utils/rag_chain.py

import os
from langchain_openai import ChatOpenAI
from langchain.chains import create_retrieval_chain, create_history_aware_retriever
from langchain.chains.combine_documents import create_stuff_documents_chain
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
from langchain_core.runnables.history import RunnableWithMessageHistory
from langchain_community.chat_message_histories import ChatMessageHistory

def get_model():
    return ChatOpenAI(api_key=os.getenv("OPEN_API_KEY"))

def create_contextualize_q_prompt():
    contextualize_q_system_prompt = (
        "Given a chat history and the latest user question "
        "which might reference context in the chat history, "
        "formulate a standalone question which can be understood "
        "without the chat history. Do NOT answer the question, "
        "just reformulate it if needed and otherwise return it as is."
    )
    return ChatPromptTemplate.from_messages([
        ("system", contextualize_q_system_prompt),
        MessagesPlaceholder("chat_history"),
        ("human", "{input}"),
    ])

def create_qa_prompt():
    qa_system_prompt = """You are an assistant for question-answering tasks. Use the following pieces of retrieved context to answer the question. \
    The retrieved content belongs to subject textbooks present. You will receive different chunks, each of which belongs to a single page of a textbook. \
    Using the chunk given, think logically and answer the questions from the user. \
    If you are not able to identify the relevant information regarding to the user's question in the retrieved chunks, then just return 'No data found'.\
    Use three sentences maximum and keep the answer concise. \

    {context}"""
    return ChatPromptTemplate.from_messages([
        ("system", qa_system_prompt),
        MessagesPlaceholder("chat_history"),
        ("human", "{input}"),
    ])

def create_rag_chain(model, retriever):
    contextualize_q_prompt = create_contextualize_q_prompt()
    qa_prompt = create_qa_prompt()

    history_aware_retriever = create_history_aware_retriever(model, retriever, contextualize_q_prompt)
    question_answer_chain = create_stuff_documents_chain(model, qa_prompt)
    return create_retrieval_chain(history_aware_retriever, question_answer_chain)

def get_conversational_rag_chain(rag_chain):
    store = {}

    def get_session_history(session_id: str):
        if session_id not in store:
            store[session_id] = ChatMessageHistory()
        return store[session_id]

    return RunnableWithMessageHistory(
        rag_chain,
        get_session_history,
        input_messages_key="input",
        history_messages_key="chat_history",
        output_messages_key="answer",
    )