app.py

import openai
import gradio as gr
import os
import time
from datetime import datetime
from pydub import AudioSegment
import io
import yt_dlp
import PyPDF2

# Define your OpenAI API key using environment variable (recommended for Hugging Face Spaces)
openai.api_key = os.getenv("OPENAI_API_KEY")

# Function to convert audio file to text using OpenAI's Whisper
def transcribe_audio(audio_file):
    try:
        # Load the audio file
        audio = AudioSegment.from_file(audio_file)
        # Export as WAV, which Whisper accepts
        buffer = io.BytesIO()
        audio.export(buffer, format="wav")
        buffer.seek(0)

        response = openai.Audio.transcribe(
            "whisper-1",
            file=buffer,
            model='whisper',
            response_format='verbose_json'
        )
        return response
    except Exception as e:
        print(f"Error in transcribe_audio: {str(e)}")
        raise

# Function to download audio from YouTube URL
def download_youtube_audio(url):
    try:
        ydl_opts = {
            'format': 'bestaudio/best',
            'outtmpl': 'downloaded_audio.%(ext)s',
            'postprocessors': [{
                'key': 'FFmpegExtractAudio',
                'preferredcodec': 'mp3',
                'preferredquality': '192',
            }],
        }
        with yt_dlp.YoutubeDL(ydl_opts) as ydl:
            ydl.download([url])
        return 'downloaded_audio.mp3'
    except Exception as e:
        print(f"Error in download_youtube_audio: {str(e)}")
        raise

# Function to extract text from PDF
def extract_text_from_pdf(pdf_file):
    try:
        pdf_reader = PyPDF2.PdfFileReader(pdf_file)
        text = ""
        for page_num in range(pdf_reader.numPages):
            text += pdf_reader.getPage(page_num).extract_text() + "\n"
        return text
    except Exception as e:
        print(f"Error in extract_text_from_pdf: {str(e)}")
        raise

# Function to generate summarised lecture notes using GPT-3.5
def generate_summary(transcription):
    try:
        transcription_text = "\n".join([f"{segment['start']:.2f}-{segment['end']:.2f}: {segment['text']}" for segment in transcription['segments']])
        prompt = f"""
        You are an intelligent assistant that will summarize the transcription below.
        The transcription text is:
        {transcription_text}

        Summarize the content into 1000 tokens or less, focusing on the key topics and main points.
        """
        response = openai.ChatCompletion.create(
            model="gpt-3.5-turbo",
            messages=[
                {"role": "system", "content": "You are an expert summarizer."},
                {"role": "user", "content": prompt}
            ]
        )
        summary = response['choices'][0]['message']['content']
        return summary
    except Exception as e:
        print(f"Error in generate_summary: {str(e)}")
        raise

# Define the main function to handle transcription and summary generation
def process_lecture(input_type, audio_input, pdf_input, youtube_input, lesson_plan):
    transcription = ""
    try:
        if input_type == "Audio File" and audio_input is not None:
            transcription = transcribe_audio(audio_input)
        elif input_type == "YouTube URL" and youtube_input:
            audio_path = download_youtube_audio(youtube_input)
            with open(audio_path, "rb") as f:
                transcription = transcribe_audio(f)
        elif input_type == "PDF Document" and pdf_input is not None:
            transcription = extract_text_from_pdf(pdf_input)
    except Exception as e:
        return f"Error during processing: {str(e)}", "No summary available."
    
    if transcription:
        try:
            transcription_text = "\n".join([f"{segment['start']:.2f}-{segment['end']:.2f}: {segment['text']}" for segment in transcription['segments']])
            summary = generate_summary(transcription)
            return transcription_text, summary
        except Exception as e:
            return f"Transcription generated, but error during summary generation: {str(e)}", "No summary available."
    else:
        return "No transcription available.", "No summary available."

# Set up Gradio interface
with gr.Blocks() as demo:
    gr.Markdown("# Lecture Notes Generation Tool\nConvert your lecture recordings, PDFs, or YouTube links into detailed and structured notes easily!")

    with gr.Tab("Upload"):
        input_type = gr.Radio(label="Select Input Type", choices=["Audio File", "PDF Document", "YouTube URL"], type="value")
        audio_input = gr.Audio(label="Upload your lecture audio file (mp3/wav)", visible=False)
        pdf_input = gr.File(label="Upload PDF Document", visible=False)
        youtube_input = gr.Textbox(label="Enter YouTube URL", visible=False)
        lesson_plan_input = gr.Textbox(label="Enter the lesson plan for context", lines=5)

        def update_visibility(input_choice):
            return gr.update(visible=input_choice == "Audio File"), gr.update(visible=input_choice == "PDF Document"), gr.update(visible=input_choice == "YouTube URL")

        input_type.change(fn=update_visibility, inputs=[input_type], outputs=[audio_input, pdf_input, youtube_input])

        submit_btn = gr.Button("Generate Notes")

        with gr.Tab("Transcription and Notes"):
            transcription_output = gr.Textbox(label="Lecture Transcription with Timestamps", interactive=False)
            summary_output = gr.Textbox(label="Summarized Lecture Notes", interactive=False)

    submit_btn.click(fn=process_lecture, inputs=[input_type, audio_input, pdf_input, youtube_input, lesson_plan_input], outputs=[transcription_output, summary_output])

# Launch the interface
if __name__ == "__main__":
    demo.launch(share=True)