handler.py

import torch
import locale
from typing import Dict, List, Any
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
from langchain.llms import HuggingFacePipeline
from langchain.retrievers.document_compressors import LLMChainExtractor
from langchain.retrievers import ContextualCompressionRetriever
from langchain.vectorstores import Chroma
from langchain import PromptTemplate, LLMChain
from langchain.chains import RetrievalQA, ConversationalRetrievalChain
from langchain.prompts import PromptTemplate
from langchain.prompts.prompt import PromptTemplate
from langchain.memory import ConversationBufferMemory
from langchain.embeddings import HuggingFaceBgeEmbeddings
from langchain.document_loaders import WebBaseLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.chains.qa_with_sources import load_qa_with_sources_chain
from langchain.chains.combine_documents import create_stuff_documents_chain
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
from langchain_core.messages import HumanMessage
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnableLambda, RunnableBranch, RunnablePassthrough
from operator import itemgetter
from langchain.schema import format_document
from langchain.memory import ConversationBufferMemory
from langchain_core.messages import AIMessage, HumanMessage, get_buffer_string
    
class EndpointHandler():
    def __init__(self, path=""):

        # Config LangChain
        # os.environ["LANGCHAIN_TRACING_V2"] = "true"
        # os.environ["LANGCHAIN_API_KEY"] = 

        # Create LLM
        model_id = path
        
        tokenizer = AutoTokenizer.from_pretrained(
            model_id,
            trust_remote_code=True,
            padding_side="left",
            add_eos_token=True,
            use_fast=False
        )
        tokenizer.pad_token = tokenizer.eos_token

        model = AutoModelForCausalLM.from_pretrained(model_id)
        pipe = pipeline("text-generation", model=model, tokenizer=tokenizer, max_new_tokens=1024)
        chat = HuggingFacePipeline(pipeline=pipe)

        # Create Text-Embedding Model
        embedding_function = HuggingFaceBgeEmbeddings(
            model_name="BAAI/bge-large-zh",
            model_kwargs={'device': 'cuda'},
            encode_kwargs={'normalize_embeddings': True}
        )
        
        # Load Vector db
        urls = [
            "https://hk.on.cc/hk/bkn/cnt/news/20221019/bkn-20221019040039334-1019_00822_001.html",
            "https://www.hk01.com/%E7%A4%BE%E6%9C%83%E6%96%B0%E8%81%9E/822848/%E5%89%B5%E7%A7%91%E7%B2%BE%E8%8B%B1-%E5%87%BA%E6%88%B02022%E4%B8%96%E7%95%8C%E6%8A%80%E8%83%BD%E5%A4%A7%E8%B3%BD%E7%89%B9%E5%88%A5%E8%B3%BD",
            "https://www.wenweipo.com/epaper/view/newsDetail/1582436861224292352.html",
            "https://www.thinkhk.com/article/2023-03/24/59874.html"
        ]

        loader = WebBaseLoader(urls)
        data = loader.load()

        text_splitter = RecursiveCharacterTextSplitter(chunk_size = 1000, chunk_overlap = 16)
        all_splits = text_splitter.split_documents(data)
        
        vectorstore = Chroma.from_documents(documents=all_splits, embedding=embedding_function)
        retriever = vectorstore.as_retriever(search_kwargs={"k": 4})
        
        compressor = LLMChainExtractor.from_llm(chat)
        retriever = ContextualCompressionRetriever(base_compressor=compressor, base_retriever=retriever)

        _template = """[INST] Given the following conversation and a follow up question, rephrase the follow up question to be a standalone question, in its original language.
        Chat History:
        {chat_history}
        Follow Up Input: {question}
        Standalone question: [/INST]"""
        
        CONDENSE_QUESTION_PROMPT = PromptTemplate.from_template(_template)
        
        template = """[INST] Answer the question based only on the following context:
        {context}
        
        Question: {question} [/INST]
        """

        ANSWER_PROMPT = ChatPromptTemplate.from_template(template)

        self.memory = ConversationBufferMemory(
            return_messages=True, output_key="answer", input_key="question"
        )
        
        # First we add a step to load memory
        # This adds a "memory" key to the input object
        loaded_memory = RunnablePassthrough.assign(
            chat_history=RunnableLambda(self.memory.load_memory_variables) | itemgetter("history"),
        )
        # Now we calculate the standalone question
        standalone_question = {
            "standalone_question": {
                "question": lambda x: x["question"],
                "chat_history": lambda x: get_buffer_string(x["chat_history"]),
            }
            | CONDENSE_QUESTION_PROMPT
            | chat
            | StrOutputParser(),
        }

        DEFAULT_DOCUMENT_PROMPT = PromptTemplate.from_template(template="{page_content}")

        def _combine_documents(
            docs, document_prompt=DEFAULT_DOCUMENT_PROMPT, document_separator="\n\n"
        ):
            doc_strings = [format_document(doc, document_prompt) for doc in docs]
            return document_separator.join(doc_strings)
  
        # Now we retrieve the documents
        retrieved_documents = {
            "docs": itemgetter("standalone_question") | retriever,
            "question": lambda x: x["standalone_question"],
        }
        # Now we construct the inputs for the final prompt
        final_inputs = {
            "context": lambda x: _combine_documents(x["docs"]),
            "question": itemgetter("question"),
        }
        # And finally, we do the part that returns the answers
        answer = {
            "answer": final_inputs | ANSWER_PROMPT | chat,
            "docs": itemgetter("docs"),
        }
        # And now we put it all together!
        self.final_chain = loaded_memory | standalone_question | retrieved_documents | answer

    def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:
        # get inputs
        inputs = data.pop("inputs",data)
        date = data.pop("date", None)

        result = self.final_chain.invoke({"question": inputs})

        # Note that the memory does not save automatically
        # This will be improved in the future
        # For now you need to save it yourself

        # self.memory.save_context(inputs, {"answer": result["answer"].content})
        # self.memory.load_memory_variables({})
        
        return result