# main.py import streamlit as st from dotenv import load_dotenv from config import Config from transcript_utils import extract_video_id, fetch_transcript from vector_utils import build_vector_store, get_retriever from qa_pipeline import run_qa # Load environment variables load_dotenv() st.set_page_config(page_title="YouTube Video Chat-QA", layout="centered") st.title("YouTube Video Chat-QA App") # Initialize session state if 'ready' not in st.session_state: st.session_state.ready = False if 'messages' not in st.session_state: st.session_state.messages = [] # Setup section for video link with st.expander("Setup Video", expanded=not st.session_state.ready): url = st.text_input("YouTube Video Link", key='input_url') if st.button("Submit Link"): vid = extract_video_id(url) if not vid: st.error("Invalid YouTube URL.") else: st.session_state.video_id = vid st.session_state.ready = True # Chat interface after video is ready def chat_interface(): st.video(f"https://www.youtube.com/watch?v={st.session_state.video_id}") # Input new message new_q = st.chat_input("Ask a question about this video...") if new_q: st.session_state.messages.append({'role': 'user', 'content': new_q}) transcript = fetch_transcript(st.session_state.video_id, languages=["en", "hi"]) if not transcript: st.session_state.messages.append({'role': 'assistant', 'content': "Transcript not available."}) else: vector_store = build_vector_store(transcript, None, Config.EMBEDDING_MODEL) retriever = get_retriever(vector_store) answer = run_qa(retriever, new_q) st.session_state.messages.append({'role': 'assistant', 'content': answer}) # Display chat history for msg in st.session_state.messages: st.chat_message(msg['role']).write(msg['content']) if st.button("Exit Chat"): for key in ['ready', 'video_id', 'messages', 'input_url']: if key in st.session_state: del st.session_state[key] if st.session_state.ready: chat_interface()