app.py

import streamlit as st
import os
import json
from utils.ingestion import DocumentProcessor
from utils.llm import LLMProcessor
from utils.qa import QAEngine

st.set_page_config(page_title="AI-Powered Document QA", layout="wide")
st.title("📄 AI-Powered Document QA")

# Initialize processors
document_processor = DocumentProcessor()
llm_processor = LLMProcessor()
qa_engine = QAEngine()

# Ensure temp directory exists
os.makedirs("temp", exist_ok=True)

# Sidebar for file upload
st.sidebar.header("📂 Upload a PDF")
uploaded_file = st.sidebar.file_uploader("Choose a PDF file", type=["pdf"])

# Document upload & processing
if uploaded_file:
    pdf_path = os.path.join("temp", uploaded_file.name)
    
    with open(pdf_path, "wb") as f:
        f.write(uploaded_file.read())

    st.sidebar.success("✅ File uploaded successfully!")

    with st.spinner("🔄 Processing document..."):
        document_processor.process_document(pdf_path)

    st.sidebar.success("✅ Document processed successfully!")
    st.session_state["document_uploaded"] = True
else:
    st.session_state["document_uploaded"] = False

# Divider between sections
st.markdown("---")

# Q&A Section
st.header("🔍 Ask a Question")

question = st.text_input("Ask a question:", placeholder="What are the key insights?")

if st.button("💡 Get Answer"):
    if question:
        with st.spinner("🧠 Generating response..."):
            if st.session_state["document_uploaded"]:
                # Use document-based QA if a file is uploaded
                answer = qa_engine.query(question)
            else:
                # Use AI-based response if no document is uploaded
                answer = llm_processor.generate_answer("", question)
                st.warning("⚠️ No document uploaded. This response is generated from general AI knowledge and may not be document-specific.")

        st.subheader("📝 Answer:")
        st.write(answer)
    else:
        st.warning("⚠️ Please enter a question.")

st.markdown("---")
st.caption("🤖 Powered by ChromaDB + Groq LLM | Built with ❤️ using Streamlit")