app.py

import streamlit as st
from dotenv import load_dotenv
from PyPDF2 import PdfReader
from langchain.text_splitter import CharacterTextSplitter
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.vectorstores import FAISS
# 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
import os
# from transformers import T5Tokenizer, T5ForConditionalGeneration
# from langchain.callbacks import get_openai_callback
hub_token = os.environ["HUGGINGFACE_HUB_TOKEN"]

def split_pdfs(pdf_docs):
    """Splits a list of PDF documents into smaller chunks.
    
    Args:
    pdf_docs: A list of PDF documents.
    
    Returns:
    A list of lists of PDF documents, where each sublist contains a smaller chunk of the original PDF documents.
    """
    
    pdf_chunks = []
    # Split the PDF document into pages.
    pdf_reader = PdfReader(pdf_doc)
    pdf_pages = pdf_reader.pages

    # Split the PDF pages into chunks.
    pdf_chunks.append([])
    for pdf_page in pdf_pages:
      # Add the PDF page to the current chunk.
      pdf_chunks[-1].append(pdf_page)

      # If the chunk is too large, start a new chunk.
      if len(pdf_chunks[-1]) >= 10:
        pdf_chunks.append([])

    return pdf_chunks

def generate_response(pdf_chunks, llm_model):
  """Generates a response to a query using a list of PDF documents and an LLM model.

  Args:
    pdf_chunks: A list of lists of PDF documents, where each sublist contains a smaller chunk of the original PDF documents.
    llm_model: An LLM model.

  Returns:
    A response to the query.
  """

  # Generate a summary of each PDF chunk.
  pdf_summaries = []
  for pdf_chunk in pdf_chunks:
    # Generate a summary of the PDF chunk.
    pdf_summary = llm_model.generate(
      prompt=f"Summarize the following text:\n{get_pdf_text(pdf_chunk)}",
      max_new_tokens=100
    )

    # Add the summary to the list of summaries.
    pdf_summaries.append(pdf_summary)

  # Generate a response to the query using the summaries of the PDF chunks.
  response = llm_model.generate(
    prompt=f"Answer the following question using the following summaries:\n{get_text_chunks(pdf_summaries)}\n\nQuestion:",
    max_new_tokens=200
  )

  return response

def main():
  load_dotenv()
  st.set_page_config(page_title="Chat with multiple PDFs", page_icon=":books:")
  st.write(css, unsafe_allow_html=True)

  # Load the LLM model.
  llm_model = HuggingFaceHub(repo_id="mistralai/Mistral-7B-v0.1", huggingfacehub_api_token=hub_token, verbose=True)

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

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

  st.header("Chat with multiple PDFs :books:")
  user_question = st.text_input("Ask a question about your documents:")

  # If the user asked a question, generate a response.
  if user_question:
    # Split the PDF documents into smaller chunks.
    pdf_chunks = split_pdfs("Geeta.pdf")

    # Generate a response to the query.
    response = generate_response(pdf_chunks, llm_model)

    st.write(response)



main()