app.py

import os
import streamlit as st
from PyPDF2 import PdfReader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.embeddings import HuggingFaceEmbeddings

from langchain.vectorstores import FAISS
from langchain_google_genai import ChatGoogleGenerativeAI
from langchain.chains.question_answering import load_qa_chain
from langchain.prompts import PromptTemplate
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
  

# Load a Hugging Face model (e.g., LLaMA or Falcon)
model_name = "mixedbread-ai/mxbai-embed-2d-large-v1"  # Replace with your preferred model
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")



def get_pdf_text(pdf_docs):
    text = ""
    for pdf in pdf_docs:
        pdf_reader = PdfReader(pdf)
        for page in pdf_reader.pages:
            text += page.extract_text()
    return text

# chuck_size = 1000, chunk_overlap = 200 (for shorted PDFs)
def get_text_chunks(text):
  text_splitter= RecursiveCharacterTextSplitter(
    chunk_size=10000,
    chunk_overlap=1000,
    # length_function=len
  )
  chunks=text_splitter.split_text(text)
  return chunks

# Converting into Vector data/store (can also be stored)
def get_vector_store(text_chunks):
  # embeddings = GoogleGenerativeAIEmbeddings(model='embedding-gecko-001')
  # embeddings = GoogleGenerativeAIEmbeddings(model='models/embedding-001')
  embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
  vector_store = FAISS.from_texts(texts=text_chunks, embedding=embeddings)
  vector_store.save_local("faiss_index")
  # return vector_store


  
def chat_with_huggingface(context, query):
    prompt_template = """
    Answer the query as detailed as possible from the provided context. 
    If the answer is not in the context, just say, "Answer is not available in the provided documents". 
    Context: {context}
    Query: {query}
    Answer:
    """
    inputs = tokenizer(prompt_template, return_tensors="pt").to(model.device)
    outputs = model.generate(**inputs, max_length=500, temperature=0.3)
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

def get_conversation_chain():
    def huggingface_chain(inputs):
        context = inputs["input_documents"][0].page_content  # Extract context from FAISS search
        query = inputs["question"]
        return {"output_text": chat_with_huggingface(context, query)}

    return huggingface_chain

def user_input(user_question):
  # embeddings = GoogleGenerativeAIEmbeddings(model='embedding-gecko-001')
  embeddings = GoogleGenerativeAIEmbeddings(model='models/embedding-001')

  # Loading the embeddings
  new_db = FAISS.load_local("faiss_index", embeddings, allow_dangerous_deserialization=True)
  docs = new_db.similarity_search(user_question)

  chain=get_conversation_chain()

  response = chain(
      {"input_documents": docs, "question": user_question}
      , return_only_outputs=True)
  
  print(response)
  st.write("Reply: ", response["output_text"])

# Frontend page Processor
def main():
  st.set_page_config(page_title="PDF Chatbot")
  st.header("PDF Chatbot made for Pooja")

  user_question = st.text_input("Puchiye kuch apne documents se:")

  if user_question:
    user_input(user_question)

  with st.sidebar:
    st.title("Menu:")
    pdf_docs = st.file_uploader(
        "Apne PDFs yaha pe upload karo then click on 'Process'", accept_multiple_files=True)
    if st.button("Submit & Process"):
      with st.spinner("Ruko Padh raha hu..."):
        raw_text = get_pdf_text(pdf_docs)
        text_chunks = get_text_chunks(raw_text)
        get_vector_store(text_chunks)
        st.success("Saare documents padh liya. Ab swaal pucho 😤")


if __name__ == '__main__':
  main()