os.path.exists(self.persist_path): backup_path = str(self.persist_path) + "-backup" os.rename(self.persist_path, backup_path) try: self.pq.write_table(table, self.persist_path) except Exception as exc: os.rename(backup_path, self.persist_path) raise exc else: os.remove(backup_path) else: self.pq.write_table(table, self.persist_path) def load(self) -> Any: table = self.pq.read_table(self.persist_path) df = table.to_pandas() return {col: series.tolist() for col, series in df.items()} SERIALIZER_MAP: Dict[str, Type[BaseSerializer]] = { "json": JsonSerializer, "bson": BsonSerializer, "parquet": ParquetSerializer, } class SKLearnVectorStoreException(RuntimeError): """Exception raised by SKLearnVectorStore.""" pass [docs]class SKLearnVectorStore(VectorStore): """A simple in-memory vector store based on the scikit-learn library NearestNeighbors implementation.""" def __init__( self, embedding: Embeddings, *, persist_path: Optional[str] = None, serializer: Literal["json", "bson", "parquet"] = "json", metric: str = "cosine", **kwargs: Any, ) -> None: np = guard_import("numpy") sklearn_neighbors = guard_import("sklearn.neighbors", pip_name="scikit-learn") # non-persistent properties self._np = np self._neighbors = sklearn_neighbors.NearestNeighbors(metric=metric, **kwargs) self._neighbors_fitted = False self._embedding_function = embedding self._persist_path = persist_path self._serializer: Optional[BaseSerializer] = None if self._persist_path is not None: serializer_cls = SERIALIZER_MAP[serializer] self._serializer = serializer_cls(persist_path=self._persist_path) # data properties self._embeddings: List[List[float]] = [] self._texts: List[str] = [] self._metadatas: List[dict] = [] self._ids: List[str] = [] # cache properties self._embeddings_np: Any = np.asarray([]) if self._persist_path is not None and os.path.isfile(self._persist_path): self._load() [docs] def persist(self) -> None: if self._serializer is None: raise SKLearnVectorStoreException( "You must specify a persist_path on creation to persist the " "collection." ) data = { "ids": self._ids, "texts": self._texts, "metadatas": self._metadatas, "embeddings": self._embeddings, } self._serializer.save(data) def _load(self) -> None: if self._serializer is None: raise SKLearnVectorStoreException( "You must specify a persist_path on creation to load the " "collection." ) data = self._serializer.load() self._embeddings = data["embeddings"] self._texts = data["texts"] self._metadatas = data["metadatas"] self._ids = data["ids"] self._update_neighbors() [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, **kwargs: Any, ) -> List[str]: _texts = list(texts) _ids = ids or [str(uuid4()) for _ in _texts] self._texts.extend(_texts) self._embeddings.extend(self._embedding_function.embed_documents(_texts)) self._metadatas.extend(metadatas or ([{}] * len(_texts))) self._ids.extend(_ids) self._update_neighbors() return _ids def _update_neighbors(self) -> None: if len(self._embeddings) == 0: raise SKLearnVectorStoreException( "No data was added to SKLearnVectorStore." ) self._embeddings_np = self._np.asarray(self._embeddings) self._neighbors.fit(self._embeddings_np) self._neighbors_fitted = True def _similarity_index_search_with_score( self, query_embedding: List[float], *, k: int = DEFAULT_K, **kwargs: Any ) -> List[Tuple[int, float]]: """Search k embeddings similar to the query embedding. Returns a list of (index, distance) tuples.""" if not self._neighbors_fitted: raise SKLearnVectorStoreException( "No data was added to SKLearnVectorStore." ) neigh_dists, neigh_idxs = self._neighbors.kneighbors( [query_embedding], n_neighbors=k ) return list(zip(neigh_idxs[0], neigh_dists[0])) [docs] def similarity_search_with_score( self, query: str, *, k: int = DEFAULT_K, **kwargs: Any ) -> List[Tuple[Document, float]]: query_embedding = self._embedding_function.embed_query(query) indices_dists = self._similarity_index_search_with_score( query_embedding, k=k, **kwargs ) return [ ( Document( page_content=self._texts[idx], metadata={"id": self._ids[idx], **self._metadatas[idx]}, ), dist, ) for idx, dist in indices_dists ] [docs] def similarity_search( self, query: str, k: int = DEFAULT_K, **kwargs: Any ) -> List[Document]: docs_scores = self.similarity_search_with_score(query, k=k, **kwargs) return [doc for doc, _ in docs_scores] def _similarity_search_with_relevance_scores( self, query: str, k: int = DEFAULT_K, **kwargs: Any ) -> List[Tuple[Document, float]]: docs_dists = self.similarity_search_with_score(query, k=k, **kwargs) docs, dists = zip(*docs_dists) scores = [1 / math.exp(dist) for dist in dists] return list(zip(list(docs), scores)) [docs] def max_marginal_relevance_search_by_vector( self, embedding: List[float], k: int = DEFAULT_K, fetch_k: int = DEFAULT_FETCH_K, lambda_mult: float = 0.5, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: embedding: Embedding to look up documents similar to. k: Number of Documents to return. Defaults to 4. fetch_k: Number of Documents to fetch to pass to MMR algorithm. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. Returns: List of Documents selected by maximal marginal relevance. """ indices_dists = self._similarity_index_search_with_score( embedding, k=fetch_k, **kwargs ) indices, _ = zip(*indices_dists) result_embeddings = self._embeddings_np[indices,] mmr_selected = maximal_marginal_relevance( self._np.array(embedding, dtype=self._np.float32), result_embeddings, k=k, lambda_mult=lambda_mult, ) mmr_indices = [indices[i] for i in mmr_selected] return [ Document( page_content=self._texts[idx], metadata={"id": self._ids[idx], **self._metadatas[idx]}, ) for idx in mmr_indices ] [docs] def max_marginal_relevance_search( self, query: str, k: int = DEFAULT_K, fetch_k: int = DEFAULT_FETCH_K, lambda_mult: float = 0.5, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. fetch_k: Number of Documents to fetch to pass to MMR algorithm. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. Returns: List of Documents selected by maximal marginal relevance. """ if self._embedding_function is None: raise ValueError( "For MMR search, you must specify an embedding function on creation." ) embedding = self._embedding_function.embed_query(query) docs = self.max_marginal_relevance_search_by_vector( embedding, k, fetch_k, lambda_mul=lambda_mult ) return docs [docs] @classmethod def from_texts( cls, texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, persist_path: Optional[str] = None, **kwargs: Any, ) -> "SKLearnVectorStore": vs = SKLearnVectorStore(embedding, persist_path=persist_path, **kwargs) vs.add_texts(texts, metadatas=metadatas, ids=ids) return vs
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/sklearn.html
dc38a84a-1164-4ee2-bc0e-75f3d25ef9b9
Source code for langchain.vectorstores.analyticdb """VectorStore wrapper around a Postgres/PGVector database.""" from __future__ import annotations import logging import uuid from typing import Any, Dict, Iterable, List, Optional, Sequence, Tuple, Type from sqlalchemy import REAL, Column, String, Table, create_engine, insert, text from sqlalchemy.dialects.postgresql import ARRAY, JSON, TEXT from sqlalchemy.engine import Row try: from sqlalchemy.orm import declarative_base except ImportError: from sqlalchemy.ext.declarative import declarative_base from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.utils import get_from_dict_or_env from langchain.vectorstores.base import VectorStore _LANGCHAIN_DEFAULT_EMBEDDING_DIM = 1536 _LANGCHAIN_DEFAULT_COLLECTION_NAME = "langchain_document" Base = declarative_base() # type: Any [docs]class AnalyticDB(VectorStore): """VectorStore implementation using AnalyticDB. AnalyticDB is a distributed full PostgresSQL syntax cloud-native database. - `connection_string` is a postgres connection string. - `embedding_function` any embedding function implementing `langchain.embeddings.base.Embeddings` interface. - `collection_name` is the name of the collection to use. (default: langchain) - NOTE: This is not the name of the table, but the name of the collection. The tables will be created when initializing the store (if not exists) So, make sure the user has the right permissions to create tables. - `pre_delete_collection` if True, will delete the collection if it exists. (default: False) - Useful for testing. """ def __init__( self, connection_string: str, embedding_function: Embeddings, embedding_dimension: int = _LANGCHAIN_DEFAULT_EMBEDDING_DIM, collection_name: str = _LANGCHAIN_DEFAULT_COLLECTION_NAME, pre_delete_collection: bool = False, logger: Optional[logging.Logger] = None, ) -> None: self.connection_string = connection_string self.embedding_function = embedding_function self.embedding_dimension = embedding_dimension self.collection_name = collection_name self.pre_delete_collection = pre_delete_collection self.logger = logger or logging.getLogger(__name__) self.__post_init__() def __post_init__( self, ) -> None: """ Initialize the store. """ self.engine = create_engine(self.connection_string) self.create_collection() [docs] def create_table_if_not_exists(self) -> None: # Define the dynamic table Table( self.collection_name, Base.metadata, Column("id", TEXT, primary_key=True, default=uuid.uuid4), Column("embedding", ARRAY(REAL)), Column("document", String, nullable=True), Column("metadata", JSON, nullable=True), extend_existing=True, ) with self.engine.connect() as conn: with conn.begin(): # Create the table Base.metadata.create_all(conn) # Check if the index exists index_name = f"{self.collection_name}_embedding_idx" index_query = text( f""" SELECT 1 FROM pg_indexes WHERE indexname = '{index_name}'; """ ) result = conn.execute(index_query).scalar() # Create the index if it doesn't exist if not result: index_statement = text( f""" CREATE INDEX {index_name} ON {self.collection_name} USING ann(embedding) WITH ( "dim" = {self.embedding_dimension}, "hnsw_m" = 100 ); """ ) conn.execute(index_statement) [docs] def create_collection(self) -> None: if self.pre_delete_collection: self.delete_collection() self.create_table_if_not_exists() [docs] def delete_collection(self) -> None: self.logger.debug("Trying to delete collection") drop_statement = text(f"DROP TABLE IF EXISTS {self.collection_name};") with self.engine.connect() as conn: with conn.begin(): conn.execute(drop_statement) [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, batch_size: int = 500, **kwargs: Any, ) -> List[str]: """Run more texts through the embeddings and add to the vectorstore. Args: texts: Iterable of strings to add to the vectorstore. metadatas: Optional list of metadatas associated with the texts. kwargs: vectorstore specific parameters Returns: List of ids from adding the texts into the vectorstore. """ if ids is None: ids = [str(uuid.uuid1()) for _ in texts] embeddings = self.embedding_function.embed_documents(list(texts)) if not metadatas: metadatas = [{} for _ in texts] # Define the table schema chunks_table = Table( self.collection_name, Base.metadata, Column("id", TEXT, primary_key=True), Column("embedding", ARRAY(REAL)), Column("document", String, nullable=True), Column("metadata", JSON, nullable=True), extend_existing=True, ) chunks_table_data = [] with self.engine.connect() as conn: with conn.begin(): for document, metadata, chunk_id, embedding in zip( texts, metadatas, ids, embeddings ): chunks_table_data.append( { "id": chunk_id, "embedding": embedding, "document": document, "metadata": metadata, } ) # Execute the batch insert when the batch size is reached if len(chunks_table_data) == batch_size: conn.execute(insert(chunks_table).values(chunks_table_data)) # Clear the chunks_table_data list for the next batch chunks_table_data.clear() # Insert any remaining records that didn't make up a full batch if chunks_table_data: conn.execute(insert(chunks_table).values(chunks_table_data)) return ids [docs] def similarity_search( self, query: str, k: int = 4, filter: Optional[dict] = None, **kwargs: Any, ) -> List[Document]: """Run similarity search with AnalyticDB with distance. Args: query (str): Query text to search for. k (int): Number of results to return. Defaults to 4. filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. Returns: List of Documents most similar to the query. """ embedding = self.embedding_function.embed_query(text=query) return self.similarity_search_by_vector( embedding=embedding, k=k, filter=filter, ) [docs] def similarity_search_with_score( self, query: str, k: int = 4, filter: Optional[dict] = None, ) -> List[Tuple[Document, float]]: """Return docs most similar to query. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. Returns: List of Documents most similar to the query and score for each """ embedding = self.embedding_function.embed_query(query) docs = self.similarity_search_with_score_by_vector( embedding=embedding, k=k, filter=filter ) return docs def _similarity_search_with_relevance_scores( self, query: str, k: int = 4, **kwargs: Any, ) -> List[Tuple[Document, float]]: """Return docs and relevance scores in the range [0, 1]. 0 is dissimilar, 1 is most similar. Args: query: input text k: Number of Documents to return. Defaults to 4. **kwargs: kwargs to be passed to similarity search. Should include: score_threshold: Optional, a floating point value between 0 to 1 to filter the resulting set of retrieved docs Returns: List of Tuples of (doc, similarity_score) """ return self.similarity_search_with_score(query, k, **kwargs) [docs] def similarity_search_with_score_by_vector( self, embedding: List[float], k: int = 4, filter: Optional[dict] = None, ) -> List[Tuple[Document, float]]: # Add the filter if provided filter_condition = "" if filter is not None: conditions = [ f"metadata->>{key!r} = {value!r}" for key, value in filter.items() ] filter_condition = f"WHERE {' AND '.join(conditions)}" # Define the base query sql_query = f""" SELECT *, l2_distance(embedding, :embedding) as distance FROM {self.collection_name} {filter_condition} ORDER BY embedding <-> :embedding LIMIT :k """ # Set up the query parameters params = {"embedding": embedding, "k": k} # Execute the query and fetch the results with self.engine.connect() as conn: results: Sequence[Row] = conn.execute(text(sql_query), params).fetchall() documents_with_scores = [ ( Document( page_content=result.document, metadata=result.metadata, ), result.distance if self.embedding_function is not None else None, ) for result in results ] return documents_with_scores [docs] def similarity_search_by_vector( self, embedding: List[float], k: int = 4, filter: Optional[dict] = None, **kwargs: Any, ) -> List[Document]: """Return docs most similar to embedding vector. Args: embedding: Embedding to look up documents similar to. k: Number of Documents to return. Defaults to 4. filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. Returns: List of Documents most similar to the query vector. """ docs_and_scores = self.similarity_search_with_score_by_vector( embedding=embedding, k=k, filter=filter ) return [doc for doc, _ in docs_and_scores] [docs] @classmethod def from_texts( cls: Type[AnalyticDB], texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, embedding_dimension: int = _LANGCHAIN_DEFAULT_EMBEDDING_DIM, collection_name: str = _LANGCHAIN_DEFAULT_COLLECTION_NAME, ids: Optional[List[str]] = None, pre_delete_collection: bool = False, **kwargs: Any, ) -> AnalyticDB: """ Return VectorStore initialized from texts and embeddings. Postgres Connection string is required Either pass it as a parameter or set the PG_CONNECTION_STRING environment variable. """ connection_string = cls.get_connection_string(kwargs) store = cls( connection_string=connection_string, collection_name=collection_name, embedding_function=embedding, embedding_dimension=embedding_dimension, pre_delete_collection=pre_delete_collection, ) store.add_texts(texts=texts, metadatas=metadatas, ids=ids, **kwargs) return store [docs] @classmethod def get_connection_string(cls, kwargs: Dict[str, Any]) -> str: connection_string: str = get_from_dict_or_env( data=kwargs, key="connection_string", env_key="PG_CONNECTION_STRING", ) if not connection_string: raise ValueError( "Postgres connection string is required" "Either pass it as a parameter" "or set the PG_CONNECTION_STRING environment variable." ) return connection_string [docs] @classmethod def from_documents( cls: Type[AnalyticDB], documents: List[Document], embedding: Embeddings, embedding_dimension: int = _LANGCHAIN_DEFAULT_EMBEDDING_DIM, collection_name: str = _LANGCHAIN_DEFAULT_COLLECTION_NAME, ids: Optional[List[str]] = None, pre_delete_collection: bool = False, **kwargs: Any, ) -> AnalyticDB: """ Return VectorStore initialized from documents and embeddings. Postgres Connection string is required Either pass it as a parameter or set the PG_CONNECTION_STRING environment variable. """ texts = [d.page_content for d in documents] metadatas = [d.metadata for d in documents] connection_string = cls.get_connection_string(kwargs) kwargs["connection_string"] = connection_string return cls.from_texts( texts=texts, pre_delete_collection=pre_delete_collection, embedding=embedding, embedding_dimension=embedding_dimension, metadatas=metadatas, ids=ids, collection_name=collection_name, **kwargs, ) [docs] @classmethod def connection_string_from_db_params( cls, driver: str, host: str, port: int, database: str, user: str, password: str, ) -> str: """Return connection string from database parameters.""" return f"postgresql+{driver}://{user}:{password}@{host}:{port}/{database}"
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/analyticdb.html
31effff9-3e6c-44d4-a659-a0c7bad8fff4
Source code for langchain.vectorstores.opensearch_vector_search """Wrapper around OpenSearch vector database.""" from __future__ import annotations import uuid from typing import Any, Dict, Iterable, List, Optional, Tuple import numpy as np from langchain.embeddings.base import Embeddings from langchain.schema import Document from langchain.utils import get_from_dict_or_env from langchain.vectorstores.base import VectorStore from langchain.vectorstores.utils import maximal_marginal_relevance IMPORT_OPENSEARCH_PY_ERROR = ( "Could not import OpenSearch. Please install it with `pip install opensearch-py`." ) SCRIPT_SCORING_SEARCH = "script_scoring" PAINLESS_SCRIPTING_SEARCH = "painless_scripting" MATCH_ALL_QUERY = {"match_all": {}} # type: Dict def _import_opensearch() -> Any: """Import OpenSearch if available, otherwise raise error.""" try: from opensearchpy import OpenSearch except ImportError: raise ValueError(IMPORT_OPENSEARCH_PY_ERROR) return OpenSearch def _import_bulk() -> Any: """Import bulk if available, otherwise raise error.""" try: from opensearchpy.helpers import bulk except ImportError: raise ValueError(IMPORT_OPENSEARCH_PY_ERROR) return bulk def _import_not_found_error() -> Any: """Import not found error if available, otherwise raise error.""" try: from opensearchpy.exceptions import NotFoundError except ImportError: raise ValueError(IMPORT_OPENSEARCH_PY_ERROR) return NotFoundError def _get_opensearch_client(opensearch_url: str, **kwargs: Any) -> Any: """Get OpenSearch client from the opensearch_url, otherwise raise error.""" try: opensearch = _import_opensearch() client = opensearch(opensearch_url, **kwargs) except ValueError as e: raise ValueError( f"OpenSearch client string provided is not in proper format. " f"Got error: {e} " ) return client def _validate_embeddings_and_bulk_size(embeddings_length: int, bulk_size: int) -> None: """Validate Embeddings Length and Bulk Size.""" if embeddings_length == 0: raise RuntimeError("Embeddings size is zero") if bulk_size < embeddings_length: raise RuntimeError( f"The embeddings count, {embeddings_length} is more than the " f"[bulk_size], {bulk_size}. Increase the value of [bulk_size]." ) def _bulk_ingest_embeddings( client: Any, index_name: str, embeddings: List[List[float]], texts: Iterable[str], metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, vector_field: str = "vector_field", text_field: str = "text", mapping: Optional[Dict] = None, ) -> List[str]: """Bulk Ingest Embeddings into given index.""" if not mapping: mapping = dict() bulk = _import_bulk() not_found_error = _import_not_found_error() requests = [] return_ids = [] mapping = mapping try: client.indices.get(index=index_name) except not_found_error: client.indices.create(index=index_name, body=mapping) for i, text in enumerate(texts): metadata = metadatas[i] if metadatas else {} _id = ids[i] if ids else str(uuid.uuid4()) request = { "_op_type": "index", "_index": index_name, vector_field: embeddings[i], text_field: text, "metadata": metadata, "_id": _id, } requests.append(request) return_ids.append(_id) bulk(client, requests) client.indices.refresh(index=index_name) return return_ids def _default_scripting_text_mapping( dim: int, vector_field: str = "vector_field", ) -> Dict: """For Painless Scripting or Script Scoring,the default mapping to create index.""" return { "mappings": { "properties": { vector_field: {"type": "knn_vector", "dimension": dim}, } } } def _default_text_mapping( dim: int, engine: str = "nmslib", space_type: str = "l2", ef_search: int = 512, ef_construction: int = 512, m: int = 16, vector_field: str = "vector_field", ) -> Dict: """For Approximate k-NN Search, this is the default mapping to create index.""" return { "settings": {"index": {"knn": True, "knn.algo_param.ef_search": ef_search}}, "mappings": { "properties": { vector_field: { "type": "knn_vector", "dimension": dim, "method": { "name": "hnsw", "space_type": space_type, "engine": engine, "parameters": {"ef_construction": ef_construction, "m": m}, }, } } }, } def _default_approximate_search_query( query_vector: List[float], k: int = 4, vector_field: str = "vector_field", ) -> Dict: """For Approximate k-NN Search, this is the default query.""" return { "size": k, "query": {"knn": {vector_field: {"vector": query_vector, "k": k}}}, } def _approximate_search_query_with_boolean_filter( query_vector: List[float], boolean_filter: Dict, k: int = 4, vector_field: str = "vector_field", subquery_clause: str = "must", ) -> Dict: """For Approximate k-NN Search, with Boolean Filter.""" return { "size": k, "query": { "bool": { "filter": boolean_filter, subquery_clause: [ {"knn": {vector_field: {"vector": query_vector, "k": k}}} ], } }, } def _approximate_search_query_with_lucene_filter( query_vector: List[float], lucene_filter: Dict, k: int = 4, vector_field: str = "vector_field", ) -> Dict: """For Approximate k-NN Search, with Lucene Filter.""" search_query = _default_approximate_search_query( query_vector, k=k, vector_field=vector_field ) search_query["query"]["knn"][vector_field]["filter"] = lucene_filter return search_query def _default_script_query( query_vector: List[float], space_type: str = "l2", pre_filter: Optional[Dict] = None, vector_field: str = "vector_field", ) -> Dict: """For Script Scoring Search, this is the default query.""" if not pre_filter: pre_filter = MATCH_ALL_QUERY return { "query": { "script_score": { "query": pre_filter, "script": { "source": "knn_score", "lang": "knn", "params": { "field": vector_field, "query_value": query_vector, "space_type": space_type, }, }, } } } def __get_painless_scripting_source( space_type: str, query_vector: List[float], vector_field: str = "vector_field" ) -> str: """For Painless Scripting, it returns the script source based on space type.""" source_value = ( "(1.0 + " + space_type + "(" + str(query_vector) + ", doc['" + vector_field + "']))" ) if space_type == "cosineSimilarity": return source_value else: return "1/" + source_value def _default_painless_scripting_query( query_vector: List[float], space_type: str = "l2Squared", pre_filter: Optional[Dict] = None, vector_field: str = "vector_field", ) -> Dict: """For Painless Scripting Search, this is the default query.""" if not pre_filter: pre_filter = MATCH_ALL_QUERY source = __get_painless_scripting_source(space_type, query_vector) return { "query": { "script_score": { "query": pre_filter, "script": { "source": source, "params": { "field": vector_field, "query_value": query_vector, }, }, } } } def _get_kwargs_value(kwargs: Any, key: str, default_value: Any) -> Any: """Get the value of the key if present. Else get the default_value.""" if key in kwargs: return kwargs.get(key) return default_value [docs]class OpenSearchVectorSearch(VectorStore): """Wrapper around OpenSearch as a vector database. Example: .. code-block:: python from langchain import OpenSearchVectorSearch opensearch_vector_search = OpenSearchVectorSearch( "http://localhost:9200", "embeddings", embedding_function ) """ def __init__( self, opensearch_url: str, index_name: str, embedding_function: Embeddings, **kwargs: Any, ): """Initialize with necessary components.""" self.embedding_function = embedding_function self.index_name = index_name self.client = _get_opensearch_client(opensearch_url, **kwargs) [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, bulk_size: int = 500, **kwargs: Any, ) -> List[str]: """Run more texts through the embeddings and add to the vectorstore. Args: texts: Iterable of strings to add to the vectorstore. metadatas: Optional list of metadatas associated with the texts. ids: Optional list of ids to associate with the texts. bulk_size: Bulk API request count; Default: 500 Returns: List of ids from adding the texts into the vectorstore. Optional Args: vector_field: Document field embeddings are stored in. Defaults to "vector_field". text_field: Document field the text of the document is stored in. Defaults to "text". """ embeddings = self.embedding_function.embed_documents(list(texts)) _validate_embeddings_and_bulk_size(len(embeddings), bulk_size) text_field = _get_kwargs_value(kwargs, "text_field", "text") dim = len(embeddings[0]) engine = _get_kwargs_value(kwargs, "engine", "nmslib") space_type = _get_kwargs_value(kwargs, "space_type", "l2") ef_search = _get_kwargs_value(kwargs, "ef_search", 512) ef_construction = _get_kwargs_value(kwargs, "ef_construction", 512) m = _get_kwargs_value(kwargs, "m", 16) vector_field = _get_kwargs_value(kwargs, "vector_field", "vector_field") mapping = _default_text_mapping( dim, engine, space_type, ef_search, ef_construction, m, vector_field ) return _bulk_ingest_embeddings( self.client, self.index_name, embeddings, texts, metadatas=metadatas, ids=ids, vector_field=vector_field, text_field=text_field, mapping=mapping, ) [docs] def similarity_search( self, query: str, k: int = 4, **kwargs: Any ) -> List[Document]: """Return docs most similar to query. By default, supports Approximate Search. Also supports Script Scoring and Painless Scripting. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. Returns: List of Documents most similar to the query. Optional Args: vector_field: Document field embeddings are stored in. Defaults to "vector_field". text_field: Document field the text of the document is stored in. Defaults to "text". metadata_field: Document field that metadata is stored in. Defaults to "metadata". Can be set to a special value "*" to include the entire document. Optional Args for Approximate Search: search_type: "approximate_search"; default: "approximate_search" boolean_filter: A Boolean filter consists of a Boolean query that contains a k-NN query and a filter. subquery_clause: Query clause on the knn vector field; default: "must" lucene_filter: the Lucene algorithm decides whether to perform an exact k-NN search with pre-filtering or an approximate search with modified post-filtering. Optional Args for Script Scoring Search: search_type: "script_scoring"; default: "approximate_search" space_type: "l2", "l1", "linf", "cosinesimil", "innerproduct", "hammingbit"; default: "l2" pre_filter: script_score query to pre-filter documents before identifying nearest neighbors; default: {"match_all": {}} Optional Args for Painless Scripting Search: search_type: "painless_scripting"; default: "approximate_search" space_type: "l2Squared", "l1Norm", "cosineSimilarity"; default: "l2Squared" pre_filter: script_score query to pre-filter documents before identifying nearest neighbors; default: {"match_all": {}} """ docs_with_scores = self.similarity_search_with_score(query, k, **kwargs) return [doc[0] for doc in docs_with_scores] [docs] def similarity_search_with_score( self, query: str, k: int = 4, **kwargs: Any ) -> List[Tuple[Document, float]]: """Return docs and it's scores most similar to query. By default, supports Approximate Search. Also supports Script Scoring and Painless Scripting. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. Returns: List of Documents along with its scores most similar to the query. Optional Args: same as `similarity_search` """ text_field = _get_kwargs_value(kwargs, "text_field", "text") metadata_field = _get_kwargs_value(kwargs, "metadata_field", "metadata") hits = self._raw_similarity_search_with_score(query=query, k=k, **kwargs) documents_with_scores = [ ( Document( page_content=hit["_source"][text_field], metadata=hit["_source"] if metadata_field == "*" or metadata_field not in hit["_source"] else hit["_source"][metadata_field], ), hit["_score"], ) for hit in hits ] return documents_with_scores def _raw_similarity_search_with_score( self, query: str, k: int = 4, **kwargs: Any ) -> List[dict]: """Return raw opensearch documents (dict) including vectors, scores most similar to query. By default, supports Approximate Search. Also supports Script Scoring and Painless Scripting. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. Returns: List of dict with its scores most similar to the query. Optional Args: same as `similarity_search` """ embedding = self.embedding_function.embed_query(query) search_type = _get_kwargs_value(kwargs, "search_type", "approximate_search") vector_field = _get_kwargs_value(kwargs, "vector_field", "vector_field") if search_type == "approximate_search": boolean_filter = _get_kwargs_value(kwargs, "boolean_filter", {}) subquery_clause = _get_kwargs_value(kwargs, "subquery_clause", "must") lucene_filter = _get_kwargs_value(kwargs, "lucene_filter", {}) if boolean_filter != {} and lucene_filter != {}: raise ValueError( "Both `boolean_filter` and `lucene_filter` are provided which " "is invalid" ) if boolean_filter != {}: search_query = _approximate_search_query_with_boolean_filter( embedding, boolean_filter, k=k, vector_field=vector_field, subquery_clause=subquery_clause, ) elif lucene_filter != {}: search_query = _approximate_search_query_with_lucene_filter( embedding, lucene_filter, k=k, vector_field=vector_field ) else: search_query = _default_approximate_search_query( embedding, k=k, vector_field=vector_field ) elif search_type == SCRIPT_SCORING_SEARCH: space_type = _get_kwargs_value(kwargs, "space_type", "l2") pre_filter = _get_kwargs_value(kwargs, "pre_filter", MATCH_ALL_QUERY) search_query = _default_script_query( embedding, space_type, pre_filter, vector_field ) elif search_type == PAINLESS_SCRIPTING_SEARCH: space_type = _get_kwargs_value(kwargs, "space_type", "l2Squared") pre_filter = _get_kwargs_value(kwargs, "pre_filter", MATCH_ALL_QUERY) search_query = _default_painless_scripting_query( embedding, space_type, pre_filter, vector_field ) else: raise ValueError("Invalid `search_type` provided as an argument") response = self.client.search(index=self.index_name, body=search_query) return [hit for hit in response["hits"]["hits"][:k]] [docs] def max_marginal_relevance_search( self, query: str, k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, **kwargs: Any, ) -> list[Document]: """Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. fetch_k: Number of Documents to fetch to pass to MMR algorithm. Defaults to 20. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. Returns: List of Documents selected by maximal marginal relevance. """ vector_field = _get_kwargs_value(kwargs, "vector_field", "vector_field") text_field = _get_kwargs_value(kwargs, "text_field", "text") metadata_field = _get_kwargs_value(kwargs, "metadata_field", "metadata") # Get embedding of the user query embedding = self.embedding_function.embed_query(query) # Do ANN/KNN search to get top fetch_k results where fetch_k >= k results = self._raw_similarity_search_with_score(query, fetch_k, **kwargs) embeddings = [result["_source"][vector_field] for result in results] # Rerank top k results using MMR, (mmr_selected is a list of indices) mmr_selected = maximal_marginal_relevance( np.array(embedding), embeddings, k=k, lambda_mult=lambda_mult ) return [ Document( page_content=results[i]["_source"][text_field], metadata=results[i]["_source"][metadata_field], ) for i in mmr_selected ] [docs] @classmethod def from_texts( cls, texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, bulk_size: int = 500, **kwargs: Any, ) -> OpenSearchVectorSearch: """Construct OpenSearchVectorSearch wrapper from raw documents. Example: .. code-block:: python from langchain import OpenSearchVectorSearch from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() opensearch_vector_search = OpenSearchVectorSearch.from_texts( texts, embeddings, opensearch_url="http://localhost:9200" ) OpenSearch by default supports Approximate Search powered by nmslib, faiss and lucene engines recommended for large datasets. Also supports brute force search through Script Scoring and Painless Scripting. Optional Args: vector_field: Document field embeddings are stored in. Defaults to "vector_field". text_field: Document field the text of the document is stored in. Defaults to "text". Optional Keyword Args for Approximate Search: engine: "nmslib", "faiss", "lucene"; default: "nmslib" space_type: "l2", "l1", "cosinesimil", "linf", "innerproduct"; default: "l2" ef_search: Size of the dynamic list used during k-NN searches. Higher values lead to more accurate but slower searches; default: 512 ef_construction: Size of the dynamic list used during k-NN graph creation. Higher values lead to more accurate graph but slower indexing speed; default: 512 m: Number of bidirectional links created for each new element. Large impact on memory consumption. Between 2 and 100; default: 16 Keyword Args for Script Scoring or Painless Scripting: is_appx_search: False """ opensearch_url = get_from_dict_or_env( kwargs, "opensearch_url", "OPENSEARCH_URL" ) # List of arguments that needs to be removed from kwargs # before passing kwargs to get opensearch client keys_list = [ "opensearch_url", "index_name", "is_appx_search", "vector_field", "text_field", "engine", "space_type", "ef_search", "ef_construction", "m", ] embeddings = embedding.embed_documents(texts) _validate_embeddings_and_bulk_size(len(embeddings), bulk_size) dim = len(embeddings[0]) # Get the index name from either from kwargs or ENV Variable # before falling back to random generation index_name = get_from_dict_or_env( kwargs, "index_name", "OPENSEARCH_INDEX_NAME", default=uuid.uuid4().hex ) is_appx_search = _get_kwargs_value(kwargs, "is_appx_search", True) vector_field = _get_kwargs_value(kwargs, "vector_field", "vector_field") text_field = _get_kwargs_value(kwargs, "text_field", "text") if is_appx_search: engine = _get_kwargs_value(kwargs, "engine", "nmslib") space_type = _get_kwargs_value(kwargs, "space_type", "l2") ef_search = _get_kwargs_value(kwargs, "ef_search", 512) ef_construction = _get_kwargs_value(kwargs, "ef_construction", 512) m = _get_kwargs_value(kwargs, "m", 16) mapping = _default_text_mapping( dim, engine, space_type, ef_search, ef_construction, m, vector_field ) else: mapping = _default_scripting_text_mapping(dim) [kwargs.pop(key, None) for key in keys_list] client = _get_opensearch_client(opensearch_url, **kwargs) _bulk_ingest_embeddings( client, index_name, embeddings, texts, metadatas=metadatas, vector_field=vector_field, text_field=text_field, mapping=mapping, ) return cls(opensearch_url, index_name, embedding, **kwargs)
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/opensearch_vector_search.html
2add156e-e6f3-4741-96e0-0478fbeeff90
Source code for langchain.vectorstores.faiss """Wrapper around FAISS vector database.""" from __future__ import annotations import math import os import pickle import uuid from pathlib import Path from typing import Any, Callable, Dict, Iterable, List, Optional, Tuple import numpy as np from langchain.docstore.base import AddableMixin, Docstore from langchain.docstore.document import Document from langchain.docstore.in_memory import InMemoryDocstore from langchain.embeddings.base import Embeddings from langchain.vectorstores.base import VectorStore from langchain.vectorstores.utils import maximal_marginal_relevance def dependable_faiss_import(no_avx2: Optional[bool] = None) -> Any: """ Import faiss if available, otherwise raise error. If FAISS_NO_AVX2 environment variable is set, it will be considered to load FAISS with no AVX2 optimization. Args: no_avx2: Load FAISS strictly with no AVX2 optimization so that the vectorstore is portable and compatible with other devices. """ if no_avx2 is None and "FAISS_NO_AVX2" in os.environ: no_avx2 = bool(os.getenv("FAISS_NO_AVX2")) try: if no_avx2: from faiss import swigfaiss as faiss else: import faiss except ImportError: raise ValueError( "Could not import faiss python package. " "Please install it with `pip install faiss` " "or `pip install faiss-cpu` (depending on Python version)." ) return faiss def _default_relevance_score_fn(score: float) -> float: """Return a similarity score on a scale [0, 1].""" # The 'correct' relevance function # may differ depending on a few things, including: # - the distance / similarity metric used by the VectorStore # - the scale of your embeddings (OpenAI's are unit normed. Many others are not!) # - embedding dimensionality # - etc. # This function converts the euclidean norm of normalized embeddings # (0 is most similar, sqrt(2) most dissimilar) # to a similarity function (0 to 1) return 1.0 - score / math.sqrt(2) [docs]class FAISS(VectorStore): """Wrapper around FAISS vector database. To use, you should have the ``faiss`` python package installed. Example: .. code-block:: python from langchain import FAISS faiss = FAISS(embedding_function, index, docstore, index_to_docstore_id) """ def __init__( self, embedding_function: Callable, index: Any, docstore: Docstore, index_to_docstore_id: Dict[int, str], relevance_score_fn: Optional[ Callable[[float], float] ] = _default_relevance_score_fn, normalize_L2: bool = False, ): """Initialize with necessary components.""" self.embedding_function = embedding_function self.index = index self.docstore = docstore self.index_to_docstore_id = index_to_docstore_id self.relevance_score_fn = relevance_score_fn self._normalize_L2 = normalize_L2 def __add( self, texts: Iterable[str], embeddings: Iterable[List[float]], metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, **kwargs: Any, ) -> List[str]: if not isinstance(self.docstore, AddableMixin): raise ValueError( "If trying to add texts, the underlying docstore should support " f"adding items, which {self.docstore} does not" ) documents = [] for i, text in enumerate(texts): metadata = metadatas[i] if metadatas else {} documents.append(Document(page_content=text, metadata=metadata)) if ids is None: ids = [str(uuid.uuid4()) for _ in texts] # Add to the index, the index_to_id mapping, and the docstore. starting_len = len(self.index_to_docstore_id) faiss = dependable_faiss_import() vector = np.array(embeddings, dtype=np.float32) if self._normalize_L2: faiss.normalize_L2(vector) self.index.add(vector) # Get list of index, id, and docs. full_info = [(starting_len + i, ids[i], doc) for i, doc in enumerate(documents)] # Add information to docstore and index. self.docstore.add({_id: doc for _, _id, doc in full_info}) index_to_id = {index: _id for index, _id, _ in full_info} self.index_to_docstore_id.update(index_to_id) return [_id for _, _id, _ in full_info] [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, **kwargs: Any, ) -> List[str]: """Run more texts through the embeddings and add to the vectorstore. Args: texts: Iterable of strings to add to the vectorstore. metadatas: Optional list of metadatas associated with the texts. ids: Optional list of unique IDs. Returns: List of ids from adding the texts into the vectorstore. """ if not isinstance(self.docstore, AddableMixin): raise ValueError( "If trying to add texts, the underlying docstore should support " f"adding items, which {self.docstore} does not" ) # Embed and create the documents. embeddings = [self.embedding_function(text) for text in texts] return self.__add(texts, embeddings, metadatas=metadatas, ids=ids, **kwargs) [docs] def add_embeddings( self, text_embeddings: Iterable[Tuple[str, List[float]]], metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, **kwargs: Any, ) -> List[str]: """Run more texts through the embeddings and add to the vectorstore. Args: text_embeddings: Iterable pairs of string and embedding to add to the vectorstore. metadatas: Optional list of metadatas associated with the texts. ids: Optional list of unique IDs. Returns: List of ids from adding the texts into the vectorstore. """ if not isinstance(self.docstore, AddableMixin): raise ValueError( "If trying to add texts, the underlying docstore should support " f"adding items, which {self.docstore} does not" ) # Embed and create the documents. texts, embeddings = zip(*text_embeddings) return self.__add(texts, embeddings, metadatas=metadatas, ids=ids, **kwargs) [docs] def similarity_search_with_score_by_vector( self, embedding: List[float], k: int = 4, filter: Optional[Dict[str, Any]] = None, fetch_k: int = 20, **kwargs: Any, ) -> List[Tuple[Document, float]]: """Return docs most similar to query. Args: embedding: Embedding vector to look up documents similar to. k: Number of Documents to return. Defaults to 4. filter (Optional[Dict[str, Any]]): Filter by metadata. Defaults to None. fetch_k: (Optional[int]) Number of Documents to fetch before filtering. Defaults to 20. **kwargs: kwargs to be passed to similarity search. Can include: score_threshold: Optional, a floating point value between 0 to 1 to filter the resulting set of retrieved docs Returns: List of documents most similar to the query text and L2 distance in float for each. Lower score represents more similarity. """ faiss = dependable_faiss_import() vector = np.array([embedding], dtype=np.float32) if self._normalize_L2: faiss.normalize_L2(vector) scores, indices = self.index.search(vector, k if filter is None else fetch_k) docs = [] for j, i in enumerate(indices[0]): if i == -1: # This happens when not enough docs are returned. continue _id = self.index_to_docstore_id[i] doc = self.docstore.search(_id) if not isinstance(doc, Document): raise ValueError(f"Could not find document for id {_id}, got {doc}") if filter is not None: filter = { key: [value] if not isinstance(value, list) else value for key, value in filter.items() } if all(doc.metadata.get(key) in value for key, value in filter.items()): docs.append((doc, scores[0][j])) else: docs.append((doc, scores[0][j])) score_threshold = kwargs.get("score_threshold") if score_threshold is not None: docs = [ (doc, similarity) for doc, similarity in docs if similarity >= score_threshold ] return docs[:k] [docs] def similarity_search_with_score( self, query: str, k: int = 4, filter: Optional[Dict[str, Any]] = None, fetch_k: int = 20, **kwargs: Any, ) -> List[Tuple[Document, float]]: """Return docs most similar to query. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. fetch_k: (Optional[int]) Number of Documents to fetch before filtering. Defaults to 20. Returns: List of documents most similar to the query text with L2 distance in float. Lower score represents more similarity. """ embedding = self.embedding_function(query) docs = self.similarity_search_with_score_by_vector( embedding, k, filter=filter, fetch_k=fetch_k, **kwargs, ) return docs [docs] def similarity_search_by_vector( self, embedding: List[float], k: int = 4, filter: Optional[Dict[str, Any]] = None, fetch_k: int = 20, **kwargs: Any, ) -> List[Document]: """Return docs most similar to embedding vector. Args: embedding: Embedding to look up documents similar to. k: Number of Documents to return. Defaults to 4. filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. fetch_k: (Optional[int]) Number of Documents to fetch before filtering. Defaults to 20. Returns: List of Documents most similar to the embedding. """ docs_and_scores = self.similarity_search_with_score_by_vector( embedding, k, filter=filter, fetch_k=fetch_k, **kwargs, ) return [doc for doc, _ in docs_and_scores] [docs] def similarity_search( self, query: str, k: int = 4, filter: Optional[Dict[str, Any]] = None, fetch_k: int = 20, **kwargs: Any, ) -> List[Document]: """Return docs most similar to query. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. filter: (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. fetch_k: (Optional[int]) Number of Documents to fetch before filtering. Defaults to 20. Returns: List of Documents most similar to the query. """ docs_and_scores = self.similarity_search_with_score( query, k, filter=filter, fetch_k=fetch_k, **kwargs ) return [doc for doc, _ in docs_and_scores] [docs] def max_marginal_relevance_search_by_vector( self, embedding: List[float], k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, filter: Optional[Dict[str, Any]] = None, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: embedding: Embedding to look up documents similar to. k: Number of Documents to return. Defaults to 4. fetch_k: Number of Documents to fetch before filtering to pass to MMR algorithm. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. Returns: List of Documents selected by maximal marginal relevance. """ _, indices = self.index.search( np.array([embedding], dtype=np.float32), fetch_k if filter is None else fetch_k * 2, ) if filter is not None: filtered_indices = [] for i in indices[0]: if i == -1: # This happens when not enough docs are returned. continue _id = self.index_to_docstore_id[i] doc = self.docstore.search(_id) if not isinstance(doc, Document): raise ValueError(f"Could not find document for id {_id}, got {doc}") if all(doc.metadata.get(key) == value for key, value in filter.items()): filtered_indices.append(i) indices = np.array([filtered_indices]) # -1 happens when not enough docs are returned. embeddings = [self.index.reconstruct(int(i)) for i in indices[0] if i != -1] mmr_selected = maximal_marginal_relevance( np.array([embedding], dtype=np.float32), embeddings, k=k, lambda_mult=lambda_mult, ) selected_indices = [indices[0][i] for i in mmr_selected] docs = [] for i in selected_indices: if i == -1: # This happens when not enough docs are returned. continue _id = self.index_to_docstore_id[i] doc = self.docstore.search(_id) if not isinstance(doc, Document): raise ValueError(f"Could not find document for id {_id}, got {doc}") docs.append(doc) return docs [docs] def max_marginal_relevance_search( self, query: str, k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, filter: Optional[Dict[str, Any]] = None, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. fetch_k: Number of Documents to fetch before filtering (if needed) to pass to MMR algorithm. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. Returns: List of Documents selected by maximal marginal relevance. """ embedding = self.embedding_function(query) docs = self.max_marginal_relevance_search_by_vector( embedding, k, fetch_k, lambda_mult=lambda_mult, filter=filter, **kwargs, ) return docs [docs] def merge_from(self, target: FAISS) -> None: """Merge another FAISS object with the current one. Add the target FAISS to the current one. Args: target: FAISS object you wish to merge into the current one Returns: None. """ if not isinstance(self.docstore, AddableMixin): raise ValueError("Cannot merge with this type of docstore") # Numerical index for target docs are incremental on existing ones starting_len = len(self.index_to_docstore_id) # Merge two IndexFlatL2 self.index.merge_from(target.index) # Get id and docs from target FAISS object full_info = [] for i, target_id in target.index_to_docstore_id.items(): doc = target.docstore.search(target_id) if not isinstance(doc, Document): raise ValueError("Document should be returned") full_info.append((starting_len + i, target_id, doc)) # Add information to docstore and index_to_docstore_id. self.docstore.add({_id: doc for _, _id, doc in full_info}) index_to_id = {index: _id for index, _id, _ in full_info} self.index_to_docstore_id.update(index_to_id) @classmethod def __from( cls, texts: List[str], embeddings: List[List[float]], embedding: Embeddings, metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, normalize_L2: bool = False, **kwargs: Any, ) -> FAISS: faiss = dependable_faiss_import() index = faiss.IndexFlatL2(len(embeddings[0])) vector = np.array(embeddings, dtype=np.float32) if normalize_L2: faiss.normalize_L2(vector) index.add(vector) documents = [] if ids is None: ids = [str(uuid.uuid4()) for _ in texts] for i, text in enumerate(texts): metadata = metadatas[i] if metadatas else {} documents.append(Document(page_content=text, metadata=metadata)) index_to_id = dict(enumerate(ids)) docstore = InMemoryDocstore(dict(zip(index_to_id.values(), documents))) return cls( embedding.embed_query, index, docstore, index_to_id, normalize_L2=normalize_L2, **kwargs, ) [docs] @classmethod def from_texts( cls, texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, **kwargs: Any, ) -> FAISS: """Construct FAISS wrapper from raw documents. This is a user friendly interface that: 1. Embeds documents. 2. Creates an in memory docstore 3. Initializes the FAISS database This is intended to be a quick way to get started. Example: .. code-block:: python from langchain import FAISS from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() faiss = FAISS.from_texts(texts, embeddings) """ embeddings = embedding.embed_documents(texts) return cls.__from( texts, embeddings, embedding, metadatas=metadatas, ids=ids, **kwargs, ) [docs] @classmethod def from_embeddings( cls, text_embeddings: List[Tuple[str, List[float]]], embedding: Embeddings, metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, **kwargs: Any, ) -> FAISS: """Construct FAISS wrapper from raw documents. This is a user friendly interface that: 1. Embeds documents. 2. Creates an in memory docstore 3. Initializes the FAISS database This is intended to be a quick way to get started. Example: .. code-block:: python from langchain import FAISS from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() text_embeddings = embeddings.embed_documents(texts) text_embedding_pairs = list(zip(texts, text_embeddings)) faiss = FAISS.from_embeddings(text_embedding_pairs, embeddings) """ texts = [t[0] for t in text_embeddings] embeddings = [t[1] for t in text_embeddings] return cls.__from( texts, embeddings, embedding, metadatas=metadatas, ids=ids, **kwargs, ) [docs] def save_local(self, folder_path: str, index_name: str = "index") -> None: """Save FAISS index, docstore, and index_to_docstore_id to disk. Args: folder_path: folder path to save index, docstore, and index_to_docstore_id to. index_name: for saving with a specific index file name """ path = Path(folder_path) path.mkdir(exist_ok=True, parents=True) # save index separately since it is not picklable faiss = dependable_faiss_import() faiss.write_index( self.index, str(path / "{index_name}.faiss".format(index_name=index_name)) ) # save docstore and index_to_docstore_id with open(path / "{index_name}.pkl".format(index_name=index_name), "wb") as f: pickle.dump((self.docstore, self.index_to_docstore_id), f) [docs] @classmethod def load_local( cls, folder_path: str, embeddings: Embeddings, index_name: str = "index" ) -> FAISS: """Load FAISS index, docstore, and index_to_docstore_id from disk. Args: folder_path: folder path to load index, docstore, and index_to_docstore_id from. embeddings: Embeddings to use when generating queries index_name: for saving with a specific index file name """ path = Path(folder_path) # load index separately since it is not picklable faiss = dependable_faiss_import() index = faiss.read_index( str(path / "{index_name}.faiss".format(index_name=index_name)) ) # load docstore and index_to_docstore_id with open(path / "{index_name}.pkl".format(index_name=index_name), "rb") as f: docstore, index_to_docstore_id = pickle.load(f) return cls(embeddings.embed_query, index, docstore, index_to_docstore_id) def _similarity_search_with_relevance_scores( self, query: str, k: int = 4, filter: Optional[Dict[str, Any]] = None, fetch_k: int = 20, **kwargs: Any, ) -> List[Tuple[Document, float]]: """Return docs and their similarity scores on a scale from 0 to 1.""" if self.relevance_score_fn is None: raise ValueError( "normalize_score_fn must be provided to" " FAISS constructor to normalize scores" ) docs_and_scores = self.similarity_search_with_score( query, k=k, filter=filter, fetch_k=fetch_k, **kwargs, ) return [(doc, self.relevance_score_fn(score)) for doc, score in docs_and_scores]
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/faiss.html
bd885a6c-4316-45c1-8acf-488b52baea1f
Source code for langchain.vectorstores.matching_engine """Vertex Matching Engine implementation of the vector store.""" from __future__ import annotations import json import logging import time import uuid from typing import TYPE_CHECKING, Any, Iterable, List, Optional, Type from langchain.docstore.document import Document from langchain.embeddings import TensorflowHubEmbeddings from langchain.embeddings.base import Embeddings from langchain.vectorstores.base import VectorStore if TYPE_CHECKING: from google.cloud import storage from google.cloud.aiplatform import MatchingEngineIndex, MatchingEngineIndexEndpoint from google.oauth2.service_account import Credentials logger = logging.getLogger() [docs]class MatchingEngine(VectorStore): """Vertex Matching Engine implementation of the vector store. While the embeddings are stored in the Matching Engine, the embedded documents will be stored in GCS. An existing Index and corresponding Endpoint are preconditions for using this module. See usage in docs/modules/indexes/vectorstores/examples/matchingengine.ipynb Note that this implementation is mostly meant for reading if you are planning to do a real time implementation. While reading is a real time operation, updating the index takes close to one hour.""" def __init__( self, project_id: str, index: MatchingEngineIndex, endpoint: MatchingEngineIndexEndpoint, embedding: Embeddings, gcs_client: storage.Client, gcs_bucket_name: str, credentials: Optional[Credentials] = None, ): """Vertex Matching Engine implementation of the vector store. While the embeddings are stored in the Matching Engine, the embedded documents will be stored in GCS. An existing Index and corresponding Endpoint are preconditions for using this module. See usage in docs/modules/indexes/vectorstores/examples/matchingengine.ipynb. Note that this implementation is mostly meant for reading if you are planning to do a real time implementation. While reading is a real time operation, updating the index takes close to one hour. Attributes: project_id: The GCS project id. index: The created index class. See ~:func:`MatchingEngine.from_components`. endpoint: The created endpoint class. See ~:func:`MatchingEngine.from_components`. embedding: A :class:`Embeddings` that will be used for embedding the text sent. If none is sent, then the multilingual Tensorflow Universal Sentence Encoder will be used. gcs_client: The GCS client. gcs_bucket_name: The GCS bucket name. credentials (Optional): Created GCP credentials. """ super().__init__() self._validate_google_libraries_installation() self.project_id = project_id self.index = index self.endpoint = endpoint self.embedding = embedding self.gcs_client = gcs_client self.credentials = credentials self.gcs_bucket_name = gcs_bucket_name def _validate_google_libraries_installation(self) -> None: """Validates that Google libraries that are needed are installed.""" try: from google.cloud import aiplatform, storage # noqa: F401 from google.oauth2 import service_account # noqa: F401 except ImportError: raise ImportError( "You must run `pip install --upgrade " "google-cloud-aiplatform google-cloud-storage`" "to use the MatchingEngine Vectorstore." ) [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, **kwargs: Any, ) -> List[str]: """Run more texts through the embeddings and add to the vectorstore. Args: texts: Iterable of strings to add to the vectorstore. metadatas: Optional list of metadatas associated with the texts. kwargs: vectorstore specific parameters. Returns: List of ids from adding the texts into the vectorstore. """ logger.debug("Embedding documents.") embeddings = self.embedding.embed_documents(list(texts)) jsons = [] ids = [] # Could be improved with async. for embedding, text in zip(embeddings, texts): id = str(uuid.uuid4()) ids.append(id) jsons.append({"id": id, "embedding": embedding}) self._upload_to_gcs(text, f"documents/{id}") logger.debug(f"Uploaded {len(ids)} documents to GCS.") # Creating json lines from the embedded documents. result_str = "\n".join([json.dumps(x) for x in jsons]) filename_prefix = f"indexes/{uuid.uuid4()}" filename = f"{filename_prefix}/{time.time()}.json" self._upload_to_gcs(result_str, filename) logger.debug( f"Uploaded updated json with embeddings to " f"{self.gcs_bucket_name}/{filename}." ) self.index = self.index.update_embeddings( contents_delta_uri=f"gs://{self.gcs_bucket_name}/{filename_prefix}/" ) logger.debug("Updated index with new configuration.") return ids def _upload_to_gcs(self, data: str, gcs_location: str) -> None: """Uploads data to gcs_location. Args: data: The data that will be stored. gcs_location: The location where the data will be stored. """ bucket = self.gcs_client.get_bucket(self.gcs_bucket_name) blob = bucket.blob(gcs_location) blob.upload_from_string(data) [docs] def similarity_search( self, query: str, k: int = 4, **kwargs: Any ) -> List[Document]: """Return docs most similar to query. Args: query: The string that will be used to search for similar documents. k: The amount of neighbors that will be retrieved. Returns: A list of k matching documents. """ logger.debug(f"Embedding query {query}.") embedding_query = self.embedding.embed_documents([query]) response = self.endpoint.match( deployed_index_id=self._get_index_id(), queries=embedding_query, num_neighbors=k, ) if len(response) == 0: return [] logger.debug(f"Found {len(response)} matches for the query {query}.") results = [] # I'm only getting the first one because queries receives an array # and the similarity_search method only recevies one query. This # means that the match method will always return an array with only # one element. for doc in response[0]: page_content = self._download_from_gcs(f"documents/{doc.id}") results.append(Document(page_content=page_content)) logger.debug("Downloaded documents for query.") return results def _get_index_id(self) -> str: """Gets the correct index id for the endpoint. Returns: The index id if found (which should be found) or throws ValueError otherwise. """ for index in self.endpoint.deployed_indexes: if index.index == self.index.resource_name: return index.id raise ValueError( f"No index with id {self.index.resource_name} " f"deployed on endpoint " f"{self.endpoint.display_name}." ) def _download_from_gcs(self, gcs_location: str) -> str: """Downloads from GCS in text format. Args: gcs_location: The location where the file is located. Returns: The string contents of the file. """ bucket = self.gcs_client.get_bucket(self.gcs_bucket_name) blob = bucket.blob(gcs_location) return blob.download_as_string() [docs] @classmethod def from_texts( cls: Type["MatchingEngine"], texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, **kwargs: Any, ) -> "MatchingEngine": """Use from components instead.""" raise NotImplementedError( "This method is not implemented. Instead, you should initialize the class" " with `MatchingEngine.from_components(...)` and then call " "`add_texts`" ) [docs] @classmethod def from_components( cls: Type["MatchingEngine"], project_id: str, region: str, gcs_bucket_name: str, index_id: str, endpoint_id: str, credentials_path: Optional[str] = None, embedding: Optional[Embeddings] = None, ) -> "MatchingEngine": """Takes the object creation out of the constructor. Args: project_id: The GCP project id. region: The default location making the API calls. It must have the same location as the GCS bucket and must be regional. gcs_bucket_name: The location where the vectors will be stored in order for the index to be created. index_id: The id of the created index. endpoint_id: The id of the created endpoint. credentials_path: (Optional) The path of the Google credentials on the local file system. embedding: The :class:`Embeddings` that will be used for embedding the texts. Returns: A configured MatchingEngine with the texts added to the index. """ gcs_bucket_name = cls._validate_gcs_bucket(gcs_bucket_name) credentials = cls._create_credentials_from_file(credentials_path) index = cls._create_index_by_id(index_id, project_id, region, credentials) endpoint = cls._create_endpoint_by_id( endpoint_id, project_id, region, credentials ) gcs_client = cls._get_gcs_client(credentials, project_id) cls._init_aiplatform(project_id, region, gcs_bucket_name, credentials) return cls( project_id=project_id, index=index, endpoint=endpoint, embedding=embedding or cls._get_default_embeddings(), gcs_client=gcs_client, credentials=credentials, gcs_bucket_name=gcs_bucket_name, ) @classmethod def _validate_gcs_bucket(cls, gcs_bucket_name: str) -> str: """Validates the gcs_bucket_name as a bucket name. Args: gcs_bucket_name: The received bucket uri. Returns: A valid gcs_bucket_name or throws ValueError if full path is provided. """ gcs_bucket_name = gcs_bucket_name.replace("gs://", "") if "/" in gcs_bucket_name: raise ValueError( f"The argument gcs_bucket_name should only be " f"the bucket name. Received {gcs_bucket_name}" ) return gcs_bucket_name @classmethod def _create_credentials_from_file( cls, json_credentials_path: Optional[str] ) -> Optional[Credentials]: """Creates credentials for GCP. Args: json_credentials_path: The path on the file system where the credentials are stored. Returns: An optional of Credentials or None, in which case the default will be used. """ from google.oauth2 import service_account credentials = None if json_credentials_path is not None: credentials = service_account.Credentials.from_service_account_file( json_credentials_path ) return credentials @classmethod def _create_index_by_id( cls, index_id: str, project_id: str, region: str, credentials: "Credentials" ) -> MatchingEngineIndex: """Creates a MatchingEngineIndex object by id. Args: index_id: The created index id. project_id: The project to retrieve index from. region: Location to retrieve index from. credentials: GCS credentials. Returns: A configured MatchingEngineIndex. """ from google.cloud import aiplatform logger.debug(f"Creating matching engine index with id {index_id}.") return aiplatform.MatchingEngineIndex( index_name=index_id, project=project_id, location=region, credentials=credentials, ) @classmethod def _create_endpoint_by_id( cls, endpoint_id: str, project_id: str, region: str, credentials: "Credentials" ) -> MatchingEngineIndexEndpoint: """Creates a MatchingEngineIndexEndpoint object by id. Args: endpoint_id: The created endpoint id. project_id: The project to retrieve index from. region: Location to retrieve index from. credentials: GCS credentials. Returns: A configured MatchingEngineIndexEndpoint. """ from google.cloud import aiplatform logger.debug(f"Creating endpoint with id {endpoint_id}.") return aiplatform.MatchingEngineIndexEndpoint( index_endpoint_name=endpoint_id, project=project_id, location=region, credentials=credentials, ) @classmethod def _get_gcs_client( cls, credentials: "Credentials", project_id: str ) -> "storage.Client": """Lazily creates a GCS client. Returns: A configured GCS client. """ from google.cloud import storage return storage.Client(credentials=credentials, project=project_id) @classmethod def _init_aiplatform( cls, project_id: str, region: str, gcs_bucket_name: str, credentials: "Credentials", ) -> None: """Configures the aiplatform library. Args: project_id: The GCP project id. region: The default location making the API calls. It must have the same location as the GCS bucket and must be regional. gcs_bucket_name: GCS staging location. credentials: The GCS Credentials object. """ from google.cloud import aiplatform logger.debug( f"Initializing AI Platform for project {project_id} on " f"{region} and for {gcs_bucket_name}." ) aiplatform.init( project=project_id, location=region, staging_bucket=gcs_bucket_name, credentials=credentials, ) @classmethod def _get_default_embeddings(cls) -> TensorflowHubEmbeddings: """This function returns the default embedding. Returns: Default TensorflowHubEmbeddings to use. """ return TensorflowHubEmbeddings()
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/matching_engine.html
1c2b565f-cb8e-479d-83eb-43d09441a166
Source code for langchain.vectorstores.tair """Wrapper around Tair Vector.""" from __future__ import annotations import json import logging import uuid from typing import Any, Iterable, List, Optional, Type from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.utils import get_from_dict_or_env from langchain.vectorstores.base import VectorStore logger = logging.getLogger(__name__) def _uuid_key() -> str: return uuid.uuid4().hex [docs]class Tair(VectorStore): """Wrapper around Tair Vector store.""" def __init__( self, embedding_function: Embeddings, url: str, index_name: str, content_key: str = "content", metadata_key: str = "metadata", search_params: Optional[dict] = None, **kwargs: Any, ): self.embedding_function = embedding_function self.index_name = index_name try: from tair import Tair as TairClient except ImportError: raise ImportError( "Could not import tair python package. " "Please install it with `pip install tair`." ) try: # connect to tair from url client = TairClient.from_url(url, **kwargs) except ValueError as e: raise ValueError(f"Tair failed to connect: {e}") self.client = client self.content_key = content_key self.metadata_key = metadata_key self.search_params = search_params [docs] def create_index_if_not_exist( self, dim: int, distance_type: str, index_type: str, data_type: str, **kwargs: Any, ) -> bool: index = self.client.tvs_get_index(self.index_name) if index is not None: logger.info("Index already exists") return False self.client.tvs_create_index( self.index_name, dim, distance_type, index_type, data_type, **kwargs, ) return True [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, **kwargs: Any, ) -> List[str]: """Add texts data to an existing index.""" ids = [] keys = kwargs.get("keys", None) # Write data to tair pipeline = self.client.pipeline(transaction=False) embeddings = self.embedding_function.embed_documents(list(texts)) for i, text in enumerate(texts): # Use provided key otherwise use default key key = keys[i] if keys else _uuid_key() metadata = metadatas[i] if metadatas else {} pipeline.tvs_hset( self.index_name, key, embeddings[i], False, **{ self.content_key: text, self.metadata_key: json.dumps(metadata), }, ) ids.append(key) pipeline.execute() return ids [docs] def similarity_search( self, query: str, k: int = 4, **kwargs: Any ) -> List[Document]: """ Returns the most similar indexed documents to the query text. Args: query (str): The query text for which to find similar documents. k (int): The number of documents to return. Default is 4. Returns: List[Document]: A list of documents that are most similar to the query text. """ # Creates embedding vector from user query embedding = self.embedding_function.embed_query(query) keys_and_scores = self.client.tvs_knnsearch( self.index_name, k, embedding, False, None, **kwargs ) pipeline = self.client.pipeline(transaction=False) for key, _ in keys_and_scores: pipeline.tvs_hmget( self.index_name, key, self.metadata_key, self.content_key ) docs = pipeline.execute() return [ Document( page_content=d[1], metadata=json.loads(d[0]), ) for d in docs ] [docs] @classmethod def from_texts( cls: Type[Tair], texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, index_name: str = "langchain", content_key: str = "content", metadata_key: str = "metadata", **kwargs: Any, ) -> Tair: try: from tair import tairvector except ImportError: raise ValueError( "Could not import tair python package. " "Please install it with `pip install tair`." ) url = get_from_dict_or_env(kwargs, "tair_url", "TAIR_URL") if "tair_url" in kwargs: kwargs.pop("tair_url") distance_type = tairvector.DistanceMetric.InnerProduct if "distance_type" in kwargs: distance_type = kwargs.pop("distance_typ") index_type = tairvector.IndexType.HNSW if "index_type" in kwargs: index_type = kwargs.pop("index_type") data_type = tairvector.DataType.Float32 if "data_type" in kwargs: data_type = kwargs.pop("data_type") index_params = {} if "index_params" in kwargs: index_params = kwargs.pop("index_params") search_params = {} if "search_params" in kwargs: search_params = kwargs.pop("search_params") keys = None if "keys" in kwargs: keys = kwargs.pop("keys") try: tair_vector_store = cls( embedding, url, index_name, content_key=content_key, metadata_key=metadata_key, search_params=search_params, **kwargs, ) except ValueError as e: raise ValueError(f"tair failed to connect: {e}") # Create embeddings for documents embeddings = embedding.embed_documents(texts) tair_vector_store.create_index_if_not_exist( len(embeddings[0]), distance_type, index_type, data_type, **index_params, ) tair_vector_store.add_texts(texts, metadatas, keys=keys) return tair_vector_store [docs] @classmethod def from_documents( cls, documents: List[Document], embedding: Embeddings, metadatas: Optional[List[dict]] = None, index_name: str = "langchain", content_key: str = "content", metadata_key: str = "metadata", **kwargs: Any, ) -> Tair: texts = [d.page_content for d in documents] metadatas = [d.metadata for d in documents] return cls.from_texts( texts, embedding, metadatas, index_name, content_key, metadata_key, **kwargs ) [docs] @staticmethod def drop_index( index_name: str = "langchain", **kwargs: Any, ) -> bool: """ Drop an existing index. Args: index_name (str): Name of the index to drop. Returns: bool: True if the index is dropped successfully. """ try: from tair import Tair as TairClient except ImportError: raise ValueError( "Could not import tair python package. " "Please install it with `pip install tair`." ) url = get_from_dict_or_env(kwargs, "tair_url", "TAIR_URL") try: if "tair_url" in kwargs: kwargs.pop("tair_url") client = TairClient.from_url(url=url, **kwargs) except ValueError as e: raise ValueError(f"Tair connection error: {e}") # delete index ret = client.tvs_del_index(index_name) if ret == 0: # index not exist logger.info("Index does not exist") return False return True [docs] @classmethod def from_existing_index( cls, embedding: Embeddings, index_name: str = "langchain", content_key: str = "content", metadata_key: str = "metadata", **kwargs: Any, ) -> Tair: """Connect to an existing Tair index.""" url = get_from_dict_or_env(kwargs, "tair_url", "TAIR_URL") search_params = {} if "search_params" in kwargs: search_params = kwargs.pop("search_params") return cls( embedding, url, index_name, content_key=content_key, metadata_key=metadata_key, search_params=search_params, **kwargs, )
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/tair.html
cc933f7b-c839-473a-8576-4d098bc34605
Source code for langchain.vectorstores.atlas """Wrapper around Atlas by Nomic.""" from __future__ import annotations import logging import uuid from typing import Any, Iterable, List, Optional, Type import numpy as np from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.vectorstores.base import VectorStore logger = logging.getLogger(__name__) [docs]class AtlasDB(VectorStore): """Wrapper around Atlas: Nomic's neural database and rhizomatic instrument. To use, you should have the ``nomic`` python package installed. Example: .. code-block:: python from langchain.vectorstores import AtlasDB from langchain.embeddings.openai import OpenAIEmbeddings embeddings = OpenAIEmbeddings() vectorstore = AtlasDB("my_project", embeddings.embed_query) """ _ATLAS_DEFAULT_ID_FIELD = "atlas_id" def __init__( self, name: str, embedding_function: Optional[Embeddings] = None, api_key: Optional[str] = None, description: str = "A description for your project", is_public: bool = True, reset_project_if_exists: bool = False, ) -> None: """ Initialize the Atlas Client Args: name (str): The name of your project. If the project already exists, it will be loaded. embedding_function (Optional[Callable]): An optional function used for embedding your data. If None, data will be embedded with Nomic's embed model. api_key (str): Your nomic API key description (str): A description for your project. is_public (bool): Whether your project is publicly accessible. True by default. reset_project_if_exists (bool): Whether to reset this project if it already exists. Default False. Generally userful during development and testing. """ try: import nomic from nomic import AtlasProject except ImportError: raise ValueError( "Could not import nomic python package. " "Please install it with `pip install nomic`." ) if api_key is None: raise ValueError("No API key provided. Sign up at atlas.nomic.ai!") nomic.login(api_key) self._embedding_function = embedding_function modality = "text" if self._embedding_function is not None: modality = "embedding" # Check if the project exists, create it if not self.project = AtlasProject( name=name, description=description, modality=modality, is_public=is_public, reset_project_if_exists=reset_project_if_exists, unique_id_field=AtlasDB._ATLAS_DEFAULT_ID_FIELD, ) self.project._latest_project_state() [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, refresh: bool = True, **kwargs: Any, ) -> List[str]: """Run more texts through the embeddings and add to the vectorstore. Args: texts (Iterable[str]): Texts to add to the vectorstore. metadatas (Optional[List[dict]], optional): Optional list of metadatas. ids (Optional[List[str]]): An optional list of ids. refresh(bool): Whether or not to refresh indices with the updated data. Default True. Returns: List[str]: List of IDs of the added texts. """ if ( metadatas is not None and len(metadatas) > 0 and "text" in metadatas[0].keys() ): raise ValueError("Cannot accept key text in metadata!") texts = list(texts) if ids is None: ids = [str(uuid.uuid1()) for _ in texts] # Embedding upload case if self._embedding_function is not None: _embeddings = self._embedding_function.embed_documents(texts) embeddings = np.stack(_embeddings) if metadatas is None: data = [ {AtlasDB._ATLAS_DEFAULT_ID_FIELD: ids[i], "text": texts[i]} for i, _ in enumerate(texts) ] else: for i in range(len(metadatas)): metadatas[i][AtlasDB._ATLAS_DEFAULT_ID_FIELD] = ids[i] metadatas[i]["text"] = texts[i] data = metadatas self.project._validate_map_data_inputs( [], id_field=AtlasDB._ATLAS_DEFAULT_ID_FIELD, data=data ) with self.project.wait_for_project_lock(): self.project.add_embeddings(embeddings=embeddings, data=data) # Text upload case else: if metadatas is None: data = [ {"text": text, AtlasDB._ATLAS_DEFAULT_ID_FIELD: ids[i]} for i, text in enumerate(texts) ] else: for i, text in enumerate(texts): metadatas[i]["text"] = texts metadatas[i][AtlasDB._ATLAS_DEFAULT_ID_FIELD] = ids[i] data = metadatas self.project._validate_map_data_inputs( [], id_field=AtlasDB._ATLAS_DEFAULT_ID_FIELD, data=data ) with self.project.wait_for_project_lock(): self.project.add_text(data) if refresh: if len(self.project.indices) > 0: with self.project.wait_for_project_lock(): self.project.rebuild_maps() return ids [docs] def create_index(self, **kwargs: Any) -> Any: """Creates an index in your project. See https://docs.nomic.ai/atlas_api.html#nomic.project.AtlasProject.create_index for full detail. """ with self.project.wait_for_project_lock(): return self.project.create_index(**kwargs) [docs] def similarity_search( self, query: str, k: int = 4, **kwargs: Any, ) -> List[Document]: """Run similarity search with AtlasDB Args: query (str): Query text to search for. k (int): Number of results to return. Defaults to 4. Returns: List[Document]: List of documents most similar to the query text. """ if self._embedding_function is None: raise NotImplementedError( "AtlasDB requires an embedding_function for text similarity search!" ) _embedding = self._embedding_function.embed_documents([query])[0] embedding = np.array(_embedding).reshape(1, -1) with self.project.wait_for_project_lock(): neighbors, _ = self.project.projections[0].vector_search( queries=embedding, k=k ) datas = self.project.get_data(ids=neighbors[0]) docs = [ Document(page_content=datas[i]["text"], metadata=datas[i]) for i, neighbor in enumerate(neighbors) ] return docs [docs] @classmethod def from_texts( cls: Type[AtlasDB], texts: List[str], embedding: Optional[Embeddings] = None, metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, name: Optional[str] = None, api_key: Optional[str] = None, description: str = "A description for your project", is_public: bool = True, reset_project_if_exists: bool = False, index_kwargs: Optional[dict] = None, **kwargs: Any, ) -> AtlasDB: """Create an AtlasDB vectorstore from a raw documents. Args: texts (List[str]): The list of texts to ingest. name (str): Name of the project to create. api_key (str): Your nomic API key, embedding (Optional[Embeddings]): Embedding function. Defaults to None. metadatas (Optional[List[dict]]): List of metadatas. Defaults to None. ids (Optional[List[str]]): Optional list of document IDs. If None, ids will be auto created description (str): A description for your project. is_public (bool): Whether your project is publicly accessible. True by default. reset_project_if_exists (bool): Whether to reset this project if it already exists. Default False. Generally userful during development and testing. index_kwargs (Optional[dict]): Dict of kwargs for index creation. See https://docs.nomic.ai/atlas_api.html Returns: AtlasDB: Nomic's neural database and finest rhizomatic instrument """ if name is None or api_key is None: raise ValueError("`name` and `api_key` cannot be None.") # Inject relevant kwargs all_index_kwargs = {"name": name + "_index", "indexed_field": "text"} if index_kwargs is not None: for k, v in index_kwargs.items(): all_index_kwargs[k] = v # Build project atlasDB = cls( name, embedding_function=embedding, api_key=api_key, description="A description for your project", is_public=is_public, reset_project_if_exists=reset_project_if_exists, ) with atlasDB.project.wait_for_project_lock(): atlasDB.add_texts(texts=texts, metadatas=metadatas, ids=ids) atlasDB.create_index(**all_index_kwargs) return atlasDB [docs] @classmethod def from_documents( cls: Type[AtlasDB], documents: List[Document], embedding: Optional[Embeddings] = None, ids: Optional[List[str]] = None, name: Optional[str] = None, api_key: Optional[str] = None, persist_directory: Optional[str] = None, description: str = "A description for your project", is_public: bool = True, reset_project_if_exists: bool = False, index_kwargs: Optional[dict] = None, **kwargs: Any, ) -> AtlasDB: """Create an AtlasDB vectorstore from a list of documents. Args: name (str): Name of the collection to create. api_key (str): Your nomic API key, documents (List[Document]): List of documents to add to the vectorstore. embedding (Optional[Embeddings]): Embedding function. Defaults to None. ids (Optional[List[str]]): Optional list of document IDs. If None, ids will be auto created description (str): A description for your project. is_public (bool): Whether your project is publicly accessible. True by default. reset_project_if_exists (bool): Whether to reset this project if it already exists. Default False. Generally userful during development and testing. index_kwargs (Optional[dict]): Dict of kwargs for index creation. See https://docs.nomic.ai/atlas_api.html Returns: AtlasDB: Nomic's neural database and finest rhizomatic instrument """ if name is None or api_key is None: raise ValueError("`name` and `api_key` cannot be None.") texts = [doc.page_content for doc in documents] metadatas = [doc.metadata for doc in documents] return cls.from_texts( name=name, api_key=api_key, texts=texts, embedding=embedding, metadatas=metadatas, ids=ids, description=description, is_public=is_public, reset_project_if_exists=reset_project_if_exists, index_kwargs=index_kwargs, )
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/atlas.html
021d3dad-6235-48c2-a7ad-042f4258d118
Source code for langchain.vectorstores.singlestoredb """Wrapper around SingleStore DB.""" from __future__ import annotations import enum import json from typing import ( Any, ClassVar, Collection, Iterable, List, Optional, Tuple, Type, ) from sqlalchemy.pool import QueuePool from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.vectorstores.base import VectorStore, VectorStoreRetriever class DistanceStrategy(str, enum.Enum): """Enumerator of the Distance strategies for SingleStoreDB.""" EUCLIDEAN_DISTANCE = "EUCLIDEAN_DISTANCE" DOT_PRODUCT = "DOT_PRODUCT" DEFAULT_DISTANCE_STRATEGY = DistanceStrategy.DOT_PRODUCT ORDERING_DIRECTIVE: dict = { DistanceStrategy.EUCLIDEAN_DISTANCE: "", DistanceStrategy.DOT_PRODUCT: "DESC", } [docs]class SingleStoreDB(VectorStore): """ This class serves as a Pythonic interface to the SingleStore DB database. The prerequisite for using this class is the installation of the ``singlestoredb`` Python package. The SingleStoreDB vectorstore can be created by providing an embedding function and the relevant parameters for the database connection, connection pool, and optionally, the names of the table and the fields to use. """ def _get_connection(self: SingleStoreDB) -> Any: try: import singlestoredb as s2 except ImportError: raise ImportError( "Could not import singlestoredb python package. " "Please install it with `pip install singlestoredb`." ) return s2.connect(**self.connection_kwargs) def __init__( self, embedding: Embeddings, *, distance_strategy: DistanceStrategy = DEFAULT_DISTANCE_STRATEGY, table_name: str = "embeddings", content_field: str = "content", metadata_field: str = "metadata", vector_field: str = "vector", pool_size: int = 5, max_overflow: int = 10, timeout: float = 30, **kwargs: Any, ): """Initialize with necessary components. Args: embedding (Embeddings): A text embedding model. distance_strategy (DistanceStrategy, optional): Determines the strategy employed for calculating the distance between vectors in the embedding space. Defaults to DOT_PRODUCT. Available options are: - DOT_PRODUCT: Computes the scalar product of two vectors. This is the default behavior - EUCLIDEAN_DISTANCE: Computes the Euclidean distance between two vectors. This metric considers the geometric distance in the vector space, and might be more suitable for embeddings that rely on spatial relationships. table_name (str, optional): Specifies the name of the table in use. Defaults to "embeddings". content_field (str, optional): Specifies the field to store the content. Defaults to "content". metadata_field (str, optional): Specifies the field to store metadata. Defaults to "metadata". vector_field (str, optional): Specifies the field to store the vector. Defaults to "vector". Following arguments pertain to the connection pool: pool_size (int, optional): Determines the number of active connections in the pool. Defaults to 5. max_overflow (int, optional): Determines the maximum number of connections allowed beyond the pool_size. Defaults to 10. timeout (float, optional): Specifies the maximum wait time in seconds for establishing a connection. Defaults to 30. Following arguments pertain to the database connection: host (str, optional): Specifies the hostname, IP address, or URL for the database connection. The default scheme is "mysql". user (str, optional): Database username. password (str, optional): Database password. port (int, optional): Database port. Defaults to 3306 for non-HTTP connections, 80 for HTTP connections, and 443 for HTTPS connections. database (str, optional): Database name. Additional optional arguments provide further customization over the database connection: pure_python (bool, optional): Toggles the connector mode. If True, operates in pure Python mode. local_infile (bool, optional): Allows local file uploads. charset (str, optional): Specifies the character set for string values. ssl_key (str, optional): Specifies the path of the file containing the SSL key. ssl_cert (str, optional): Specifies the path of the file containing the SSL certificate. ssl_ca (str, optional): Specifies the path of the file containing the SSL certificate authority. ssl_cipher (str, optional): Sets the SSL cipher list. ssl_disabled (bool, optional): Disables SSL usage. ssl_verify_cert (bool, optional): Verifies the server's certificate. Automatically enabled if ``ssl_ca`` is specified. ssl_verify_identity (bool, optional): Verifies the server's identity. conv (dict[int, Callable], optional): A dictionary of data conversion functions. credential_type (str, optional): Specifies the type of authentication to use: auth.PASSWORD, auth.JWT, or auth.BROWSER_SSO. autocommit (bool, optional): Enables autocommits. results_type (str, optional): Determines the structure of the query results: tuples, namedtuples, dicts. results_format (str, optional): Deprecated. This option has been renamed to results_type. Examples: Basic Usage: .. code-block:: python from langchain.embeddings import OpenAIEmbeddings from langchain.vectorstores import SingleStoreDB vectorstore = SingleStoreDB( OpenAIEmbeddings(), host="https://user:password@127.0.0.1:3306/database" ) Advanced Usage: .. code-block:: python from langchain.embeddings import OpenAIEmbeddings from langchain.vectorstores import SingleStoreDB vectorstore = SingleStoreDB( OpenAIEmbeddings(), distance_strategy=DistanceStrategy.EUCLIDEAN_DISTANCE, host="127.0.0.1", port=3306, user="user", password="password", database="db", table_name="my_custom_table", pool_size=10, timeout=60, ) Using environment variables: .. code-block:: python from langchain.embeddings import OpenAIEmbeddings from langchain.vectorstores import SingleStoreDB os.environ['SINGLESTOREDB_URL'] = 'me:p455w0rd@s2-host.com/my_db' vectorstore = SingleStoreDB(OpenAIEmbeddings()) """ self.embedding = embedding self.distance_strategy = distance_strategy self.table_name = table_name self.content_field = content_field self.metadata_field = metadata_field self.vector_field = vector_field """Pass the rest of the kwargs to the connection.""" self.connection_kwargs = kwargs """Add program name and version to connection attributes.""" if "conn_attrs" not in self.connection_kwargs: self.connection_kwargs["conn_attrs"] = dict() if "program_name" not in self.connection_kwargs["conn_attrs"]: self.connection_kwargs["conn_attrs"][ "program_name" ] = "langchain python sdk" self.connection_kwargs["conn_attrs"][ "program_version" ] = "0.0.205" # the version of SingleStoreDB VectorStore implementation """Create connection pool.""" self.connection_pool = QueuePool( self._get_connection, max_overflow=max_overflow, pool_size=pool_size, timeout=timeout, ) self._create_table() def _create_table(self: SingleStoreDB) -> None: """Create table if it doesn't exist.""" conn = self.connection_pool.connect() try: cur = conn.cursor() try: cur.execute( """CREATE TABLE IF NOT EXISTS {} ({} TEXT CHARACTER SET utf8mb4 COLLATE utf8mb4_general_ci, {} BLOB, {} JSON);""".format( self.table_name, self.content_field, self.vector_field, self.metadata_field, ), ) finally: cur.close() finally: conn.close() [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, embeddings: Optional[List[List[float]]] = None, **kwargs: Any, ) -> List[str]: """Add more texts to the vectorstore. Args: texts (Iterable[str]): Iterable of strings/text to add to the vectorstore. metadatas (Optional[List[dict]], optional): Optional list of metadatas. Defaults to None. embeddings (Optional[List[List[float]]], optional): Optional pre-generated embeddings. Defaults to None. Returns: List[str]: empty list """ conn = self.connection_pool.connect() try: cur = conn.cursor() try: # Write data to singlestore db for i, text in enumerate(texts): # Use provided values by default or fallback metadata = metadatas[i] if metadatas else {} embedding = ( embeddings[i] if embeddings else self.embedding.embed_documents([text])[0] ) cur.execute( "INSERT INTO {} VALUES (%s, JSON_ARRAY_PACK(%s), %s)".format( self.table_name ), ( text, "[{}]".format(",".join(map(str, embedding))), json.dumps(metadata), ), ) finally: cur.close() finally: conn.close() return [] [docs] def similarity_search( self, query: str, k: int = 4, filter: Optional[dict] = None, **kwargs: Any ) -> List[Document]: """Returns the most similar indexed documents to the query text. Uses cosine similarity. Args: query (str): The query text for which to find similar documents. k (int): The number of documents to return. Default is 4. filter (dict): A dictionary of metadata fields and values to filter by. Returns: List[Document]: A list of documents that are most similar to the query text. Examples: .. code-block:: python from langchain.vectorstores import SingleStoreDB from langchain.embeddings import OpenAIEmbeddings s2 = SingleStoreDB.from_documents( docs, OpenAIEmbeddings(), host="username:password@localhost:3306/database" ) s2.similarity_search("query text", 1, {"metadata_field": "metadata_value"}) """ docs_and_scores = self.similarity_search_with_score( query=query, k=k, filter=filter ) return [doc for doc, _ in docs_and_scores] [docs] def similarity_search_with_score( self, query: str, k: int = 4, filter: Optional[dict] = None ) -> List[Tuple[Document, float]]: """Return docs most similar to query. Uses cosine similarity. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. filter: A dictionary of metadata fields and values to filter by. Defaults to None. Returns: List of Documents most similar to the query and score for each """ # Creates embedding vector from user query embedding = self.embedding.embed_query(query) conn = self.connection_pool.connect() result = [] where_clause: str = "" where_clause_values: List[Any] = [] if filter: where_clause = "WHERE " arguments = [] def build_where_clause( where_clause_values: List[Any], sub_filter: dict, prefix_args: List[str] = [], ) -> None: for key in sub_filter.keys(): if isinstance(sub_filter[key], dict): build_where_clause( where_clause_values, sub_filter[key], prefix_args + [key] ) else: arguments.append( "JSON_EXTRACT_JSON({}, {}) = %s".format( self.metadata_field, ", ".join(["%s"] * (len(prefix_args) + 1)), ) ) where_clause_values += prefix_args + [key] where_clause_values.append(json.dumps(sub_filter[key])) build_where_clause(where_clause_values, filter) where_clause += " AND ".join(arguments) try: cur = conn.cursor() try: cur.execute( """SELECT {}, {}, {}({}, JSON_ARRAY_PACK(%s)) as __score FROM {} {} ORDER BY __score {} LIMIT %s""".format( self.content_field, self.metadata_field, self.distance_strategy, self.vector_field, self.table_name, where_clause, ORDERING_DIRECTIVE[self.distance_strategy], ), ("[{}]".format(",".join(map(str, embedding))),) + tuple(where_clause_values) + (k,), ) for row in cur.fetchall(): doc = Document(page_content=row[0], metadata=row[1]) result.append((doc, float(row[2]))) finally: cur.close() finally: conn.close() return result [docs] @classmethod def from_texts( cls: Type[SingleStoreDB], texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, distance_strategy: DistanceStrategy = DEFAULT_DISTANCE_STRATEGY, table_name: str = "embeddings", content_field: str = "content", metadata_field: str = "metadata", vector_field: str = "vector", pool_size: int = 5, max_overflow: int = 10, timeout: float = 30, **kwargs: Any, ) -> SingleStoreDB: """Create a SingleStoreDB vectorstore from raw documents. This is a user-friendly interface that: 1. Embeds documents. 2. Creates a new table for the embeddings in SingleStoreDB. 3. Adds the documents to the newly created table. This is intended to be a quick way to get started. Example: .. code-block:: python from langchain.vectorstores import SingleStoreDB from langchain.embeddings import OpenAIEmbeddings s2 = SingleStoreDB.from_texts( texts, OpenAIEmbeddings(), host="username:password@localhost:3306/database" ) """ instance = cls( embedding, distance_strategy=distance_strategy, table_name=table_name, content_field=content_field, metadata_field=metadata_field, vector_field=vector_field, pool_size=pool_size, max_overflow=max_overflow, timeout=timeout, **kwargs, ) instance.add_texts(texts, metadatas, embedding.embed_documents(texts), **kwargs) return instance [docs] def as_retriever(self, **kwargs: Any) -> SingleStoreDBRetriever: return SingleStoreDBRetriever(vectorstore=self, **kwargs) class SingleStoreDBRetriever(VectorStoreRetriever): """Retriever for SingleStoreDB vector stores.""" vectorstore: SingleStoreDB k: int = 4 allowed_search_types: ClassVar[Collection[str]] = ("similarity",) def get_relevant_documents(self, query: str) -> List[Document]: if self.search_type == "similarity": docs = self.vectorstore.similarity_search(query, k=self.k) else: raise ValueError(f"search_type of {self.search_type} not allowed.") return docs async def aget_relevant_documents(self, query: str) -> List[Document]: raise NotImplementedError( "SingleStoreDBVectorStoreRetriever does not support async" )
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/singlestoredb.html
9ed70bfe-e90e-47c2-801b-d021d45adfcc
Source code for langchain.vectorstores.weaviate """Wrapper around weaviate vector database.""" from __future__ import annotations import datetime from typing import Any, Callable, Dict, Iterable, List, Optional, Tuple, Type from uuid import uuid4 import numpy as np from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.utils import get_from_dict_or_env from langchain.vectorstores.base import VectorStore from langchain.vectorstores.utils import maximal_marginal_relevance def _default_schema(index_name: str) -> Dict: return { "class": index_name, "properties": [ { "name": "text", "dataType": ["text"], } ], } def _create_weaviate_client(**kwargs: Any) -> Any: client = kwargs.get("client") if client is not None: return client weaviate_url = get_from_dict_or_env(kwargs, "weaviate_url", "WEAVIATE_URL") try: # the weaviate api key param should not be mandatory weaviate_api_key = get_from_dict_or_env( kwargs, "weaviate_api_key", "WEAVIATE_API_KEY", None ) except ValueError: weaviate_api_key = None try: import weaviate except ImportError: raise ValueError( "Could not import weaviate python package. " "Please install it with `pip install weaviate-client`" ) auth = ( weaviate.auth.AuthApiKey(api_key=weaviate_api_key) if weaviate_api_key is not None else None ) client = weaviate.Client(weaviate_url, auth_client_secret=auth) return client def _default_score_normalizer(val: float) -> float: return 1 - 1 / (1 + np.exp(val)) def _json_serializable(value: Any) -> Any: if isinstance(value, datetime.datetime): return value.isoformat() return value [docs]class Weaviate(VectorStore): """Wrapper around Weaviate vector database. To use, you should have the ``weaviate-client`` python package installed. Example: .. code-block:: python import weaviate from langchain.vectorstores import Weaviate client = weaviate.Client(url=os.environ["WEAVIATE_URL"], ...) weaviate = Weaviate(client, index_name, text_key) """ def __init__( self, client: Any, index_name: str, text_key: str, embedding: Optional[Embeddings] = None, attributes: Optional[List[str]] = None, relevance_score_fn: Optional[ Callable[[float], float] ] = _default_score_normalizer, by_text: bool = True, ): """Initialize with Weaviate client.""" try: import weaviate except ImportError: raise ValueError( "Could not import weaviate python package. " "Please install it with `pip install weaviate-client`." ) if not isinstance(client, weaviate.Client): raise ValueError( f"client should be an instance of weaviate.Client, got {type(client)}" ) self._client = client self._index_name = index_name self._embedding = embedding self._text_key = text_key self._query_attrs = [self._text_key] self._relevance_score_fn = relevance_score_fn self._by_text = by_text if attributes is not None: self._query_attrs.extend(attributes) [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, **kwargs: Any, ) -> List[str]: """Upload texts with metadata (properties) to Weaviate.""" from weaviate.util import get_valid_uuid ids = [] with self._client.batch as batch: for i, text in enumerate(texts): data_properties = {self._text_key: text} if metadatas is not None: for key, val in metadatas[i].items(): data_properties[key] = _json_serializable(val) # Allow for ids (consistent w/ other methods) # # Or uuids (backwards compatble w/ existing arg) # If the UUID of one of the objects already exists # then the existing object will be replaced by the new object. _id = get_valid_uuid(uuid4()) if "uuids" in kwargs: _id = kwargs["uuids"][i] elif "ids" in kwargs: _id = kwargs["ids"][i] if self._embedding is not None: vector = self._embedding.embed_documents([text])[0] else: vector = None batch.add_data_object( data_object=data_properties, class_name=self._index_name, uuid=_id, vector=vector, ) ids.append(_id) return ids [docs] def similarity_search( self, query: str, k: int = 4, **kwargs: Any ) -> List[Document]: """Return docs most similar to query. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. Returns: List of Documents most similar to the query. """ if self._by_text: return self.similarity_search_by_text(query, k, **kwargs) else: if self._embedding is None: raise ValueError( "_embedding cannot be None for similarity_search when " "_by_text=False" ) embedding = self._embedding.embed_query(query) return self.similarity_search_by_vector(embedding, k, **kwargs) [docs] def similarity_search_by_text( self, query: str, k: int = 4, **kwargs: Any ) -> List[Document]: """Return docs most similar to query. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. Returns: List of Documents most similar to the query. """ content: Dict[str, Any] = {"concepts": [query]} if kwargs.get("search_distance"): content["certainty"] = kwargs.get("search_distance") query_obj = self._client.query.get(self._index_name, self._query_attrs) if kwargs.get("where_filter"): query_obj = query_obj.with_where(kwargs.get("where_filter")) if kwargs.get("additional"): query_obj = query_obj.with_additional(kwargs.get("additional")) result = query_obj.with_near_text(content).with_limit(k).do() if "errors" in result: raise ValueError(f"Error during query: {result['errors']}") docs = [] for res in result["data"]["Get"][self._index_name]: text = res.pop(self._text_key) docs.append(Document(page_content=text, metadata=res)) return docs [docs] def similarity_search_by_vector( self, embedding: List[float], k: int = 4, **kwargs: Any ) -> List[Document]: """Look up similar documents by embedding vector in Weaviate.""" vector = {"vector": embedding} query_obj = self._client.query.get(self._index_name, self._query_attrs) if kwargs.get("where_filter"): query_obj = query_obj.with_where(kwargs.get("where_filter")) if kwargs.get("additional"): query_obj = query_obj.with_additional(kwargs.get("additional")) result = query_obj.with_near_vector(vector).with_limit(k).do() if "errors" in result: raise ValueError(f"Error during query: {result['errors']}") docs = [] for res in result["data"]["Get"][self._index_name]: text = res.pop(self._text_key) docs.append(Document(page_content=text, metadata=res)) return docs [docs] def max_marginal_relevance_search( self, query: str, k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. fetch_k: Number of Documents to fetch to pass to MMR algorithm. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. Returns: List of Documents selected by maximal marginal relevance. """ if self._embedding is not None: embedding = self._embedding.embed_query(query) else: raise ValueError( "max_marginal_relevance_search requires a suitable Embeddings object" ) return self.max_marginal_relevance_search_by_vector( embedding, k=k, fetch_k=fetch_k, lambda_mult=lambda_mult, **kwargs ) [docs] def max_marginal_relevance_search_by_vector( self, embedding: List[float], k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: embedding: Embedding to look up documents similar to. k: Number of Documents to return. Defaults to 4. fetch_k: Number of Documents to fetch to pass to MMR algorithm. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. Returns: List of Documents selected by maximal marginal relevance. """ vector = {"vector": embedding} query_obj = self._client.query.get(self._index_name, self._query_attrs) if kwargs.get("where_filter"): query_obj = query_obj.with_where(kwargs.get("where_filter")) results = ( query_obj.with_additional("vector") .with_near_vector(vector) .with_limit(fetch_k) .do() ) payload = results["data"]["Get"][self._index_name] embeddings = [result["_additional"]["vector"] for result in payload] mmr_selected = maximal_marginal_relevance( np.array(embedding), embeddings, k=k, lambda_mult=lambda_mult ) docs = [] for idx in mmr_selected: text = payload[idx].pop(self._text_key) payload[idx].pop("_additional") meta = payload[idx] docs.append(Document(page_content=text, metadata=meta)) return docs [docs] def similarity_search_with_score( self, query: str, k: int = 4, **kwargs: Any ) -> List[Tuple[Document, float]]: """ Return list of documents most similar to the query text and cosine distance in float for each. Lower score represents more similarity. """ if self._embedding is None: raise ValueError( "_embedding cannot be None for similarity_search_with_score" ) content: Dict[str, Any] = {"concepts": [query]} if kwargs.get("search_distance"): content["certainty"] = kwargs.get("search_distance") query_obj = self._client.query.get(self._index_name, self._query_attrs) if not self._by_text: embedding = self._embedding.embed_query(query) vector = {"vector": embedding} result = ( query_obj.with_near_vector(vector) .with_limit(k) .with_additional("vector") .do() ) else: result = ( query_obj.with_near_text(content) .with_limit(k) .with_additional("vector") .do() ) if "errors" in result: raise ValueError(f"Error during query: {result['errors']}") docs_and_scores = [] for res in result["data"]["Get"][self._index_name]: text = res.pop(self._text_key) score = np.dot( res["_additional"]["vector"], self._embedding.embed_query(query) ) docs_and_scores.append((Document(page_content=text, metadata=res), score)) return docs_and_scores def _similarity_search_with_relevance_scores( self, query: str, k: int = 4, **kwargs: Any, ) -> List[Tuple[Document, float]]: """Return docs and relevance scores, normalized on a scale from 0 to 1. 0 is dissimilar, 1 is most similar. """ if self._relevance_score_fn is None: raise ValueError( "relevance_score_fn must be provided to" " Weaviate constructor to normalize scores" ) docs_and_scores = self.similarity_search_with_score(query, k=k, **kwargs) return [ (doc, self._relevance_score_fn(score)) for doc, score in docs_and_scores ] [docs] @classmethod def from_texts( cls: Type[Weaviate], texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, **kwargs: Any, ) -> Weaviate: """Construct Weaviate wrapper from raw documents. This is a user-friendly interface that: 1. Embeds documents. 2. Creates a new index for the embeddings in the Weaviate instance. 3. Adds the documents to the newly created Weaviate index. This is intended to be a quick way to get started. Example: .. code-block:: python from langchain.vectorstores.weaviate import Weaviate from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() weaviate = Weaviate.from_texts( texts, embeddings, weaviate_url="http://localhost:8080" ) """ client = _create_weaviate_client(**kwargs) from weaviate.util import get_valid_uuid index_name = kwargs.get("index_name", f"LangChain_{uuid4().hex}") embeddings = embedding.embed_documents(texts) if embedding else None text_key = "text" schema = _default_schema(index_name) attributes = list(metadatas[0].keys()) if metadatas else None # check whether the index already exists if not client.schema.contains(schema): client.schema.create_class(schema) with client.batch as batch: for i, text in enumerate(texts): data_properties = { text_key: text, } if metadatas is not None: for key in metadatas[i].keys(): data_properties[key] = metadatas[i][key] # If the UUID of one of the objects already exists # then the existing objectwill be replaced by the new object. if "uuids" in kwargs: _id = kwargs["uuids"][i] else: _id = get_valid_uuid(uuid4()) # if an embedding strategy is not provided, we let # weaviate create the embedding. Note that this will only # work if weaviate has been installed with a vectorizer module # like text2vec-contextionary for example params = { "uuid": _id, "data_object": data_properties, "class_name": index_name, } if embeddings is not None: params["vector"] = embeddings[i] batch.add_data_object(**params) batch.flush() relevance_score_fn = kwargs.get("relevance_score_fn") by_text: bool = kwargs.get("by_text", False) return cls( client, index_name, text_key, embedding=embedding, attributes=attributes, relevance_score_fn=relevance_score_fn, by_text=by_text, ) [docs] def delete(self, ids: List[str]) -> None: """Delete by vector IDs. Args: ids: List of ids to delete. """ # TODO: Check if this can be done in bulk for id in ids: self._client.data_object.delete(uuid=id)
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/weaviate.html
6182d2e5-b824-4e42-82d5-957546a1ead0
Source code for langchain.vectorstores.myscale """Wrapper around MyScale vector database.""" from __future__ import annotations import json import logging from hashlib import sha1 from threading import Thread from typing import Any, Dict, Iterable, List, Optional, Tuple from pydantic import BaseSettings from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.vectorstores.base import VectorStore logger = logging.getLogger() def has_mul_sub_str(s: str, *args: Any) -> bool: """ Check if a string contains multiple substrings. Args: s: string to check. *args: substrings to check. Returns: True if all substrings are in the string, False otherwise. """ for a in args: if a not in s: return False return True [docs]class MyScaleSettings(BaseSettings): """MyScale Client Configuration Attribute: myscale_host (str) : An URL to connect to MyScale backend. Defaults to 'localhost'. myscale_port (int) : URL port to connect with HTTP. Defaults to 8443. username (str) : Username to login. Defaults to None. password (str) : Password to login. Defaults to None. index_type (str): index type string. index_param (dict): index build parameter. database (str) : Database name to find the table. Defaults to 'default'. table (str) : Table name to operate on. Defaults to 'vector_table'. metric (str) : Metric to compute distance, supported are ('l2', 'cosine', 'ip'). Defaults to 'cosine'. column_map (Dict) : Column type map to project column name onto langchain semantics. Must have keys: `text`, `id`, `vector`, must be same size to number of columns. For example: .. code-block:: python { 'id': 'text_id', 'vector': 'text_embedding', 'text': 'text_plain', 'metadata': 'metadata_dictionary_in_json', } Defaults to identity map. """ host: str = "localhost" port: int = 8443 username: Optional[str] = None password: Optional[str] = None index_type: str = "IVFFLAT" index_param: Optional[Dict[str, str]] = None column_map: Dict[str, str] = { "id": "id", "text": "text", "vector": "vector", "metadata": "metadata", } database: str = "default" table: str = "langchain" metric: str = "cosine" def __getitem__(self, item: str) -> Any: return getattr(self, item) class Config: env_file = ".env" env_prefix = "myscale_" env_file_encoding = "utf-8" [docs]class MyScale(VectorStore): """Wrapper around MyScale vector database You need a `clickhouse-connect` python package, and a valid account to connect to MyScale. MyScale can not only search with simple vector indexes, it also supports complex query with multiple conditions, constraints and even sub-queries. For more information, please visit [myscale official site](https://docs.myscale.com/en/overview/) """ def __init__( self, embedding: Embeddings, config: Optional[MyScaleSettings] = None, **kwargs: Any, ) -> None: """MyScale Wrapper to LangChain embedding_function (Embeddings): config (MyScaleSettings): Configuration to MyScale Client Other keyword arguments will pass into [clickhouse-connect](https://docs.myscale.com/) """ try: from clickhouse_connect import get_client except ImportError: raise ValueError( "Could not import clickhouse connect python package. " "Please install it with `pip install clickhouse-connect`." ) try: from tqdm import tqdm self.pgbar = tqdm except ImportError: # Just in case if tqdm is not installed self.pgbar = lambda x: x super().__init__() if config is not None: self.config = config else: self.config = MyScaleSettings() assert self.config assert self.config.host and self.config.port assert ( self.config.column_map and self.config.database and self.config.table and self.config.metric ) for k in ["id", "vector", "text", "metadata"]: assert k in self.config.column_map assert self.config.metric in ["ip", "cosine", "l2"] # initialize the schema dim = len(embedding.embed_query("try this out")) index_params = ( ", " + ",".join([f"'{k}={v}'" for k, v in self.config.index_param.items()]) if self.config.index_param else "" ) schema_ = f""" CREATE TABLE IF NOT EXISTS {self.config.database}.{self.config.table}( {self.config.column_map['id']} String, {self.config.column_map['text']} String, {self.config.column_map['vector']} Array(Float32), {self.config.column_map['metadata']} JSON, CONSTRAINT cons_vec_len CHECK length(\ {self.config.column_map['vector']}) = {dim}, VECTOR INDEX vidx {self.config.column_map['vector']} \ TYPE {self.config.index_type}(\ 'metric_type={self.config.metric}'{index_params}) ) ENGINE = MergeTree ORDER BY {self.config.column_map['id']} """ self.dim = dim self.BS = "\\" self.must_escape = ("\\", "'") self.embedding_function = embedding.embed_query self.dist_order = "ASC" if self.config.metric in ["cosine", "l2"] else "DESC" # Create a connection to myscale self.client = get_client( host=self.config.host, port=self.config.port, username=self.config.username, password=self.config.password, **kwargs, ) self.client.command("SET allow_experimental_object_type=1") self.client.command(schema_) [docs] def escape_str(self, value: str) -> str: return "".join(f"{self.BS}{c}" if c in self.must_escape else c for c in value) def _build_istr(self, transac: Iterable, column_names: Iterable[str]) -> str: ks = ",".join(column_names) _data = [] for n in transac: n = ",".join([f"'{self.escape_str(str(_n))}'" for _n in n]) _data.append(f"({n})") i_str = f""" INSERT INTO TABLE {self.config.database}.{self.config.table}({ks}) VALUES {','.join(_data)} """ return i_str def _insert(self, transac: Iterable, column_names: Iterable[str]) -> None: _i_str = self._build_istr(transac, column_names) self.client.command(_i_str) [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, batch_size: int = 32, ids: Optional[Iterable[str]] = None, **kwargs: Any, ) -> List[str]: """Run more texts through the embeddings and add to the vectorstore. Args: texts: Iterable of strings to add to the vectorstore. ids: Optional list of ids to associate with the texts. batch_size: Batch size of insertion metadata: Optional column data to be inserted Returns: List of ids from adding the texts into the vectorstore. """ # Embed and create the documents ids = ids or [sha1(t.encode("utf-8")).hexdigest() for t in texts] colmap_ = self.config.column_map transac = [] column_names = { colmap_["id"]: ids, colmap_["text"]: texts, colmap_["vector"]: map(self.embedding_function, texts), } metadatas = metadatas or [{} for _ in texts] column_names[colmap_["metadata"]] = map(json.dumps, metadatas) assert len(set(colmap_) - set(column_names)) >= 0 keys, values = zip(*column_names.items()) try: t = None for v in self.pgbar( zip(*values), desc="Inserting data...", total=len(metadatas) ): assert len(v[keys.index(self.config.column_map["vector"])]) == self.dim transac.append(v) if len(transac) == batch_size: if t: t.join() t = Thread(target=self._insert, args=[transac, keys]) t.start() transac = [] if len(transac) > 0: if t: t.join() self._insert(transac, keys) return [i for i in ids] except Exception as e: logger.error(f"\033[91m\033[1m{type(e)}\033[0m \033[95m{str(e)}\033[0m") return [] [docs] @classmethod def from_texts( cls, texts: List[str], embedding: Embeddings, metadatas: Optional[List[Dict[Any, Any]]] = None, config: Optional[MyScaleSettings] = None, text_ids: Optional[Iterable[str]] = None, batch_size: int = 32, **kwargs: Any, ) -> MyScale: """Create Myscale wrapper with existing texts Args: embedding_function (Embeddings): Function to extract text embedding texts (Iterable[str]): List or tuple of strings to be added config (MyScaleSettings, Optional): Myscale configuration text_ids (Optional[Iterable], optional): IDs for the texts. Defaults to None. batch_size (int, optional): Batchsize when transmitting data to MyScale. Defaults to 32. metadata (List[dict], optional): metadata to texts. Defaults to None. Other keyword arguments will pass into [clickhouse-connect](https://clickhouse.com/docs/en/integrations/python#clickhouse-connect-driver-api) Returns: MyScale Index """ ctx = cls(embedding, config, **kwargs) ctx.add_texts(texts, ids=text_ids, batch_size=batch_size, metadatas=metadatas) return ctx def __repr__(self) -> str: """Text representation for myscale, prints backends, username and schemas. Easy to use with `str(Myscale())` Returns: repr: string to show connection info and data schema """ _repr = f"\033[92m\033[1m{self.config.database}.{self.config.table} @ " _repr += f"{self.config.host}:{self.config.port}\033[0m\n\n" _repr += f"\033[1musername: {self.config.username}\033[0m\n\nTable Schema:\n" _repr += "-" * 51 + "\n" for r in self.client.query( f"DESC {self.config.database}.{self.config.table}" ).named_results(): _repr += ( f"|\033[94m{r['name']:24s}\033[0m|\033[96m{r['type']:24s}\033[0m|\n" ) _repr += "-" * 51 + "\n" return _repr def _build_qstr( self, q_emb: List[float], topk: int, where_str: Optional[str] = None ) -> str: q_emb_str = ",".join(map(str, q_emb)) if where_str: where_str = f"PREWHERE {where_str}" else: where_str = "" q_str = f""" SELECT {self.config.column_map['text']}, {self.config.column_map['metadata']}, dist FROM {self.config.database}.{self.config.table} {where_str} ORDER BY distance({self.config.column_map['vector']}, [{q_emb_str}]) AS dist {self.dist_order} LIMIT {topk} """ return q_str [docs] def similarity_search( self, query: str, k: int = 4, where_str: Optional[str] = None, **kwargs: Any ) -> List[Document]: """Perform a similarity search with MyScale Args: query (str): query string k (int, optional): Top K neighbors to retrieve. Defaults to 4. where_str (Optional[str], optional): where condition string. Defaults to None. NOTE: Please do not let end-user to fill this and always be aware of SQL injection. When dealing with metadatas, remember to use `{self.metadata_column}.attribute` instead of `attribute` alone. The default name for it is `metadata`. Returns: List[Document]: List of Documents """ return self.similarity_search_by_vector( self.embedding_function(query), k, where_str, **kwargs ) [docs] def similarity_search_by_vector( self, embedding: List[float], k: int = 4, where_str: Optional[str] = None, **kwargs: Any, ) -> List[Document]: """Perform a similarity search with MyScale by vectors Args: query (str): query string k (int, optional): Top K neighbors to retrieve. Defaults to 4. where_str (Optional[str], optional): where condition string. Defaults to None. NOTE: Please do not let end-user to fill this and always be aware of SQL injection. When dealing with metadatas, remember to use `{self.metadata_column}.attribute` instead of `attribute` alone. The default name for it is `metadata`. Returns: List[Document]: List of (Document, similarity) """ q_str = self._build_qstr(embedding, k, where_str) try: return [ Document( page_content=r[self.config.column_map["text"]], metadata=r[self.config.column_map["metadata"]], ) for r in self.client.query(q_str).named_results() ] except Exception as e: logger.error(f"\033[91m\033[1m{type(e)}\033[0m \033[95m{str(e)}\033[0m") return [] [docs] def similarity_search_with_relevance_scores( self, query: str, k: int = 4, where_str: Optional[str] = None, **kwargs: Any ) -> List[Tuple[Document, float]]: """Perform a similarity search with MyScale Args: query (str): query string k (int, optional): Top K neighbors to retrieve. Defaults to 4. where_str (Optional[str], optional): where condition string. Defaults to None. NOTE: Please do not let end-user to fill this and always be aware of SQL injection. When dealing with metadatas, remember to use `{self.metadata_column}.attribute` instead of `attribute` alone. The default name for it is `metadata`. Returns: List[Document]: List of documents most similar to the query text and cosine distance in float for each. Lower score represents more similarity. """ q_str = self._build_qstr(self.embedding_function(query), k, where_str) try: return [ ( Document( page_content=r[self.config.column_map["text"]], metadata=r[self.config.column_map["metadata"]], ), r["dist"], ) for r in self.client.query(q_str).named_results() ] except Exception as e: logger.error(f"\033[91m\033[1m{type(e)}\033[0m \033[95m{str(e)}\033[0m") return [] [docs] def drop(self) -> None: """ Helper function: Drop data """ self.client.command( f"DROP TABLE IF EXISTS {self.config.database}.{self.config.table}" ) @property def metadata_column(self) -> str: return self.config.column_map["metadata"]
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/myscale.html
e792d280-4f29-4858-a8c7-aaa8d27403f3
Source code for langchain.vectorstores.deeplake """Wrapper around Activeloop Deep Lake.""" from __future__ import annotations import logging from typing import Any, Callable, Dict, Iterable, List, Optional, Tuple, Union import numpy as np try: import deeplake from deeplake.core.fast_forwarding import version_compare from deeplake.core.vectorstore import DeepLakeVectorStore _DEEPLAKE_INSTALLED = True except ImportError: _DEEPLAKE_INSTALLED = False from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.vectorstores.base import VectorStore from langchain.vectorstores.utils import maximal_marginal_relevance logger = logging.getLogger(__name__) [docs]class DeepLake(VectorStore): """Wrapper around Deep Lake, a data lake for deep learning applications. We integrated deeplake's similarity search and filtering for fast prototyping, Now, it supports Tensor Query Language (TQL) for production use cases over billion rows. Why Deep Lake? - Not only stores embeddings, but also the original data with version control. - Serverless, doesn't require another service and can be used with major cloud providers (S3, GCS, etc.) - More than just a multi-modal vector store. You can use the dataset to fine-tune your own LLM models. To use, you should have the ``deeplake`` python package installed. Example: .. code-block:: python from langchain.vectorstores import DeepLake from langchain.embeddings.openai import OpenAIEmbeddings embeddings = OpenAIEmbeddings() vectorstore = DeepLake("langchain_store", embeddings.embed_query) """ _LANGCHAIN_DEFAULT_DEEPLAKE_PATH = "./deeplake/" def __init__( self, dataset_path: str = _LANGCHAIN_DEFAULT_DEEPLAKE_PATH, token: Optional[str] = None, embedding_function: Optional[Embeddings] = None, read_only: bool = False, ingestion_batch_size: int = 1000, num_workers: int = 0, verbose: bool = True, exec_option: str = "python", **kwargs: Any, ) -> None: """Creates an empty DeepLakeVectorStore or loads an existing one. The DeepLakeVectorStore is located at the specified ``path``. Examples: >>> # Create a vector store with default tensors >>> deeplake_vectorstore = DeepLake( ... path = <path_for_storing_Data>, ... ) >>> >>> # Create a vector store in the Deep Lake Managed Tensor Database >>> data = DeepLake( ... path = "hub://org_id/dataset_name", ... exec_option = "tensor_db", ... ) Args: dataset_path (str): Path to existing dataset or where to create a new one. Defaults to _LANGCHAIN_DEFAULT_DEEPLAKE_PATH. token (str, optional): Activeloop token, for fetching credentials to the dataset at path if it is a Deep Lake dataset. Tokens are normally autogenerated. Optional. embedding_function (str, optional): Function to convert either documents or query. Optional. read_only (bool): Open dataset in read-only mode. Default is False. ingestion_batch_size (int): During data ingestion, data is divided into batches. Batch size is the size of each batch. Default is 1000. num_workers (int): Number of workers to use during data ingestion. Default is 0. verbose (bool): Print dataset summary after each operation. Default is True. exec_option (str): DeepLakeVectorStore supports 3 ways to perform searching - "python", "compute_engine", "tensor_db". Default is "python". - ``python`` - Pure-python implementation that runs on the client. WARNING: using this with big datasets can lead to memory issues. Data can be stored anywhere. - ``compute_engine`` - C++ implementation of the Deep Lake Compute Engine that runs on the client. Can be used for any data stored in or connected to Deep Lake. Not for in-memory or local datasets. - ``tensor_db`` - Hosted Managed Tensor Database that is responsible for storage and query execution. Only for data stored in the Deep Lake Managed Database. Use runtime = {"db_engine": True} during dataset creation. **kwargs: Other optional keyword arguments. Raises: ValueError: If some condition is not met. """ self.ingestion_batch_size = ingestion_batch_size self.num_workers = num_workers self.verbose = verbose if _DEEPLAKE_INSTALLED is False: raise ValueError( "Could not import deeplake python package. " "Please install it with `pip install deeplake`." ) if version_compare(deeplake.__version__, "3.6.2") == -1: raise ValueError( "deeplake version should be >= 3.6.3, but you've installed" f" {deeplake.__version__}. Consider upgrading deeplake version \ pip install --upgrade deeplake." ) self.dataset_path = dataset_path self.vectorstore = DeepLakeVectorStore( path=self.dataset_path, embedding_function=embedding_function, read_only=read_only, token=token, exec_option=exec_option, verbose=verbose, **kwargs, ) self._embedding_function = embedding_function self._id_tensor_name = "ids" if "ids" in self.vectorstore.tensors() else "id" [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, **kwargs: Any, ) -> List[str]: """Run more texts through the embeddings and add to the vectorstore. Examples: >>> ids = deeplake_vectorstore.add_texts( ... texts = <list_of_texts>, ... metadatas = <list_of_metadata_jsons>, ... ids = <list_of_ids>, ... ) Args: texts (Iterable[str]): Texts to add to the vectorstore. metadatas (Optional[List[dict]], optional): Optional list of metadatas. ids (Optional[List[str]], optional): Optional list of IDs. **kwargs: other optional keyword arguments. Returns: List[str]: List of IDs of the added texts. """ kwargs = {} if ids: if self._id_tensor_name == "ids": # for backwards compatibility kwargs["ids"] = ids else: kwargs["id"] = ids if metadatas is None: metadatas = [{}] * len(list(texts)) return self.vectorstore.add( text=texts, metadata=metadatas, embedding_data=texts, embedding_tensor="embedding", embedding_function=kwargs.get("embedding_function") or self._embedding_function.embed_documents, # type: ignore return_ids=True, **kwargs, ) def _search_tql( self, tql_query: Optional[str], exec_option: Optional[str] = None, return_score: bool = False, ) -> Any[List[Document], List[Tuple[Document, float]]]: """Function for performing tql_search. Args: tql_query (str): TQL Query string for direct evaluation. Available only for `compute_engine` and `tensor_db`. exec_option (str, optional): Supports 3 ways to search. Could be "python", "compute_engine" or "tensor_db". Default is "python". - ``python`` - Pure-python implementation for the client. WARNING: not recommended for big datasets due to potential memory issues. - ``compute_engine`` - C++ implementation of Deep Lake Compute Engine for the client. Not for in-memory or local datasets. - ``tensor_db`` - Hosted Managed Tensor Database for storage and query execution. Only for data in Deep Lake Managed Database. Use runtime = {"db_engine": True} during dataset creation. return_score (bool): Return score with document. Default is False. Returns: List[Document] - A list of documents Raises: ValueError: If return_score is True but some condition is not met. """ result = self.vectorstore.search( query=tql_query, exec_option=exec_option, ) metadatas = result["metadata"] texts = result["text"] docs = [ Document( page_content=text, metadata=metadata, ) for text, metadata in zip(texts, metadatas) ] if return_score: raise ValueError("scores can't be returned with tql search") return docs def _search( self, query: Optional[str] = None, embedding: Optional[Union[List[float], np.ndarray]] = None, embedding_function: Optional[Callable] = None, k: int = 4, distance_metric: str = "L2", use_maximal_marginal_relevance: bool = False, fetch_k: Optional[int] = 20, filter: Optional[Union[Dict, Callable]] = None, return_score: bool = False, exec_option: Optional[str] = None, **kwargs: Any, ) -> Any[List[Document], List[Tuple[Document, float]]]: """ Return docs similar to query. Args: query (str, optional): Text to look up similar docs. embedding (Union[List[float], np.ndarray], optional): Query's embedding. embedding_function (Callable, optional): Function to convert `query` into embedding. k (int): Number of Documents to return. distance_metric (str): `L2` for Euclidean, `L1` for Nuclear, `max` for L-infinity distance, `cos` for cosine similarity, 'dot' for dot product. filter (Union[Dict, Callable], optional): Additional filter prior to the embedding search. - ``Dict`` - Key-value search on tensors of htype json, on an AND basis (a sample must satisfy all key-value filters to be True) Dict = {"tensor_name_1": {"key": value}, "tensor_name_2": {"key": value}} - ``Function`` - Any function compatible with `deeplake.filter`. use_maximal_marginal_relevance (bool): Use maximal marginal relevance. fetch_k (int): Number of Documents for MMR algorithm. return_score (bool): Return the score. exec_option (str, optional): Supports 3 ways to perform searching. Could be "python", "compute_engine" or "tensor_db". - ``python`` - Pure-python implementation for the client. WARNING: not recommended for big datasets. - ``compute_engine`` - C++ implementation of Deep Lake Compute Engine for the client. Not for in-memory or local datasets. - ``tensor_db`` - Hosted Managed Tensor Database for storage and query execution. Only for data in Deep Lake Managed Database. Use runtime = {"db_engine": True} during dataset creation. **kwargs: Additional keyword arguments. Returns: List of Documents by the specified distance metric, if return_score True, return a tuple of (Document, score) Raises: ValueError: if both `embedding` and `embedding_function` are not specified. """ if kwargs.get("tql_query"): return self._search_tql( tql_query=kwargs["tql_query"], exec_option=exec_option, return_score=return_score, ) if embedding_function: if isinstance(embedding_function, Embeddings): _embedding_function = embedding_function.embed_query else: _embedding_function = embedding_function elif self._embedding_function: _embedding_function = self._embedding_function.embed_query else: _embedding_function = None if embedding is None: if _embedding_function is None: raise ValueError( "Either `embedding` or `embedding_function` needs to be" " specified." ) embedding = _embedding_function(query) if query else None if isinstance(embedding, list): embedding = np.array(embedding, dtype=np.float32) if len(embedding.shape) > 1: embedding = embedding[0] result = self.vectorstore.search( embedding=embedding, k=fetch_k if use_maximal_marginal_relevance else k, distance_metric=distance_metric, filter=filter, exec_option=exec_option, return_tensors=["embedding", "metadata", "text"], ) scores = result["score"] embeddings = result["embedding"] metadatas = result["metadata"] texts = result["text"] if use_maximal_marginal_relevance: lambda_mult = kwargs.get("lambda_mult", 0.5) indices = maximal_marginal_relevance( # type: ignore embedding, # type: ignore embeddings, k=min(k, len(texts)), lambda_mult=lambda_mult, ) scores = [scores[i] for i in indices] texts = [texts[i] for i in indices] metadatas = [metadatas[i] for i in indices] docs = [ Document( page_content=text, metadata=metadata, ) for text, metadata in zip(texts, metadatas) ] if return_score: return [(doc, score) for doc, score in zip(docs, scores)] return docs [docs] def similarity_search( self, query: str, k: int = 4, **kwargs: Any, ) -> List[Document]: """ Return docs most similar to query. Examples: >>> # Search using an embedding >>> data = vector_store.similarity_search( ... query=<your_query>, ... k=<num_items>, ... exec_option=<preferred_exec_option>, ... ) >>> # Run tql search: >>> data = vector_store.tql_search( ... tql_query="SELECT * WHERE id == <id>", ... exec_option="compute_engine", ... ) Args: k (int): Number of Documents to return. Defaults to 4. query (str): Text to look up similar documents. **kwargs: Additional keyword arguments include: embedding (Callable): Embedding function to use. Defaults to None. distance_metric (str): 'L2' for Euclidean, 'L1' for Nuclear, 'max' for L-infinity, 'cos' for cosine, 'dot' for dot product. Defaults to 'L2'. filter (Union[Dict, Callable], optional): Additional filter before embedding search. - Dict: Key-value search on tensors of htype json, (sample must satisfy all key-value filters) Dict = {"tensor_1": {"key": value}, "tensor_2": {"key": value}} - Function: Compatible with `deeplake.filter`. Defaults to None. exec_option (str): Supports 3 ways to perform searching. 'python', 'compute_engine', or 'tensor_db'. Defaults to 'python'. - 'python': Pure-python implementation for the client. WARNING: not recommended for big datasets. - 'compute_engine': C++ implementation of the Compute Engine for the client. Not for in-memory or local datasets. - 'tensor_db': Managed Tensor Database for storage and query. Only for data in Deep Lake Managed Database. Use `runtime = {"db_engine": True}` during dataset creation. Returns: List[Document]: List of Documents most similar to the query vector. """ return self._search( query=query, k=k, use_maximal_marginal_relevance=False, return_score=False, **kwargs, ) [docs] def similarity_search_by_vector( self, embedding: Union[List[float], np.ndarray], k: int = 4, **kwargs: Any, ) -> List[Document]: """ Return docs most similar to embedding vector. Examples: >>> # Search using an embedding >>> data = vector_store.similarity_search_by_vector( ... embedding=<your_embedding>, ... k=<num_items_to_return>, ... exec_option=<preferred_exec_option>, ... ) Args: embedding (Union[List[float], np.ndarray]): Embedding to find similar docs. k (int): Number of Documents to return. Defaults to 4. **kwargs: Additional keyword arguments including: filter (Union[Dict, Callable], optional): Additional filter before embedding search. - ``Dict`` - Key-value search on tensors of htype json. True if all key-value filters are satisfied. Dict = {"tensor_name_1": {"key": value}, "tensor_name_2": {"key": value}} - ``Function`` - Any function compatible with `deeplake.filter`. Defaults to None. exec_option (str): Options for search execution include "python", "compute_engine", or "tensor_db". Defaults to "python". - "python" - Pure-python implementation running on the client. Can be used for data stored anywhere. WARNING: using this option with big datasets is discouraged due to potential memory issues. - "compute_engine" - Performant C++ implementation of the Deep Lake Compute Engine. Runs on the client and can be used for any data stored in or connected to Deep Lake. It cannot be used with in-memory or local datasets. - "tensor_db" - Performant, fully-hosted Managed Tensor Database. Responsible for storage and query execution. Only available for data stored in the Deep Lake Managed Database. To store datasets in this database, specify `runtime = {"db_engine": True}` during dataset creation. distance_metric (str): `L2` for Euclidean, `L1` for Nuclear, `max` for L-infinity distance, `cos` for cosine similarity, 'dot' for dot product. Defaults to `L2`. Returns: List[Document]: List of Documents most similar to the query vector. """ return self._search( embedding=embedding, k=k, use_maximal_marginal_relevance=False, return_score=False, **kwargs, ) [docs] def similarity_search_with_score( self, query: str, k: int = 4, **kwargs: Any, ) -> List[Tuple[Document, float]]: """ Run similarity search with Deep Lake with distance returned. Examples: >>> data = vector_store.similarity_search_with_score( ... query=<your_query>, ... embedding=<your_embedding_function> ... k=<number_of_items_to_return>, ... exec_option=<preferred_exec_option>, ... ) Args: query (str): Query text to search for. k (int): Number of results to return. Defaults to 4. **kwargs: Additional keyword arguments. Some of these arguments are: distance_metric: `L2` for Euclidean, `L1` for Nuclear, `max` L-infinity distance, `cos` for cosine similarity, 'dot' for dot product. Defaults to `L2`. filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. embedding_function (Callable): Embedding function to use. Defaults to None. exec_option (str): DeepLakeVectorStore supports 3 ways to perform searching. It could be either "python", "compute_engine" or "tensor_db". Defaults to "python". - "python" - Pure-python implementation running on the client. Can be used for data stored anywhere. WARNING: using this option with big datasets is discouraged due to potential memory issues. - "compute_engine" - Performant C++ implementation of the Deep Lake Compute Engine. Runs on the client and can be used for any data stored in or connected to Deep Lake. It cannot be used with in-memory or local datasets. - "tensor_db" - Performant, fully-hosted Managed Tensor Database. Responsible for storage and query execution. Only available for data stored in the Deep Lake Managed Database. To store datasets in this database, specify `runtime = {"db_engine": True}` during dataset creation. Returns: List[Tuple[Document, float]]: List of documents most similar to the query text with distance in float.""" return self._search( query=query, k=k, return_score=True, **kwargs, ) [docs] def max_marginal_relevance_search_by_vector( self, embedding: List[float], k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, exec_option: Optional[str] = None, **kwargs: Any, ) -> List[Document]: """ Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected docs. Examples: >>> data = vector_store.max_marginal_relevance_search_by_vector( ... embedding=<your_embedding>, ... fetch_k=<elements_to_fetch_before_mmr_search>, ... k=<number_of_items_to_return>, ... exec_option=<preferred_exec_option>, ... ) Args: embedding: Embedding to look up documents similar to. k: Number of Documents to return. Defaults to 4. fetch_k: Number of Documents to fetch for MMR algorithm. lambda_mult: Number between 0 and 1 determining the degree of diversity. 0 corresponds to max diversity and 1 to min diversity. Defaults to 0.5. exec_option (str): DeepLakeVectorStore supports 3 ways for searching. Could be "python", "compute_engine" or "tensor_db". Defaults to "python". - "python" - Pure-python implementation running on the client. Can be used for data stored anywhere. WARNING: using this option with big datasets is discouraged due to potential memory issues. - "compute_engine" - Performant C++ implementation of the Deep Lake Compute Engine. Runs on the client and can be used for any data stored in or connected to Deep Lake. It cannot be used with in-memory or local datasets. - "tensor_db" - Performant, fully-hosted Managed Tensor Database. Responsible for storage and query execution. Only available for data stored in the Deep Lake Managed Database. To store datasets in this database, specify `runtime = {"db_engine": True}` during dataset creation. **kwargs: Additional keyword arguments. Returns: List[Documents] - A list of documents. """ return self._search( embedding=embedding, k=k, fetch_k=fetch_k, use_maximal_marginal_relevance=True, lambda_mult=lambda_mult, exec_option=exec_option, **kwargs, ) [docs] def max_marginal_relevance_search( self, query: str, k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, exec_option: Optional[str] = None, **kwargs: Any, ) -> List[Document]: """Return docs selected using maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Examples: >>> # Search using an embedding >>> data = vector_store.max_marginal_relevance_search( ... query = <query_to_search>, ... embedding_function = <embedding_function_for_query>, ... k = <number_of_items_to_return>, ... exec_option = <preferred_exec_option>, ... ) Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. fetch_k: Number of Documents for MMR algorithm. lambda_mult: Value between 0 and 1. 0 corresponds to maximum diversity and 1 to minimum. Defaults to 0.5. exec_option (str): Supports 3 ways to perform searching. - "python" - Pure-python implementation running on the client. Can be used for data stored anywhere. WARNING: using this option with big datasets is discouraged due to potential memory issues. - "compute_engine" - Performant C++ implementation of the Deep Lake Compute Engine. Runs on the client and can be used for any data stored in or connected to Deep Lake. It cannot be used with in-memory or local datasets. - "tensor_db" - Performant, fully-hosted Managed Tensor Database. Responsible for storage and query execution. Only available for data stored in the Deep Lake Managed Database. To store datasets in this database, specify `runtime = {"db_engine": True}` during dataset creation. **kwargs: Additional keyword arguments Returns: List of Documents selected by maximal marginal relevance. Raises: ValueError: when MRR search is on but embedding function is not specified. """ embedding_function = kwargs.get("embedding") or self._embedding_function if embedding_function is None: raise ValueError( "For MMR search, you must specify an embedding function on" " `creation` or during add call." ) return self._search( query=query, k=k, fetch_k=fetch_k, use_maximal_marginal_relevance=True, lambda_mult=lambda_mult, exec_option=exec_option, embedding_function=embedding_function, # type: ignore **kwargs, ) [docs] @classmethod def from_texts( cls, texts: List[str], embedding: Optional[Embeddings] = None, metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, dataset_path: str = _LANGCHAIN_DEFAULT_DEEPLAKE_PATH, **kwargs: Any, ) -> DeepLake: """Create a Deep Lake dataset from a raw documents. If a dataset_path is specified, the dataset will be persisted in that location, otherwise by default at `./deeplake` Examples: >>> # Search using an embedding >>> vector_store = DeepLake.from_texts( ... texts = <the_texts_that_you_want_to_embed>, ... embedding_function = <embedding_function_for_query>, ... k = <number_of_items_to_return>, ... exec_option = <preferred_exec_option>, ... ) Args: dataset_path (str): - The full path to the dataset. Can be: - Deep Lake cloud path of the form ``hub://username/dataset_name``. To write to Deep Lake cloud datasets, ensure that you are logged in to Deep Lake (use 'activeloop login' from command line) - AWS S3 path of the form ``s3://bucketname/path/to/dataset``. Credentials are required in either the environment - Google Cloud Storage path of the form ``gcs://bucketname/path/to/dataset`` Credentials are required in either the environment - Local file system path of the form ``./path/to/dataset`` or ``~/path/to/dataset`` or ``path/to/dataset``. - In-memory path of the form ``mem://path/to/dataset`` which doesn't save the dataset, but keeps it in memory instead. Should be used only for testing as it does not persist. texts (List[Document]): List of documents to add. embedding (Optional[Embeddings]): Embedding function. Defaults to None. Note, in other places, it is called embedding_function. metadatas (Optional[List[dict]]): List of metadatas. Defaults to None. ids (Optional[List[str]]): List of document IDs. Defaults to None. **kwargs: Additional keyword arguments. Returns: DeepLake: Deep Lake dataset. Raises: ValueError: If 'embedding' is provided in kwargs. This is deprecated, please use `embedding_function` instead. """ if kwargs.get("embedding"): raise ValueError( "using embedding as embedidng_functions is deprecated. " "Please use `embedding_function` instead." ) deeplake_dataset = cls( dataset_path=dataset_path, embedding_function=embedding, **kwargs ) deeplake_dataset.add_texts( texts=texts, metadatas=metadatas, ids=ids, embedding_function=embedding.embed_documents, # type: ignore ) return deeplake_dataset [docs] def delete( self, ids: Any[List[str], None] = None, filter: Any[Dict[str, str], None] = None, delete_all: Any[bool, None] = None, ) -> bool: """Delete the entities in the dataset. Args: ids (Optional[List[str]], optional): The document_ids to delete. Defaults to None. filter (Optional[Dict[str, str]], optional): The filter to delete by. Defaults to None. delete_all (Optional[bool], optional): Whether to drop the dataset. Defaults to None. Returns: bool: Whether the delete operation was successful. """ self.vectorstore.delete( ids=ids, filter=filter, delete_all=delete_all, ) return True [docs] @classmethod def force_delete_by_path(cls, path: str) -> None: """Force delete dataset by path. Args: path (str): path of the dataset to delete. Raises: ValueError: if deeplake is not installed. """ try: import deeplake except ImportError: raise ValueError( "Could not import deeplake python package. " "Please install it with `pip install deeplake`." ) deeplake.delete(path, large_ok=True, force=True) [docs] def delete_dataset(self) -> None: """Delete the collection.""" self.delete(delete_all=True)
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/deeplake.html
ac87b9c4-9202-4c49-a189-f5178d6556db
Source code for langchain.vectorstores.annoy """Wrapper around Annoy vector database.""" from __future__ import annotations import os import pickle import uuid from configparser import ConfigParser from pathlib import Path from typing import Any, Callable, Dict, Iterable, List, Optional, Tuple import numpy as np from langchain.docstore.base import Docstore from langchain.docstore.document import Document from langchain.docstore.in_memory import InMemoryDocstore from langchain.embeddings.base import Embeddings from langchain.vectorstores.base import VectorStore from langchain.vectorstores.utils import maximal_marginal_relevance INDEX_METRICS = frozenset(["angular", "euclidean", "manhattan", "hamming", "dot"]) DEFAULT_METRIC = "angular" def dependable_annoy_import() -> Any: """Import annoy if available, otherwise raise error.""" try: import annoy except ImportError: raise ValueError( "Could not import annoy python package. " "Please install it with `pip install --user annoy` " ) return annoy [docs]class Annoy(VectorStore): """Wrapper around Annoy vector database. To use, you should have the ``annoy`` python package installed. Example: .. code-block:: python from langchain import Annoy db = Annoy(embedding_function, index, docstore, index_to_docstore_id) """ def __init__( self, embedding_function: Callable, index: Any, metric: str, docstore: Docstore, index_to_docstore_id: Dict[int, str], ): """Initialize with necessary components.""" self.embedding_function = embedding_function self.index = index self.metric = metric self.docstore = docstore self.index_to_docstore_id = index_to_docstore_id [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, **kwargs: Any, ) -> List[str]: raise NotImplementedError( "Annoy does not allow to add new data once the index is build." ) [docs] def process_index_results( self, idxs: List[int], dists: List[float] ) -> List[Tuple[Document, float]]: """Turns annoy results into a list of documents and scores. Args: idxs: List of indices of the documents in the index. dists: List of distances of the documents in the index. Returns: List of Documents and scores. """ docs = [] for idx, dist in zip(idxs, dists): _id = self.index_to_docstore_id[idx] doc = self.docstore.search(_id) if not isinstance(doc, Document): raise ValueError(f"Could not find document for id {_id}, got {doc}") docs.append((doc, dist)) return docs [docs] def similarity_search_with_score_by_vector( self, embedding: List[float], k: int = 4, search_k: int = -1 ) -> List[Tuple[Document, float]]: """Return docs most similar to query. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. search_k: inspect up to search_k nodes which defaults to n_trees * n if not provided Returns: List of Documents most similar to the query and score for each """ idxs, dists = self.index.get_nns_by_vector( embedding, k, search_k=search_k, include_distances=True ) return self.process_index_results(idxs, dists) [docs] def similarity_search_with_score_by_index( self, docstore_index: int, k: int = 4, search_k: int = -1 ) -> List[Tuple[Document, float]]: """Return docs most similar to query. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. search_k: inspect up to search_k nodes which defaults to n_trees * n if not provided Returns: List of Documents most similar to the query and score for each """ idxs, dists = self.index.get_nns_by_item( docstore_index, k, search_k=search_k, include_distances=True ) return self.process_index_results(idxs, dists) [docs] def similarity_search_with_score( self, query: str, k: int = 4, search_k: int = -1 ) -> List[Tuple[Document, float]]: """Return docs most similar to query. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. search_k: inspect up to search_k nodes which defaults to n_trees * n if not provided Returns: List of Documents most similar to the query and score for each """ embedding = self.embedding_function(query) docs = self.similarity_search_with_score_by_vector(embedding, k, search_k) return docs [docs] def similarity_search_by_vector( self, embedding: List[float], k: int = 4, search_k: int = -1, **kwargs: Any ) -> List[Document]: """Return docs most similar to embedding vector. Args: embedding: Embedding to look up documents similar to. k: Number of Documents to return. Defaults to 4. search_k: inspect up to search_k nodes which defaults to n_trees * n if not provided Returns: List of Documents most similar to the embedding. """ docs_and_scores = self.similarity_search_with_score_by_vector( embedding, k, search_k ) return [doc for doc, _ in docs_and_scores] [docs] def similarity_search_by_index( self, docstore_index: int, k: int = 4, search_k: int = -1, **kwargs: Any ) -> List[Document]: """Return docs most similar to docstore_index. Args: docstore_index: Index of document in docstore k: Number of Documents to return. Defaults to 4. search_k: inspect up to search_k nodes which defaults to n_trees * n if not provided Returns: List of Documents most similar to the embedding. """ docs_and_scores = self.similarity_search_with_score_by_index( docstore_index, k, search_k ) return [doc for doc, _ in docs_and_scores] [docs] def similarity_search( self, query: str, k: int = 4, search_k: int = -1, **kwargs: Any ) -> List[Document]: """Return docs most similar to query. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. search_k: inspect up to search_k nodes which defaults to n_trees * n if not provided Returns: List of Documents most similar to the query. """ docs_and_scores = self.similarity_search_with_score(query, k, search_k) return [doc for doc, _ in docs_and_scores] [docs] def max_marginal_relevance_search_by_vector( self, embedding: List[float], k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: embedding: Embedding to look up documents similar to. fetch_k: Number of Documents to fetch to pass to MMR algorithm. k: Number of Documents to return. Defaults to 4. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. Returns: List of Documents selected by maximal marginal relevance. """ idxs = self.index.get_nns_by_vector( embedding, fetch_k, search_k=-1, include_distances=False ) embeddings = [self.index.get_item_vector(i) for i in idxs] mmr_selected = maximal_marginal_relevance( np.array([embedding], dtype=np.float32), embeddings, k=k, lambda_mult=lambda_mult, ) # ignore the -1's if not enough docs are returned/indexed selected_indices = [idxs[i] for i in mmr_selected if i != -1] docs = [] for i in selected_indices: _id = self.index_to_docstore_id[i] doc = self.docstore.search(_id) if not isinstance(doc, Document): raise ValueError(f"Could not find document for id {_id}, got {doc}") docs.append(doc) return docs [docs] def max_marginal_relevance_search( self, query: str, k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. fetch_k: Number of Documents to fetch to pass to MMR algorithm. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. Returns: List of Documents selected by maximal marginal relevance. """ embedding = self.embedding_function(query) docs = self.max_marginal_relevance_search_by_vector( embedding, k, fetch_k, lambda_mult=lambda_mult ) return docs @classmethod def __from( cls, texts: List[str], embeddings: List[List[float]], embedding: Embeddings, metadatas: Optional[List[dict]] = None, metric: str = DEFAULT_METRIC, trees: int = 100, n_jobs: int = -1, **kwargs: Any, ) -> Annoy: if metric not in INDEX_METRICS: raise ValueError( ( f"Unsupported distance metric: {metric}. " f"Expected one of {list(INDEX_METRICS)}" ) ) annoy = dependable_annoy_import() if not embeddings: raise ValueError("embeddings must be provided to build AnnoyIndex") f = len(embeddings[0]) index = annoy.AnnoyIndex(f, metric=metric) for i, emb in enumerate(embeddings): index.add_item(i, emb) index.build(trees, n_jobs=n_jobs) documents = [] for i, text in enumerate(texts): metadata = metadatas[i] if metadatas else {} documents.append(Document(page_content=text, metadata=metadata)) index_to_id = {i: str(uuid.uuid4()) for i in range(len(documents))} docstore = InMemoryDocstore( {index_to_id[i]: doc for i, doc in enumerate(documents)} ) return cls(embedding.embed_query, index, metric, docstore, index_to_id) [docs] @classmethod def from_texts( cls, texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, metric: str = DEFAULT_METRIC, trees: int = 100, n_jobs: int = -1, **kwargs: Any, ) -> Annoy: """Construct Annoy wrapper from raw documents. Args: texts: List of documents to index. embedding: Embedding function to use. metadatas: List of metadata dictionaries to associate with documents. metric: Metric to use for indexing. Defaults to "angular". trees: Number of trees to use for indexing. Defaults to 100. n_jobs: Number of jobs to use for indexing. Defaults to -1. This is a user friendly interface that: 1. Embeds documents. 2. Creates an in memory docstore 3. Initializes the Annoy database This is intended to be a quick way to get started. Example: .. code-block:: python from langchain import Annoy from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() index = Annoy.from_texts(texts, embeddings) """ embeddings = embedding.embed_documents(texts) return cls.__from( texts, embeddings, embedding, metadatas, metric, trees, n_jobs, **kwargs ) [docs] @classmethod def from_embeddings( cls, text_embeddings: List[Tuple[str, List[float]]], embedding: Embeddings, metadatas: Optional[List[dict]] = None, metric: str = DEFAULT_METRIC, trees: int = 100, n_jobs: int = -1, **kwargs: Any, ) -> Annoy: """Construct Annoy wrapper from embeddings. Args: text_embeddings: List of tuples of (text, embedding) embedding: Embedding function to use. metadatas: List of metadata dictionaries to associate with documents. metric: Metric to use for indexing. Defaults to "angular". trees: Number of trees to use for indexing. Defaults to 100. n_jobs: Number of jobs to use for indexing. Defaults to -1 This is a user friendly interface that: 1. Creates an in memory docstore with provided embeddings 2. Initializes the Annoy database This is intended to be a quick way to get started. Example: .. code-block:: python from langchain import Annoy from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() text_embeddings = embeddings.embed_documents(texts) text_embedding_pairs = list(zip(texts, text_embeddings)) db = Annoy.from_embeddings(text_embedding_pairs, embeddings) """ texts = [t[0] for t in text_embeddings] embeddings = [t[1] for t in text_embeddings] return cls.__from( texts, embeddings, embedding, metadatas, metric, trees, n_jobs, **kwargs ) [docs] def save_local(self, folder_path: str, prefault: bool = False) -> None: """Save Annoy index, docstore, and index_to_docstore_id to disk. Args: folder_path: folder path to save index, docstore, and index_to_docstore_id to. prefault: Whether to pre-load the index into memory. """ path = Path(folder_path) os.makedirs(path, exist_ok=True) # save index, index config, docstore and index_to_docstore_id config_object = ConfigParser() config_object["ANNOY"] = { "f": self.index.f, "metric": self.metric, } self.index.save(str(path / "index.annoy"), prefault=prefault) with open(path / "index.pkl", "wb") as file: pickle.dump((self.docstore, self.index_to_docstore_id, config_object), file) [docs] @classmethod def load_local( cls, folder_path: str, embeddings: Embeddings, ) -> Annoy: """Load Annoy index, docstore, and index_to_docstore_id to disk. Args: folder_path: folder path to load index, docstore, and index_to_docstore_id from. embeddings: Embeddings to use when generating queries. """ path = Path(folder_path) # load index separately since it is not picklable annoy = dependable_annoy_import() # load docstore and index_to_docstore_id with open(path / "index.pkl", "rb") as file: docstore, index_to_docstore_id, config_object = pickle.load(file) f = int(config_object["ANNOY"]["f"]) metric = config_object["ANNOY"]["metric"] index = annoy.AnnoyIndex(f, metric=metric) index.load(str(path / "index.annoy")) return cls( embeddings.embed_query, index, metric, docstore, index_to_docstore_id )
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/annoy.html
482a9cdc-7488-4b33-b2cd-472c6c9b817b
Source code for langchain.vectorstores.typesense """Wrapper around Typesense vector search""" from __future__ import annotations import uuid from typing import TYPE_CHECKING, Any, Iterable, List, Optional, Tuple, Union from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.utils import get_from_env from langchain.vectorstores.base import VectorStore if TYPE_CHECKING: from typesense.client import Client from typesense.collection import Collection [docs]class Typesense(VectorStore): """Wrapper around Typesense vector search. To use, you should have the ``typesense`` python package installed. Example: .. code-block:: python from langchain.embedding.openai import OpenAIEmbeddings from langchain.vectorstores import Typesense import typesense node = { "host": "localhost", # For Typesense Cloud use xxx.a1.typesense.net "port": "8108", # For Typesense Cloud use 443 "protocol": "http" # For Typesense Cloud use https } typesense_client = typesense.Client( { "nodes": [node], "api_key": "<API_KEY>", "connection_timeout_seconds": 2 } ) typesense_collection_name = "langchain-memory" embedding = OpenAIEmbeddings() vectorstore = Typesense( typesense_client=typesense_client, embedding=embedding, typesense_collection_name=typesense_collection_name, text_key="text", ) """ def __init__( self, typesense_client: Client, embedding: Embeddings, *, typesense_collection_name: Optional[str] = None, text_key: str = "text", ): """Initialize with Typesense client.""" try: from typesense import Client except ImportError: raise ValueError( "Could not import typesense python package. " "Please install it with `pip install typesense`." ) if not isinstance(typesense_client, Client): raise ValueError( f"typesense_client should be an instance of typesense.Client, " f"got {type(typesense_client)}" ) self._typesense_client = typesense_client self._embedding = embedding self._typesense_collection_name = ( typesense_collection_name or f"langchain-{str(uuid.uuid4())}" ) self._text_key = text_key @property def _collection(self) -> Collection: return self._typesense_client.collections[self._typesense_collection_name] def _prep_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]], ids: Optional[List[str]], ) -> List[dict]: """Embed and create the documents""" _ids = ids or (str(uuid.uuid4()) for _ in texts) _metadatas: Iterable[dict] = metadatas or ({} for _ in texts) embedded_texts = self._embedding.embed_documents(list(texts)) return [ {"id": _id, "vec": vec, f"{self._text_key}": text, "metadata": metadata} for _id, vec, text, metadata in zip(_ids, embedded_texts, texts, _metadatas) ] def _create_collection(self, num_dim: int) -> None: fields = [ {"name": "vec", "type": "float[]", "num_dim": num_dim}, {"name": f"{self._text_key}", "type": "string"}, {"name": ".*", "type": "auto"}, ] self._typesense_client.collections.create( {"name": self._typesense_collection_name, "fields": fields} ) [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, **kwargs: Any, ) -> List[str]: """Run more texts through the embedding and add to the vectorstore. Args: texts: Iterable of strings to add to the vectorstore. metadatas: Optional list of metadatas associated with the texts. ids: Optional list of ids to associate with the texts. Returns: List of ids from adding the texts into the vectorstore. """ from typesense.exceptions import ObjectNotFound docs = self._prep_texts(texts, metadatas, ids) try: self._collection.documents.import_(docs, {"action": "upsert"}) except ObjectNotFound: # Create the collection if it doesn't already exist self._create_collection(len(docs[0]["vec"])) self._collection.documents.import_(docs, {"action": "upsert"}) return [doc["id"] for doc in docs] [docs] def similarity_search_with_score( self, query: str, k: int = 10, filter: Optional[str] = "", ) -> List[Tuple[Document, float]]: """Return typesense documents most similar to query, along with scores. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 10. Minimum 10 results would be returned. filter: typesense filter_by expression to filter documents on Returns: List of Documents most similar to the query and score for each """ embedded_query = [str(x) for x in self._embedding.embed_query(query)] query_obj = { "q": "*", "vector_query": f'vec:([{",".join(embedded_query)}], k:{k})', "filter_by": filter, "collection": self._typesense_collection_name, } docs = [] response = self._typesense_client.multi_search.perform( {"searches": [query_obj]}, {} ) for hit in response["results"][0]["hits"]: document = hit["document"] metadata = document["metadata"] text = document[self._text_key] score = hit["vector_distance"] docs.append((Document(page_content=text, metadata=metadata), score)) return docs [docs] def similarity_search( self, query: str, k: int = 10, filter: Optional[str] = "", **kwargs: Any, ) -> List[Document]: """Return typesense documents most similar to query. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 10. Minimum 10 results would be returned. filter: typesense filter_by expression to filter documents on Returns: List of Documents most similar to the query and score for each """ docs_and_score = self.similarity_search_with_score(query, k=k, filter=filter) return [doc for doc, _ in docs_and_score] [docs] @classmethod def from_client_params( cls, embedding: Embeddings, *, host: str = "localhost", port: Union[str, int] = "8108", protocol: str = "http", typesense_api_key: Optional[str] = None, connection_timeout_seconds: int = 2, **kwargs: Any, ) -> Typesense: """Initialize Typesense directly from client parameters. Example: .. code-block:: python from langchain.embedding.openai import OpenAIEmbeddings from langchain.vectorstores import Typesense # Pass in typesense_api_key as kwarg or set env var "TYPESENSE_API_KEY". vectorstore = Typesense( OpenAIEmbeddings(), host="localhost", port="8108", protocol="http", typesense_collection_name="langchain-memory", ) """ try: from typesense import Client except ImportError: raise ValueError( "Could not import typesense python package. " "Please install it with `pip install typesense`." ) node = { "host": host, "port": str(port), "protocol": protocol, } typesense_api_key = typesense_api_key or get_from_env( "typesense_api_key", "TYPESENSE_API_KEY" ) client_config = { "nodes": [node], "api_key": typesense_api_key, "connection_timeout_seconds": connection_timeout_seconds, } return cls(Client(client_config), embedding, **kwargs) [docs] @classmethod def from_texts( cls, texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, typesense_client: Optional[Client] = None, typesense_client_params: Optional[dict] = None, typesense_collection_name: Optional[str] = None, text_key: str = "text", **kwargs: Any, ) -> Typesense: """Construct Typesense wrapper from raw text.""" if typesense_client: vectorstore = cls(typesense_client, embedding, **kwargs) elif typesense_client_params: vectorstore = cls.from_client_params( embedding, **typesense_client_params, **kwargs ) else: raise ValueError( "Must specify one of typesense_client or typesense_client_params." ) vectorstore.add_texts(texts, metadatas=metadatas, ids=ids) return vectorstore
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/typesense.html
f44e9ce0-6cbc-49e2-87b7-86b774a21511
Source code for langchain.vectorstores.pinecone """Wrapper around Pinecone vector database.""" from __future__ import annotations import logging import uuid from typing import Any, Callable, Iterable, List, Optional, Tuple import numpy as np from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.vectorstores.base import VectorStore from langchain.vectorstores.utils import maximal_marginal_relevance logger = logging.getLogger(__name__) [docs]class Pinecone(VectorStore): """Wrapper around Pinecone vector database. To use, you should have the ``pinecone-client`` python package installed. Example: .. code-block:: python from langchain.vectorstores import Pinecone from langchain.embeddings.openai import OpenAIEmbeddings import pinecone # The environment should be the one specified next to the API key # in your Pinecone console pinecone.init(api_key="***", environment="...") index = pinecone.Index("langchain-demo") embeddings = OpenAIEmbeddings() vectorstore = Pinecone(index, embeddings.embed_query, "text") """ def __init__( self, index: Any, embedding_function: Callable, text_key: str, namespace: Optional[str] = None, ): """Initialize with Pinecone client.""" try: import pinecone except ImportError: raise ValueError( "Could not import pinecone python package. " "Please install it with `pip install pinecone-client`." ) if not isinstance(index, pinecone.index.Index): raise ValueError( f"client should be an instance of pinecone.index.Index, " f"got {type(index)}" ) self._index = index self._embedding_function = embedding_function self._text_key = text_key self._namespace = namespace [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, namespace: Optional[str] = None, batch_size: int = 32, **kwargs: Any, ) -> List[str]: """Run more texts through the embeddings and add to the vectorstore. Args: texts: Iterable of strings to add to the vectorstore. metadatas: Optional list of metadatas associated with the texts. ids: Optional list of ids to associate with the texts. namespace: Optional pinecone namespace to add the texts to. Returns: List of ids from adding the texts into the vectorstore. """ if namespace is None: namespace = self._namespace # Embed and create the documents docs = [] ids = ids or [str(uuid.uuid4()) for _ in texts] for i, text in enumerate(texts): embedding = self._embedding_function(text) metadata = metadatas[i] if metadatas else {} metadata[self._text_key] = text docs.append((ids[i], embedding, metadata)) # upsert to Pinecone self._index.upsert(vectors=docs, namespace=namespace, batch_size=batch_size) return ids [docs] def similarity_search_with_score( self, query: str, k: int = 4, filter: Optional[dict] = None, namespace: Optional[str] = None, ) -> List[Tuple[Document, float]]: """Return pinecone documents most similar to query, along with scores. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. filter: Dictionary of argument(s) to filter on metadata namespace: Namespace to search in. Default will search in '' namespace. Returns: List of Documents most similar to the query and score for each """ if namespace is None: namespace = self._namespace query_obj = self._embedding_function(query) docs = [] results = self._index.query( [query_obj], top_k=k, include_metadata=True, namespace=namespace, filter=filter, ) for res in results["matches"]: metadata = res["metadata"] if self._text_key in metadata: text = metadata.pop(self._text_key) score = res["score"] docs.append((Document(page_content=text, metadata=metadata), score)) else: logger.warning( f"Found document with no `{self._text_key}` key. Skipping." ) return docs [docs] def similarity_search( self, query: str, k: int = 4, filter: Optional[dict] = None, namespace: Optional[str] = None, **kwargs: Any, ) -> List[Document]: """Return pinecone documents most similar to query. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. filter: Dictionary of argument(s) to filter on metadata namespace: Namespace to search in. Default will search in '' namespace. Returns: List of Documents most similar to the query and score for each """ docs_and_scores = self.similarity_search_with_score( query, k=k, filter=filter, namespace=namespace, **kwargs ) return [doc for doc, _ in docs_and_scores] def _similarity_search_with_relevance_scores( self, query: str, k: int = 4, **kwargs: Any, ) -> List[Tuple[Document, float]]: return self.similarity_search_with_score(query, k) [docs] def max_marginal_relevance_search_by_vector( self, embedding: List[float], k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, filter: Optional[dict] = None, namespace: Optional[str] = None, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: embedding: Embedding to look up documents similar to. k: Number of Documents to return. Defaults to 4. fetch_k: Number of Documents to fetch to pass to MMR algorithm. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. Returns: List of Documents selected by maximal marginal relevance. """ if namespace is None: namespace = self._namespace results = self._index.query( [embedding], top_k=fetch_k, include_values=True, include_metadata=True, namespace=namespace, filter=filter, ) mmr_selected = maximal_marginal_relevance( np.array([embedding], dtype=np.float32), [item["values"] for item in results["matches"]], k=k, lambda_mult=lambda_mult, ) selected = [results["matches"][i]["metadata"] for i in mmr_selected] return [ Document(page_content=metadata.pop((self._text_key)), metadata=metadata) for metadata in selected ] [docs] def max_marginal_relevance_search( self, query: str, k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, filter: Optional[dict] = None, namespace: Optional[str] = None, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. fetch_k: Number of Documents to fetch to pass to MMR algorithm. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. Returns: List of Documents selected by maximal marginal relevance. """ embedding = self._embedding_function(query) return self.max_marginal_relevance_search_by_vector( embedding, k, fetch_k, lambda_mult, filter, namespace ) [docs] @classmethod def from_texts( cls, texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, batch_size: int = 32, text_key: str = "text", index_name: Optional[str] = None, namespace: Optional[str] = None, **kwargs: Any, ) -> Pinecone: """Construct Pinecone wrapper from raw documents. This is a user friendly interface that: 1. Embeds documents. 2. Adds the documents to a provided Pinecone index This is intended to be a quick way to get started. Example: .. code-block:: python from langchain import Pinecone from langchain.embeddings import OpenAIEmbeddings import pinecone # The environment should be the one specified next to the API key # in your Pinecone console pinecone.init(api_key="***", environment="...") embeddings = OpenAIEmbeddings() pinecone = Pinecone.from_texts( texts, embeddings, index_name="langchain-demo" ) """ try: import pinecone except ImportError: raise ValueError( "Could not import pinecone python package. " "Please install it with `pip install pinecone-client`." ) indexes = pinecone.list_indexes() # checks if provided index exists if index_name in indexes: index = pinecone.Index(index_name) elif len(indexes) == 0: raise ValueError( "No active indexes found in your Pinecone project, " "are you sure you're using the right API key and environment?" ) else: raise ValueError( f"Index '{index_name}' not found in your Pinecone project. " f"Did you mean one of the following indexes: {', '.join(indexes)}" ) for i in range(0, len(texts), batch_size): # set end position of batch i_end = min(i + batch_size, len(texts)) # get batch of texts and ids lines_batch = texts[i:i_end] # create ids if not provided if ids: ids_batch = ids[i:i_end] else: ids_batch = [str(uuid.uuid4()) for n in range(i, i_end)] # create embeddings embeds = embedding.embed_documents(lines_batch) # prep metadata and upsert batch if metadatas: metadata = metadatas[i:i_end] else: metadata = [{} for _ in range(i, i_end)] for j, line in enumerate(lines_batch): metadata[j][text_key] = line to_upsert = zip(ids_batch, embeds, metadata) # upsert to Pinecone index.upsert(vectors=list(to_upsert), namespace=namespace) return cls(index, embedding.embed_query, text_key, namespace) [docs] @classmethod def from_existing_index( cls, index_name: str, embedding: Embeddings, text_key: str = "text", namespace: Optional[str] = None, ) -> Pinecone: """Load pinecone vectorstore from index name.""" try: import pinecone except ImportError: raise ValueError( "Could not import pinecone python package. " "Please install it with `pip install pinecone-client`." ) return cls( pinecone.Index(index_name), embedding.embed_query, text_key, namespace ) [docs] def delete(self, ids: List[str]) -> None: """Delete by vector IDs. Args: ids: List of ids to delete. """ # This is the maximum number of IDs that can be deleted chunk_size = 1000 for i in range(0, len(ids), chunk_size): chunk = ids[i : i + chunk_size] self._index.delete(ids=chunk)
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/pinecone.html
e50fe63e-768d-4199-a74a-dfc314f1633e
Source code for langchain.vectorstores.tigris from __future__ import annotations import itertools from typing import TYPE_CHECKING, Any, Iterable, List, Optional, Tuple from langchain.embeddings.base import Embeddings from langchain.schema import Document from langchain.vectorstores import VectorStore if TYPE_CHECKING: from tigrisdb import TigrisClient from tigrisdb import VectorStore as TigrisVectorStore from tigrisdb.types.filters import Filter as TigrisFilter from tigrisdb.types.vector import Document as TigrisDocument [docs]class Tigris(VectorStore): def __init__(self, client: TigrisClient, embeddings: Embeddings, index_name: str): """Initialize Tigris vector store""" try: import tigrisdb # noqa: F401 except ImportError: raise ValueError( "Could not import tigrisdb python package. " "Please install it with `pip install tigrisdb`" ) self._embed_fn = embeddings self._vector_store = TigrisVectorStore(client.get_search(), index_name) @property def search_index(self) -> TigrisVectorStore: return self._vector_store [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, **kwargs: Any, ) -> List[str]: """Run more texts through the embeddings and add to the vectorstore. Args: texts: Iterable of strings to add to the vectorstore. metadatas: Optional list of metadatas associated with the texts. ids: Optional list of ids for documents. Ids will be autogenerated if not provided. kwargs: vectorstore specific parameters Returns: List of ids from adding the texts into the vectorstore. """ docs = self._prep_docs(texts, metadatas, ids) result = self.search_index.add_documents(docs) return [r.id for r in result] [docs] def similarity_search( self, query: str, k: int = 4, filter: Optional[TigrisFilter] = None, **kwargs: Any, ) -> List[Document]: """Return docs most similar to query.""" docs_with_scores = self.similarity_search_with_score(query, k, filter) return [doc for doc, _ in docs_with_scores] [docs] def similarity_search_with_score( self, query: str, k: int = 4, filter: Optional[TigrisFilter] = None, ) -> List[Tuple[Document, float]]: """Run similarity search with Chroma with distance. Args: query (str): Query text to search for. k (int): Number of results to return. Defaults to 4. filter (Optional[TigrisFilter]): Filter by metadata. Defaults to None. Returns: List[Tuple[Document, float]]: List of documents most similar to the query text with distance in float. """ vector = self._embed_fn.embed_query(query) result = self.search_index.similarity_search( vector=vector, k=k, filter_by=filter ) docs: List[Tuple[Document, float]] = [] for r in result: docs.append( ( Document( page_content=r.doc["text"], metadata=r.doc.get("metadata") ), r.score, ) ) return docs [docs] @classmethod def from_texts( cls, texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, client: Optional[TigrisClient] = None, index_name: Optional[str] = None, **kwargs: Any, ) -> Tigris: """Return VectorStore initialized from texts and embeddings.""" if not index_name: raise ValueError("`index_name` is required") if not client: client = TigrisClient() store = cls(client, embedding, index_name) store.add_texts(texts=texts, metadatas=metadatas, ids=ids) return store def _prep_docs( self, texts: Iterable[str], metadatas: Optional[List[dict]], ids: Optional[List[str]], ) -> List[TigrisDocument]: embeddings: List[List[float]] = self._embed_fn.embed_documents(list(texts)) docs: List[TigrisDocument] = [] for t, m, e, _id in itertools.zip_longest( texts, metadatas or [], embeddings or [], ids or [] ): doc: TigrisDocument = { "text": t, "embeddings": e or [], "metadata": m or {}, } if _id: doc["id"] = _id docs.append(doc) return docs
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/tigris.html
442359b8-1b63-4d9a-bfc7-8924d6a16625
Source code for langchain.vectorstores.starrocks """Wrapper around open source StarRocks VectorSearch capability.""" from __future__ import annotations import json import logging from hashlib import sha1 from threading import Thread from typing import Any, Dict, Iterable, List, Optional, Tuple from pydantic import BaseSettings from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.vectorstores.base import VectorStore logger = logging.getLogger() DEBUG = False def has_mul_sub_str(s: str, *args: Any) -> bool: """ Check if a string has multiple substrings. Args: s: The string to check *args: The substrings to check for in the string Returns: bool: True if all substrings are present in the string, False otherwise """ for a in args: if a not in s: return False return True def debug_output(s: Any) -> None: """ Print a debug message if DEBUG is True. Args: s: The message to print """ if DEBUG: print(s) def get_named_result(connection: Any, query: str) -> List[dict[str, Any]]: """ Get a named result from a query. Args: connection: The connection to the database query: The query to execute Returns: List[dict[str, Any]]: The result of the query """ cursor = connection.cursor() cursor.execute(query) columns = cursor.description result = [] for value in cursor.fetchall(): r = {} for idx, datum in enumerate(value): k = columns[idx][0] r[k] = datum result.append(r) debug_output(result) cursor.close() return result class StarRocksSettings(BaseSettings): """StarRocks Client Configuration Attribute: StarRocks_host (str) : An URL to connect to MyScale backend. Defaults to 'localhost'. StarRocks_port (int) : URL port to connect with HTTP. Defaults to 8443. username (str) : Username to login. Defaults to None. password (str) : Password to login. Defaults to None. database (str) : Database name to find the table. Defaults to 'default'. table (str) : Table name to operate on. Defaults to 'vector_table'. column_map (Dict) : Column type map to project column name onto langchain semantics. Must have keys: `text`, `id`, `vector`, must be same size to number of columns. For example: .. code-block:: python { 'id': 'text_id', 'embedding': 'text_embedding', 'document': 'text_plain', 'metadata': 'metadata_dictionary_in_json', } Defaults to identity map. """ host: str = "localhost" port: int = 9030 username: str = "root" password: str = "" column_map: Dict[str, str] = { "id": "id", "document": "document", "embedding": "embedding", "metadata": "metadata", } database: str = "default" table: str = "langchain" def __getitem__(self, item: str) -> Any: return getattr(self, item) class Config: env_file = ".env" env_prefix = "starrocks_" env_file_encoding = "utf-8" [docs]class StarRocks(VectorStore): """Wrapper around StarRocks vector database You need a `pymysql` python package, and a valid account to connect to StarRocks. Right now StarRocks has only implemented `cosine_similarity` function to compute distance between two vectors. And there is no vector inside right now, so we have to iterate all vectors and compute spatial distance. For more information, please visit [StarRocks official site](https://www.starrocks.io/) [StarRocks github](https://github.com/StarRocks/starrocks) """ def __init__( self, embedding: Embeddings, config: Optional[StarRocksSettings] = None, **kwargs: Any, ) -> None: """StarRocks Wrapper to LangChain embedding_function (Embeddings): config (StarRocksSettings): Configuration to StarRocks Client """ try: import pymysql # type: ignore[import] except ImportError: raise ImportError( "Could not import pymysql python package. " "Please install it with `pip install pymysql`." ) try: from tqdm import tqdm self.pgbar = tqdm except ImportError: # Just in case if tqdm is not installed self.pgbar = lambda x, **kwargs: x super().__init__() if config is not None: self.config = config else: self.config = StarRocksSettings() assert self.config assert self.config.host and self.config.port assert self.config.column_map and self.config.database and self.config.table for k in ["id", "embedding", "document", "metadata"]: assert k in self.config.column_map # initialize the schema dim = len(embedding.embed_query("test")) self.schema = f"""\ CREATE TABLE IF NOT EXISTS {self.config.database}.{self.config.table}( {self.config.column_map['id']} string, {self.config.column_map['document']} string, {self.config.column_map['embedding']} array<float>, {self.config.column_map['metadata']} string ) ENGINE = OLAP PRIMARY KEY(id) DISTRIBUTED BY HASH(id) \ PROPERTIES ("replication_num" = "1")\ """ self.dim = dim self.BS = "\\" self.must_escape = ("\\", "'") self.embedding_function = embedding self.dist_order = "DESC" debug_output(self.config) # Create a connection to StarRocks self.connection = pymysql.connect( host=self.config.host, port=self.config.port, user=self.config.username, password=self.config.password, database=self.config.database, **kwargs, ) debug_output(self.schema) get_named_result(self.connection, self.schema) [docs] def escape_str(self, value: str) -> str: return "".join(f"{self.BS}{c}" if c in self.must_escape else c for c in value) def _build_insert_sql(self, transac: Iterable, column_names: Iterable[str]) -> str: ks = ",".join(column_names) embed_tuple_index = tuple(column_names).index( self.config.column_map["embedding"] ) _data = [] for n in transac: n = ",".join( [ f"'{self.escape_str(str(_n))}'" if idx != embed_tuple_index else f"array<float>{str(_n)}" for (idx, _n) in enumerate(n) ] ) _data.append(f"({n})") i_str = f""" INSERT INTO {self.config.database}.{self.config.table}({ks}) VALUES {','.join(_data)} """ return i_str def _insert(self, transac: Iterable, column_names: Iterable[str]) -> None: _insert_query = self._build_insert_sql(transac, column_names) debug_output(_insert_query) get_named_result(self.connection, _insert_query) [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, batch_size: int = 32, ids: Optional[Iterable[str]] = None, **kwargs: Any, ) -> List[str]: """Insert more texts through the embeddings and add to the VectorStore. Args: texts: Iterable of strings to add to the VectorStore. ids: Optional list of ids to associate with the texts. batch_size: Batch size of insertion metadata: Optional column data to be inserted Returns: List of ids from adding the texts into the VectorStore. """ # Embed and create the documents ids = ids or [sha1(t.encode("utf-8")).hexdigest() for t in texts] colmap_ = self.config.column_map transac = [] column_names = { colmap_["id"]: ids, colmap_["document"]: texts, colmap_["embedding"]: self.embedding_function.embed_documents(list(texts)), } metadatas = metadatas or [{} for _ in texts] column_names[colmap_["metadata"]] = map(json.dumps, metadatas) assert len(set(colmap_) - set(column_names)) >= 0 keys, values = zip(*column_names.items()) try: t = None for v in self.pgbar( zip(*values), desc="Inserting data...", total=len(metadatas) ): assert ( len(v[keys.index(self.config.column_map["embedding"])]) == self.dim ) transac.append(v) if len(transac) == batch_size: if t: t.join() t = Thread(target=self._insert, args=[transac, keys]) t.start() transac = [] if len(transac) > 0: if t: t.join() self._insert(transac, keys) return [i for i in ids] except Exception as e: logger.error(f"\033[91m\033[1m{type(e)}\033[0m \033[95m{str(e)}\033[0m") return [] [docs] @classmethod def from_texts( cls, texts: List[str], embedding: Embeddings, metadatas: Optional[List[Dict[Any, Any]]] = None, config: Optional[StarRocksSettings] = None, text_ids: Optional[Iterable[str]] = None, batch_size: int = 32, **kwargs: Any, ) -> StarRocks: """Create StarRocks wrapper with existing texts Args: embedding_function (Embeddings): Function to extract text embedding texts (Iterable[str]): List or tuple of strings to be added config (StarRocksSettings, Optional): StarRocks configuration text_ids (Optional[Iterable], optional): IDs for the texts. Defaults to None. batch_size (int, optional): Batchsize when transmitting data to StarRocks. Defaults to 32. metadata (List[dict], optional): metadata to texts. Defaults to None. Returns: StarRocks Index """ ctx = cls(embedding, config, **kwargs) ctx.add_texts(texts, ids=text_ids, batch_size=batch_size, metadatas=metadatas) return ctx def __repr__(self) -> str: """Text representation for StarRocks Vector Store, prints backends, username and schemas. Easy to use with `str(StarRocks())` Returns: repr: string to show connection info and data schema """ _repr = f"\033[92m\033[1m{self.config.database}.{self.config.table} @ " _repr += f"{self.config.host}:{self.config.port}\033[0m\n\n" _repr += f"\033[1musername: {self.config.username}\033[0m\n\nTable Schema:\n" width = 25 fields = 3 _repr += "-" * (width * fields + 1) + "\n" columns = ["name", "type", "key"] _repr += f"|\033[94m{columns[0]:24s}\033[0m|\033[96m{columns[1]:24s}" _repr += f"\033[0m|\033[96m{columns[2]:24s}\033[0m|\n" _repr += "-" * (width * fields + 1) + "\n" q_str = f"DESC {self.config.database}.{self.config.table}" debug_output(q_str) rs = get_named_result(self.connection, q_str) for r in rs: _repr += f"|\033[94m{r['Field']:24s}\033[0m|\033[96m{r['Type']:24s}" _repr += f"\033[0m|\033[96m{r['Key']:24s}\033[0m|\n" _repr += "-" * (width * fields + 1) + "\n" return _repr def _build_query_sql( self, q_emb: List[float], topk: int, where_str: Optional[str] = None ) -> str: q_emb_str = ",".join(map(str, q_emb)) if where_str: where_str = f"WHERE {where_str}" else: where_str = "" q_str = f""" SELECT {self.config.column_map['document']}, {self.config.column_map['metadata']}, cosine_similarity_norm(array<float>[{q_emb_str}], {self.config.column_map['embedding']}) as dist FROM {self.config.database}.{self.config.table} {where_str} ORDER BY dist {self.dist_order} LIMIT {topk} """ debug_output(q_str) return q_str [docs] def similarity_search( self, query: str, k: int = 4, where_str: Optional[str] = None, **kwargs: Any ) -> List[Document]: """Perform a similarity search with StarRocks Args: query (str): query string k (int, optional): Top K neighbors to retrieve. Defaults to 4. where_str (Optional[str], optional): where condition string. Defaults to None. NOTE: Please do not let end-user to fill this and always be aware of SQL injection. When dealing with metadatas, remember to use `{self.metadata_column}.attribute` instead of `attribute` alone. The default name for it is `metadata`. Returns: List[Document]: List of Documents """ return self.similarity_search_by_vector( self.embedding_function.embed_query(query), k, where_str, **kwargs ) [docs] def similarity_search_by_vector( self, embedding: List[float], k: int = 4, where_str: Optional[str] = None, **kwargs: Any, ) -> List[Document]: """Perform a similarity search with StarRocks by vectors Args: query (str): query string k (int, optional): Top K neighbors to retrieve. Defaults to 4. where_str (Optional[str], optional): where condition string. Defaults to None. NOTE: Please do not let end-user to fill this and always be aware of SQL injection. When dealing with metadatas, remember to use `{self.metadata_column}.attribute` instead of `attribute` alone. The default name for it is `metadata`. Returns: List[Document]: List of (Document, similarity) """ q_str = self._build_query_sql(embedding, k, where_str) try: return [ Document( page_content=r[self.config.column_map["document"]], metadata=json.loads(r[self.config.column_map["metadata"]]), ) for r in get_named_result(self.connection, q_str) ] except Exception as e: logger.error(f"\033[91m\033[1m{type(e)}\033[0m \033[95m{str(e)}\033[0m") return [] [docs] def similarity_search_with_relevance_scores( self, query: str, k: int = 4, where_str: Optional[str] = None, **kwargs: Any ) -> List[Tuple[Document, float]]: """Perform a similarity search with StarRocks Args: query (str): query string k (int, optional): Top K neighbors to retrieve. Defaults to 4. where_str (Optional[str], optional): where condition string. Defaults to None. NOTE: Please do not let end-user to fill this and always be aware of SQL injection. When dealing with metadatas, remember to use `{self.metadata_column}.attribute` instead of `attribute` alone. The default name for it is `metadata`. Returns: List[Document]: List of documents """ q_str = self._build_query_sql( self.embedding_function.embed_query(query), k, where_str ) try: return [ ( Document( page_content=r[self.config.column_map["document"]], metadata=json.loads(r[self.config.column_map["metadata"]]), ), r["dist"], ) for r in get_named_result(self.connection, q_str) ] except Exception as e: logger.error(f"\033[91m\033[1m{type(e)}\033[0m \033[95m{str(e)}\033[0m") return [] [docs] def drop(self) -> None: """ Helper function: Drop data """ get_named_result( self.connection, f"DROP TABLE IF EXISTS {self.config.database}.{self.config.table}", ) @property def metadata_column(self) -> str: return self.config.column_map["metadata"]
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/starrocks.html
42a885e8-e2c4-471f-ad50-0e72902b6ef1
Source code for langchain.vectorstores.vectara """Wrapper around Vectara vector database.""" from __future__ import annotations import json import logging import os from hashlib import md5 from typing import Any, Iterable, List, Optional, Tuple, Type import requests from pydantic import Field from langchain.embeddings.base import Embeddings from langchain.schema import Document from langchain.vectorstores.base import VectorStore, VectorStoreRetriever [docs]class Vectara(VectorStore): """Implementation of Vector Store using Vectara (https://vectara.com). Example: .. code-block:: python from langchain.vectorstores import Vectara vectorstore = Vectara( vectara_customer_id=vectara_customer_id, vectara_corpus_id=vectara_corpus_id, vectara_api_key=vectara_api_key ) """ def __init__( self, vectara_customer_id: Optional[str] = None, vectara_corpus_id: Optional[str] = None, vectara_api_key: Optional[str] = None, ): """Initialize with Vectara API.""" self._vectara_customer_id = vectara_customer_id or os.environ.get( "VECTARA_CUSTOMER_ID" ) self._vectara_corpus_id = vectara_corpus_id or os.environ.get( "VECTARA_CORPUS_ID" ) self._vectara_api_key = vectara_api_key or os.environ.get("VECTARA_API_KEY") if ( self._vectara_customer_id is None or self._vectara_corpus_id is None or self._vectara_api_key is None ): logging.warning( "Cant find Vectara credentials, customer_id or corpus_id in " "environment." ) else: logging.debug(f"Using corpus id {self._vectara_corpus_id}") self._session = requests.Session() # to reuse connections adapter = requests.adapters.HTTPAdapter(max_retries=3) self._session.mount("http://", adapter) def _get_post_headers(self) -> dict: """Returns headers that should be attached to each post request.""" return { "x-api-key": self._vectara_api_key, "customer-id": self._vectara_customer_id, "Content-Type": "application/json", } def _delete_doc(self, doc_id: str) -> bool: """ Delete a document from the Vectara corpus. Args: url (str): URL of the page to delete. doc_id (str): ID of the document to delete. Returns: bool: True if deletion was successful, False otherwise. """ body = { "customer_id": self._vectara_customer_id, "corpus_id": self._vectara_corpus_id, "document_id": doc_id, } response = self._session.post( "https://api.vectara.io/v1/delete-doc", data=json.dumps(body), verify=True, headers=self._get_post_headers(), ) if response.status_code != 200: logging.error( f"Delete request failed for doc_id = {doc_id} with status code " f"{response.status_code}, reason {response.reason}, text " f"{response.text}" ) return False return True def _index_doc(self, doc: dict) -> bool: request: dict[str, Any] = {} request["customer_id"] = self._vectara_customer_id request["corpus_id"] = self._vectara_corpus_id request["document"] = doc response = self._session.post( headers=self._get_post_headers(), url="https://api.vectara.io/v1/core/index", data=json.dumps(request), timeout=30, verify=True, ) status_code = response.status_code result = response.json() status_str = result["status"]["code"] if "status" in result else None if status_code == 409 or (status_str and status_str == "ALREADY_EXISTS"): return False else: return True [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, **kwargs: Any, ) -> List[str]: """Run more texts through the embeddings and add to the vectorstore. Args: texts: Iterable of strings to add to the vectorstore. metadatas: Optional list of metadatas associated with the texts. Returns: List of ids from adding the texts into the vectorstore. """ doc_hash = md5() for t in texts: doc_hash.update(t.encode()) doc_id = doc_hash.hexdigest() if metadatas is None: metadatas = [{} for _ in texts] doc = { "document_id": doc_id, "metadataJson": json.dumps({"source": "langchain"}), "parts": [ {"text": text, "metadataJson": json.dumps(md)} for text, md in zip(texts, metadatas) ], } succeeded = self._index_doc(doc) if not succeeded: self._delete_doc(doc_id) self._index_doc(doc) return [doc_id] [docs] def similarity_search_with_score( self, query: str, k: int = 5, lambda_val: float = 0.025, filter: Optional[str] = None, n_sentence_context: int = 0, **kwargs: Any, ) -> List[Tuple[Document, float]]: """Return Vectara documents most similar to query, along with scores. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 5. lambda_val: lexical match parameter for hybrid search. filter: Dictionary of argument(s) to filter on metadata. For example a filter can be "doc.rating > 3.0 and part.lang = 'deu'"} see https://docs.vectara.com/docs/search-apis/sql/filter-overview for more details. n_sentence_context: number of sentences before/after the matching segment to add Returns: List of Documents most similar to the query and score for each. """ data = json.dumps( { "query": [ { "query": query, "start": 0, "num_results": k, "context_config": { "sentences_before": n_sentence_context, "sentences_after": n_sentence_context, }, "corpus_key": [ { "customer_id": self._vectara_customer_id, "corpus_id": self._vectara_corpus_id, "metadataFilter": filter, "lexical_interpolation_config": {"lambda": lambda_val}, } ], } ] } ) response = self._session.post( headers=self._get_post_headers(), url="https://api.vectara.io/v1/query", data=data, timeout=10, ) if response.status_code != 200: logging.error( "Query failed %s", f"(code {response.status_code}, reason {response.reason}, details " f"{response.text})", ) return [] result = response.json() responses = result["responseSet"][0]["response"] vectara_default_metadata = ["lang", "len", "offset"] docs = [ ( Document( page_content=x["text"], metadata={ m["name"]: m["value"] for m in x["metadata"] if m["name"] not in vectara_default_metadata }, ), x["score"], ) for x in responses ] return docs [docs] def similarity_search( self, query: str, k: int = 5, lambda_val: float = 0.025, filter: Optional[str] = None, n_sentence_context: int = 0, **kwargs: Any, ) -> List[Document]: """Return Vectara documents most similar to query, along with scores. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 5. filter: Dictionary of argument(s) to filter on metadata. For example a filter can be "doc.rating > 3.0 and part.lang = 'deu'"} see https://docs.vectara.com/docs/search-apis/sql/filter-overview for more details. n_sentence_context: number of sentences before/after the matching segment to add Returns: List of Documents most similar to the query """ docs_and_scores = self.similarity_search_with_score( query, k=k, lambda_val=lambda_val, filter=filter, n_sentence_context=n_sentence_context, **kwargs, ) return [doc for doc, _ in docs_and_scores] [docs] @classmethod def from_texts( cls: Type[Vectara], texts: List[str], embedding: Optional[Embeddings] = None, metadatas: Optional[List[dict]] = None, **kwargs: Any, ) -> Vectara: """Construct Vectara wrapper from raw documents. This is intended to be a quick way to get started. Example: .. code-block:: python from langchain import Vectara vectara = Vectara.from_texts( texts, vectara_customer_id=customer_id, vectara_corpus_id=corpus_id, vectara_api_key=api_key, ) """ # Note: Vectara generates its own embeddings, so we ignore the provided # embeddings (required by interface) vectara = cls(**kwargs) vectara.add_texts(texts, metadatas) return vectara [docs] def as_retriever(self, **kwargs: Any) -> VectaraRetriever: return VectaraRetriever(vectorstore=self, **kwargs) class VectaraRetriever(VectorStoreRetriever): vectorstore: Vectara search_kwargs: dict = Field( default_factory=lambda: { "lambda_val": 0.025, "k": 5, "filter": "", "n_sentence_context": "0", } ) """Search params. k: Number of Documents to return. Defaults to 5. lambda_val: lexical match parameter for hybrid search. filter: Dictionary of argument(s) to filter on metadata. For example a filter can be "doc.rating > 3.0 and part.lang = 'deu'"} see https://docs.vectara.com/docs/search-apis/sql/filter-overview for more details. n_sentence_context: number of sentences before/after the matching segment to add """ def add_texts( self, texts: List[str], metadatas: Optional[List[dict]] = None ) -> None: """Add text to the Vectara vectorstore. Args: texts (List[str]): The text metadatas (List[dict]): Metadata dicts, must line up with existing store """ self.vectorstore.add_texts(texts, metadatas)
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/vectara.html
4bdb6b26-7ece-428a-be62-8e5e7ea525aa
Source code for langchain.vectorstores.hologres """VectorStore wrapper around a Hologres database.""" from __future__ import annotations import json import logging import uuid from typing import Any, Dict, Iterable, List, Optional, Tuple, Type from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.utils import get_from_dict_or_env from langchain.vectorstores.base import VectorStore ADA_TOKEN_COUNT = 1536 _LANGCHAIN_DEFAULT_TABLE_NAME = "langchain_pg_embedding" class HologresWrapper: def __init__(self, connection_string: str, ndims: int, table_name: str) -> None: import psycopg2 self.table_name = table_name self.conn = psycopg2.connect(connection_string) self.cursor = self.conn.cursor() self.conn.autocommit = False self.ndims = ndims def create_vector_extension(self) -> None: self.cursor.execute("create extension if not exists proxima") self.conn.commit() def create_table(self, drop_if_exist: bool = True) -> None: if drop_if_exist: self.cursor.execute(f"drop table if exists {self.table_name}") self.conn.commit() self.cursor.execute( f"""create table if not exists {self.table_name} ( id text, embedding float4[] check(array_ndims(embedding) = 1 and \ array_length(embedding, 1) = {self.ndims}), metadata json, document text);""" ) self.cursor.execute( f"call set_table_property('{self.table_name}'" + """, 'proxima_vectors', '{"embedding":{"algorithm":"Graph", "distance_method":"SquaredEuclidean", "build_params":{"min_flush_proxima_row_count" : 1, "min_compaction_proxima_row_count" : 1, "max_total_size_to_merge_mb" : 2000}}}');""" ) self.conn.commit() def get_by_id(self, id: str) -> List[Tuple]: statement = ( f"select id, embedding, metadata, " f"document from {self.table_name} where id = %s;" ) self.cursor.execute( statement, (id), ) self.conn.commit() return self.cursor.fetchall() def insert( self, embedding: List[float], metadata: dict, document: str, id: Optional[str] = None, ) -> None: self.cursor.execute( f'insert into "{self.table_name}" ' f"values (%s, array{json.dumps(embedding)}::float4[], %s, %s)", (id if id is not None else "null", json.dumps(metadata), document), ) self.conn.commit() def query_nearest_neighbours( self, embedding: List[float], k: int, filter: Optional[Dict[str, str]] = None ) -> List[Tuple[str, str, float]]: params = [] filter_clause = "" if filter is not None: conjuncts = [] for key, val in filter.items(): conjuncts.append("metadata->>%s=%s") params.append(key) params.append(val) filter_clause = "where " + " and ".join(conjuncts) sql = ( f"select document, metadata::text, " f"pm_approx_squared_euclidean_distance(array{json.dumps(embedding)}" f"::float4[], embedding) as distance from" f" {self.table_name} {filter_clause} order by distance asc limit {k};" ) self.cursor.execute(sql, tuple(params)) self.conn.commit() return self.cursor.fetchall() [docs]class Hologres(VectorStore): """VectorStore implementation using Hologres. - `connection_string` is a hologres connection string. - `embedding_function` any embedding function implementing `langchain.embeddings.base.Embeddings` interface. - `ndims` is the number of dimensions of the embedding output. - `table_name` is the name of the table to store embeddings and data. (default: langchain_pg_embedding) - NOTE: The table will be created when initializing the store (if not exists) So, make sure the user has the right permissions to create tables. - `pre_delete_table` if True, will delete the table if it exists. (default: False) - Useful for testing. """ def __init__( self, connection_string: str, embedding_function: Embeddings, ndims: int = ADA_TOKEN_COUNT, table_name: str = _LANGCHAIN_DEFAULT_TABLE_NAME, pre_delete_table: bool = False, logger: Optional[logging.Logger] = None, ) -> None: self.connection_string = connection_string self.ndims = ndims self.table_name = table_name self.embedding_function = embedding_function self.pre_delete_table = pre_delete_table self.logger = logger or logging.getLogger(__name__) self.__post_init__() def __post_init__( self, ) -> None: """ Initialize the store. """ self.storage = HologresWrapper( self.connection_string, self.ndims, self.table_name ) self.create_vector_extension() self.create_table() [docs] def create_vector_extension(self) -> None: try: self.storage.create_vector_extension() except Exception as e: self.logger.exception(e) raise e [docs] def create_table(self) -> None: self.storage.create_table(self.pre_delete_table) @classmethod def __from( cls, texts: List[str], embeddings: List[List[float]], embedding_function: Embeddings, metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, ndims: int = ADA_TOKEN_COUNT, table_name: str = _LANGCHAIN_DEFAULT_TABLE_NAME, pre_delete_table: bool = False, **kwargs: Any, ) -> Hologres: if ids is None: ids = [str(uuid.uuid1()) for _ in texts] if not metadatas: metadatas = [{} for _ in texts] connection_string = cls.get_connection_string(kwargs) store = cls( connection_string=connection_string, embedding_function=embedding_function, ndims=ndims, table_name=table_name, pre_delete_table=pre_delete_table, ) store.add_embeddings( texts=texts, embeddings=embeddings, metadatas=metadatas, ids=ids, **kwargs ) return store [docs] def add_embeddings( self, texts: Iterable[str], embeddings: List[List[float]], metadatas: List[dict], ids: List[str], **kwargs: Any, ) -> None: """Add embeddings to the vectorstore. Args: texts: Iterable of strings to add to the vectorstore. embeddings: List of list of embedding vectors. metadatas: List of metadatas associated with the texts. kwargs: vectorstore specific parameters """ try: for text, metadata, embedding, id in zip(texts, metadatas, embeddings, ids): self.storage.insert(embedding, metadata, text, id) except Exception as e: self.logger.exception(e) self.storage.conn.commit() [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, **kwargs: Any, ) -> List[str]: """Run more texts through the embeddings and add to the vectorstore. Args: texts: Iterable of strings to add to the vectorstore. metadatas: Optional list of metadatas associated with the texts. kwargs: vectorstore specific parameters Returns: List of ids from adding the texts into the vectorstore. """ if ids is None: ids = [str(uuid.uuid1()) for _ in texts] embeddings = self.embedding_function.embed_documents(list(texts)) if not metadatas: metadatas = [{} for _ in texts] self.add_embeddings(texts, embeddings, metadatas, ids, **kwargs) return ids [docs] def similarity_search( self, query: str, k: int = 4, filter: Optional[dict] = None, **kwargs: Any, ) -> List[Document]: """Run similarity search with Hologres with distance. Args: query (str): Query text to search for. k (int): Number of results to return. Defaults to 4. filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. Returns: List of Documents most similar to the query. """ embedding = self.embedding_function.embed_query(text=query) return self.similarity_search_by_vector( embedding=embedding, k=k, filter=filter, ) [docs] def similarity_search_by_vector( self, embedding: List[float], k: int = 4, filter: Optional[dict] = None, **kwargs: Any, ) -> List[Document]: """Return docs most similar to embedding vector. Args: embedding: Embedding to look up documents similar to. k: Number of Documents to return. Defaults to 4. filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. Returns: List of Documents most similar to the query vector. """ docs_and_scores = self.similarity_search_with_score_by_vector( embedding=embedding, k=k, filter=filter ) return [doc for doc, _ in docs_and_scores] [docs] def similarity_search_with_score( self, query: str, k: int = 4, filter: Optional[dict] = None, ) -> List[Tuple[Document, float]]: """Return docs most similar to query. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. Returns: List of Documents most similar to the query and score for each """ embedding = self.embedding_function.embed_query(query) docs = self.similarity_search_with_score_by_vector( embedding=embedding, k=k, filter=filter ) return docs [docs] def similarity_search_with_score_by_vector( self, embedding: List[float], k: int = 4, filter: Optional[dict] = None, ) -> List[Tuple[Document, float]]: results: List[Tuple[str, str, float]] = self.storage.query_nearest_neighbours( embedding, k, filter ) docs = [ ( Document( page_content=result[0], metadata=json.loads(result[1]), ), result[2], ) for result in results ] return docs [docs] @classmethod def from_texts( cls: Type[Hologres], texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, ndims: int = ADA_TOKEN_COUNT, table_name: str = _LANGCHAIN_DEFAULT_TABLE_NAME, ids: Optional[List[str]] = None, pre_delete_table: bool = False, **kwargs: Any, ) -> Hologres: """ Return VectorStore initialized from texts and embeddings. Postgres connection string is required "Either pass it as a parameter or set the HOLOGRES_CONNECTION_STRING environment variable. """ embeddings = embedding.embed_documents(list(texts)) return cls.__from( texts, embeddings, embedding, metadatas=metadatas, ids=ids, ndims=ndims, table_name=table_name, pre_delete_table=pre_delete_table, **kwargs, ) [docs] @classmethod def from_embeddings( cls, text_embeddings: List[Tuple[str, List[float]]], embedding: Embeddings, metadatas: Optional[List[dict]] = None, ndims: int = ADA_TOKEN_COUNT, table_name: str = _LANGCHAIN_DEFAULT_TABLE_NAME, ids: Optional[List[str]] = None, pre_delete_table: bool = False, **kwargs: Any, ) -> Hologres: """Construct Hologres wrapper from raw documents and pre- generated embeddings. Return VectorStore initialized from documents and embeddings. Postgres connection string is required "Either pass it as a parameter or set the HOLOGRES_CONNECTION_STRING environment variable. Example: .. code-block:: python from langchain import Hologres from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() text_embeddings = embeddings.embed_documents(texts) text_embedding_pairs = list(zip(texts, text_embeddings)) faiss = Hologres.from_embeddings(text_embedding_pairs, embeddings) """ texts = [t[0] for t in text_embeddings] embeddings = [t[1] for t in text_embeddings] return cls.__from( texts, embeddings, embedding, metadatas=metadatas, ids=ids, ndims=ndims, table_name=table_name, pre_delete_table=pre_delete_table, **kwargs, ) [docs] @classmethod def from_existing_index( cls: Type[Hologres], embedding: Embeddings, ndims: int = ADA_TOKEN_COUNT, table_name: str = _LANGCHAIN_DEFAULT_TABLE_NAME, pre_delete_table: bool = False, **kwargs: Any, ) -> Hologres: """ Get intsance of an existing Hologres store.This method will return the instance of the store without inserting any new embeddings """ connection_string = cls.get_connection_string(kwargs) store = cls( connection_string=connection_string, ndims=ndims, table_name=table_name, embedding_function=embedding, pre_delete_table=pre_delete_table, ) return store [docs] @classmethod def get_connection_string(cls, kwargs: Dict[str, Any]) -> str: connection_string: str = get_from_dict_or_env( data=kwargs, key="connection_string", env_key="HOLOGRES_CONNECTION_STRING", ) if not connection_string: raise ValueError( "Postgres connection string is required" "Either pass it as a parameter" "or set the HOLOGRES_CONNECTION_STRING environment variable." ) return connection_string [docs] @classmethod def from_documents( cls: Type[Hologres], documents: List[Document], embedding: Embeddings, ndims: int = ADA_TOKEN_COUNT, table_name: str = _LANGCHAIN_DEFAULT_TABLE_NAME, ids: Optional[List[str]] = None, pre_delete_collection: bool = False, **kwargs: Any, ) -> Hologres: """ Return VectorStore initialized from documents and embeddings. Postgres connection string is required "Either pass it as a parameter or set the HOLOGRES_CONNECTION_STRING environment variable. """ texts = [d.page_content for d in documents] metadatas = [d.metadata for d in documents] connection_string = cls.get_connection_string(kwargs) kwargs["connection_string"] = connection_string return cls.from_texts( texts=texts, pre_delete_collection=pre_delete_collection, embedding=embedding, metadatas=metadatas, ids=ids, ndims=ndims, table_name=table_name, **kwargs, ) [docs] @classmethod def connection_string_from_db_params( cls, host: str, port: int, database: str, user: str, password: str, ) -> str: """Return connection string from database parameters.""" return ( f"dbname={database} user={user} password={password} host={host} port={port}" )
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/hologres.html
954a436d-7f80-49e2-945b-995b28e13c49
Source code for langchain.vectorstores.redis """Wrapper around Redis vector database.""" from __future__ import annotations import json import logging import uuid from typing import ( TYPE_CHECKING, Any, Callable, Dict, Iterable, List, Literal, Mapping, Optional, Tuple, Type, ) import numpy as np from pydantic import BaseModel, root_validator from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.utils import get_from_dict_or_env from langchain.vectorstores.base import VectorStore, VectorStoreRetriever logger = logging.getLogger(__name__) if TYPE_CHECKING: from redis.client import Redis as RedisType from redis.commands.search.query import Query # required modules REDIS_REQUIRED_MODULES = [ {"name": "search", "ver": 20400}, {"name": "searchlight", "ver": 20400}, ] # distance mmetrics REDIS_DISTANCE_METRICS = Literal["COSINE", "IP", "L2"] def _check_redis_module_exist(client: RedisType, required_modules: List[dict]) -> None: """Check if the correct Redis modules are installed.""" installed_modules = client.module_list() installed_modules = { module[b"name"].decode("utf-8"): module for module in installed_modules } for module in required_modules: if module["name"] in installed_modules and int( installed_modules[module["name"]][b"ver"] ) >= int(module["ver"]): return # otherwise raise error error_message = ( "Redis cannot be used as a vector database without RediSearch >=2.4" "Please head to https://redis.io/docs/stack/search/quick_start/" "to know more about installing the RediSearch module within Redis Stack." ) logging.error(error_message) raise ValueError(error_message) def _check_index_exists(client: RedisType, index_name: str) -> bool: """Check if Redis index exists.""" try: client.ft(index_name).info() except: # noqa: E722 logger.info("Index does not exist") return False logger.info("Index already exists") return True def _redis_key(prefix: str) -> str: """Redis key schema for a given prefix.""" return f"{prefix}:{uuid.uuid4().hex}" def _redis_prefix(index_name: str) -> str: """Redis key prefix for a given index.""" return f"doc:{index_name}" def _default_relevance_score(val: float) -> float: return 1 - val [docs]class Redis(VectorStore): """Wrapper around Redis vector database. To use, you should have the ``redis`` python package installed. Example: .. code-block:: python from langchain.vectorstores import Redis from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() vectorstore = Redis( redis_url="redis://username:password@localhost:6379" index_name="my-index", embedding_function=embeddings.embed_query, ) """ def __init__( self, redis_url: str, index_name: str, embedding_function: Callable, content_key: str = "content", metadata_key: str = "metadata", vector_key: str = "content_vector", relevance_score_fn: Optional[ Callable[[float], float] ] = _default_relevance_score, **kwargs: Any, ): """Initialize with necessary components.""" try: import redis except ImportError: raise ValueError( "Could not import redis python package. " "Please install it with `pip install redis>=4.1.0`." ) self.embedding_function = embedding_function self.index_name = index_name try: # connect to redis from url redis_client = redis.from_url(redis_url, **kwargs) # check if redis has redisearch module installed _check_redis_module_exist(redis_client, REDIS_REQUIRED_MODULES) except ValueError as e: raise ValueError(f"Redis failed to connect: {e}") self.client = redis_client self.content_key = content_key self.metadata_key = metadata_key self.vector_key = vector_key self.relevance_score_fn = relevance_score_fn def _create_index( self, dim: int = 1536, distance_metric: REDIS_DISTANCE_METRICS = "COSINE" ) -> None: try: from redis.commands.search.field import TextField, VectorField from redis.commands.search.indexDefinition import IndexDefinition, IndexType except ImportError: raise ValueError( "Could not import redis python package. " "Please install it with `pip install redis`." ) # Check if index exists if not _check_index_exists(self.client, self.index_name): # Define schema schema = ( TextField(name=self.content_key), TextField(name=self.metadata_key), VectorField( self.vector_key, "FLAT", { "TYPE": "FLOAT32", "DIM": dim, "DISTANCE_METRIC": distance_metric, }, ), ) prefix = _redis_prefix(self.index_name) # Create Redis Index self.client.ft(self.index_name).create_index( fields=schema, definition=IndexDefinition(prefix=[prefix], index_type=IndexType.HASH), ) [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, embeddings: Optional[List[List[float]]] = None, batch_size: int = 1000, **kwargs: Any, ) -> List[str]: """Add more texts to the vectorstore. Args: texts (Iterable[str]): Iterable of strings/text to add to the vectorstore. metadatas (Optional[List[dict]], optional): Optional list of metadatas. Defaults to None. embeddings (Optional[List[List[float]]], optional): Optional pre-generated embeddings. Defaults to None. keys (List[str]) or ids (List[str]): Identifiers of entries. Defaults to None. batch_size (int, optional): Batch size to use for writes. Defaults to 1000. Returns: List[str]: List of ids added to the vectorstore """ ids = [] prefix = _redis_prefix(self.index_name) # Get keys or ids from kwargs # Other vectorstores use ids keys_or_ids = kwargs.get("keys", kwargs.get("ids")) # Write data to redis pipeline = self.client.pipeline(transaction=False) for i, text in enumerate(texts): # Use provided values by default or fallback key = keys_or_ids[i] if keys_or_ids else _redis_key(prefix) metadata = metadatas[i] if metadatas else {} embedding = embeddings[i] if embeddings else self.embedding_function(text) pipeline.hset( key, mapping={ self.content_key: text, self.vector_key: np.array(embedding, dtype=np.float32).tobytes(), self.metadata_key: json.dumps(metadata), }, ) ids.append(key) # Write batch if i % batch_size == 0: pipeline.execute() # Cleanup final batch pipeline.execute() return ids [docs] def similarity_search( self, query: str, k: int = 4, **kwargs: Any ) -> List[Document]: """ Returns the most similar indexed documents to the query text. Args: query (str): The query text for which to find similar documents. k (int): The number of documents to return. Default is 4. Returns: List[Document]: A list of documents that are most similar to the query text. """ docs_and_scores = self.similarity_search_with_score(query, k=k) return [doc for doc, _ in docs_and_scores] [docs] def similarity_search_limit_score( self, query: str, k: int = 4, score_threshold: float = 0.2, **kwargs: Any ) -> List[Document]: """ Returns the most similar indexed documents to the query text within the score_threshold range. Args: query (str): The query text for which to find similar documents. k (int): The number of documents to return. Default is 4. score_threshold (float): The minimum matching score required for a document to be considered a match. Defaults to 0.2. Because the similarity calculation algorithm is based on cosine similarity, the smaller the angle, the higher the similarity. Returns: List[Document]: A list of documents that are most similar to the query text, including the match score for each document. Note: If there are no documents that satisfy the score_threshold value, an empty list is returned. """ docs_and_scores = self.similarity_search_with_score(query, k=k) return [doc for doc, score in docs_and_scores if score < score_threshold] def _prepare_query(self, k: int) -> Query: try: from redis.commands.search.query import Query except ImportError: raise ValueError( "Could not import redis python package. " "Please install it with `pip install redis`." ) # Prepare the Query hybrid_fields = "*" base_query = ( f"{hybrid_fields}=>[KNN {k} @{self.vector_key} $vector AS vector_score]" ) return_fields = [self.metadata_key, self.content_key, "vector_score"] return ( Query(base_query) .return_fields(*return_fields) .sort_by("vector_score") .paging(0, k) .dialect(2) ) [docs] def similarity_search_with_score( self, query: str, k: int = 4 ) -> List[Tuple[Document, float]]: """Return docs most similar to query. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. Returns: List of Documents most similar to the query and score for each """ # Creates embedding vector from user query embedding = self.embedding_function(query) # Creates Redis query redis_query = self._prepare_query(k) params_dict: Mapping[str, str] = { "vector": np.array(embedding) # type: ignore .astype(dtype=np.float32) .tobytes() } # Perform vector search results = self.client.ft(self.index_name).search(redis_query, params_dict) # Prepare document results docs = [ ( Document( page_content=result.content, metadata=json.loads(result.metadata) ), float(result.vector_score), ) for result in results.docs ] return docs def _similarity_search_with_relevance_scores( self, query: str, k: int = 4, **kwargs: Any, ) -> List[Tuple[Document, float]]: """Return docs and relevance scores, normalized on a scale from 0 to 1. 0 is dissimilar, 1 is most similar. """ if self.relevance_score_fn is None: raise ValueError( "relevance_score_fn must be provided to" " Redis constructor to normalize scores" ) docs_and_scores = self.similarity_search_with_score(query, k=k) return [(doc, self.relevance_score_fn(score)) for doc, score in docs_and_scores] [docs] @classmethod def from_texts_return_keys( cls, texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, index_name: Optional[str] = None, content_key: str = "content", metadata_key: str = "metadata", vector_key: str = "content_vector", distance_metric: REDIS_DISTANCE_METRICS = "COSINE", **kwargs: Any, ) -> Tuple[Redis, List[str]]: """Create a Redis vectorstore from raw documents. This is a user-friendly interface that: 1. Embeds documents. 2. Creates a new index for the embeddings in Redis. 3. Adds the documents to the newly created Redis index. 4. Returns the keys of the newly created documents. This is intended to be a quick way to get started. Example: .. code-block:: python from langchain.vectorstores import Redis from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() redisearch, keys = RediSearch.from_texts_return_keys( texts, embeddings, redis_url="redis://username:password@localhost:6379" ) """ redis_url = get_from_dict_or_env(kwargs, "redis_url", "REDIS_URL") if "redis_url" in kwargs: kwargs.pop("redis_url") # Name of the search index if not given if not index_name: index_name = uuid.uuid4().hex # Create instance instance = cls( redis_url, index_name, embedding.embed_query, content_key=content_key, metadata_key=metadata_key, vector_key=vector_key, **kwargs, ) # Create embeddings over documents embeddings = embedding.embed_documents(texts) # Create the search index instance._create_index(dim=len(embeddings[0]), distance_metric=distance_metric) # Add data to Redis keys = instance.add_texts(texts, metadatas, embeddings) return instance, keys [docs] @classmethod def from_texts( cls: Type[Redis], texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, index_name: Optional[str] = None, content_key: str = "content", metadata_key: str = "metadata", vector_key: str = "content_vector", **kwargs: Any, ) -> Redis: """Create a Redis vectorstore from raw documents. This is a user-friendly interface that: 1. Embeds documents. 2. Creates a new index for the embeddings in Redis. 3. Adds the documents to the newly created Redis index. This is intended to be a quick way to get started. Example: .. code-block:: python from langchain.vectorstores import Redis from langchain.embeddings import OpenAIEmbeddings embeddings = OpenAIEmbeddings() redisearch = RediSearch.from_texts( texts, embeddings, redis_url="redis://username:password@localhost:6379" ) """ instance, _ = cls.from_texts_return_keys( texts, embedding, metadatas=metadatas, index_name=index_name, content_key=content_key, metadata_key=metadata_key, vector_key=vector_key, **kwargs, ) return instance [docs] @staticmethod def delete( ids: List[str], **kwargs: Any, ) -> bool: """ Delete a Redis entry. Args: ids: List of ids (keys) to delete. Returns: bool: Whether or not the deletions were successful. """ redis_url = get_from_dict_or_env(kwargs, "redis_url", "REDIS_URL") if ids is None: raise ValueError("'ids' (keys)() were not provided.") try: import redis except ImportError: raise ValueError( "Could not import redis python package. " "Please install it with `pip install redis`." ) try: # We need to first remove redis_url from kwargs, # otherwise passing it to Redis will result in an error. if "redis_url" in kwargs: kwargs.pop("redis_url") client = redis.from_url(url=redis_url, **kwargs) except ValueError as e: raise ValueError(f"Your redis connected error: {e}") # Check if index exists try: client.delete(*ids) logger.info("Entries deleted") return True except: # noqa: E722 # ids does not exist return False [docs] @staticmethod def drop_index( index_name: str, delete_documents: bool, **kwargs: Any, ) -> bool: """ Drop a Redis search index. Args: index_name (str): Name of the index to drop. delete_documents (bool): Whether to drop the associated documents. Returns: bool: Whether or not the drop was successful. """ redis_url = get_from_dict_or_env(kwargs, "redis_url", "REDIS_URL") try: import redis except ImportError: raise ValueError( "Could not import redis python package. " "Please install it with `pip install redis`." ) try: # We need to first remove redis_url from kwargs, # otherwise passing it to Redis will result in an error. if "redis_url" in kwargs: kwargs.pop("redis_url") client = redis.from_url(url=redis_url, **kwargs) except ValueError as e: raise ValueError(f"Your redis connected error: {e}") # Check if index exists try: client.ft(index_name).dropindex(delete_documents) logger.info("Drop index") return True except: # noqa: E722 # Index not exist return False [docs] @classmethod def from_existing_index( cls, embedding: Embeddings, index_name: str, content_key: str = "content", metadata_key: str = "metadata", vector_key: str = "content_vector", **kwargs: Any, ) -> Redis: """Connect to an existing Redis index.""" redis_url = get_from_dict_or_env(kwargs, "redis_url", "REDIS_URL") try: import redis except ImportError: raise ValueError( "Could not import redis python package. " "Please install it with `pip install redis`." ) try: # We need to first remove redis_url from kwargs, # otherwise passing it to Redis will result in an error. if "redis_url" in kwargs: kwargs.pop("redis_url") client = redis.from_url(url=redis_url, **kwargs) # check if redis has redisearch module installed _check_redis_module_exist(client, REDIS_REQUIRED_MODULES) # ensure that the index already exists assert _check_index_exists( client, index_name ), f"Index {index_name} does not exist" except Exception as e: raise ValueError(f"Redis failed to connect: {e}") return cls( redis_url, index_name, embedding.embed_query, content_key=content_key, metadata_key=metadata_key, vector_key=vector_key, **kwargs, ) [docs] def as_retriever(self, **kwargs: Any) -> RedisVectorStoreRetriever: return RedisVectorStoreRetriever(vectorstore=self, **kwargs) class RedisVectorStoreRetriever(VectorStoreRetriever, BaseModel): vectorstore: Redis search_type: str = "similarity" k: int = 4 score_threshold: float = 0.4 class Config: """Configuration for this pydantic object.""" arbitrary_types_allowed = True @root_validator() def validate_search_type(cls, values: Dict) -> Dict: """Validate search type.""" if "search_type" in values: search_type = values["search_type"] if search_type not in ("similarity", "similarity_limit"): raise ValueError(f"search_type of {search_type} not allowed.") return values def get_relevant_documents(self, query: str) -> List[Document]: if self.search_type == "similarity": docs = self.vectorstore.similarity_search(query, k=self.k) elif self.search_type == "similarity_limit": docs = self.vectorstore.similarity_search_limit_score( query, k=self.k, score_threshold=self.score_threshold ) else: raise ValueError(f"search_type of {self.search_type} not allowed.") return docs async def aget_relevant_documents(self, query: str) -> List[Document]: raise NotImplementedError("RedisVectorStoreRetriever does not support async") def add_documents(self, documents: List[Document], **kwargs: Any) -> List[str]: """Add documents to vectorstore.""" return self.vectorstore.add_documents(documents, **kwargs) async def aadd_documents( self, documents: List[Document], **kwargs: Any ) -> List[str]: """Add documents to vectorstore.""" return await self.vectorstore.aadd_documents(documents, **kwargs)
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/redis.html
7fbfe242-ade3-4417-9a1d-07739c573194
Source code for langchain.vectorstores.zilliz from __future__ import annotations import logging from typing import Any, List, Optional from langchain.embeddings.base import Embeddings from langchain.vectorstores.milvus import Milvus logger = logging.getLogger(__name__) [docs]class Zilliz(Milvus): def _create_index(self) -> None: """Create a index on the collection""" from pymilvus import Collection, MilvusException if isinstance(self.col, Collection) and self._get_index() is None: try: # If no index params, use a default AutoIndex based one if self.index_params is None: self.index_params = { "metric_type": "L2", "index_type": "AUTOINDEX", "params": {}, } try: self.col.create_index( self._vector_field, index_params=self.index_params, using=self.alias, ) # If default did not work, most likely Milvus self-hosted except MilvusException: # Use HNSW based index self.index_params = { "metric_type": "L2", "index_type": "HNSW", "params": {"M": 8, "efConstruction": 64}, } self.col.create_index( self._vector_field, index_params=self.index_params, using=self.alias, ) logger.debug( "Successfully created an index on collection: %s", self.collection_name, ) except MilvusException as e: logger.error( "Failed to create an index on collection: %s", self.collection_name ) raise e [docs] @classmethod def from_texts( cls, texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, collection_name: str = "LangChainCollection", connection_args: dict[str, Any] = {}, consistency_level: str = "Session", index_params: Optional[dict] = None, search_params: Optional[dict] = None, drop_old: bool = False, **kwargs: Any, ) -> Zilliz: """Create a Zilliz collection, indexes it with HNSW, and insert data. Args: texts (List[str]): Text data. embedding (Embeddings): Embedding function. metadatas (Optional[List[dict]]): Metadata for each text if it exists. Defaults to None. collection_name (str, optional): Collection name to use. Defaults to "LangChainCollection". connection_args (dict[str, Any], optional): Connection args to use. Defaults to DEFAULT_MILVUS_CONNECTION. consistency_level (str, optional): Which consistency level to use. Defaults to "Session". index_params (Optional[dict], optional): Which index_params to use. Defaults to None. search_params (Optional[dict], optional): Which search params to use. Defaults to None. drop_old (Optional[bool], optional): Whether to drop the collection with that name if it exists. Defaults to False. Returns: Zilliz: Zilliz Vector Store """ vector_db = cls( embedding_function=embedding, collection_name=collection_name, connection_args=connection_args, consistency_level=consistency_level, index_params=index_params, search_params=search_params, drop_old=drop_old, **kwargs, ) vector_db.add_texts(texts=texts, metadatas=metadatas) return vector_db
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/zilliz.html
1d702ba7-c4c3-4606-92ee-5e4697ce140d
Source code for langchain.vectorstores.supabase from __future__ import annotations import uuid from itertools import repeat from typing import ( TYPE_CHECKING, Any, Iterable, List, Optional, Tuple, Type, Union, ) import numpy as np from langchain.docstore.document import Document from langchain.embeddings.base import Embeddings from langchain.vectorstores.base import VectorStore from langchain.vectorstores.utils import maximal_marginal_relevance if TYPE_CHECKING: import supabase [docs]class SupabaseVectorStore(VectorStore): """VectorStore for a Supabase postgres database. Assumes you have the `pgvector` extension installed and a `match_documents` (or similar) function. For more details: https://js.langchain.com/docs/modules/indexes/vector_stores/integrations/supabase You can implement your own `match_documents` function in order to limit the search space to a subset of documents based on your own authorization or business logic. Note that the Supabase Python client does not yet support async operations. If you'd like to use `max_marginal_relevance_search`, please review the instructions below on modifying the `match_documents` function to return matched embeddings. """ _client: supabase.client.Client # This is the embedding function. Don't confuse with the embedding vectors. # We should perhaps rename the underlying Embedding base class to EmbeddingFunction # or something _embedding: Embeddings table_name: str query_name: str def __init__( self, client: supabase.client.Client, embedding: Embeddings, table_name: str, query_name: Union[str, None] = None, ) -> None: """Initialize with supabase client.""" try: import supabase # noqa: F401 except ImportError: raise ValueError( "Could not import supabase python package. " "Please install it with `pip install supabase`." ) self._client = client self._embedding: Embeddings = embedding self.table_name = table_name or "documents" self.query_name = query_name or "match_documents" [docs] def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict[Any, Any]]] = None, ids: Optional[List[str]] = None, **kwargs: Any, ) -> List[str]: ids = ids or [str(uuid.uuid4()) for _ in texts] docs = self._texts_to_documents(texts, metadatas) vectors = self._embedding.embed_documents(list(texts)) return self.add_vectors(vectors, docs, ids) [docs] @classmethod def from_texts( cls: Type["SupabaseVectorStore"], texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, client: Optional[supabase.client.Client] = None, table_name: Optional[str] = "documents", query_name: Union[str, None] = "match_documents", ids: Optional[List[str]] = None, **kwargs: Any, ) -> "SupabaseVectorStore": """Return VectorStore initialized from texts and embeddings.""" if not client: raise ValueError("Supabase client is required.") if not table_name: raise ValueError("Supabase document table_name is required.") embeddings = embedding.embed_documents(texts) ids = [str(uuid.uuid4()) for _ in texts] docs = cls._texts_to_documents(texts, metadatas) _ids = cls._add_vectors(client, table_name, embeddings, docs, ids) return cls( client=client, embedding=embedding, table_name=table_name, query_name=query_name, ) [docs] def add_vectors( self, vectors: List[List[float]], documents: List[Document], ids: List[str], ) -> List[str]: return self._add_vectors(self._client, self.table_name, vectors, documents, ids) [docs] def similarity_search( self, query: str, k: int = 4, **kwargs: Any ) -> List[Document]: vectors = self._embedding.embed_documents([query]) return self.similarity_search_by_vector(vectors[0], k) [docs] def similarity_search_by_vector( self, embedding: List[float], k: int = 4, **kwargs: Any ) -> List[Document]: result = self.similarity_search_by_vector_with_relevance_scores(embedding, k) documents = [doc for doc, _ in result] return documents [docs] def similarity_search_with_relevance_scores( self, query: str, k: int = 4, **kwargs: Any ) -> List[Tuple[Document, float]]: vectors = self._embedding.embed_documents([query]) return self.similarity_search_by_vector_with_relevance_scores(vectors[0], k) [docs] def similarity_search_by_vector_with_relevance_scores( self, query: List[float], k: int ) -> List[Tuple[Document, float]]: match_documents_params = dict(query_embedding=query, match_count=k) res = self._client.rpc(self.query_name, match_documents_params).execute() match_result = [ ( Document( metadata=search.get("metadata", {}), # type: ignore page_content=search.get("content", ""), ), search.get("similarity", 0.0), ) for search in res.data if search.get("content") ] return match_result [docs] def similarity_search_by_vector_returning_embeddings( self, query: List[float], k: int ) -> List[Tuple[Document, float, np.ndarray[np.float32, Any]]]: match_documents_params = dict(query_embedding=query, match_count=k) res = self._client.rpc(self.query_name, match_documents_params).execute() match_result = [ ( Document( metadata=search.get("metadata", {}), # type: ignore page_content=search.get("content", ""), ), search.get("similarity", 0.0), # Supabase returns a vector type as its string represation (!). # This is a hack to convert the string to numpy array. np.fromstring( search.get("embedding", "").strip("[]"), np.float32, sep="," ), ) for search in res.data if search.get("content") ] return match_result @staticmethod def _texts_to_documents( texts: Iterable[str], metadatas: Optional[Iterable[dict[Any, Any]]] = None, ) -> List[Document]: """Return list of Documents from list of texts and metadatas.""" if metadatas is None: metadatas = repeat({}) docs = [ Document(page_content=text, metadata=metadata) for text, metadata in zip(texts, metadatas) ] return docs @staticmethod def _add_vectors( client: supabase.client.Client, table_name: str, vectors: List[List[float]], documents: List[Document], ids: List[str], ) -> List[str]: """Add vectors to Supabase table.""" rows: List[dict[str, Any]] = [ { "id": ids[idx], "content": documents[idx].page_content, "embedding": embedding, "metadata": documents[idx].metadata, # type: ignore } for idx, embedding in enumerate(vectors) ] # According to the SupabaseVectorStore JS implementation, the best chunk size # is 500 chunk_size = 500 id_list: List[str] = [] for i in range(0, len(rows), chunk_size): chunk = rows[i : i + chunk_size] result = client.from_(table_name).upsert(chunk).execute() # type: ignore if len(result.data) == 0: raise Exception("Error inserting: No rows added") # VectorStore.add_vectors returns ids as strings ids = [str(i.get("id")) for i in result.data if i.get("id")] id_list.extend(ids) return id_list [docs] def max_marginal_relevance_search_by_vector( self, embedding: List[float], k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: embedding: Embedding to look up documents similar to. k: Number of Documents to return. Defaults to 4. fetch_k: Number of Documents to fetch to pass to MMR algorithm. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. Returns: List of Documents selected by maximal marginal relevance. """ result = self.similarity_search_by_vector_returning_embeddings( embedding, fetch_k ) matched_documents = [doc_tuple[0] for doc_tuple in result] matched_embeddings = [doc_tuple[2] for doc_tuple in result] mmr_selected = maximal_marginal_relevance( np.array([embedding], dtype=np.float32), matched_embeddings, k=k, lambda_mult=lambda_mult, ) filtered_documents = [matched_documents[i] for i in mmr_selected] return filtered_documents [docs] def max_marginal_relevance_search( self, query: str, k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: query: Text to look up documents similar to. k: Number of Documents to return. Defaults to 4. fetch_k: Number of Documents to fetch to pass to MMR algorithm. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. Returns: List of Documents selected by maximal marginal relevance. `max_marginal_relevance_search` requires that `query_name` returns matched embeddings alongside the match documents. The following function demonstrates how to do this: ```sql CREATE FUNCTION match_documents_embeddings(query_embedding vector(1536), match_count int) RETURNS TABLE( id bigint, content text, metadata jsonb, embedding vector(1536), similarity float) LANGUAGE plpgsql AS $$ # variable_conflict use_column BEGIN RETURN query SELECT id, content, metadata, embedding, 1 -(docstore.embedding <=> query_embedding) AS similarity FROM docstore ORDER BY docstore.embedding <=> query_embedding LIMIT match_count; END; $$; ``` """ embedding = self._embedding.embed_documents([query]) docs = self.max_marginal_relevance_search_by_vector( embedding[0], k, fetch_k, lambda_mult=lambda_mult ) return docs [docs] def delete(self, ids: List[str]) -> None: """Delete by vector IDs. Args: ids: List of ids to delete. """ rows: List[dict[str, Any]] = [ { "id": id, } for id in ids ] # TODO: Check if this can be done in bulk for row in rows: self._client.from_(self.table_name).delete().eq("id", row["id"]).execute()
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/supabase.html
4751880f-f062-4611-965a-3ca4d41952a3
Source code for langchain.vectorstores.docarray.in_memory """Wrapper around in-memory storage.""" from __future__ import annotations from typing import Any, Dict, List, Literal, Optional from langchain.embeddings.base import Embeddings from langchain.vectorstores.docarray.base import ( DocArrayIndex, _check_docarray_import, ) [docs]class DocArrayInMemorySearch(DocArrayIndex): """Wrapper around in-memory storage for exact search. To use it, you should have the ``docarray`` package with version >=0.32.0 installed. You can install it with `pip install "langchain[docarray]"`. """ [docs] @classmethod def from_params( cls, embedding: Embeddings, metric: Literal[ "cosine_sim", "euclidian_dist", "sgeuclidean_dist" ] = "cosine_sim", **kwargs: Any, ) -> DocArrayInMemorySearch: """Initialize DocArrayInMemorySearch store. Args: embedding (Embeddings): Embedding function. metric (str): metric for exact nearest-neighbor search. Can be one of: "cosine_sim", "euclidean_dist" and "sqeuclidean_dist". Defaults to "cosine_sim". **kwargs: Other keyword arguments to be passed to the get_doc_cls method. """ _check_docarray_import() from docarray.index import InMemoryExactNNIndex doc_cls = cls._get_doc_cls(space=metric, **kwargs) doc_index = InMemoryExactNNIndex[doc_cls]() # type: ignore return cls(doc_index, embedding) [docs] @classmethod def from_texts( cls, texts: List[str], embedding: Embeddings, metadatas: Optional[List[Dict[Any, Any]]] = None, **kwargs: Any, ) -> DocArrayInMemorySearch: """Create an DocArrayInMemorySearch store and insert data. Args: texts (List[str]): Text data. embedding (Embeddings): Embedding function. metadatas (Optional[List[Dict[Any, Any]]]): Metadata for each text if it exists. Defaults to None. metric (str): metric for exact nearest-neighbor search. Can be one of: "cosine_sim", "euclidean_dist" and "sqeuclidean_dist". Defaults to "cosine_sim". Returns: DocArrayInMemorySearch Vector Store """ store = cls.from_params(embedding, **kwargs) store.add_texts(texts=texts, metadatas=metadatas) return store
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/docarray/in_memory.html
5d4858bb-0ed9-4df8-8f00-e2fc6a6085ac
Source code for langchain.vectorstores.docarray.hnsw """Wrapper around Hnswlib store.""" from __future__ import annotations from typing import Any, List, Literal, Optional from langchain.embeddings.base import Embeddings from langchain.vectorstores.docarray.base import ( DocArrayIndex, _check_docarray_import, ) [docs]class DocArrayHnswSearch(DocArrayIndex): """Wrapper around HnswLib storage. To use it, you should have the ``docarray`` package with version >=0.32.0 installed. You can install it with `pip install "langchain[docarray]"`. """ [docs] @classmethod def from_params( cls, embedding: Embeddings, work_dir: str, n_dim: int, dist_metric: Literal["cosine", "ip", "l2"] = "cosine", max_elements: int = 1024, index: bool = True, ef_construction: int = 200, ef: int = 10, M: int = 16, allow_replace_deleted: bool = True, num_threads: int = 1, **kwargs: Any, ) -> DocArrayHnswSearch: """Initialize DocArrayHnswSearch store. Args: embedding (Embeddings): Embedding function. work_dir (str): path to the location where all the data will be stored. n_dim (int): dimension of an embedding. dist_metric (str): Distance metric for DocArrayHnswSearch can be one of: "cosine", "ip", and "l2". Defaults to "cosine". max_elements (int): Maximum number of vectors that can be stored. Defaults to 1024. index (bool): Whether an index should be built for this field. Defaults to True. ef_construction (int): defines a construction time/accuracy trade-off. Defaults to 200. ef (int): parameter controlling query time/accuracy trade-off. Defaults to 10. M (int): parameter that defines the maximum number of outgoing connections in the graph. Defaults to 16. allow_replace_deleted (bool): Enables replacing of deleted elements with new added ones. Defaults to True. num_threads (int): Sets the number of cpu threads to use. Defaults to 1. **kwargs: Other keyword arguments to be passed to the get_doc_cls method. """ _check_docarray_import() from docarray.index import HnswDocumentIndex doc_cls = cls._get_doc_cls( dim=n_dim, space=dist_metric, max_elements=max_elements, index=index, ef_construction=ef_construction, ef=ef, M=M, allow_replace_deleted=allow_replace_deleted, num_threads=num_threads, **kwargs, ) doc_index = HnswDocumentIndex[doc_cls](work_dir=work_dir) # type: ignore return cls(doc_index, embedding) [docs] @classmethod def from_texts( cls, texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, work_dir: Optional[str] = None, n_dim: Optional[int] = None, **kwargs: Any, ) -> DocArrayHnswSearch: """Create an DocArrayHnswSearch store and insert data. Args: texts (List[str]): Text data. embedding (Embeddings): Embedding function. metadatas (Optional[List[dict]]): Metadata for each text if it exists. Defaults to None. work_dir (str): path to the location where all the data will be stored. n_dim (int): dimension of an embedding. **kwargs: Other keyword arguments to be passed to the __init__ method. Returns: DocArrayHnswSearch Vector Store """ if work_dir is None: raise ValueError("`work_dir` parameter has not been set.") if n_dim is None: raise ValueError("`n_dim` parameter has not been set.") store = cls.from_params(embedding, work_dir, n_dim, **kwargs) store.add_texts(texts=texts, metadatas=metadatas) return store
https://api.python.langchain.com/en/latest/_modules/langchain/vectorstores/docarray/hnsw.html
662495a0-0c33-4b77-8265-773409a68b15
Source code for langchain.utilities.powerbi """Wrapper around a Power BI endpoint.""" from __future__ import annotations import asyncio import logging import os from typing import TYPE_CHECKING, Any, Dict, Iterable, List, Optional, Union import aiohttp import requests from aiohttp import ServerTimeoutError from pydantic import BaseModel, Field, root_validator, validator from requests.exceptions import Timeout _LOGGER = logging.getLogger(__name__) BASE_URL = os.getenv("POWERBI_BASE_URL", "https://api.powerbi.com/v1.0/myorg") if TYPE_CHECKING: from azure.core.credentials import TokenCredential [docs]class PowerBIDataset(BaseModel): """Create PowerBI engine from dataset ID and credential or token. Use either the credential or a supplied token to authenticate. If both are supplied the credential is used to generate a token. The impersonated_user_name is the UPN of a user to be impersonated. If the model is not RLS enabled, this will be ignored. """ dataset_id: str table_names: List[str] group_id: Optional[str] = None credential: Optional[TokenCredential] = None token: Optional[str] = None impersonated_user_name: Optional[str] = None sample_rows_in_table_info: int = Field(default=1, gt=0, le=10) schemas: Dict[str, str] = Field(default_factory=dict) aiosession: Optional[aiohttp.ClientSession] = None class Config: """Configuration for this pydantic object.""" arbitrary_types_allowed = True @validator("table_names", allow_reuse=True) def fix_table_names(cls, table_names: List[str]) -> List[str]: """Fix the table names.""" return [fix_table_name(table) for table in table_names] @root_validator(pre=True, allow_reuse=True) def token_or_credential_present(cls, values: Dict[str, Any]) -> Dict[str, Any]: """Validate that at least one of token and credentials is present.""" if "token" in values or "credential" in values: return values raise ValueError("Please provide either a credential or a token.") @property def request_url(self) -> str: """Get the request url.""" if self.group_id: return f"{BASE_URL}/groups/{self.group_id}/datasets/{self.dataset_id}/executeQueries" # noqa: E501 # pylint: disable=C0301 return f"{BASE_URL}/datasets/{self.dataset_id}/executeQueries" # noqa: E501 # pylint: disable=C0301 @property def headers(self) -> Dict[str, str]: """Get the token.""" if self.token: return { "Content-Type": "application/json", "Authorization": "Bearer " + self.token, } from azure.core.exceptions import ( ClientAuthenticationError, # pylint: disable=import-outside-toplevel ) if self.credential: try: token = self.credential.get_token( "https://analysis.windows.net/powerbi/api/.default" ).token return { "Content-Type": "application/json", "Authorization": "Bearer " + token, } except Exception as exc: # pylint: disable=broad-exception-caught raise ClientAuthenticationError( "Could not get a token from the supplied credentials." ) from exc raise ClientAuthenticationError("No credential or token supplied.") [docs] def get_table_names(self) -> Iterable[str]: """Get names of tables available.""" return self.table_names [docs] def get_schemas(self) -> str: """Get the available schema's.""" if self.schemas: return ", ".join([f"{key}: {value}" for key, value in self.schemas.items()]) return "No known schema's yet. Use the schema_powerbi tool first." @property def table_info(self) -> str: """Information about all tables in the database.""" return self.get_table_info() def _get_tables_to_query( self, table_names: Optional[Union[List[str], str]] = None ) -> Optional[List[str]]: """Get the tables names that need to be queried, after checking they exist.""" if table_names is not None: if ( isinstance(table_names, list) and len(table_names) > 0 and table_names[0] != "" ): fixed_tables = [fix_table_name(table) for table in table_names] non_existing_tables = [ table for table in fixed_tables if table not in self.table_names ] if non_existing_tables: _LOGGER.warning( "Table(s) %s not found in dataset.", ", ".join(non_existing_tables), ) tables = [ table for table in fixed_tables if table not in non_existing_tables ] return tables if tables else None if isinstance(table_names, str) and table_names != "": if table_names not in self.table_names: _LOGGER.warning("Table %s not found in dataset.", table_names) return None return [fix_table_name(table_names)] return self.table_names def _get_tables_todo(self, tables_todo: List[str]) -> List[str]: """Get the tables that still need to be queried.""" return [table for table in tables_todo if table not in self.schemas] def _get_schema_for_tables(self, table_names: List[str]) -> str: """Create a string of the table schemas for the supplied tables.""" schemas = [ schema for table, schema in self.schemas.items() if table in table_names ] return ", ".join(schemas) [docs] def get_table_info( self, table_names: Optional[Union[List[str], str]] = None ) -> str: """Get information about specified tables.""" tables_requested = self._get_tables_to_query(table_names) if tables_requested is None: return "No (valid) tables requested." tables_todo = self._get_tables_todo(tables_requested) for table in tables_todo: self._get_schema(table) return self._get_schema_for_tables(tables_requested) [docs] async def aget_table_info( self, table_names: Optional[Union[List[str], str]] = None ) -> str: """Get information about specified tables.""" tables_requested = self._get_tables_to_query(table_names) if tables_requested is None: return "No (valid) tables requested." tables_todo = self._get_tables_todo(tables_requested) await asyncio.gather(*[self._aget_schema(table) for table in tables_todo]) return self._get_schema_for_tables(tables_requested) def _get_schema(self, table: str) -> None: """Get the schema for a table.""" try: result = self.run( f"EVALUATE TOPN({self.sample_rows_in_table_info}, {table})" ) self.schemas[table] = json_to_md(result["results"][0]["tables"][0]["rows"]) except Timeout: _LOGGER.warning("Timeout while getting table info for %s", table) self.schemas[table] = "unknown" except Exception as exc: # pylint: disable=broad-exception-caught _LOGGER.warning("Error while getting table info for %s: %s", table, exc) self.schemas[table] = "unknown" async def _aget_schema(self, table: str) -> None: """Get the schema for a table.""" try: result = await self.arun( f"EVALUATE TOPN({self.sample_rows_in_table_info}, {table})" ) self.schemas[table] = json_to_md(result["results"][0]["tables"][0]["rows"]) except ServerTimeoutError: _LOGGER.warning("Timeout while getting table info for %s", table) self.schemas[table] = "unknown" except Exception as exc: # pylint: disable=broad-exception-caught _LOGGER.warning("Error while getting table info for %s: %s", table, exc) self.schemas[table] = "unknown" def _create_json_content(self, command: str) -> dict[str, Any]: """Create the json content for the request.""" return { "queries": [{"query": rf"{command}"}], "impersonatedUserName": self.impersonated_user_name, "serializerSettings": {"includeNulls": True}, } [docs] def run(self, command: str) -> Any: """Execute a DAX command and return a json representing the results.""" _LOGGER.debug("Running command: %s", command) result = requests.post( self.request_url, json=self._create_json_content(command), headers=self.headers, timeout=10, ) return result.json() [docs] async def arun(self, command: str) -> Any: """Execute a DAX command and return the result asynchronously.""" _LOGGER.debug("Running command: %s", command) if self.aiosession: async with self.aiosession.post( self.request_url, headers=self.headers, json=self._create_json_content(command), timeout=10, ) as response: response_json = await response.json(content_type=response.content_type) return response_json async with aiohttp.ClientSession() as session: async with session.post( self.request_url, headers=self.headers, json=self._create_json_content(command), timeout=10, ) as response: response_json = await response.json(content_type=response.content_type) return response_json def json_to_md( json_contents: List[Dict[str, Union[str, int, float]]], table_name: Optional[str] = None, ) -> str: """Converts a JSON object to a markdown table.""" output_md = "" headers = json_contents[0].keys() for header in headers: header.replace("[", ".").replace("]", "") if table_name: header.replace(f"{table_name}.", "") output_md += f"| {header} " output_md += "|\n" for row in json_contents: for value in row.values(): output_md += f"| {value} " output_md += "|\n" return output_md def fix_table_name(table: str) -> str: """Add single quotes around table names that contain spaces.""" if " " in table and not table.startswith("'") and not table.endswith("'"): return f"'{table}'" return table
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/powerbi.html
c1806ebd-1bb9-4b13-9431-274a5f229635
Source code for langchain.utilities.bing_search """Util that calls Bing Search. In order to set this up, follow instructions at: https://levelup.gitconnected.com/api-tutorial-how-to-use-bing-web-search-api-in-python-4165d5592a7e """ from typing import Dict, List import requests from pydantic import BaseModel, Extra, root_validator from langchain.utils import get_from_dict_or_env [docs]class BingSearchAPIWrapper(BaseModel): """Wrapper for Bing Search API. In order to set this up, follow instructions at: https://levelup.gitconnected.com/api-tutorial-how-to-use-bing-web-search-api-in-python-4165d5592a7e """ bing_subscription_key: str bing_search_url: str k: int = 10 class Config: """Configuration for this pydantic object.""" extra = Extra.forbid def _bing_search_results(self, search_term: str, count: int) -> List[dict]: headers = {"Ocp-Apim-Subscription-Key": self.bing_subscription_key} params = { "q": search_term, "count": count, "textDecorations": True, "textFormat": "HTML", } response = requests.get( self.bing_search_url, headers=headers, params=params # type: ignore ) response.raise_for_status() search_results = response.json() return search_results["webPages"]["value"] @root_validator(pre=True) def validate_environment(cls, values: Dict) -> Dict: """Validate that api key and endpoint exists in environment.""" bing_subscription_key = get_from_dict_or_env( values, "bing_subscription_key", "BING_SUBSCRIPTION_KEY" ) values["bing_subscription_key"] = bing_subscription_key bing_search_url = get_from_dict_or_env( values, "bing_search_url", "BING_SEARCH_URL", # default="https://api.bing.microsoft.com/v7.0/search", ) values["bing_search_url"] = bing_search_url return values [docs] def run(self, query: str) -> str: """Run query through BingSearch and parse result.""" snippets = [] results = self._bing_search_results(query, count=self.k) if len(results) == 0: return "No good Bing Search Result was found" for result in results: snippets.append(result["snippet"]) return " ".join(snippets) [docs] def results(self, query: str, num_results: int) -> List[Dict]: """Run query through BingSearch and return metadata. Args: query: The query to search for. num_results: The number of results to return. Returns: A list of dictionaries with the following keys: snippet - The description of the result. title - The title of the result. link - The link to the result. """ metadata_results = [] results = self._bing_search_results(query, count=num_results) if len(results) == 0: return [{"Result": "No good Bing Search Result was found"}] for result in results: metadata_result = { "snippet": result["snippet"], "title": result["name"], "link": result["url"], } metadata_results.append(metadata_result) return metadata_results
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/bing_search.html
473737ff-a320-4c4c-911d-5c3006abf4ae
Source code for langchain.utilities.serpapi """Chain that calls SerpAPI. Heavily borrowed from https://github.com/ofirpress/self-ask """ import os import sys from typing import Any, Dict, Optional, Tuple import aiohttp from pydantic import BaseModel, Extra, Field, root_validator from langchain.utils import get_from_dict_or_env class HiddenPrints: """Context manager to hide prints.""" def __enter__(self) -> None: """Open file to pipe stdout to.""" self._original_stdout = sys.stdout sys.stdout = open(os.devnull, "w") def __exit__(self, *_: Any) -> None: """Close file that stdout was piped to.""" sys.stdout.close() sys.stdout = self._original_stdout [docs]class SerpAPIWrapper(BaseModel): """Wrapper around SerpAPI. To use, you should have the ``google-search-results`` python package installed, and the environment variable ``SERPAPI_API_KEY`` set with your API key, or pass `serpapi_api_key` as a named parameter to the constructor. Example: .. code-block:: python from langchain import SerpAPIWrapper serpapi = SerpAPIWrapper() """ search_engine: Any #: :meta private: params: dict = Field( default={ "engine": "google", "google_domain": "google.com", "gl": "us", "hl": "en", } ) serpapi_api_key: Optional[str] = None aiosession: Optional[aiohttp.ClientSession] = None class Config: """Configuration for this pydantic object.""" extra = Extra.forbid arbitrary_types_allowed = True @root_validator() def validate_environment(cls, values: Dict) -> Dict: """Validate that api key and python package exists in environment.""" serpapi_api_key = get_from_dict_or_env( values, "serpapi_api_key", "SERPAPI_API_KEY" ) values["serpapi_api_key"] = serpapi_api_key try: from serpapi import GoogleSearch values["search_engine"] = GoogleSearch except ImportError: raise ValueError( "Could not import serpapi python package. " "Please install it with `pip install google-search-results`." ) return values [docs] async def arun(self, query: str, **kwargs: Any) -> str: """Run query through SerpAPI and parse result async.""" return self._process_response(await self.aresults(query)) [docs] def run(self, query: str, **kwargs: Any) -> str: """Run query through SerpAPI and parse result.""" return self._process_response(self.results(query)) [docs] def results(self, query: str) -> dict: """Run query through SerpAPI and return the raw result.""" params = self.get_params(query) with HiddenPrints(): search = self.search_engine(params) res = search.get_dict() return res [docs] async def aresults(self, query: str) -> dict: """Use aiohttp to run query through SerpAPI and return the results async.""" def construct_url_and_params() -> Tuple[str, Dict[str, str]]: params = self.get_params(query) params["source"] = "python" if self.serpapi_api_key: params["serp_api_key"] = self.serpapi_api_key params["output"] = "json" url = "https://serpapi.com/search" return url, params url, params = construct_url_and_params() if not self.aiosession: async with aiohttp.ClientSession() as session: async with session.get(url, params=params) as response: res = await response.json() else: async with self.aiosession.get(url, params=params) as response: res = await response.json() return res [docs] def get_params(self, query: str) -> Dict[str, str]: """Get parameters for SerpAPI.""" _params = { "api_key": self.serpapi_api_key, "q": query, } params = {**self.params, **_params} return params @staticmethod def _process_response(res: dict) -> str: """Process response from SerpAPI.""" if "error" in res.keys(): raise ValueError(f"Got error from SerpAPI: {res['error']}") if "answer_box" in res.keys() and type(res["answer_box"]) == list: res["answer_box"] = res["answer_box"][0] if "answer_box" in res.keys() and "answer" in res["answer_box"].keys(): toret = res["answer_box"]["answer"] elif "answer_box" in res.keys() and "snippet" in res["answer_box"].keys(): toret = res["answer_box"]["snippet"] elif ( "answer_box" in res.keys() and "snippet_highlighted_words" in res["answer_box"].keys() ): toret = res["answer_box"]["snippet_highlighted_words"][0] elif ( "sports_results" in res.keys() and "game_spotlight" in res["sports_results"].keys() ): toret = res["sports_results"]["game_spotlight"] elif ( "shopping_results" in res.keys() and "title" in res["shopping_results"][0].keys() ): toret = res["shopping_results"][:3] elif ( "knowledge_graph" in res.keys() and "description" in res["knowledge_graph"].keys() ): toret = res["knowledge_graph"]["description"] elif "snippet" in res["organic_results"][0].keys(): toret = res["organic_results"][0]["snippet"] elif "link" in res["organic_results"][0].keys(): toret = res["organic_results"][0]["link"] else: toret = "No good search result found" return toret
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/serpapi.html
3bc9c5af-9b0b-4e0b-96ef-6845cad761b5
Source code for langchain.utilities.awslambda """Util that calls Lambda.""" import json from typing import Any, Dict, Optional from pydantic import BaseModel, Extra, root_validator [docs]class LambdaWrapper(BaseModel): """Wrapper for AWS Lambda SDK. Docs for using: 1. pip install boto3 2. Create a lambda function using the AWS Console or CLI 3. Run `aws configure` and enter your AWS credentials """ lambda_client: Any #: :meta private: function_name: Optional[str] = None awslambda_tool_name: Optional[str] = None awslambda_tool_description: Optional[str] = None class Config: """Configuration for this pydantic object.""" extra = Extra.forbid @root_validator() def validate_environment(cls, values: Dict) -> Dict: """Validate that python package exists in environment.""" try: import boto3 except ImportError: raise ImportError( "boto3 is not installed. Please install it with `pip install boto3`" ) values["lambda_client"] = boto3.client("lambda") values["function_name"] = values["function_name"] return values [docs] def run(self, query: str) -> str: """Invoke Lambda function and parse result.""" res = self.lambda_client.invoke( FunctionName=self.function_name, InvocationType="RequestResponse", Payload=json.dumps({"body": query}), ) try: payload_stream = res["Payload"] payload_string = payload_stream.read().decode("utf-8") answer = json.loads(payload_string)["body"] except StopIteration: return "Failed to parse response from Lambda" if answer is None or answer == "": # We don't want to return the assumption alone if answer is empty return "Request failed." else: return f"Result: {answer}"
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/awslambda.html
c99077d1-6383-4987-b1ed-9ea508a704bb
Source code for langchain.utilities.bash """Wrapper around subprocess to run commands.""" from __future__ import annotations import platform import re import subprocess from typing import TYPE_CHECKING, List, Union from uuid import uuid4 if TYPE_CHECKING: import pexpect def _lazy_import_pexpect() -> pexpect: """Import pexpect only when needed.""" if platform.system() == "Windows": raise ValueError("Persistent bash processes are not yet supported on Windows.") try: import pexpect except ImportError: raise ImportError( "pexpect required for persistent bash processes." " To install, run `pip install pexpect`." ) return pexpect [docs]class BashProcess: """Executes bash commands and returns the output.""" def __init__( self, strip_newlines: bool = False, return_err_output: bool = False, persistent: bool = False, ): """Initialize with stripping newlines.""" self.strip_newlines = strip_newlines self.return_err_output = return_err_output self.prompt = "" self.process = None if persistent: self.prompt = str(uuid4()) self.process = self._initialize_persistent_process(self.prompt) @staticmethod def _initialize_persistent_process(prompt: str) -> pexpect.spawn: # Start bash in a clean environment # Doesn't work on windows pexpect = _lazy_import_pexpect() process = pexpect.spawn( "env", ["-i", "bash", "--norc", "--noprofile"], encoding="utf-8" ) # Set the custom prompt process.sendline("PS1=" + prompt) process.expect_exact(prompt, timeout=10) return process [docs] def run(self, commands: Union[str, List[str]]) -> str: """Run commands and return final output.""" if isinstance(commands, str): commands = [commands] commands = ";".join(commands) if self.process is not None: return self._run_persistent( commands, ) else: return self._run(commands) def _run(self, command: str) -> str: """Run commands and return final output.""" try: output = subprocess.run( command, shell=True, check=True, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, ).stdout.decode() except subprocess.CalledProcessError as error: if self.return_err_output: return error.stdout.decode() return str(error) if self.strip_newlines: output = output.strip() return output [docs] def process_output(self, output: str, command: str) -> str: # Remove the command from the output using a regular expression pattern = re.escape(command) + r"\s*\n" output = re.sub(pattern, "", output, count=1) return output.strip() def _run_persistent(self, command: str) -> str: """Run commands and return final output.""" pexpect = _lazy_import_pexpect() if self.process is None: raise ValueError("Process not initialized") self.process.sendline(command) # Clear the output with an empty string self.process.expect(self.prompt, timeout=10) self.process.sendline("") try: self.process.expect([self.prompt, pexpect.EOF], timeout=10) except pexpect.TIMEOUT: return f"Timeout error while executing command {command}" if self.process.after == pexpect.EOF: return f"Exited with error status: {self.process.exitstatus}" output = self.process.before output = self.process_output(output, command) if self.strip_newlines: return output.strip() return output
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/bash.html
0ff2eaf8-e2f2-4547-8de6-3ab87920b44f
Source code for langchain.utilities.google_search """Util that calls Google Search.""" from typing import Any, Dict, List, Optional from pydantic import BaseModel, Extra, root_validator from langchain.utils import get_from_dict_or_env [docs]class GoogleSearchAPIWrapper(BaseModel): """Wrapper for Google Search API. Adapted from: Instructions adapted from https://stackoverflow.com/questions/ 37083058/ programmatically-searching-google-in-python-using-custom-search TODO: DOCS for using it 1. Install google-api-python-client - If you don't already have a Google account, sign up. - If you have never created a Google APIs Console project, read the Managing Projects page and create a project in the Google API Console. - Install the library using pip install google-api-python-client The current version of the library is 2.70.0 at this time 2. To create an API key: - Navigate to the APIs & Servicesβ†’Credentials panel in Cloud Console. - Select Create credentials, then select API key from the drop-down menu. - The API key created dialog box displays your newly created key. - You now have an API_KEY 3. Setup Custom Search Engine so you can search the entire web - Create a custom search engine in this link. - In Sites to search, add any valid URL (i.e. www.stackoverflow.com). - That’s all you have to fill up, the rest doesn’t matter. In the left-side menu, click Edit search engine β†’ {your search engine name} β†’ Setup Set Search the entire web to ON. Remove the URL you added from the list of Sites to search. - Under Search engine ID you’ll find the search-engine-ID. 4. Enable the Custom Search API - Navigate to the APIs & Servicesβ†’Dashboard panel in Cloud Console. - Click Enable APIs and Services. - Search for Custom Search API and click on it. - Click Enable. URL for it: https://console.cloud.google.com/apis/library/customsearch.googleapis .com """ search_engine: Any #: :meta private: google_api_key: Optional[str] = None google_cse_id: Optional[str] = None k: int = 10 siterestrict: bool = False class Config: """Configuration for this pydantic object.""" extra = Extra.forbid def _google_search_results(self, search_term: str, **kwargs: Any) -> List[dict]: cse = self.search_engine.cse() if self.siterestrict: cse = cse.siterestrict() res = cse.list(q=search_term, cx=self.google_cse_id, **kwargs).execute() return res.get("items", []) @root_validator() def validate_environment(cls, values: Dict) -> Dict: """Validate that api key and python package exists in environment.""" google_api_key = get_from_dict_or_env( values, "google_api_key", "GOOGLE_API_KEY" ) values["google_api_key"] = google_api_key google_cse_id = get_from_dict_or_env(values, "google_cse_id", "GOOGLE_CSE_ID") values["google_cse_id"] = google_cse_id try: from googleapiclient.discovery import build except ImportError: raise ImportError( "google-api-python-client is not installed. " "Please install it with `pip install google-api-python-client`" ) service = build("customsearch", "v1", developerKey=google_api_key) values["search_engine"] = service return values [docs] def run(self, query: str) -> str: """Run query through GoogleSearch and parse result.""" snippets = [] results = self._google_search_results(query, num=self.k) if len(results) == 0: return "No good Google Search Result was found" for result in results: if "snippet" in result: snippets.append(result["snippet"]) return " ".join(snippets) [docs] def results(self, query: str, num_results: int) -> List[Dict]: """Run query through GoogleSearch and return metadata. Args: query: The query to search for. num_results: The number of results to return. Returns: A list of dictionaries with the following keys: snippet - The description of the result. title - The title of the result. link - The link to the result. """ metadata_results = [] results = self._google_search_results(query, num=num_results) if len(results) == 0: return [{"Result": "No good Google Search Result was found"}] for result in results: metadata_result = { "title": result["title"], "link": result["link"], } if "snippet" in result: metadata_result["snippet"] = result["snippet"] metadata_results.append(metadata_result) return metadata_results
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/google_search.html
b22c87b1-aa9b-4db4-85d5-dd4a54ed4e04
Source code for langchain.utilities.max_compute from __future__ import annotations from typing import TYPE_CHECKING, Iterator, List, Optional from langchain.utils import get_from_env if TYPE_CHECKING: from odps import ODPS [docs]class MaxComputeAPIWrapper: """Interface for querying Alibaba Cloud MaxCompute tables.""" def __init__(self, client: ODPS): """Initialize MaxCompute document loader. Args: client: odps.ODPS MaxCompute client object. """ self.client = client [docs] @classmethod def from_params( cls, endpoint: str, project: str, *, access_id: Optional[str] = None, secret_access_key: Optional[str] = None, ) -> MaxComputeAPIWrapper: """Convenience constructor that builds the odsp.ODPS MaxCompute client from given parameters. Args: endpoint: MaxCompute endpoint. project: A project is a basic organizational unit of MaxCompute, which is similar to a database. access_id: MaxCompute access ID. Should be passed in directly or set as the environment variable `MAX_COMPUTE_ACCESS_ID`. secret_access_key: MaxCompute secret access key. Should be passed in directly or set as the environment variable `MAX_COMPUTE_SECRET_ACCESS_KEY`. """ try: from odps import ODPS except ImportError as ex: raise ImportError( "Could not import pyodps python package. " "Please install it with `pip install pyodps` or refer to " "https://pyodps.readthedocs.io/." ) from ex access_id = access_id or get_from_env("access_id", "MAX_COMPUTE_ACCESS_ID") secret_access_key = secret_access_key or get_from_env( "secret_access_key", "MAX_COMPUTE_SECRET_ACCESS_KEY" ) client = ODPS( access_id=access_id, secret_access_key=secret_access_key, project=project, endpoint=endpoint, ) if not client.exist_project(project): raise ValueError(f'The project "{project}" does not exist.') return cls(client) [docs] def lazy_query(self, query: str) -> Iterator[dict]: # Execute SQL query. with self.client.execute_sql(query).open_reader() as reader: if reader.count == 0: raise ValueError("Table contains no data.") for record in reader: yield {k: v for k, v in record} [docs] def query(self, query: str) -> List[dict]: return list(self.lazy_query(query))
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/max_compute.html
1770fe1a-05eb-4786-800b-46e436fe2b30
Source code for langchain.utilities.arxiv """Util that calls Arxiv.""" import logging import os from typing import Any, Dict, List, Optional from pydantic import BaseModel, Extra, root_validator from langchain.schema import Document logger = logging.getLogger(__name__) [docs]class ArxivAPIWrapper(BaseModel): """Wrapper around ArxivAPI. To use, you should have the ``arxiv`` python package installed. https://lukasschwab.me/arxiv.py/index.html This wrapper will use the Arxiv API to conduct searches and fetch document summaries. By default, it will return the document summaries of the top-k results. It limits the Document content by doc_content_chars_max. Set doc_content_chars_max=None if you don't want to limit the content size. Parameters: top_k_results: number of the top-scored document used for the arxiv tool ARXIV_MAX_QUERY_LENGTH: the cut limit on the query used for the arxiv tool. load_max_docs: a limit to the number of loaded documents load_all_available_meta: if True: the `metadata` of the loaded Documents gets all available meta info (see https://lukasschwab.me/arxiv.py/index.html#Result), if False: the `metadata` gets only the most informative fields. """ arxiv_search: Any #: :meta private: arxiv_exceptions: Any # :meta private: top_k_results: int = 3 ARXIV_MAX_QUERY_LENGTH = 300 load_max_docs: int = 100 load_all_available_meta: bool = False doc_content_chars_max: Optional[int] = 4000 class Config: """Configuration for this pydantic object.""" extra = Extra.forbid @root_validator() def validate_environment(cls, values: Dict) -> Dict: """Validate that the python package exists in environment.""" try: import arxiv values["arxiv_search"] = arxiv.Search values["arxiv_exceptions"] = ( arxiv.ArxivError, arxiv.UnexpectedEmptyPageError, arxiv.HTTPError, ) values["arxiv_result"] = arxiv.Result except ImportError: raise ImportError( "Could not import arxiv python package. " "Please install it with `pip install arxiv`." ) return values [docs] def run(self, query: str) -> str: """ Run Arxiv search and get the article meta information. See https://lukasschwab.me/arxiv.py/index.html#Search See https://lukasschwab.me/arxiv.py/index.html#Result It uses only the most informative fields of article meta information. """ try: results = self.arxiv_search( # type: ignore query[: self.ARXIV_MAX_QUERY_LENGTH], max_results=self.top_k_results ).results() except self.arxiv_exceptions as ex: return f"Arxiv exception: {ex}" docs = [ f"Published: {result.updated.date()}\nTitle: {result.title}\n" f"Authors: {', '.join(a.name for a in result.authors)}\n" f"Summary: {result.summary}" for result in results ] if docs: return "\n\n".join(docs)[: self.doc_content_chars_max] else: return "No good Arxiv Result was found" [docs] def load(self, query: str) -> List[Document]: """ Run Arxiv search and get the article texts plus the article meta information. See https://lukasschwab.me/arxiv.py/index.html#Search Returns: a list of documents with the document.page_content in text format """ try: import fitz except ImportError: raise ImportError( "PyMuPDF package not found, please install it with " "`pip install pymupdf`" ) try: results = self.arxiv_search( # type: ignore query[: self.ARXIV_MAX_QUERY_LENGTH], max_results=self.load_max_docs ).results() except self.arxiv_exceptions as ex: logger.debug("Error on arxiv: %s", ex) return [] docs: List[Document] = [] for result in results: try: doc_file_name: str = result.download_pdf() with fitz.open(doc_file_name) as doc_file: text: str = "".join(page.get_text() for page in doc_file) except FileNotFoundError as f_ex: logger.debug(f_ex) continue if self.load_all_available_meta: extra_metadata = { "entry_id": result.entry_id, "published_first_time": str(result.published.date()), "comment": result.comment, "journal_ref": result.journal_ref, "doi": result.doi, "primary_category": result.primary_category, "categories": result.categories, "links": [link.href for link in result.links], } else: extra_metadata = {} metadata = { "Published": str(result.updated.date()), "Title": result.title, "Authors": ", ".join(a.name for a in result.authors), "Summary": result.summary, **extra_metadata, } doc = Document( page_content=text[: self.doc_content_chars_max], metadata=metadata ) docs.append(doc) os.remove(doc_file_name) return docs
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/arxiv.html
56b5fa74-4ad1-483e-b944-0d3d4021e5f1
Source code for langchain.utilities.python import sys from io import StringIO from typing import Dict, Optional from pydantic import BaseModel, Field [docs]class PythonREPL(BaseModel): """Simulates a standalone Python REPL.""" globals: Optional[Dict] = Field(default_factory=dict, alias="_globals") locals: Optional[Dict] = Field(default_factory=dict, alias="_locals") [docs] def run(self, command: str) -> str: """Run command with own globals/locals and returns anything printed.""" old_stdout = sys.stdout sys.stdout = mystdout = StringIO() try: exec(command, self.globals, self.locals) sys.stdout = old_stdout output = mystdout.getvalue() except Exception as e: sys.stdout = old_stdout output = repr(e) return output
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/python.html
47a04a73-5e12-42bf-bafb-afdfd2fcff8a
Source code for langchain.utilities.openweathermap """Util that calls OpenWeatherMap using PyOWM.""" from typing import Any, Dict, Optional from pydantic import Extra, root_validator from langchain.tools.base import BaseModel from langchain.utils import get_from_dict_or_env [docs]class OpenWeatherMapAPIWrapper(BaseModel): """Wrapper for OpenWeatherMap API using PyOWM. Docs for using: 1. Go to OpenWeatherMap and sign up for an API key 2. Save your API KEY into OPENWEATHERMAP_API_KEY env variable 3. pip install pyowm """ owm: Any openweathermap_api_key: Optional[str] = None class Config: """Configuration for this pydantic object.""" extra = Extra.forbid @root_validator(pre=True) def validate_environment(cls, values: Dict) -> Dict: """Validate that api key exists in environment.""" openweathermap_api_key = get_from_dict_or_env( values, "openweathermap_api_key", "OPENWEATHERMAP_API_KEY" ) try: import pyowm except ImportError: raise ImportError( "pyowm is not installed. Please install it with `pip install pyowm`" ) owm = pyowm.OWM(openweathermap_api_key) values["owm"] = owm return values def _format_weather_info(self, location: str, w: Any) -> str: detailed_status = w.detailed_status wind = w.wind() humidity = w.humidity temperature = w.temperature("celsius") rain = w.rain heat_index = w.heat_index clouds = w.clouds return ( f"In {location}, the current weather is as follows:\n" f"Detailed status: {detailed_status}\n" f"Wind speed: {wind['speed']} m/s, direction: {wind['deg']}Β°\n" f"Humidity: {humidity}%\n" f"Temperature: \n" f" - Current: {temperature['temp']}Β°C\n" f" - High: {temperature['temp_max']}Β°C\n" f" - Low: {temperature['temp_min']}Β°C\n" f" - Feels like: {temperature['feels_like']}Β°C\n" f"Rain: {rain}\n" f"Heat index: {heat_index}\n" f"Cloud cover: {clouds}%" ) [docs] def run(self, location: str) -> str: """Get the current weather information for a specified location.""" mgr = self.owm.weather_manager() observation = mgr.weather_at_place(location) w = observation.weather return self._format_weather_info(location, w)
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/openweathermap.html
7c933dd6-6979-4e97-b0dd-d28277e46393
Source code for langchain.utilities.metaphor_search """Util that calls Metaphor Search API. In order to set this up, follow instructions at: """ import json from typing import Dict, List, Optional import aiohttp import requests from pydantic import BaseModel, Extra, root_validator from langchain.utils import get_from_dict_or_env METAPHOR_API_URL = "https://api.metaphor.systems" [docs]class MetaphorSearchAPIWrapper(BaseModel): """Wrapper for Metaphor Search API.""" metaphor_api_key: str k: int = 10 class Config: """Configuration for this pydantic object.""" extra = Extra.forbid def _metaphor_search_results( self, query: str, num_results: int, include_domains: Optional[List[str]] = None, exclude_domains: Optional[List[str]] = None, start_crawl_date: Optional[str] = None, end_crawl_date: Optional[str] = None, start_published_date: Optional[str] = None, end_published_date: Optional[str] = None, ) -> List[dict]: headers = {"X-Api-Key": self.metaphor_api_key} params = { "numResults": num_results, "query": query, "includeDomains": include_domains, "excludeDomains": exclude_domains, "startCrawlDate": start_crawl_date, "endCrawlDate": end_crawl_date, "startPublishedDate": start_published_date, "endPublishedDate": end_published_date, } response = requests.post( # type: ignore f"{METAPHOR_API_URL}/search", headers=headers, json=params, ) response.raise_for_status() search_results = response.json() print(search_results) return search_results["results"] @root_validator(pre=True) def validate_environment(cls, values: Dict) -> Dict: """Validate that api key and endpoint exists in environment.""" metaphor_api_key = get_from_dict_or_env( values, "metaphor_api_key", "METAPHOR_API_KEY" ) values["metaphor_api_key"] = metaphor_api_key return values [docs] def results( self, query: str, num_results: int, include_domains: Optional[List[str]] = None, exclude_domains: Optional[List[str]] = None, start_crawl_date: Optional[str] = None, end_crawl_date: Optional[str] = None, start_published_date: Optional[str] = None, end_published_date: Optional[str] = None, ) -> List[Dict]: """Run query through Metaphor Search and return metadata. Args: query: The query to search for. num_results: The number of results to return. Returns: A list of dictionaries with the following keys: title - The title of the url - The url author - Author of the content, if applicable. Otherwise, None. published_date - Estimated date published in YYYY-MM-DD format. Otherwise, None. """ raw_search_results = self._metaphor_search_results( query, num_results=num_results, include_domains=include_domains, exclude_domains=exclude_domains, start_crawl_date=start_crawl_date, end_crawl_date=end_crawl_date, start_published_date=start_published_date, end_published_date=end_published_date, ) return self._clean_results(raw_search_results) [docs] async def results_async( self, query: str, num_results: int, include_domains: Optional[List[str]] = None, exclude_domains: Optional[List[str]] = None, start_crawl_date: Optional[str] = None, end_crawl_date: Optional[str] = None, start_published_date: Optional[str] = None, end_published_date: Optional[str] = None, ) -> List[Dict]: """Get results from the Metaphor Search API asynchronously.""" # Function to perform the API call async def fetch() -> str: headers = {"X-Api-Key": self.metaphor_api_key} params = { "numResults": num_results, "query": query, "includeDomains": include_domains, "excludeDomains": exclude_domains, "startCrawlDate": start_crawl_date, "endCrawlDate": end_crawl_date, "startPublishedDate": start_published_date, "endPublishedDate": end_published_date, } async with aiohttp.ClientSession() as session: async with session.post( f"{METAPHOR_API_URL}/search", json=params, headers=headers ) as res: if res.status == 200: data = await res.text() return data else: raise Exception(f"Error {res.status}: {res.reason}") results_json_str = await fetch() results_json = json.loads(results_json_str) return self._clean_results(results_json["results"]) def _clean_results(self, raw_search_results: List[Dict]) -> List[Dict]: cleaned_results = [] for result in raw_search_results: cleaned_results.append( { "title": result["title"], "url": result["url"], "author": result["author"], "published_date": result["publishedDate"], } ) return cleaned_results
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/metaphor_search.html
cfacfae8-e824-429e-9346-631f7197f444
Source code for langchain.utilities.bibtex """Util that calls bibtexparser.""" import logging from typing import Any, Dict, List, Mapping from pydantic import BaseModel, Extra, root_validator logger = logging.getLogger(__name__) OPTIONAL_FIELDS = [ "annotate", "booktitle", "editor", "howpublished", "journal", "keywords", "note", "organization", "publisher", "school", "series", "type", "doi", "issn", "isbn", ] [docs]class BibtexparserWrapper(BaseModel): """Wrapper around bibtexparser. To use, you should have the ``bibtexparser`` python package installed. https://bibtexparser.readthedocs.io/en/master/ This wrapper will use bibtexparser to load a collection of references from a bibtex file and fetch document summaries. """ class Config: """Configuration for this pydantic object.""" extra = Extra.forbid @root_validator() def validate_environment(cls, values: Dict) -> Dict: """Validate that the python package exists in environment.""" try: import bibtexparser # noqa except ImportError: raise ImportError( "Could not import bibtexparser python package. " "Please install it with `pip install bibtexparser`." ) return values [docs] def load_bibtex_entries(self, path: str) -> List[Dict[str, Any]]: """Load bibtex entries from the bibtex file at the given path.""" import bibtexparser with open(path) as file: entries = bibtexparser.load(file).entries return entries [docs] def get_metadata( self, entry: Mapping[str, Any], load_extra: bool = False ) -> Dict[str, Any]: """Get metadata for the given entry.""" publication = entry.get("journal") or entry.get("booktitle") if "url" in entry: url = entry["url"] elif "doi" in entry: url = f'https://doi.org/{entry["doi"]}' else: url = None meta = { "id": entry.get("ID"), "published_year": entry.get("year"), "title": entry.get("title"), "publication": publication, "authors": entry.get("author"), "abstract": entry.get("abstract"), "url": url, } if load_extra: for field in OPTIONAL_FIELDS: meta[field] = entry.get(field) return {k: v for k, v in meta.items() if v is not None}
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/bibtex.html
34bb4b5b-0bc0-4726-a0f1-3481669e9b4b
Source code for langchain.utilities.searx_search """Utility for using SearxNG meta search API. SearxNG is a privacy-friendly free metasearch engine that aggregates results from `multiple search engines <https://docs.searxng.org/admin/engines/configured_engines.html>`_ and databases and supports the `OpenSearch <https://github.com/dewitt/opensearch/blob/master/opensearch-1-1-draft-6.md>`_ specification. More details on the installation instructions `here. <../../integrations/searx.html>`_ For the search API refer to https://docs.searxng.org/dev/search_api.html Quick Start ----------- In order to use this utility you need to provide the searx host. This can be done by passing the named parameter :attr:`searx_host <SearxSearchWrapper.searx_host>` or exporting the environment variable SEARX_HOST. Note: this is the only required parameter. Then create a searx search instance like this: .. code-block:: python from langchain.utilities import SearxSearchWrapper # when the host starts with `http` SSL is disabled and the connection # is assumed to be on a private network searx_host='http://self.hosted' search = SearxSearchWrapper(searx_host=searx_host) You can now use the ``search`` instance to query the searx API. Searching --------- Use the :meth:`run() <SearxSearchWrapper.run>` and :meth:`results() <SearxSearchWrapper.results>` methods to query the searx API. Other methods are available for convenience. :class:`SearxResults` is a convenience wrapper around the raw json result. Example usage of the ``run`` method to make a search: .. code-block:: python s.run(query="what is the best search engine?") Engine Parameters ----------------- You can pass any `accepted searx search API <https://docs.searxng.org/dev/search_api.html>`_ parameters to the :py:class:`SearxSearchWrapper` instance. In the following example we are using the :attr:`engines <SearxSearchWrapper.engines>` and the ``language`` parameters: .. code-block:: python # assuming the searx host is set as above or exported as an env variable s = SearxSearchWrapper(engines=['google', 'bing'], language='es') Search Tips ----------- Searx offers a special `search syntax <https://docs.searxng.org/user/index.html#search-syntax>`_ that can also be used instead of passing engine parameters. For example the following query: .. code-block:: python s = SearxSearchWrapper("langchain library", engines=['github']) # can also be written as: s = SearxSearchWrapper("langchain library !github") # or even: s = SearxSearchWrapper("langchain library !gh") In some situations you might want to pass an extra string to the search query. For example when the `run()` method is called by an agent. The search suffix can also be used as a way to pass extra parameters to searx or the underlying search engines. .. code-block:: python # select the github engine and pass the search suffix s = SearchWrapper("langchain library", query_suffix="!gh") s = SearchWrapper("langchain library") # select github the conventional google search syntax s.run("large language models", query_suffix="site:github.com") *NOTE*: A search suffix can be defined on both the instance and the method level. The resulting query will be the concatenation of the two with the former taking precedence. See `SearxNG Configured Engines <https://docs.searxng.org/admin/engines/configured_engines.html>`_ and `SearxNG Search Syntax <https://docs.searxng.org/user/index.html#id1>`_ for more details. Notes ----- This wrapper is based on the SearxNG fork https://github.com/searxng/searxng which is better maintained than the original Searx project and offers more features. Public searxNG instances often use a rate limiter for API usage, so you might want to use a self hosted instance and disable the rate limiter. If you are self-hosting an instance you can customize the rate limiter for your own network as described `here <https://docs.searxng.org/src/searx.botdetection.html#limiter-src>`_. For a list of public SearxNG instances see https://searx.space/ """ import json from typing import Any, Dict, List, Optional import aiohttp import requests from pydantic import BaseModel, Extra, Field, PrivateAttr, root_validator, validator from langchain.utils import get_from_dict_or_env def _get_default_params() -> dict: return {"language": "en", "format": "json"} class SearxResults(dict): """Dict like wrapper around search api results.""" _data = "" def __init__(self, data: str): """Take a raw result from Searx and make it into a dict like object.""" json_data = json.loads(data) super().__init__(json_data) self.__dict__ = self def __str__(self) -> str: """Text representation of searx result.""" return self._data @property def results(self) -> Any: """Silence mypy for accessing this field. :meta private: """ return self.get("results") @property def answers(self) -> Any: """Helper accessor on the json result.""" return self.get("answers") [docs]class SearxSearchWrapper(BaseModel): """Wrapper for Searx API. To use you need to provide the searx host by passing the named parameter ``searx_host`` or exporting the environment variable ``SEARX_HOST``. In some situations you might want to disable SSL verification, for example if you are running searx locally. You can do this by passing the named parameter ``unsecure``. You can also pass the host url scheme as ``http`` to disable SSL. Example: .. code-block:: python from langchain.utilities import SearxSearchWrapper searx = SearxSearchWrapper(searx_host="http://localhost:8888") Example with SSL disabled: .. code-block:: python from langchain.utilities import SearxSearchWrapper # note the unsecure parameter is not needed if you pass the url scheme as # http searx = SearxSearchWrapper(searx_host="http://localhost:8888", unsecure=True) """ _result: SearxResults = PrivateAttr() searx_host: str = "" unsecure: bool = False params: dict = Field(default_factory=_get_default_params) headers: Optional[dict] = None engines: Optional[List[str]] = [] categories: Optional[List[str]] = [] query_suffix: Optional[str] = "" k: int = 10 aiosession: Optional[Any] = None @validator("unsecure") def disable_ssl_warnings(cls, v: bool) -> bool: """Disable SSL warnings.""" if v: # requests.urllib3.disable_warnings() try: import urllib3 urllib3.disable_warnings() except ImportError as e: print(e) return v @root_validator() def validate_params(cls, values: Dict) -> Dict: """Validate that custom searx params are merged with default ones.""" user_params = values["params"] default = _get_default_params() values["params"] = {**default, **user_params} engines = values.get("engines") if engines: values["params"]["engines"] = ",".join(engines) categories = values.get("categories") if categories: values["params"]["categories"] = ",".join(categories) searx_host = get_from_dict_or_env(values, "searx_host", "SEARX_HOST") if not searx_host.startswith("http"): print( f"Warning: missing the url scheme on host \ ! assuming secure https://{searx_host} " ) searx_host = "https://" + searx_host elif searx_host.startswith("http://"): values["unsecure"] = True cls.disable_ssl_warnings(True) values["searx_host"] = searx_host return values class Config: """Configuration for this pydantic object.""" extra = Extra.forbid def _searx_api_query(self, params: dict) -> SearxResults: """Actual request to searx API.""" raw_result = requests.get( self.searx_host, headers=self.headers, params=params, verify=not self.unsecure, ) # test if http result is ok if not raw_result.ok: raise ValueError("Searx API returned an error: ", raw_result.text) res = SearxResults(raw_result.text) self._result = res return res async def _asearx_api_query(self, params: dict) -> SearxResults: if not self.aiosession: async with aiohttp.ClientSession() as session: async with session.get( self.searx_host, headers=self.headers, params=params, ssl=(lambda: False if self.unsecure else None)(), ) as response: if not response.ok: raise ValueError("Searx API returned an error: ", response.text) result = SearxResults(await response.text()) self._result = result else: async with self.aiosession.get( self.searx_host, headers=self.headers, params=params, verify=not self.unsecure, ) as response: if not response.ok: raise ValueError("Searx API returned an error: ", response.text) result = SearxResults(await response.text()) self._result = result return result [docs] def run( self, query: str, engines: Optional[List[str]] = None, categories: Optional[List[str]] = None, query_suffix: Optional[str] = "", **kwargs: Any, ) -> str: """Run query through Searx API and parse results. You can pass any other params to the searx query API. Args: query: The query to search for. query_suffix: Extra suffix appended to the query. engines: List of engines to use for the query. categories: List of categories to use for the query. **kwargs: extra parameters to pass to the searx API. Returns: str: The result of the query. Raises: ValueError: If an error occured with the query. Example: This will make a query to the qwant engine: .. code-block:: python from langchain.utilities import SearxSearchWrapper searx = SearxSearchWrapper(searx_host="http://my.searx.host") searx.run("what is the weather in France ?", engine="qwant") # the same result can be achieved using the `!` syntax of searx # to select the engine using `query_suffix` searx.run("what is the weather in France ?", query_suffix="!qwant") """ _params = { "q": query, } params = {**self.params, **_params, **kwargs} if self.query_suffix and len(self.query_suffix) > 0: params["q"] += " " + self.query_suffix if isinstance(query_suffix, str) and len(query_suffix) > 0: params["q"] += " " + query_suffix if isinstance(engines, list) and len(engines) > 0: params["engines"] = ",".join(engines) if isinstance(categories, list) and len(categories) > 0: params["categories"] = ",".join(categories) res = self._searx_api_query(params) if len(res.answers) > 0: toret = res.answers[0] # only return the content of the results list elif len(res.results) > 0: toret = "\n\n".join([r.get("content", "") for r in res.results[: self.k]]) else: toret = "No good search result found" return toret [docs] async def arun( self, query: str, engines: Optional[List[str]] = None, query_suffix: Optional[str] = "", **kwargs: Any, ) -> str: """Asynchronously version of `run`.""" _params = { "q": query, } params = {**self.params, **_params, **kwargs} if self.query_suffix and len(self.query_suffix) > 0: params["q"] += " " + self.query_suffix if isinstance(query_suffix, str) and len(query_suffix) > 0: params["q"] += " " + query_suffix if isinstance(engines, list) and len(engines) > 0: params["engines"] = ",".join(engines) res = await self._asearx_api_query(params) if len(res.answers) > 0: toret = res.answers[0] # only return the content of the results list elif len(res.results) > 0: toret = "\n\n".join([r.get("content", "") for r in res.results[: self.k]]) else: toret = "No good search result found" return toret [docs] def results( self, query: str, num_results: int, engines: Optional[List[str]] = None, categories: Optional[List[str]] = None, query_suffix: Optional[str] = "", **kwargs: Any, ) -> List[Dict]: """Run query through Searx API and returns the results with metadata. Args: query: The query to search for. query_suffix: Extra suffix appended to the query. num_results: Limit the number of results to return. engines: List of engines to use for the query. categories: List of categories to use for the query. **kwargs: extra parameters to pass to the searx API. Returns: Dict with the following keys: { snippet: The description of the result. title: The title of the result. link: The link to the result. engines: The engines used for the result. category: Searx category of the result. } """ _params = { "q": query, } params = {**self.params, **_params, **kwargs} if self.query_suffix and len(self.query_suffix) > 0: params["q"] += " " + self.query_suffix if isinstance(query_suffix, str) and len(query_suffix) > 0: params["q"] += " " + query_suffix if isinstance(engines, list) and len(engines) > 0: params["engines"] = ",".join(engines) if isinstance(categories, list) and len(categories) > 0: params["categories"] = ",".join(categories) results = self._searx_api_query(params).results[:num_results] if len(results) == 0: return [{"Result": "No good Search Result was found"}] return [ { "snippet": result.get("content", ""), "title": result["title"], "link": result["url"], "engines": result["engines"], "category": result["category"], } for result in results ] [docs] async def aresults( self, query: str, num_results: int, engines: Optional[List[str]] = None, query_suffix: Optional[str] = "", **kwargs: Any, ) -> List[Dict]: """Asynchronously query with json results. Uses aiohttp. See `results` for more info. """ _params = { "q": query, } params = {**self.params, **_params, **kwargs} if self.query_suffix and len(self.query_suffix) > 0: params["q"] += " " + self.query_suffix if isinstance(query_suffix, str) and len(query_suffix) > 0: params["q"] += " " + query_suffix if isinstance(engines, list) and len(engines) > 0: params["engines"] = ",".join(engines) results = (await self._asearx_api_query(params)).results[:num_results] if len(results) == 0: return [{"Result": "No good Search Result was found"}] return [ { "snippet": result.get("content", ""), "title": result["title"], "link": result["url"], "engines": result["engines"], "category": result["category"], } for result in results ]
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/searx_search.html
96522c26-15f1-470e-90eb-d5f487cd7ac3
Source code for langchain.utilities.graphql import json from typing import Any, Callable, Dict, Optional from pydantic import BaseModel, Extra, root_validator [docs]class GraphQLAPIWrapper(BaseModel): """Wrapper around GraphQL API. To use, you should have the ``gql`` python package installed. This wrapper will use the GraphQL API to conduct queries. """ custom_headers: Optional[Dict[str, str]] = None graphql_endpoint: str gql_client: Any #: :meta private: gql_function: Callable[[str], Any] #: :meta private: class Config: """Configuration for this pydantic object.""" extra = Extra.forbid @root_validator(pre=True) def validate_environment(cls, values: Dict) -> Dict: """Validate that the python package exists in the environment.""" try: from gql import Client, gql from gql.transport.requests import RequestsHTTPTransport except ImportError as e: raise ImportError( "Could not import gql python package. " f"Try installing it with `pip install gql`. Received error: {e}" ) headers = values.get("custom_headers") transport = RequestsHTTPTransport( url=values["graphql_endpoint"], headers=headers, ) client = Client(transport=transport, fetch_schema_from_transport=True) values["gql_client"] = client values["gql_function"] = gql return values [docs] def run(self, query: str) -> str: """Run a GraphQL query and get the results.""" result = self._execute_query(query) return json.dumps(result, indent=2) def _execute_query(self, query: str) -> Dict[str, Any]: """Execute a GraphQL query and return the results.""" document_node = self.gql_function(query) result = self.gql_client.execute(document_node) return result
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/graphql.html
ab07615c-3462-4621-8f1e-74c08e9cfb71
Source code for langchain.utilities.brave_search import json import requests from pydantic import BaseModel, Field [docs]class BraveSearchWrapper(BaseModel): api_key: str search_kwargs: dict = Field(default_factory=dict) [docs] def run(self, query: str) -> str: headers = { "X-Subscription-Token": self.api_key, "Accept": "application/json", } base_url = "https://api.search.brave.com/res/v1/web/search" req = requests.PreparedRequest() params = {**self.search_kwargs, **{"q": query}} req.prepare_url(base_url, params) if req.url is None: raise ValueError("prepared url is None, this should not happen") response = requests.get(req.url, headers=headers) if not response.ok: raise Exception(f"HTTP error {response.status_code}") parsed_response = response.json() web_search_results = parsed_response.get("web", {}).get("results", []) final_results = [] if isinstance(web_search_results, list): for item in web_search_results: final_results.append( { "title": item.get("title"), "link": item.get("url"), "snippet": item.get("description"), } ) return json.dumps(final_results)
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/brave_search.html
efe7d237-1512-4a63-8dc9-a9bda9f0ce20
Source code for langchain.utilities.jira """Util that calls Jira.""" from typing import Any, Dict, List, Optional from pydantic import BaseModel, Extra, root_validator from langchain.tools.jira.prompt import ( JIRA_CATCH_ALL_PROMPT, JIRA_CONFLUENCE_PAGE_CREATE_PROMPT, JIRA_GET_ALL_PROJECTS_PROMPT, JIRA_ISSUE_CREATE_PROMPT, JIRA_JQL_PROMPT, ) from langchain.utils import get_from_dict_or_env # TODO: think about error handling, more specific api specs, and jql/project limits [docs]class JiraAPIWrapper(BaseModel): """Wrapper for Jira API.""" jira: Any #: :meta private: confluence: Any jira_username: Optional[str] = None jira_api_token: Optional[str] = None jira_instance_url: Optional[str] = None operations: List[Dict] = [ { "mode": "jql", "name": "JQL Query", "description": JIRA_JQL_PROMPT, }, { "mode": "get_projects", "name": "Get Projects", "description": JIRA_GET_ALL_PROJECTS_PROMPT, }, { "mode": "create_issue", "name": "Create Issue", "description": JIRA_ISSUE_CREATE_PROMPT, }, { "mode": "other", "name": "Catch all Jira API call", "description": JIRA_CATCH_ALL_PROMPT, }, { "mode": "create_page", "name": "Create confluence page", "description": JIRA_CONFLUENCE_PAGE_CREATE_PROMPT, }, ] class Config: """Configuration for this pydantic object.""" extra = Extra.forbid [docs] def list(self) -> List[Dict]: return self.operations @root_validator() def validate_environment(cls, values: Dict) -> Dict: """Validate that api key and python package exists in environment.""" jira_username = get_from_dict_or_env(values, "jira_username", "JIRA_USERNAME") values["jira_username"] = jira_username jira_api_token = get_from_dict_or_env( values, "jira_api_token", "JIRA_API_TOKEN" ) values["jira_api_token"] = jira_api_token jira_instance_url = get_from_dict_or_env( values, "jira_instance_url", "JIRA_INSTANCE_URL" ) values["jira_instance_url"] = jira_instance_url try: from atlassian import Confluence, Jira except ImportError: raise ImportError( "atlassian-python-api is not installed. " "Please install it with `pip install atlassian-python-api`" ) jira = Jira( url=jira_instance_url, username=jira_username, password=jira_api_token, cloud=True, ) confluence = Confluence( url=jira_instance_url, username=jira_username, password=jira_api_token, cloud=True, ) values["jira"] = jira values["confluence"] = confluence return values [docs] def parse_issues(self, issues: Dict) -> List[dict]: parsed = [] for issue in issues["issues"]: key = issue["key"] summary = issue["fields"]["summary"] created = issue["fields"]["created"][0:10] priority = issue["fields"]["priority"]["name"] status = issue["fields"]["status"]["name"] try: assignee = issue["fields"]["assignee"]["displayName"] except Exception: assignee = "None" rel_issues = {} for related_issue in issue["fields"]["issuelinks"]: if "inwardIssue" in related_issue.keys(): rel_type = related_issue["type"]["inward"] rel_key = related_issue["inwardIssue"]["key"] rel_summary = related_issue["inwardIssue"]["fields"]["summary"] if "outwardIssue" in related_issue.keys(): rel_type = related_issue["type"]["outward"] rel_key = related_issue["outwardIssue"]["key"] rel_summary = related_issue["outwardIssue"]["fields"]["summary"] rel_issues = {"type": rel_type, "key": rel_key, "summary": rel_summary} parsed.append( { "key": key, "summary": summary, "created": created, "assignee": assignee, "priority": priority, "status": status, "related_issues": rel_issues, } ) return parsed [docs] def parse_projects(self, projects: List[dict]) -> List[dict]: parsed = [] for project in projects: id = project["id"] key = project["key"] name = project["name"] type = project["projectTypeKey"] style = project["style"] parsed.append( {"id": id, "key": key, "name": name, "type": type, "style": style} ) return parsed [docs] def search(self, query: str) -> str: issues = self.jira.jql(query) parsed_issues = self.parse_issues(issues) parsed_issues_str = ( "Found " + str(len(parsed_issues)) + " issues:\n" + str(parsed_issues) ) return parsed_issues_str [docs] def project(self) -> str: projects = self.jira.projects() parsed_projects = self.parse_projects(projects) parsed_projects_str = ( "Found " + str(len(parsed_projects)) + " projects:\n" + str(parsed_projects) ) return parsed_projects_str [docs] def issue_create(self, query: str) -> str: try: import json except ImportError: raise ImportError( "json is not installed. Please install it with `pip install json`" ) params = json.loads(query) return self.jira.issue_create(fields=dict(params)) [docs] def page_create(self, query: str) -> str: try: import json except ImportError: raise ImportError( "json is not installed. Please install it with `pip install json`" ) params = json.loads(query) return self.confluence.create_page(**dict(params)) [docs] def other(self, query: str) -> str: context = {"self": self} exec(f"result = {query}", context) result = context["result"] return str(result) [docs] def run(self, mode: str, query: str) -> str: if mode == "jql": return self.search(query) elif mode == "get_projects": return self.project() elif mode == "create_issue": return self.issue_create(query) elif mode == "other": return self.other(query) elif mode == "create_page": return self.page_create(query) else: raise ValueError(f"Got unexpected mode {mode}")
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/jira.html
be098da0-8768-4654-9db0-dbab83cea814
Source code for langchain.utilities.spark_sql from __future__ import annotations from typing import TYPE_CHECKING, Any, Iterable, List, Optional if TYPE_CHECKING: from pyspark.sql import DataFrame, Row, SparkSession [docs]class SparkSQL: def __init__( self, spark_session: Optional[SparkSession] = None, catalog: Optional[str] = None, schema: Optional[str] = None, ignore_tables: Optional[List[str]] = None, include_tables: Optional[List[str]] = None, sample_rows_in_table_info: int = 3, ): try: from pyspark.sql import SparkSession except ImportError: raise ValueError( "pyspark is not installed. Please install it with `pip install pyspark`" ) self._spark = ( spark_session if spark_session else SparkSession.builder.getOrCreate() ) if catalog is not None: self._spark.catalog.setCurrentCatalog(catalog) if schema is not None: self._spark.catalog.setCurrentDatabase(schema) self._all_tables = set(self._get_all_table_names()) self._include_tables = set(include_tables) if include_tables else set() if self._include_tables: missing_tables = self._include_tables - self._all_tables if missing_tables: raise ValueError( f"include_tables {missing_tables} not found in database" ) self._ignore_tables = set(ignore_tables) if ignore_tables else set() if self._ignore_tables: missing_tables = self._ignore_tables - self._all_tables if missing_tables: raise ValueError( f"ignore_tables {missing_tables} not found in database" ) usable_tables = self.get_usable_table_names() self._usable_tables = set(usable_tables) if usable_tables else self._all_tables if not isinstance(sample_rows_in_table_info, int): raise TypeError("sample_rows_in_table_info must be an integer") self._sample_rows_in_table_info = sample_rows_in_table_info [docs] @classmethod def from_uri( cls, database_uri: str, engine_args: Optional[dict] = None, **kwargs: Any ) -> SparkSQL: """Creating a remote Spark Session via Spark connect. For example: SparkSQL.from_uri("sc://localhost:15002") """ try: from pyspark.sql import SparkSession except ImportError: raise ValueError( "pyspark is not installed. Please install it with `pip install pyspark`" ) spark = SparkSession.builder.remote(database_uri).getOrCreate() return cls(spark, **kwargs) [docs] def get_usable_table_names(self) -> Iterable[str]: """Get names of tables available.""" if self._include_tables: return self._include_tables # sorting the result can help LLM understanding it. return sorted(self._all_tables - self._ignore_tables) def _get_all_table_names(self) -> Iterable[str]: rows = self._spark.sql("SHOW TABLES").select("tableName").collect() return list(map(lambda row: row.tableName, rows)) def _get_create_table_stmt(self, table: str) -> str: statement = ( self._spark.sql(f"SHOW CREATE TABLE {table}").collect()[0].createtab_stmt ) # Ignore the data source provider and options to reduce the number of tokens. using_clause_index = statement.find("USING") return statement[:using_clause_index] + ";" [docs] def get_table_info(self, table_names: Optional[List[str]] = None) -> str: all_table_names = self.get_usable_table_names() if table_names is not None: missing_tables = set(table_names).difference(all_table_names) if missing_tables: raise ValueError(f"table_names {missing_tables} not found in database") all_table_names = table_names tables = [] for table_name in all_table_names: table_info = self._get_create_table_stmt(table_name) if self._sample_rows_in_table_info: table_info += "\n\n/*" table_info += f"\n{self._get_sample_spark_rows(table_name)}\n" table_info += "*/" tables.append(table_info) final_str = "\n\n".join(tables) return final_str def _get_sample_spark_rows(self, table: str) -> str: query = f"SELECT * FROM {table} LIMIT {self._sample_rows_in_table_info}" df = self._spark.sql(query) columns_str = "\t".join(list(map(lambda f: f.name, df.schema.fields))) try: sample_rows = self._get_dataframe_results(df) # save the sample rows in string format sample_rows_str = "\n".join(["\t".join(row) for row in sample_rows]) except Exception: sample_rows_str = "" return ( f"{self._sample_rows_in_table_info} rows from {table} table:\n" f"{columns_str}\n" f"{sample_rows_str}" ) def _convert_row_as_tuple(self, row: Row) -> tuple: return tuple(map(str, row.asDict().values())) def _get_dataframe_results(self, df: DataFrame) -> list: return list(map(self._convert_row_as_tuple, df.collect())) [docs] def run(self, command: str, fetch: str = "all") -> str: df = self._spark.sql(command) if fetch == "one": df = df.limit(1) return str(self._get_dataframe_results(df)) [docs] def get_table_info_no_throw(self, table_names: Optional[List[str]] = None) -> str: """Get information about specified tables. Follows best practices as specified in: Rajkumar et al, 2022 (https://arxiv.org/abs/2204.00498) If `sample_rows_in_table_info`, the specified number of sample rows will be appended to each table description. This can increase performance as demonstrated in the paper. """ try: return self.get_table_info(table_names) except ValueError as e: """Format the error message""" return f"Error: {e}" [docs] def run_no_throw(self, command: str, fetch: str = "all") -> str: """Execute a SQL command and return a string representing the results. If the statement returns rows, a string of the results is returned. If the statement returns no rows, an empty string is returned. If the statement throws an error, the error message is returned. """ try: from pyspark.errors import PySparkException except ImportError: raise ValueError( "pyspark is not installed. Please install it with `pip install pyspark`" ) try: return self.run(command, fetch) except PySparkException as e: """Format the error message""" return f"Error: {e}"
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/spark_sql.html
25cc11fc-9474-4b07-a527-62c8d56208ef
Source code for langchain.utilities.wikipedia """Util that calls Wikipedia.""" import logging from typing import Any, Dict, List, Optional from pydantic import BaseModel, Extra, root_validator from langchain.schema import Document logger = logging.getLogger(__name__) WIKIPEDIA_MAX_QUERY_LENGTH = 300 [docs]class WikipediaAPIWrapper(BaseModel): """Wrapper around WikipediaAPI. To use, you should have the ``wikipedia`` python package installed. This wrapper will use the Wikipedia API to conduct searches and fetch page summaries. By default, it will return the page summaries of the top-k results. It limits the Document content by doc_content_chars_max. """ wiki_client: Any #: :meta private: top_k_results: int = 3 lang: str = "en" load_all_available_meta: bool = False doc_content_chars_max: int = 4000 class Config: """Configuration for this pydantic object.""" extra = Extra.forbid @root_validator() def validate_environment(cls, values: Dict) -> Dict: """Validate that the python package exists in environment.""" try: import wikipedia wikipedia.set_lang(values["lang"]) values["wiki_client"] = wikipedia except ImportError: raise ImportError( "Could not import wikipedia python package. " "Please install it with `pip install wikipedia`." ) return values [docs] def run(self, query: str) -> str: """Run Wikipedia search and get page summaries.""" page_titles = self.wiki_client.search(query[:WIKIPEDIA_MAX_QUERY_LENGTH]) summaries = [] for page_title in page_titles[: self.top_k_results]: if wiki_page := self._fetch_page(page_title): if summary := self._formatted_page_summary(page_title, wiki_page): summaries.append(summary) if not summaries: return "No good Wikipedia Search Result was found" return "\n\n".join(summaries)[: self.doc_content_chars_max] @staticmethod def _formatted_page_summary(page_title: str, wiki_page: Any) -> Optional[str]: return f"Page: {page_title}\nSummary: {wiki_page.summary}" def _page_to_document(self, page_title: str, wiki_page: Any) -> Document: main_meta = { "title": page_title, "summary": wiki_page.summary, "source": wiki_page.url, } add_meta = ( { "categories": wiki_page.categories, "page_url": wiki_page.url, "image_urls": wiki_page.images, "related_titles": wiki_page.links, "parent_id": wiki_page.parent_id, "references": wiki_page.references, "revision_id": wiki_page.revision_id, "sections": wiki_page.sections, } if self.load_all_available_meta else {} ) doc = Document( page_content=wiki_page.content[: self.doc_content_chars_max], metadata={ **main_meta, **add_meta, }, ) return doc def _fetch_page(self, page: str) -> Optional[str]: try: return self.wiki_client.page(title=page, auto_suggest=False) except ( self.wiki_client.exceptions.PageError, self.wiki_client.exceptions.DisambiguationError, ): return None [docs] def load(self, query: str) -> List[Document]: """ Run Wikipedia search and get the article text plus the meta information. See Returns: a list of documents. """ page_titles = self.wiki_client.search(query[:WIKIPEDIA_MAX_QUERY_LENGTH]) docs = [] for page_title in page_titles[: self.top_k_results]: if wiki_page := self._fetch_page(page_title): if doc := self._page_to_document(page_title, wiki_page): docs.append(doc) return docs
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/wikipedia.html
868942d1-bfcf-4c0e-b344-7759ae0477c1
Source code for langchain.utilities.apify from typing import Any, Callable, Dict, Optional from pydantic import BaseModel, root_validator from langchain.document_loaders import ApifyDatasetLoader from langchain.document_loaders.base import Document from langchain.utils import get_from_dict_or_env [docs]class ApifyWrapper(BaseModel): """Wrapper around Apify. To use, you should have the ``apify-client`` python package installed, and the environment variable ``APIFY_API_TOKEN`` set with your API key, or pass `apify_api_token` as a named parameter to the constructor. """ apify_client: Any apify_client_async: Any @root_validator() def validate_environment(cls, values: Dict) -> Dict: """Validate environment. Validate that an Apify API token is set and the apify-client Python package exists in the current environment. """ apify_api_token = get_from_dict_or_env( values, "apify_api_token", "APIFY_API_TOKEN" ) try: from apify_client import ApifyClient, ApifyClientAsync values["apify_client"] = ApifyClient(apify_api_token) values["apify_client_async"] = ApifyClientAsync(apify_api_token) except ImportError: raise ValueError( "Could not import apify-client Python package. " "Please install it with `pip install apify-client`." ) return values [docs] def call_actor( self, actor_id: str, run_input: Dict, dataset_mapping_function: Callable[[Dict], Document], *, build: Optional[str] = None, memory_mbytes: Optional[int] = None, timeout_secs: Optional[int] = None, ) -> ApifyDatasetLoader: """Run an Actor on the Apify platform and wait for results to be ready. Args: actor_id (str): The ID or name of the Actor on the Apify platform. run_input (Dict): The input object of the Actor that you're trying to run. dataset_mapping_function (Callable): A function that takes a single dictionary (an Apify dataset item) and converts it to an instance of the Document class. build (str, optional): Optionally specifies the actor build to run. It can be either a build tag or build number. memory_mbytes (int, optional): Optional memory limit for the run, in megabytes. timeout_secs (int, optional): Optional timeout for the run, in seconds. Returns: ApifyDatasetLoader: A loader that will fetch the records from the Actor run's default dataset. """ actor_call = self.apify_client.actor(actor_id).call( run_input=run_input, build=build, memory_mbytes=memory_mbytes, timeout_secs=timeout_secs, ) return ApifyDatasetLoader( dataset_id=actor_call["defaultDatasetId"], dataset_mapping_function=dataset_mapping_function, ) [docs] async def acall_actor( self, actor_id: str, run_input: Dict, dataset_mapping_function: Callable[[Dict], Document], *, build: Optional[str] = None, memory_mbytes: Optional[int] = None, timeout_secs: Optional[int] = None, ) -> ApifyDatasetLoader: """Run an Actor on the Apify platform and wait for results to be ready. Args: actor_id (str): The ID or name of the Actor on the Apify platform. run_input (Dict): The input object of the Actor that you're trying to run. dataset_mapping_function (Callable): A function that takes a single dictionary (an Apify dataset item) and converts it to an instance of the Document class. build (str, optional): Optionally specifies the actor build to run. It can be either a build tag or build number. memory_mbytes (int, optional): Optional memory limit for the run, in megabytes. timeout_secs (int, optional): Optional timeout for the run, in seconds. Returns: ApifyDatasetLoader: A loader that will fetch the records from the Actor run's default dataset. """ actor_call = await self.apify_client_async.actor(actor_id).call( run_input=run_input, build=build, memory_mbytes=memory_mbytes, timeout_secs=timeout_secs, ) return ApifyDatasetLoader( dataset_id=actor_call["defaultDatasetId"], dataset_mapping_function=dataset_mapping_function, )
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/apify.html
60e598e2-aebf-4932-b73a-efbb639097c8
Source code for langchain.utilities.pupmed import json import logging import time import urllib.error import urllib.request from typing import List from pydantic import BaseModel, Extra from langchain.schema import Document logger = logging.getLogger(__name__) [docs]class PubMedAPIWrapper(BaseModel): """ Wrapper around PubMed API. This wrapper will use the PubMed API to conduct searches and fetch document summaries. By default, it will return the document summaries of the top-k results of an input search. Parameters: top_k_results: number of the top-scored document used for the PubMed tool load_max_docs: a limit to the number of loaded documents load_all_available_meta: if True: the `metadata` of the loaded Documents gets all available meta info (see https://www.ncbi.nlm.nih.gov/books/NBK25499/#chapter4.ESearch) if False: the `metadata` gets only the most informative fields. """ base_url_esearch = "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?" base_url_efetch = "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?" max_retry = 5 sleep_time = 0.2 # Default values for the parameters top_k_results: int = 3 load_max_docs: int = 25 ARXIV_MAX_QUERY_LENGTH = 300 doc_content_chars_max: int = 2000 load_all_available_meta: bool = False email: str = "your_email@example.com" class Config: """Configuration for this pydantic object.""" extra = Extra.forbid [docs] def run(self, query: str) -> str: """ Run PubMed search and get the article meta information. See https://www.ncbi.nlm.nih.gov/books/NBK25499/#chapter4.ESearch It uses only the most informative fields of article meta information. """ try: # Retrieve the top-k results for the query docs = [ f"Published: {result['pub_date']}\nTitle: {result['title']}\n" f"Summary: {result['summary']}" for result in self.load(query[: self.ARXIV_MAX_QUERY_LENGTH]) ] # Join the results and limit the character count return ( "\n\n".join(docs)[: self.doc_content_chars_max] if docs else "No good PubMed Result was found" ) except Exception as ex: return f"PubMed exception: {ex}" [docs] def load(self, query: str) -> List[dict]: """ Search PubMed for documents matching the query. Return a list of dictionaries containing the document metadata. """ url = ( self.base_url_esearch + "db=pubmed&term=" + str({urllib.parse.quote(query)}) + f"&retmode=json&retmax={self.top_k_results}&usehistory=y" ) result = urllib.request.urlopen(url) text = result.read().decode("utf-8") json_text = json.loads(text) articles = [] webenv = json_text["esearchresult"]["webenv"] for uid in json_text["esearchresult"]["idlist"]: article = self.retrieve_article(uid, webenv) articles.append(article) # Convert the list of articles to a JSON string return articles def _transform_doc(self, doc: dict) -> Document: summary = doc.pop("summary") return Document(page_content=summary, metadata=doc) [docs] def load_docs(self, query: str) -> List[Document]: document_dicts = self.load(query=query) return [self._transform_doc(d) for d in document_dicts] [docs] def retrieve_article(self, uid: str, webenv: str) -> dict: url = ( self.base_url_efetch + "db=pubmed&retmode=xml&id=" + uid + "&webenv=" + webenv ) retry = 0 while True: try: result = urllib.request.urlopen(url) break except urllib.error.HTTPError as e: if e.code == 429 and retry < self.max_retry: # Too Many Requests error # wait for an exponentially increasing amount of time print( f"Too Many Requests, " f"waiting for {self.sleep_time:.2f} seconds..." ) time.sleep(self.sleep_time) self.sleep_time *= 2 retry += 1 else: raise e xml_text = result.read().decode("utf-8") # Get title title = "" if "<ArticleTitle>" in xml_text and "</ArticleTitle>" in xml_text: start_tag = "<ArticleTitle>" end_tag = "</ArticleTitle>" title = xml_text[ xml_text.index(start_tag) + len(start_tag) : xml_text.index(end_tag) ] # Get abstract abstract = "" if "<AbstractText>" in xml_text and "</AbstractText>" in xml_text: start_tag = "<AbstractText>" end_tag = "</AbstractText>" abstract = xml_text[ xml_text.index(start_tag) + len(start_tag) : xml_text.index(end_tag) ] # Get publication date pub_date = "" if "<PubDate>" in xml_text and "</PubDate>" in xml_text: start_tag = "<PubDate>" end_tag = "</PubDate>" pub_date = xml_text[ xml_text.index(start_tag) + len(start_tag) : xml_text.index(end_tag) ] # Return article as dictionary article = { "uid": uid, "title": title, "summary": abstract, "pub_date": pub_date, } return article
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/pupmed.html
14d09fd8-2156-4876-ba8b-2abd23eb1055
Source code for langchain.utilities.duckduckgo_search """Util that calls DuckDuckGo Search. No setup required. Free. https://pypi.org/project/duckduckgo-search/ """ from typing import Dict, List, Optional from pydantic import BaseModel, Extra from pydantic.class_validators import root_validator [docs]class DuckDuckGoSearchAPIWrapper(BaseModel): """Wrapper for DuckDuckGo Search API. Free and does not require any setup """ k: int = 10 region: Optional[str] = "wt-wt" safesearch: str = "moderate" time: Optional[str] = "y" max_results: int = 5 class Config: """Configuration for this pydantic object.""" extra = Extra.forbid @root_validator() def validate_environment(cls, values: Dict) -> Dict: """Validate that python package exists in environment.""" try: from duckduckgo_search import DDGS # noqa: F401 except ImportError: raise ValueError( "Could not import duckduckgo-search python package. " "Please install it with `pip install duckduckgo-search`." ) return values [docs] def get_snippets(self, query: str) -> List[str]: """Run query through DuckDuckGo and return concatenated results.""" from duckduckgo_search import DDGS with DDGS() as ddgs: results = ddgs.text( query, region=self.region, safesearch=self.safesearch, timelimit=self.time, ) if results is None: return ["No good DuckDuckGo Search Result was found"] snippets = [] for i, res in enumerate(results, 1): if res is not None: snippets.append(res["body"]) if len(snippets) == self.max_results: break return snippets [docs] def run(self, query: str) -> str: snippets = self.get_snippets(query) return " ".join(snippets) [docs] def results(self, query: str, num_results: int) -> List[Dict[str, str]]: """Run query through DuckDuckGo and return metadata. Args: query: The query to search for. num_results: The number of results to return. Returns: A list of dictionaries with the following keys: snippet - The description of the result. title - The title of the result. link - The link to the result. """ from duckduckgo_search import DDGS with DDGS() as ddgs: results = ddgs.text( query, region=self.region, safesearch=self.safesearch, timelimit=self.time, ) if results is None: return [{"Result": "No good DuckDuckGo Search Result was found"}] def to_metadata(result: Dict) -> Dict[str, str]: return { "snippet": result["body"], "title": result["title"], "link": result["href"], } formatted_results = [] for i, res in enumerate(results, 1): if res is not None: formatted_results.append(to_metadata(res)) if len(formatted_results) == num_results: break return formatted_results
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/duckduckgo_search.html
cdd185f5-ae08-47a8-8319-2af42c889aec
Source code for langchain.utilities.openapi """Utility functions for parsing an OpenAPI spec.""" import copy import json import logging import re from enum import Enum from pathlib import Path from typing import Dict, List, Optional, Union import requests import yaml from openapi_schema_pydantic import ( Components, OpenAPI, Operation, Parameter, PathItem, Paths, Reference, RequestBody, Schema, ) from pydantic import ValidationError logger = logging.getLogger(__name__) class HTTPVerb(str, Enum): """HTTP verbs.""" GET = "get" PUT = "put" POST = "post" DELETE = "delete" OPTIONS = "options" HEAD = "head" PATCH = "patch" TRACE = "trace" @classmethod def from_str(cls, verb: str) -> "HTTPVerb": """Parse an HTTP verb.""" try: return cls(verb) except ValueError: raise ValueError(f"Invalid HTTP verb. Valid values are {cls.__members__}") [docs]class OpenAPISpec(OpenAPI): """OpenAPI Model that removes misformatted parts of the spec.""" @property def _paths_strict(self) -> Paths: if not self.paths: raise ValueError("No paths found in spec") return self.paths def _get_path_strict(self, path: str) -> PathItem: path_item = self._paths_strict.get(path) if not path_item: raise ValueError(f"No path found for {path}") return path_item @property def _components_strict(self) -> Components: """Get components or err.""" if self.components is None: raise ValueError("No components found in spec. ") return self.components @property def _parameters_strict(self) -> Dict[str, Union[Parameter, Reference]]: """Get parameters or err.""" parameters = self._components_strict.parameters if parameters is None: raise ValueError("No parameters found in spec. ") return parameters @property def _schemas_strict(self) -> Dict[str, Schema]: """Get the dictionary of schemas or err.""" schemas = self._components_strict.schemas if schemas is None: raise ValueError("No schemas found in spec. ") return schemas @property def _request_bodies_strict(self) -> Dict[str, Union[RequestBody, Reference]]: """Get the request body or err.""" request_bodies = self._components_strict.requestBodies if request_bodies is None: raise ValueError("No request body found in spec. ") return request_bodies def _get_referenced_parameter(self, ref: Reference) -> Union[Parameter, Reference]: """Get a parameter (or nested reference) or err.""" ref_name = ref.ref.split("/")[-1] parameters = self._parameters_strict if ref_name not in parameters: raise ValueError(f"No parameter found for {ref_name}") return parameters[ref_name] def _get_root_referenced_parameter(self, ref: Reference) -> Parameter: """Get the root reference or err.""" parameter = self._get_referenced_parameter(ref) while isinstance(parameter, Reference): parameter = self._get_referenced_parameter(parameter) return parameter [docs] def get_referenced_schema(self, ref: Reference) -> Schema: """Get a schema (or nested reference) or err.""" ref_name = ref.ref.split("/")[-1] schemas = self._schemas_strict if ref_name not in schemas: raise ValueError(f"No schema found for {ref_name}") return schemas[ref_name] [docs] def get_schema(self, schema: Union[Reference, Schema]) -> Schema: if isinstance(schema, Reference): return self.get_referenced_schema(schema) return schema def _get_root_referenced_schema(self, ref: Reference) -> Schema: """Get the root reference or err.""" schema = self.get_referenced_schema(ref) while isinstance(schema, Reference): schema = self.get_referenced_schema(schema) return schema def _get_referenced_request_body( self, ref: Reference ) -> Optional[Union[Reference, RequestBody]]: """Get a request body (or nested reference) or err.""" ref_name = ref.ref.split("/")[-1] request_bodies = self._request_bodies_strict if ref_name not in request_bodies: raise ValueError(f"No request body found for {ref_name}") return request_bodies[ref_name] def _get_root_referenced_request_body( self, ref: Reference ) -> Optional[RequestBody]: """Get the root request Body or err.""" request_body = self._get_referenced_request_body(ref) while isinstance(request_body, Reference): request_body = self._get_referenced_request_body(request_body) return request_body @staticmethod def _alert_unsupported_spec(obj: dict) -> None: """Alert if the spec is not supported.""" warning_message = ( " This may result in degraded performance." + " Convert your OpenAPI spec to 3.1.* spec" + " for better support." ) swagger_version = obj.get("swagger") openapi_version = obj.get("openapi") if isinstance(openapi_version, str): if openapi_version != "3.1.0": logger.warning( f"Attempting to load an OpenAPI {openapi_version}" f" spec. {warning_message}" ) else: pass elif isinstance(swagger_version, str): logger.warning( f"Attempting to load a Swagger {swagger_version}" f" spec. {warning_message}" ) else: raise ValueError( "Attempting to load an unsupported spec:" f"\n\n{obj}\n{warning_message}" ) [docs] @classmethod def parse_obj(cls, obj: dict) -> "OpenAPISpec": try: cls._alert_unsupported_spec(obj) return super().parse_obj(obj) except ValidationError as e: # We are handling possibly misconfigured specs and want to do a best-effort # job to get a reasonable interface out of it. new_obj = copy.deepcopy(obj) for error in e.errors(): keys = error["loc"] item = new_obj for key in keys[:-1]: item = item[key] item.pop(keys[-1], None) return cls.parse_obj(new_obj) [docs] @classmethod def from_spec_dict(cls, spec_dict: dict) -> "OpenAPISpec": """Get an OpenAPI spec from a dict.""" return cls.parse_obj(spec_dict) [docs] @classmethod def from_text(cls, text: str) -> "OpenAPISpec": """Get an OpenAPI spec from a text.""" try: spec_dict = json.loads(text) except json.JSONDecodeError: spec_dict = yaml.safe_load(text) return cls.from_spec_dict(spec_dict) [docs] @classmethod def from_file(cls, path: Union[str, Path]) -> "OpenAPISpec": """Get an OpenAPI spec from a file path.""" path_ = path if isinstance(path, Path) else Path(path) if not path_.exists(): raise FileNotFoundError(f"{path} does not exist") with path_.open("r") as f: return cls.from_text(f.read()) [docs] @classmethod def from_url(cls, url: str) -> "OpenAPISpec": """Get an OpenAPI spec from a URL.""" response = requests.get(url) return cls.from_text(response.text) @property def base_url(self) -> str: """Get the base url.""" return self.servers[0].url [docs] def get_methods_for_path(self, path: str) -> List[str]: """Return a list of valid methods for the specified path.""" path_item = self._get_path_strict(path) results = [] for method in HTTPVerb: operation = getattr(path_item, method.value, None) if isinstance(operation, Operation): results.append(method.value) return results [docs] def get_parameters_for_path(self, path: str) -> List[Parameter]: path_item = self._get_path_strict(path) parameters = [] if not path_item.parameters: return [] for parameter in path_item.parameters: if isinstance(parameter, Reference): parameter = self._get_root_referenced_parameter(parameter) parameters.append(parameter) return parameters [docs] def get_operation(self, path: str, method: str) -> Operation: """Get the operation object for a given path and HTTP method.""" path_item = self._get_path_strict(path) operation_obj = getattr(path_item, method, None) if not isinstance(operation_obj, Operation): raise ValueError(f"No {method} method found for {path}") return operation_obj [docs] def get_parameters_for_operation(self, operation: Operation) -> List[Parameter]: """Get the components for a given operation.""" parameters = [] if operation.parameters: for parameter in operation.parameters: if isinstance(parameter, Reference): parameter = self._get_root_referenced_parameter(parameter) parameters.append(parameter) return parameters [docs] def get_request_body_for_operation( self, operation: Operation ) -> Optional[RequestBody]: """Get the request body for a given operation.""" request_body = operation.requestBody if isinstance(request_body, Reference): request_body = self._get_root_referenced_request_body(request_body) return request_body [docs] @staticmethod def get_cleaned_operation_id(operation: Operation, path: str, method: str) -> str: """Get a cleaned operation id from an operation id.""" operation_id = operation.operationId if operation_id is None: # Replace all punctuation of any kind with underscore path = re.sub(r"[^a-zA-Z0-9]", "_", path.lstrip("/")) operation_id = f"{path}_{method}" return operation_id.replace("-", "_").replace(".", "_").replace("/", "_")
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/openapi.html
5e833693-9d36-4683-be68-a76808bcb0db
Source code for langchain.utilities.zapier """Util that can interact with Zapier NLA. Full docs here: https://nla.zapier.com/api/v1/docs Note: this wrapper currently only implemented the `api_key` auth method for testing and server-side production use cases (using the developer's connected accounts on Zapier.com) For use-cases where LangChain + Zapier NLA is powering a user-facing application, and LangChain needs access to the end-user's connected accounts on Zapier.com, you'll need to use oauth. Review the full docs above and reach out to nla@zapier.com for developer support. """ import json from typing import Dict, List, Optional import requests from pydantic import BaseModel, Extra, root_validator from requests import Request, Session from langchain.utils import get_from_dict_or_env [docs]class ZapierNLAWrapper(BaseModel): """Wrapper for Zapier NLA. Full docs here: https://nla.zapier.com/api/v1/docs Note: this wrapper currently only implemented the `api_key` auth method for testingand server-side production use cases (using the developer's connected accounts on Zapier.com) For use-cases where LangChain + Zapier NLA is powering a user-facing application, and LangChain needs access to the end-user's connected accounts on Zapier.com, you'll need to use oauth. Review the full docs above and reach out to nla@zapier.com for developer support. """ zapier_nla_api_key: str zapier_nla_oauth_access_token: str zapier_nla_api_base: str = "https://nla.zapier.com/api/v1/" class Config: """Configuration for this pydantic object.""" extra = Extra.forbid def _get_session(self) -> Session: session = requests.Session() session.headers.update( { "Accept": "application/json", "Content-Type": "application/json", } ) if self.zapier_nla_oauth_access_token: session.headers.update( {"Authorization": f"Bearer {self.zapier_nla_oauth_access_token}"} ) else: session.params = {"api_key": self.zapier_nla_api_key} return session def _get_action_request( self, action_id: str, instructions: str, params: Optional[Dict] = None ) -> Request: data = params if params else {} data.update( { "instructions": instructions, } ) return Request( "POST", self.zapier_nla_api_base + f"exposed/{action_id}/execute/", json=data, ) @root_validator(pre=True) def validate_environment(cls, values: Dict) -> Dict: """Validate that api key exists in environment.""" zapier_nla_api_key_default = None # If there is a oauth_access_key passed in the values # we don't need a nla_api_key it can be blank if "zapier_nla_oauth_access_token" in values: zapier_nla_api_key_default = "" else: values["zapier_nla_oauth_access_token"] = "" # we require at least one API Key zapier_nla_api_key = get_from_dict_or_env( values, "zapier_nla_api_key", "ZAPIER_NLA_API_KEY", zapier_nla_api_key_default, ) values["zapier_nla_api_key"] = zapier_nla_api_key return values [docs] def list(self) -> List[Dict]: """Returns a list of all exposed (enabled) actions associated with current user (associated with the set api_key). Change your exposed actions here: https://nla.zapier.com/demo/start/ The return list can be empty if no actions exposed. Else will contain a list of action objects: [{ "id": str, "description": str, "params": Dict[str, str] }] `params` will always contain an `instructions` key, the only required param. All others optional and if provided will override any AI guesses (see "understanding the AI guessing flow" here: https://nla.zapier.com/api/v1/docs) """ session = self._get_session() response = session.get(self.zapier_nla_api_base + "exposed/") response.raise_for_status() return response.json()["results"] [docs] def run( self, action_id: str, instructions: str, params: Optional[Dict] = None ) -> Dict: """Executes an action that is identified by action_id, must be exposed (enabled) by the current user (associated with the set api_key). Change your exposed actions here: https://nla.zapier.com/demo/start/ The return JSON is guaranteed to be less than ~500 words (350 tokens) making it safe to inject into the prompt of another LLM call. """ session = self._get_session() request = self._get_action_request(action_id, instructions, params) response = session.send(session.prepare_request(request)) response.raise_for_status() return response.json()["result"] [docs] def preview( self, action_id: str, instructions: str, params: Optional[Dict] = None ) -> Dict: """Same as run, but instead of actually executing the action, will instead return a preview of params that have been guessed by the AI in case you need to explicitly review before executing.""" session = self._get_session() params = params if params else {} params.update({"preview_only": True}) request = self._get_action_request(action_id, instructions, params) response = session.send(session.prepare_request(request)) response.raise_for_status() return response.json()["input_params"] [docs] def run_as_str(self, *args, **kwargs) -> str: # type: ignore[no-untyped-def] """Same as run, but returns a stringified version of the JSON for insertting back into an LLM.""" data = self.run(*args, **kwargs) return json.dumps(data) [docs] def preview_as_str(self, *args, **kwargs) -> str: # type: ignore[no-untyped-def] """Same as preview, but returns a stringified version of the JSON for insertting back into an LLM.""" data = self.preview(*args, **kwargs) return json.dumps(data) [docs] def list_as_str(self) -> str: # type: ignore[no-untyped-def] """Same as list, but returns a stringified version of the JSON for insertting back into an LLM.""" actions = self.list() return json.dumps(actions)
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/zapier.html
944d4eb5-2f27-4243-b44d-408202f320b7
Source code for langchain.utilities.scenexplain """Util that calls SceneXplain. In order to set this up, you need API key for the SceneXplain API. You can obtain a key by following the steps below. - Sign up for a free account at https://scenex.jina.ai/. - Navigate to the API Access page (https://scenex.jina.ai/api) and create a new API key. """ from typing import Dict import requests from pydantic import BaseModel, BaseSettings, Field, root_validator from langchain.utils import get_from_dict_or_env [docs]class SceneXplainAPIWrapper(BaseSettings, BaseModel): """Wrapper for SceneXplain API. In order to set this up, you need API key for the SceneXplain API. You can obtain a key by following the steps below. - Sign up for a free account at https://scenex.jina.ai/. - Navigate to the API Access page (https://scenex.jina.ai/api) and create a new API key. """ scenex_api_key: str = Field(..., env="SCENEX_API_KEY") scenex_api_url: str = ( "https://us-central1-causal-diffusion.cloudfunctions.net/describe" ) def _describe_image(self, image: str) -> str: headers = { "x-api-key": f"token {self.scenex_api_key}", "content-type": "application/json", } payload = { "data": [ { "image": image, "algorithm": "Ember", "languages": ["en"], } ] } response = requests.post(self.scenex_api_url, headers=headers, json=payload) response.raise_for_status() result = response.json().get("result", []) img = result[0] if result else {} return img.get("text", "") [docs] @root_validator(pre=True) def validate_environment(cls, values: Dict) -> Dict: """Validate that api key exists in environment.""" scenex_api_key = get_from_dict_or_env( values, "scenex_api_key", "SCENEX_API_KEY" ) values["scenex_api_key"] = scenex_api_key return values [docs] def run(self, image: str) -> str: """Run SceneXplain image explainer.""" description = self._describe_image(image) if not description: return "No description found." return description
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/scenexplain.html
3b45a8d1-1b02-48cd-8e98-46fbbd049552
Source code for langchain.utilities.wolfram_alpha """Util that calls WolframAlpha.""" from typing import Any, Dict, Optional from pydantic import BaseModel, Extra, root_validator from langchain.utils import get_from_dict_or_env [docs]class WolframAlphaAPIWrapper(BaseModel): """Wrapper for Wolfram Alpha. Docs for using: 1. Go to wolfram alpha and sign up for a developer account 2. Create an app and get your APP ID 3. Save your APP ID into WOLFRAM_ALPHA_APPID env variable 4. pip install wolframalpha """ wolfram_client: Any #: :meta private: wolfram_alpha_appid: Optional[str] = None class Config: """Configuration for this pydantic object.""" extra = Extra.forbid @root_validator() def validate_environment(cls, values: Dict) -> Dict: """Validate that api key and python package exists in environment.""" wolfram_alpha_appid = get_from_dict_or_env( values, "wolfram_alpha_appid", "WOLFRAM_ALPHA_APPID" ) values["wolfram_alpha_appid"] = wolfram_alpha_appid try: import wolframalpha except ImportError: raise ImportError( "wolframalpha is not installed. " "Please install it with `pip install wolframalpha`" ) client = wolframalpha.Client(wolfram_alpha_appid) values["wolfram_client"] = client return values [docs] def run(self, query: str) -> str: """Run query through WolframAlpha and parse result.""" res = self.wolfram_client.query(query) try: assumption = next(res.pods).text answer = next(res.results).text except StopIteration: return "Wolfram Alpha wasn't able to answer it" if answer is None or answer == "": # We don't want to return the assumption alone if answer is empty return "No good Wolfram Alpha Result was found" else: return f"Assumption: {assumption} \nAnswer: {answer}"
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/wolfram_alpha.html
4ee1ccdc-5c0f-430b-ac81-1951c131fdc6
Source code for langchain.utilities.google_serper """Util that calls Google Search using the Serper.dev API.""" from typing import Any, Dict, List, Optional import aiohttp import requests from pydantic.class_validators import root_validator from pydantic.main import BaseModel from typing_extensions import Literal from langchain.utils import get_from_dict_or_env [docs]class GoogleSerperAPIWrapper(BaseModel): """Wrapper around the Serper.dev Google Search API. You can create a free API key at https://serper.dev. To use, you should have the environment variable ``SERPER_API_KEY`` set with your API key, or pass `serper_api_key` as a named parameter to the constructor. Example: .. code-block:: python from langchain import GoogleSerperAPIWrapper google_serper = GoogleSerperAPIWrapper() """ k: int = 10 gl: str = "us" hl: str = "en" # "places" and "images" is available from Serper but not implemented in the # parser of run(). They can be used in results() type: Literal["news", "search", "places", "images"] = "search" result_key_for_type = { "news": "news", "places": "places", "images": "images", "search": "organic", } tbs: Optional[str] = None serper_api_key: Optional[str] = None aiosession: Optional[aiohttp.ClientSession] = None class Config: """Configuration for this pydantic object.""" arbitrary_types_allowed = True @root_validator() def validate_environment(cls, values: Dict) -> Dict: """Validate that api key exists in environment.""" serper_api_key = get_from_dict_or_env( values, "serper_api_key", "SERPER_API_KEY" ) values["serper_api_key"] = serper_api_key return values [docs] def results(self, query: str, **kwargs: Any) -> Dict: """Run query through GoogleSearch.""" return self._google_serper_api_results( query, gl=self.gl, hl=self.hl, num=self.k, tbs=self.tbs, search_type=self.type, **kwargs, ) [docs] def run(self, query: str, **kwargs: Any) -> str: """Run query through GoogleSearch and parse result.""" results = self._google_serper_api_results( query, gl=self.gl, hl=self.hl, num=self.k, tbs=self.tbs, search_type=self.type, **kwargs, ) return self._parse_results(results) [docs] async def aresults(self, query: str, **kwargs: Any) -> Dict: """Run query through GoogleSearch.""" results = await self._async_google_serper_search_results( query, gl=self.gl, hl=self.hl, num=self.k, search_type=self.type, tbs=self.tbs, **kwargs, ) return results [docs] async def arun(self, query: str, **kwargs: Any) -> str: """Run query through GoogleSearch and parse result async.""" results = await self._async_google_serper_search_results( query, gl=self.gl, hl=self.hl, num=self.k, search_type=self.type, tbs=self.tbs, **kwargs, ) return self._parse_results(results) def _parse_snippets(self, results: dict) -> List[str]: snippets = [] if results.get("answerBox"): answer_box = results.get("answerBox", {}) if answer_box.get("answer"): return [answer_box.get("answer")] elif answer_box.get("snippet"): return [answer_box.get("snippet").replace("\n", " ")] elif answer_box.get("snippetHighlighted"): return answer_box.get("snippetHighlighted") if results.get("knowledgeGraph"): kg = results.get("knowledgeGraph", {}) title = kg.get("title") entity_type = kg.get("type") if entity_type: snippets.append(f"{title}: {entity_type}.") description = kg.get("description") if description: snippets.append(description) for attribute, value in kg.get("attributes", {}).items(): snippets.append(f"{title} {attribute}: {value}.") for result in results[self.result_key_for_type[self.type]][: self.k]: if "snippet" in result: snippets.append(result["snippet"]) for attribute, value in result.get("attributes", {}).items(): snippets.append(f"{attribute}: {value}.") if len(snippets) == 0: return ["No good Google Search Result was found"] return snippets def _parse_results(self, results: dict) -> str: return " ".join(self._parse_snippets(results)) def _google_serper_api_results( self, search_term: str, search_type: str = "search", **kwargs: Any ) -> dict: headers = { "X-API-KEY": self.serper_api_key or "", "Content-Type": "application/json", } params = { "q": search_term, **{key: value for key, value in kwargs.items() if value is not None}, } response = requests.post( f"https://google.serper.dev/{search_type}", headers=headers, params=params ) response.raise_for_status() search_results = response.json() return search_results async def _async_google_serper_search_results( self, search_term: str, search_type: str = "search", **kwargs: Any ) -> dict: headers = { "X-API-KEY": self.serper_api_key or "", "Content-Type": "application/json", } url = f"https://google.serper.dev/{search_type}" params = { "q": search_term, **{key: value for key, value in kwargs.items() if value is not None}, } if not self.aiosession: async with aiohttp.ClientSession() as session: async with session.post( url, params=params, headers=headers, raise_for_status=False ) as response: search_results = await response.json() else: async with self.aiosession.post( url, params=params, headers=headers, raise_for_status=True ) as response: search_results = await response.json() return search_results
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/google_serper.html
83c2fcd0-0a7d-4972-9988-3bb136950bd8
Source code for langchain.utilities.twilio """Util that calls Twilio.""" from typing import Any, Dict, Optional from pydantic import BaseModel, Extra, root_validator from langchain.utils import get_from_dict_or_env [docs]class TwilioAPIWrapper(BaseModel): """Messaging Client using Twilio. To use, you should have the ``twilio`` python package installed, and the environment variables ``TWILIO_ACCOUNT_SID``, ``TWILIO_AUTH_TOKEN``, and ``TWILIO_FROM_NUMBER``, or pass `account_sid`, `auth_token`, and `from_number` as named parameters to the constructor. Example: .. code-block:: python from langchain.utilities.twilio import TwilioAPIWrapper twilio = TwilioAPIWrapper( account_sid="ACxxx", auth_token="xxx", from_number="+10123456789" ) twilio.run('test', '+12484345508') """ client: Any #: :meta private: account_sid: Optional[str] = None """Twilio account string identifier.""" auth_token: Optional[str] = None """Twilio auth token.""" from_number: Optional[str] = None """A Twilio phone number in [E.164](https://www.twilio.com/docs/glossary/what-e164) format, an [alphanumeric sender ID](https://www.twilio.com/docs/sms/send-messages#use-an-alphanumeric-sender-id), or a [Channel Endpoint address](https://www.twilio.com/docs/sms/channels#channel-addresses) that is enabled for the type of message you want to send. Phone numbers or [short codes](https://www.twilio.com/docs/sms/api/short-code) purchased from Twilio also work here. You cannot, for example, spoof messages from a private cell phone number. If you are using `messaging_service_sid`, this parameter must be empty. """ # noqa: E501 class Config: """Configuration for this pydantic object.""" extra = Extra.forbid arbitrary_types_allowed = False @root_validator() def validate_environment(cls, values: Dict) -> Dict: """Validate that api key and python package exists in environment.""" try: from twilio.rest import Client except ImportError: raise ImportError( "Could not import twilio python package. " "Please install it with `pip install twilio`." ) account_sid = get_from_dict_or_env(values, "account_sid", "TWILIO_ACCOUNT_SID") auth_token = get_from_dict_or_env(values, "auth_token", "TWILIO_AUTH_TOKEN") values["from_number"] = get_from_dict_or_env( values, "from_number", "TWILIO_FROM_NUMBER" ) values["client"] = Client(account_sid, auth_token) return values [docs] def run(self, body: str, to: str) -> str: """Run body through Twilio and respond with message sid. Args: body: The text of the message you want to send. Can be up to 1,600 characters in length. to: The destination phone number in [E.164](https://www.twilio.com/docs/glossary/what-e164) format for SMS/MMS or [Channel user address](https://www.twilio.com/docs/sms/channels#channel-addresses) for other 3rd-party channels. """ # noqa: E501 message = self.client.messages.create(to, from_=self.from_number, body=body) return message.sid
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/twilio.html
20fb05a6-b810-4862-acdd-baa537f8ce08
Source code for langchain.utilities.google_places_api """Chain that calls Google Places API. """ import logging from typing import Any, Dict, Optional from pydantic import BaseModel, Extra, root_validator from langchain.utils import get_from_dict_or_env [docs]class GooglePlacesAPIWrapper(BaseModel): """Wrapper around Google Places API. To use, you should have the ``googlemaps`` python package installed, **an API key for the google maps platform**, and the enviroment variable ''GPLACES_API_KEY'' set with your API key , or pass 'gplaces_api_key' as a named parameter to the constructor. By default, this will return the all the results on the input query. You can use the top_k_results argument to limit the number of results. Example: .. code-block:: python from langchain import GooglePlacesAPIWrapper gplaceapi = GooglePlacesAPIWrapper() """ gplaces_api_key: Optional[str] = None google_map_client: Any #: :meta private: top_k_results: Optional[int] = None class Config: """Configuration for this pydantic object.""" extra = Extra.forbid arbitrary_types_allowed = True @root_validator() def validate_environment(cls, values: Dict) -> Dict: """Validate that api key is in your environment variable.""" gplaces_api_key = get_from_dict_or_env( values, "gplaces_api_key", "GPLACES_API_KEY" ) values["gplaces_api_key"] = gplaces_api_key try: import googlemaps values["google_map_client"] = googlemaps.Client(gplaces_api_key) except ImportError: raise ImportError( "Could not import googlemaps python package. " "Please install it with `pip install googlemaps`." ) return values [docs] def run(self, query: str) -> str: """Run Places search and get k number of places that exists that match.""" search_results = self.google_map_client.places(query)["results"] num_to_return = len(search_results) places = [] if num_to_return == 0: return "Google Places did not find any places that match the description" num_to_return = ( num_to_return if self.top_k_results is None else min(num_to_return, self.top_k_results) ) for i in range(num_to_return): result = search_results[i] details = self.fetch_place_details(result["place_id"]) if details is not None: places.append(details) return "\n".join([f"{i+1}. {item}" for i, item in enumerate(places)]) [docs] def fetch_place_details(self, place_id: str) -> Optional[str]: try: place_details = self.google_map_client.place(place_id) formatted_details = self.format_place_details(place_details) return formatted_details except Exception as e: logging.error(f"An Error occurred while fetching place details: {e}") return None [docs] def format_place_details(self, place_details: Dict[str, Any]) -> Optional[str]: try: name = place_details.get("result", {}).get("name", "Unkown") address = place_details.get("result", {}).get( "formatted_address", "Unknown" ) phone_number = place_details.get("result", {}).get( "formatted_phone_number", "Unknown" ) website = place_details.get("result", {}).get("website", "Unknown") formatted_details = ( f"{name}\nAddress: {address}\n" f"Phone: {phone_number}\nWebsite: {website}\n\n" ) return formatted_details except Exception as e: logging.error(f"An error occurred while formatting place details: {e}") return None
https://api.python.langchain.com/en/latest/_modules/langchain/utilities/google_places_api.html
b18be03a-7934-4af2-bcb2-d2f8b925b682
Source code for langchain.document_loaders.tomarkdown """Loader that loads HTML to markdown using 2markdown.""" from __future__ import annotations from typing import Iterator, List import requests from langchain.docstore.document import Document from langchain.document_loaders.base import BaseLoader [docs]class ToMarkdownLoader(BaseLoader): """Loader that loads HTML to markdown using 2markdown.""" def __init__(self, url: str, api_key: str): """Initialize with url and api key.""" self.url = url self.api_key = api_key [docs] def lazy_load( self, ) -> Iterator[Document]: """Lazily load the file.""" response = requests.post( "https://2markdown.com/api/2md", headers={"X-Api-Key": self.api_key}, json={"url": self.url}, ) text = response.json()["article"] metadata = {"source": self.url} yield Document(page_content=text, metadata=metadata) [docs] def load(self) -> List[Document]: """Load file.""" return list(self.lazy_load())
https://api.python.langchain.com/en/latest/_modules/langchain/document_loaders/tomarkdown.html
e886b059-b129-4b34-8297-173d4f226281
Source code for langchain.document_loaders.conllu """Load CoNLL-U files.""" import csv from typing import List from langchain.docstore.document import Document from langchain.document_loaders.base import BaseLoader [docs]class CoNLLULoader(BaseLoader): """Load CoNLL-U files.""" def __init__(self, file_path: str): """Initialize with file path.""" self.file_path = file_path [docs] def load(self) -> List[Document]: """Load from file path.""" with open(self.file_path, encoding="utf8") as f: tsv = list(csv.reader(f, delimiter="\t")) # If len(line) > 1, the line is not a comment lines = [line for line in tsv if len(line) > 1] text = "" for i, line in enumerate(lines): # Do not add a space after a punctuation mark or at the end of the sentence if line[9] == "SpaceAfter=No" or i == len(lines) - 1: text += line[1] else: text += line[1] + " " metadata = {"source": self.file_path} return [Document(page_content=text, metadata=metadata)]
https://api.python.langchain.com/en/latest/_modules/langchain/document_loaders/conllu.html
8060e866-43c5-49aa-905e-a95da957f56f
Source code for langchain.document_loaders.toml import json from pathlib import Path from typing import Iterator, List, Union from langchain.docstore.document import Document from langchain.document_loaders.base import BaseLoader [docs]class TomlLoader(BaseLoader): """ A TOML document loader that inherits from the BaseLoader class. This class can be initialized with either a single source file or a source directory containing TOML files. """ def __init__(self, source: Union[str, Path]): """Initialize the TomlLoader with a source file or directory.""" self.source = Path(source) [docs] def load(self) -> List[Document]: """Load and return all documents.""" return list(self.lazy_load()) [docs] def lazy_load(self) -> Iterator[Document]: """Lazily load the TOML documents from the source file or directory.""" import tomli if self.source.is_file() and self.source.suffix == ".toml": files = [self.source] elif self.source.is_dir(): files = list(self.source.glob("**/*.toml")) else: raise ValueError("Invalid source path or file type") for file_path in files: with file_path.open("r", encoding="utf-8") as file: content = file.read() try: data = tomli.loads(content) doc = Document( page_content=json.dumps(data), metadata={"source": str(file_path)}, ) yield doc except tomli.TOMLDecodeError as e: print(f"Error parsing TOML file {file_path}: {e}")
https://api.python.langchain.com/en/latest/_modules/langchain/document_loaders/toml.html
5f0b5252-763f-434d-9581-959f3987edb3
Source code for langchain.document_loaders.url_playwright """Loader that uses Playwright to load a page, then uses unstructured to load the html. """ import logging from typing import List, Optional from langchain.docstore.document import Document from langchain.document_loaders.base import BaseLoader logger = logging.getLogger(__name__) [docs]class PlaywrightURLLoader(BaseLoader): """Loader that uses Playwright and to load a page and unstructured to load the html. This is useful for loading pages that require javascript to render. Attributes: urls (List[str]): List of URLs to load. continue_on_failure (bool): If True, continue loading other URLs on failure. headless (bool): If True, the browser will run in headless mode. """ def __init__( self, urls: List[str], continue_on_failure: bool = True, headless: bool = True, remove_selectors: Optional[List[str]] = None, ): """Load a list of URLs using Playwright and unstructured.""" try: import playwright # noqa:F401 except ImportError: raise ImportError( "playwright package not found, please install it with " "`pip install playwright`" ) try: import unstructured # noqa:F401 except ImportError: raise ValueError( "unstructured package not found, please install it with " "`pip install unstructured`" ) self.urls = urls self.continue_on_failure = continue_on_failure self.headless = headless self.remove_selectors = remove_selectors [docs] def load(self) -> List[Document]: """Load the specified URLs using Playwright and create Document instances. Returns: List[Document]: A list of Document instances with loaded content. """ from playwright.sync_api import sync_playwright from unstructured.partition.html import partition_html docs: List[Document] = list() with sync_playwright() as p: browser = p.chromium.launch(headless=self.headless) for url in self.urls: try: page = browser.new_page() page.goto(url) for selector in self.remove_selectors or []: elements = page.locator(selector).all() for element in elements: if element.is_visible(): element.evaluate("element => element.remove()") page_source = page.content() elements = partition_html(text=page_source) text = "\n\n".join([str(el) for el in elements]) metadata = {"source": url} docs.append(Document(page_content=text, metadata=metadata)) except Exception as e: if self.continue_on_failure: logger.error( f"Error fetching or processing {url}, exception: {e}" ) else: raise e browser.close() return docs
https://api.python.langchain.com/en/latest/_modules/langchain/document_loaders/url_playwright.html
5c8dd0b6-d119-4485-983f-a024b4c80b88
Source code for langchain.document_loaders.gcs_directory """Loading logic for loading documents from an GCS directory.""" from typing import List from langchain.docstore.document import Document from langchain.document_loaders.base import BaseLoader from langchain.document_loaders.gcs_file import GCSFileLoader [docs]class GCSDirectoryLoader(BaseLoader): """Loading logic for loading documents from GCS.""" def __init__(self, project_name: str, bucket: str, prefix: str = ""): """Initialize with bucket and key name.""" self.project_name = project_name self.bucket = bucket self.prefix = prefix [docs] def load(self) -> List[Document]: """Load documents.""" try: from google.cloud import storage except ImportError: raise ValueError( "Could not import google-cloud-storage python package. " "Please install it with `pip install google-cloud-storage`." ) client = storage.Client(project=self.project_name) docs = [] for blob in client.list_blobs(self.bucket, prefix=self.prefix): # we shall just skip directories since GCSFileLoader creates # intermediate directories on the fly if blob.name.endswith("/"): continue loader = GCSFileLoader(self.project_name, self.bucket, blob.name) docs.extend(loader.load()) return docs
https://api.python.langchain.com/en/latest/_modules/langchain/document_loaders/gcs_directory.html
2ea52d58-c3a0-458d-a4df-10fa3c3d93e3
Source code for langchain.document_loaders.joplin import json import urllib from datetime import datetime from typing import Iterator, List, Optional from langchain.document_loaders.base import BaseLoader from langchain.schema import Document from langchain.utils import get_from_env LINK_NOTE_TEMPLATE = "joplin://x-callback-url/openNote?id={id}" [docs]class JoplinLoader(BaseLoader): """ Loader that fetches notes from Joplin. In order to use this loader, you need to have Joplin running with the Web Clipper enabled (look for "Web Clipper" in the app settings). To get the access token, you need to go to the Web Clipper options and under "Advanced Options" you will find the access token. You can find more information about the Web Clipper service here: https://joplinapp.org/clipper/ """ def __init__( self, access_token: Optional[str] = None, port: int = 41184, host: str = "localhost", ) -> None: access_token = access_token or get_from_env( "access_token", "JOPLIN_ACCESS_TOKEN" ) base_url = f"http://{host}:{port}" self._get_note_url = ( f"{base_url}/notes?token={access_token}" f"&fields=id,parent_id,title,body,created_time,updated_time&page={{page}}" ) self._get_folder_url = ( f"{base_url}/folders/{{id}}?token={access_token}&fields=title" ) self._get_tag_url = ( f"{base_url}/notes/{{id}}/tags?token={access_token}&fields=title" ) def _get_notes(self) -> Iterator[Document]: has_more = True page = 1 while has_more: req_note = urllib.request.Request(self._get_note_url.format(page=page)) with urllib.request.urlopen(req_note) as response: json_data = json.loads(response.read().decode()) for note in json_data["items"]: metadata = { "source": LINK_NOTE_TEMPLATE.format(id=note["id"]), "folder": self._get_folder(note["parent_id"]), "tags": self._get_tags(note["id"]), "title": note["title"], "created_time": self._convert_date(note["created_time"]), "updated_time": self._convert_date(note["updated_time"]), } yield Document(page_content=note["body"], metadata=metadata) has_more = json_data["has_more"] page += 1 def _get_folder(self, folder_id: str) -> str: req_folder = urllib.request.Request(self._get_folder_url.format(id=folder_id)) with urllib.request.urlopen(req_folder) as response: json_data = json.loads(response.read().decode()) return json_data["title"] def _get_tags(self, note_id: str) -> List[str]: req_tag = urllib.request.Request(self._get_tag_url.format(id=note_id)) with urllib.request.urlopen(req_tag) as response: json_data = json.loads(response.read().decode()) return [tag["title"] for tag in json_data["items"]] def _convert_date(self, date: int) -> str: return datetime.fromtimestamp(date / 1000).strftime("%Y-%m-%d %H:%M:%S") [docs] def lazy_load(self) -> Iterator[Document]: yield from self._get_notes() [docs] def load(self) -> List[Document]: return list(self.lazy_load())
https://api.python.langchain.com/en/latest/_modules/langchain/document_loaders/joplin.html
32761cf4-9206-4424-9a39-a6c61093a588
Source code for langchain.document_loaders.powerpoint """Loader that loads powerpoint files.""" import os from typing import List from langchain.document_loaders.unstructured import UnstructuredFileLoader [docs]class UnstructuredPowerPointLoader(UnstructuredFileLoader): """Loader that uses unstructured to load powerpoint files.""" def _get_elements(self) -> List: from unstructured.__version__ import __version__ as __unstructured_version__ from unstructured.file_utils.filetype import FileType, detect_filetype unstructured_version = tuple( [int(x) for x in __unstructured_version__.split(".")] ) # NOTE(MthwRobinson) - magic will raise an import error if the libmagic # system dependency isn't installed. If it's not installed, we'll just # check the file extension try: import magic # noqa: F401 is_ppt = detect_filetype(self.file_path) == FileType.PPT except ImportError: _, extension = os.path.splitext(str(self.file_path)) is_ppt = extension == ".ppt" if is_ppt and unstructured_version < (0, 4, 11): raise ValueError( f"You are on unstructured version {__unstructured_version__}. " "Partitioning .ppt files is only supported in unstructured>=0.4.11. " "Please upgrade the unstructured package and try again." ) if is_ppt: from unstructured.partition.ppt import partition_ppt return partition_ppt(filename=self.file_path, **self.unstructured_kwargs) else: from unstructured.partition.pptx import partition_pptx return partition_pptx(filename=self.file_path, **self.unstructured_kwargs)
https://api.python.langchain.com/en/latest/_modules/langchain/document_loaders/powerpoint.html
7acc3dea-e63b-4edc-a881-25501a391cf8
Source code for langchain.document_loaders.snowflake_loader from __future__ import annotations from typing import Any, Dict, Iterator, List, Optional, Tuple from langchain.docstore.document import Document from langchain.document_loaders.base import BaseLoader [docs]class SnowflakeLoader(BaseLoader): """Loads a query result from Snowflake into a list of documents. Each document represents one row of the result. The `page_content_columns` are written into the `page_content` of the document. The `metadata_columns` are written into the `metadata` of the document. By default, all columns are written into the `page_content` and none into the `metadata`. """ def __init__( self, query: str, user: str, password: str, account: str, warehouse: str, role: str, database: str, schema: str, parameters: Optional[Dict[str, Any]] = None, page_content_columns: Optional[List[str]] = None, metadata_columns: Optional[List[str]] = None, ): """Initialize Snowflake document loader. Args: query: The query to run in Snowflake. user: Snowflake user. password: Snowflake password. account: Snowflake account. warehouse: Snowflake warehouse. role: Snowflake role. database: Snowflake database schema: Snowflake schema page_content_columns: Optional. Columns written to Document `page_content`. metadata_columns: Optional. Columns written to Document `metadata`. """ self.query = query self.user = user self.password = password self.account = account self.warehouse = warehouse self.role = role self.database = database self.schema = schema self.parameters = parameters self.page_content_columns = ( page_content_columns if page_content_columns is not None else ["*"] ) self.metadata_columns = metadata_columns if metadata_columns is not None else [] def _execute_query(self) -> List[Dict[str, Any]]: try: import snowflake.connector except ImportError as ex: raise ValueError( "Could not import snowflake-connector-python package. " "Please install it with `pip install snowflake-connector-python`." ) from ex conn = snowflake.connector.connect( user=self.user, password=self.password, account=self.account, warehouse=self.warehouse, role=self.role, database=self.database, schema=self.schema, parameters=self.parameters, ) try: cur = conn.cursor() cur.execute("USE DATABASE " + self.database) cur.execute("USE SCHEMA " + self.schema) cur.execute(self.query, self.parameters) query_result = cur.fetchall() column_names = [column[0] for column in cur.description] query_result = [dict(zip(column_names, row)) for row in query_result] except Exception as e: print(f"An error occurred: {e}") query_result = [] finally: cur.close() return query_result def _get_columns( self, query_result: List[Dict[str, Any]] ) -> Tuple[List[str], List[str]]: page_content_columns = ( self.page_content_columns if self.page_content_columns else [] ) metadata_columns = self.metadata_columns if self.metadata_columns else [] if page_content_columns is None and query_result: page_content_columns = list(query_result[0].keys()) if metadata_columns is None: metadata_columns = [] return page_content_columns or [], metadata_columns [docs] def lazy_load(self) -> Iterator[Document]: query_result = self._execute_query() if isinstance(query_result, Exception): print(f"An error occurred during the query: {query_result}") return [] page_content_columns, metadata_columns = self._get_columns(query_result) if "*" in page_content_columns: page_content_columns = list(query_result[0].keys()) for row in query_result: page_content = "\n".join( f"{k}: {v}" for k, v in row.items() if k in page_content_columns ) metadata = {k: v for k, v in row.items() if k in metadata_columns} doc = Document(page_content=page_content, metadata=metadata) yield doc [docs] def load(self) -> List[Document]: """Load data into document objects.""" return list(self.lazy_load())
https://api.python.langchain.com/en/latest/_modules/langchain/document_loaders/snowflake_loader.html