app.py

import os
import requests
import streamlit as st
from youtube_transcript_api import YouTubeTranscriptApi
from langchain.text_splitter import RecursiveCharacterTextSplitter
from dotenv import load_dotenv

# Load environment variables from the .env file in the project directory
load_dotenv()

# Access environment variables
API_URL = os.getenv('HUGGING_FACE_API_URL')
API_KEY = os.getenv('HUGGING_FACE_API_KEY')

def get_transcript(youtube_url):
    video_id = youtube_url.split("v=")[-1]
    transcript_list = YouTubeTranscriptApi.list_transcripts(video_id)

    try:
        # Try to fetch the manual transcript
        transcript = transcript_list.find_manually_created_transcript()
        language_code = transcript.language_code  # Save the detected language
    except:
        try:
            # If no manual transcript is found, try fetching an auto-generated transcript in a supported language
            generated_transcripts = [trans for trans in transcript_list if trans.is_generated]
            transcript = generated_transcripts[0]
            language_code = transcript.language_code  # Save the detected language
        except:
            # If no auto-generated transcript is found, raise an exception
            raise Exception("No suitable transcript found.")

    full_transcript = " ".join([part['text'] for part in transcript.fetch()])
    return full_transcript, language_code  # Return both the transcript and detected language

def summarize_with_hugging_face(transcript, language_code, model_name='meta-llama/Meta-Llama-3-8B'):
    # Split the document if it's too long
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=2000, chunk_overlap=0)
    texts = text_splitter.split_text(transcript)
    text_to_summarize = " ".join(texts[:4])  # Adjust this as needed

    # Prepare the payload for summarization
    payload = {
        "inputs": {
            "prompt": f'''Summarize the following text in {language_code}.
            Text: {text_to_summarize}

            Add a title to the summary in {language_code}. 
            Include an INTRODUCTION, BULLET POINTS if possible, and a CONCLUSION in {language_code}.'''
        }
    }

    # Start summarizing using Hugging Face
    headers = {"Authorization": f"Bearer {API_KEY}"}
    response = requests.post(API_URL, headers=headers, json=payload)

    if response.status_code == 200:
        return response.json()["generated_text"]
    else:
        raise Exception("Summarization failed.")

def main():
    st.title('YouTube Video Summarizer')
    link = st.text_input('Enter the link of the YouTube video you want to summarize:')

    if st.button('Start'):
        if link:
            try:
                progress = st.progress(0)
                status_text = st.empty()

                status_text.text('Loading the transcript...')
                progress.progress(25)

                # Get both the transcript and language_code
                transcript, language_code = get_transcript(link)

                status_text.text(f'Creating summary...')
                progress.progress(75)

                summary = summarize_with_hugging_face(transcript, language_code)

                status_text.text('Summary:')
                st.markdown(summary)
                progress.progress(100)
            except Exception as e:
                st.write(str(e))
        else:
            st.write('Please enter a valid YouTube link.')

if __name__ == "__main__":
    main()