app.py

# import streamlit as st
# import os
# import sys
# from logger import logging
# from exception import CustomException
# from dotenv import load_dotenv
# import os
# from components.transcript_retriever import TranscriptRetriever
# from components.preprocessor import Preprocessor
# from components.qa_engine import QAEngine


# load_dotenv()

# def get_summary(video_url, question):
#     api_key = os.getenv('GEMINI_API_KEY')
#     if not api_key:
#         raise RuntimeError("GEMINI_API_KEY not found in environment or .env file.")

#     # Step 1: Retrieve transcript
#     retriever = TranscriptRetriever(video_url)
#     raw_transcript = retriever.fetch_transcript(video_url)

    
#     preprocessor = Preprocessor()
#     cleaned = preprocessor.clean_transcript(raw_transcript)

#     def determine_chunk_size(transcript):
#         length = len(transcript)
#         if length < 2000:
#             return length
#         elif length < 10000:
#             return 3000
#         elif length < 40000:
#             return 8000
#         else:
#             return 12000

#     chunk_size = determine_chunk_size(cleaned)
#     chunks = preprocessor.chunk_transcript(cleaned, chunk_size=chunk_size)

#     engine = QAEngine(api_key=api_key)

#     # Determine mode: summarize if question is blank or summarization requested
#     if (not question) or "summary" in question.lower() or "summarize" in question.lower():
#         # Map-reduce summarization
#         summaries = [engine.answer_question(chunk, "Summarize this transcript section.") for chunk in chunks]
#         combined = "\n\n".join(summaries)
#         reduce_prompt = (
#             f"Given these summaries/explanations from different parts of the video, "
#             f"write a single, clear summary covering all key points:\n\n{combined}"
#         )
#         final_summary = engine.summarize_transcript(reduce_prompt)
#         return final_summary
#     else:
#         # Per-chunk Q&A
#         answers = [engine.answer_question(chunk, question) for chunk in chunks]
#         return "\n\n".join([f"Chunk {i+1}:\n{ans}" for i, ans in enumerate(answers)])


# st.set_page_config(page_title="YouTube Transcript Q&A & Summarizer")


# ICON_PATH = "yt_logo.png"
# st.image(ICON_PATH, width=100)

# st.title(" YouTube Transcript Q&A & Summarizer")
# st.markdown("Enter YouTube video link and choose an action:")

# video_url = st.text_input("YouTube Video Link", placeholder="Paste the URL here...")

# option = st.radio("Select Action:", ("Summarize Video", "Ask a Question"))

# question = ""
# if option == "Ask a Question":
#     question = st.text_input("Enter your question about this video:")

# if st.button("Go") and video_url:
#     with st.spinner("Processing..."):
#         try:
#             output_text = get_summary(video_url, question)
#         except Exception as e:
#             output_text = f"Error: {str(e)}"

#         st.subheader("Output:")
#         st.write(output_text)

#         st.download_button(
#             label="Download Output",
#             data=output_text,
#             file_name="output.txt",
#             mime="text/plain"
#         )

# st.markdown("Created by **Siddham Jain** (Student of Shiv Nadar IOE).")

import streamlit as st
import os
import sys
from logger import logging
from exception import CustomException
from dotenv import load_dotenv
from components.transcript_retriever import TranscriptRetriever
from components.preprocessor import Preprocessor
from components.qa_engine import QAEngine

# Load environment variables
load_dotenv()

def get_summary(video_url, question):
    api_key = os.getenv('GEMINI_API_KEY')
    if not api_key:
        raise RuntimeError("GEMINI_API_KEY not found in environment or .env file.")

    # Step 1: Retrieve transcript
    retriever = TranscriptRetriever(video_url)
    raw_transcript = retriever.fetch_transcript(video_url)

    # Step 2: Preprocess transcript
    preprocessor = Preprocessor()
    cleaned = preprocessor.clean_transcript(raw_transcript)

    def determine_chunk_size(transcript):
        length = len(transcript)
        if length < 2000:
            return length
        elif length < 10000:
            return 3000
        elif length < 40000:
            return 8000
        else:
            return 12000

    chunk_size = determine_chunk_size(cleaned)
    chunks = preprocessor.chunk_transcript(cleaned, chunk_size=chunk_size)

    engine = QAEngine(api_key=api_key)

    # --------- ALWAYS use map-reduce for any question (summary or specific) ---------
    per_chunk_answers = [engine.answer_question(chunk, question if question else "Summarize this transcript section.") for chunk in chunks]
    combined_text = "\n\n".join(per_chunk_answers)
    reduce_prompt = (
        f"Given these answers (from different parts of the video) to the question: '{question if question else 'Summarize the video'}', "
        f"combine everything into ONE comprehensive, concise answer—removing repetition and contradictions, and giving the user the best possible response:\n\n"
        f"{combined_text}"
    )
    final_answer = engine.summarize_transcript(reduce_prompt)
    return final_answer

# --- Streamlit UI ---
st.set_page_config(page_title="YouTube Transcript Q&A & Summarizer")
ICON_PATH = "yt_logo.png"
st.image(ICON_PATH, width=100)

st.title("YouTube Transcript Q&A & Summarizer")
st.markdown("Enter YouTube video link and choose an action:")

video_url = st.text_input("YouTube Video Link", placeholder="Paste the URL here...")

option = st.radio("Select Action:", ("Summarize Video", "Ask a Question"))

question = ""
if option == "Ask a Question":
    question = st.text_input("Enter your question about this video:")

if st.button("Go") and video_url:
    with st.spinner("Processing..."):
        try:
            output_text = get_summary(video_url, question)
        except Exception as e:
            output_text = f"Error: {str(e)}"

        st.subheader("Output:")
        st.write(output_text)

        st.download_button(
            label="Download Output",
            data=output_text,
            file_name="output.txt",
            mime="text/plain"
        )

st.markdown("Created by **Siddham Jain** (Student of Shiv Nadar IOE).")