app.py

import streamlit as st
from PyPDF2 import PdfReader
from sentence_transformers import SentenceTransformer
from transformers import pipeline
import faiss
import numpy as np

# Load models
embedder = SentenceTransformer('all-MiniLM-L6-v2')
qa_pipeline = pipeline('question-answering', model='distilbert-base-uncased-distilled-squad')

st.set_page_config(page_title="QuickLit - AI Research Assistant")
st.title("📚 QuickLit: Literature Q&A Assistant")

# File upload
uploaded_file = st.file_uploader("Upload a research paper (PDF)", type=["pdf"])

if uploaded_file:
    reader = PdfReader(uploaded_file)
    full_text = ""
    for page in reader.pages:
        full_text += page.extract_text()

    # Split text into chunks
    sentences = full_text.split('. ')
    chunks = ['. '.join(sentences[i:i+3]) for i in range(0, len(sentences), 3)]

    # Generate embeddings
    st.info("🔎 Generating embeddings...")
    embeddings = embedder.encode(chunks)
    
    # Create FAISS index
    index = faiss.IndexFlatL2(embeddings[0].shape[0])
    index.add(np.array(embeddings))

    # Input question
    question = st.text_input("Ask a question about the paper:")

    if question:
        # Embed the question
        q_embedding = embedder.encode([question])

        # Retrieve top 3 similar chunks
        D, I = index.search(np.array(q_embedding), k=3)
        retrieved_contexts = [chunks[i] for i in I[0]]
        context = " ".join(retrieved_contexts)

        # Answer using transformer
        st.info("💡 Answering with AI...")
        answer = qa_pipeline(question=question, context=context)
        st.success(f"**Answer:** {answer['answer']}")