app.py

import streamlit as st
import PyPDF2
from langchain.embeddings import SentenceTransformerEmbeddings
from langchain.vectorstores import FAISS
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.chains import RetrievalQA
from langchain.llms import HuggingFacePipeline
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline

st.set_page_config(page_title="PDF QA App (CPU)", layout="wide")
st.title("📘 Ask Questions from Uploaded PDFs (Free & CPU Friendly)")

uploaded_files = st.file_uploader("Upload multiple PDF files", type=["pdf"], accept_multiple_files=True)

@st.cache_resource
def load_llm():
    model_id = "google/flan-t5-base"
    tokenizer = AutoTokenizer.from_pretrained(model_id)
    model = AutoModelForSeq2SeqLM.from_pretrained(model_id)
    pipe = pipeline("text2text-generation", model=model, tokenizer=tokenizer)
    return HuggingFacePipeline(pipeline=pipe)

if uploaded_files:
    st.info("Reading and processing PDFs...")
    all_text = ""
    for file in uploaded_files:
        reader = PyPDF2.PdfReader(file)
        for page in reader.pages:
            text = page.extract_text()
            if text:
                all_text += text

    text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
    texts = text_splitter.split_text(all_text)

    embeddings = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2")
    db = FAISS.from_texts(texts, embeddings)

    retriever = db.as_retriever()
    llm = load_llm()
    qa_chain = RetrievalQA.from_chain_type(llm=llm, retriever=retriever)

    question = st.text_input("Ask a question based on the uploaded PDFs:")
    if question:
        with st.spinner("Generating answer..."):
            answer = qa_chain.run(question)
            st.success(answer)