app.py

"""
Question Answering with Retrieval QA and LangChain Language Models featuring FAISS vector stores.
This script uses the LangChain Language Model API to answer questions using Retrieval QA 
and FAISS vector stores. It also uses the Mistral huggingface inference endpoint to 
generate responses.
"""

import os
import streamlit as st
from dotenv import load_dotenv
from PyPDF2 import PdfReader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.document_loaders import UnstructuredPDFLoader
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.vectorstores import Chroma
from langchain.chat_models import ChatOpenAI
from langchain.memory import ConversationBufferMemory
from langchain.chains import ConversationalRetrievalChain
from htmlTemplates import css, bot_template, user_template
from langchain.llms import HuggingFaceHub


def get_pdf_pages(pdf_docs):
    """
    Extract text from a list of PDF documents.

    Parameters
    ----------
    pdf_docs : list
        List of PDF documents to extract text from.

    Returns
    -------
    str
        Extracted text from all the PDF documents.

    """
    pages = []
    import tempfile

    with tempfile.TemporaryDirectory() as tmpdirname:
        for pdf in pdf_docs:
            pdf_path=os.path.join(tmpdirname,pdf.name)
            with open(pdf_path, "wb") as f:
               f.write(pdf.getbuffer())
        
            pdf_loader = UnstructuredPDFLoader(pdf_path)
            pdf_pages = pdf_loader.load_and_split()
            pages=pages+pdf_pages
    return pages


def get_text_chunks(pages):
    """
    Split the input text into chunks.

    Parameters
    ----------
    text : str
        The input text to be split.

    Returns
    -------
    list
        List of text chunks.

    """
    text_splitter = RecursiveCharacterTextSplitter(
        chunk_size=1024, chunk_overlap=64
    )
    texts = text_splitter.split_documents(pages)
    print(str(len(texts)))
    return texts


def get_vectorstore(text_chunks):
    """
    Generate a vector store from a list of text chunks using HuggingFace BgeEmbeddings.

    Parameters
    ----------
    text_chunks : list
        List of text chunks to be embedded.

    Returns
    -------
    FAISS
        A FAISS vector store containing the embeddings of the text chunks.

    """
    MODEL_NAME = "sentence-transformers/all-MiniLM-L6-v2"
    hf_embeddings = HuggingFaceEmbeddings(model_name=MODEL_NAME)
    vectorstore = Chroma.from_documents(text_chunks, hf_embeddings, persist_directory="db")
    return vectorstore


def get_conversation_chain(vectorstore):
    """
    Create a conversational retrieval chain using a vector store and a language model.

    Parameters
    ----------
    vectorstore : FAISS
        A FAISS vector store containing the embeddings of the text chunks.

    Returns
    -------
    ConversationalRetrievalChain
        A conversational retrieval chain for generating responses.

    """
    llm = HuggingFaceHub(
        repo_id="mistralai/Mixtral-8x7B-Instruct-v0.1",
        model_kwargs={"temperature": 0.5, "max_length": 1048},
    )
    # llm = ChatOpenAI(temperature=0, model="gpt-3.5-turbo-0613")

    memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)
    conversation_chain = ConversationalRetrievalChain.from_llm(
        llm=llm, retriever=vectorstore.as_retriever(), memory=memory
    )
    return conversation_chain


def handle_userinput(user_question):
    """
    Handle user input and generate a response using the conversational retrieval chain.
    Parameters
    ----------
    user_question : str
        The user's question.
    """
    response = st.session_state.conversation({"question": user_question})
    st.session_state.chat_history = response["chat_history"]

    for i, message in enumerate(st.session_state.chat_history):
        if i % 2 == 0:
            st.write("//_^ User: " + message.content)
        else:
            st.write("🤖 ChatBot: " + message.content)


def main():
    """
    Putting it all together.
    """
    st.set_page_config(
        page_title="Chat with a Bot that tries to answer questions about multiple PDFs",
        page_icon=":books:",
    )

    st.markdown("# Chat with a Bot")
    st.markdown("This bot tries to answer questions about multiple PDFs. Let the processing of the PDF finish before adding your question. 🙏🏾")

    st.write(css, unsafe_allow_html=True)

    # set huggingface hub token in st.text_input widget
    # then hide the input
    huggingface_token = st.text_input("Enter your HuggingFace Hub token", type="password", value="DNTClESFouRJbgsoxTzdLFzYfIlGSVsWvM")
    #openai_api_key = st.text_input("Enter your OpenAI API key", type="password")
    if not huggingface_token.startswith("hf_"):
        huggingface_token = "hf_" + huggingface_token
    # set this key as an environment variable
    os.environ["HUGGINGFACEHUB_API_TOKEN"] = huggingface_token
    #os.environ["OPENAI_API_KEY"] = openai_api_key

    if "chat_history" not in st.session_state:
        st.session_state.chat_history = None

    with st.sidebar:
        st.subheader("Your documents")
        pdf_docs = st.file_uploader(
            "Upload your PDFs here and click on 'Process'", accept_multiple_files=True
        )
        if st.button("Process"):
            with st.spinner("Processing"):
                # get the raw text
                pages = get_pdf_pages(pdf_docs)

                # get the text chunks
                text_chunks = get_text_chunks(pages)

                # create vector store
                vectorstore = get_vectorstore(text_chunks)

                # create conversation chain
                st.session_state.conversation = get_conversation_chain(vectorstore)
                print(st.session_state.conversation)

    st.header("Chat with a Bot 🤖🦾 that tries to answer questions about multiple PDFs :books:")
    user_question = st.text_input("Ask a question about your documents:")
    if user_question:
        handle_userinput(user_question)


if __name__ == "__main__":
    main()