Upload 4 files

helper.py

@@ -0,0 +1,27 @@
+from langchain_community.document_loaders import PyPDFLoader, DirectoryLoader
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain_community.embeddings import HuggingFaceEmbeddings
+
+
+# Extract data from the PDF
+def load_pdf(data):
+    loader = DirectoryLoader(data, glob="*.pdf", loader_cls=PyPDFLoader)
+
+    documents = loader.load()
+
+    return documents
+
+
+# Create text chunks
+def text_split(extracted_data):
+    text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=20)
+    text_chunks = text_splitter.split_documents(extracted_data)
+
+    return text_chunks
+
+
+def download_hugging_face_embeddings():
+    embeddings = HuggingFaceEmbeddings(
+        model_name="sentence-transformers/all-MiniLM-L6-v2"
+    )
+    return embeddings

llama_call.py

@@ -0,0 +1,55 @@
+from src.helper import download_hugging_face_embeddings
+from langchain_pinecone import PineconeVectorStore
+from langchain.prompts import PromptTemplate
+from langchain_community.llms import CTransformers
+from langchain.chains import RetrievalQA
+from dotenv import load_dotenv
+from src.prompt import prompt_template
+import os
+
+load_dotenv()
+
+PINECONE_API_KEY = os.environ.get("PINECONE_API_KEY")
+PINECONE_API_ENV = os.environ.get("PINECONE_API_ENV")
+
+
+embeddings = download_hugging_face_embeddings()
+index_name = "llm-chatbot"
+
+# Initializing the Pinecone
+docsearch = PineconeVectorStore.from_existing_index(index_name, embeddings)
+
+
+PROMPT = PromptTemplate(
+    template=prompt_template, input_variables=["context", "question"]
+)
+
+chain_type_kwargs = {"prompt": PROMPT}
+
+current_dir = os.getcwd()
+llm = CTransformers(
+    model=os.path.join(current_dir, "saved_models/llama-2-7b-chat.ggmlv3.q4_0.bin"),
+    model_type="llama",
+    streaming=True,
+    config={"max_new_tokens": 256, "temperature": 0.6, "context_length": -1},
+)
+
+
+qa = RetrievalQA.from_chain_type(
+    llm=llm,
+    chain_type="stuff",
+    retriever=docsearch.as_retriever(search_kwargs={"k": 2}),
+    return_source_documents=True,
+    chain_type_kwargs=chain_type_kwargs,
+    verbose=True,
+)
+
+
+def llama_call(input):
+    result = qa.invoke({"query": input})
+    return str(result["result"])
+
+
+if __name__ == "__main__":
+    msg = "If a previous owner of a land had allowed a neighbour or neighbour to walk or drive over his land in a shortcut and this has been going on for say a decade or so can I as the new owner stop them now from using the shortcut?"
+    print(f"response: {llama_call(msg)}")

openai_call.py

@@ -0,0 +1,50 @@
+from src.helper import download_hugging_face_embeddings
+from langchain_pinecone import PineconeVectorStore
+from langchain.prompts import PromptTemplate
+from langchain_openai import OpenAI
+from langchain.chains import RetrievalQA
+from dotenv import load_dotenv
+from src.prompt import prompt_template
+import os
+
+load_dotenv()
+
+PINECONE_API_KEY = os.environ.get("PINECONE_API_KEY")
+PINECONE_API_ENV = os.environ.get("PINECONE_API_ENV")
+
+
+embeddings = download_hugging_face_embeddings()
+index_name = "llm-chatbot"
+
+# Initializing the Pinecone
+docsearch = PineconeVectorStore.from_existing_index(index_name, embeddings)
+
+
+PROMPT = PromptTemplate(
+    template=prompt_template, input_variables=["context", "question"]
+)
+
+chain_type_kwargs = {"prompt": PROMPT}
+
+current_dir = os.getcwd()
+llm = OpenAI()
+
+
+qa = RetrievalQA.from_chain_type(
+    llm=llm,
+    chain_type="stuff",
+    retriever=docsearch.as_retriever(search_kwargs={"k": 2}),
+    return_source_documents=True,
+    chain_type_kwargs=chain_type_kwargs,
+    verbose=True,
+)
+
+
+def openai_call(input):
+    result = qa.invoke({"query": input})
+    return str(result["result"])
+
+
+if __name__ == "__main__":
+    msg = "If a previous owner of a land had allowed a neighbour or neighbour to walk or drive over his land in a shortcut and this has been going on for say a decade or so can I as the new owner stop them now from using the shortcut?"
+    print(f"response: {openai_call(msg)}")

prompt.py

@@ -0,0 +1,16 @@
+prompt_template = """
+As a seasoned legal advisor, you possess deep knowledge of legal intricacies and are skilled in referencing relevant laws and regulations. Users will seek guidance on various legal matters.
+
+If a question falls outside the scope of legal expertise, kindly inform the user that your specialization is limited to legal advice.
+
+In cases where you're uncertain of the answer, it's important to uphold integrity by admitting 'I don't know' rather than providing potentially erroneous information.
+
+Below is a snippet of context from the relevant section of the constitution, although it will not be disclosed to users.
+
+Context: {context}
+Question: {question}
+
+Your response should consist solely of helpful advice without any extraneous details.
+
+Helpful advice:
+"""