app.py

import gradio as gr
import os
import tempfile
import yt_dlp
import subprocess
import requests
from huggingface_hub import InferenceClient

# ---------------------------
# Model setup
# ---------------------------
WHISPER_MODEL = "openai/whisper-large-v3"
EN_SUMMARY_MODEL = "facebook/bart-large-cnn"
UR_SUMMARY_MODEL = "openai/gpt-oss-120b"

client = InferenceClient()

# ---------------------------
# Helper Functions
# ---------------------------
def download_youtube_audio(url: str):
    """Download YouTube audio as mp3"""
    try:
        tmp = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3")
        ydl_opts = {
            "format": "bestaudio/best",
            "outtmpl": tmp.name,
            "quiet": True,
            "postprocessors": [{"key": "FFmpegExtractAudio", "preferredcodec": "mp3"}],
        }
        with yt_dlp.YoutubeDL(ydl_opts) as ydl:
            ydl.download([url])
        return tmp.name
    except Exception as e:
        raise RuntimeError(f"❌ Error downloading YouTube audio: {e}")

def convert_to_wav(audio_path: str):
    """Convert any audio to 16kHz mono WAV"""
    wav_path = tempfile.mktemp(suffix=".wav")
    try:
        subprocess.run(
            ["ffmpeg", "-y", "-i", audio_path, "-ar", "16000", "-ac", "1", wav_path],
            check=True,
            capture_output=True,
        )
        return wav_path
    except subprocess.CalledProcessError as e:
        raise RuntimeError(f"❌ ffmpeg conversion failed: {e}")

# ---------------------------
# Transcription
# ---------------------------
def transcribe_audio(audio_file=None, youtube_url=None):
    try:
        if youtube_url and youtube_url.strip():
            audio_path = download_youtube_audio(youtube_url)
        elif audio_file:
            audio_path = audio_file
        else:
            return "❌ Please record, upload, or provide a YouTube link."

        wav_path = convert_to_wav(audio_path)

        headers = {
            "Authorization": f"Bearer {os.environ.get('HUGGINGFACE_API_TOKEN','')}",
            "Content-Type": "audio/wav"
        }
        api_url = f"https://api-inference.huggingface.co/models/{WHISPER_MODEL}"

        with open(wav_path, "rb") as f:
            response = requests.post(api_url, headers=headers, data=f.read())

        if response.status_code != 200:
            raise RuntimeError(f"HF API error: {response.text}")

        data = response.json()
        if isinstance(data, dict) and "text" in data:
            return data["text"]
        elif isinstance(data, list) and "text" in data[0]:
            return data[0]["text"]
        else:
            return str(data)
    except Exception as e:
        return f"❌ Error during transcription: {e}"

# ---------------------------
# Summarization
# ---------------------------
def summarize_text(text, language):
    try:
        if language == "English":
            result = client.summarization(model=EN_SUMMARY_MODEL, inputs=text, max_new_tokens=1024)
            return result["summary_text"]
        else:
            resp = client.text_generation(
                model=UR_SUMMARY_MODEL,
                prompt=f"مندرجہ ذیل انگریزی متن کا جامع اردو خلاصہ لکھیں:\n\n{text}",
                max_new_tokens=2048,
            )
            return resp
    except Exception as e:
        return f"❌ Summarization failed: {e}"

# ---------------------------
# Tutorial Generator
# ---------------------------
def generate_tutorial(transcription, summary, language):
    try:
        prompt = (
            f"Create a comprehensive, beginner-friendly tutorial in {language} "
            f"based on the following transcript and summary.\n\n"
            f"Transcript:\n{transcription}\n\nSummary:\n{summary}"
        )
        model = UR_SUMMARY_MODEL if language == "Urdu" else EN_SUMMARY_MODEL
        result = client.text_generation(model=model, prompt=prompt, max_new_tokens=2200)
        return result
    except Exception as e:
        return f"❌ Error generating tutorial: {e}"

# ---------------------------
# Gradio UI
# ---------------------------
with gr.Blocks(title="🎙️ Smart Transcriber & Tutorial Maker") as demo:
    gr.Markdown("## 🎧 Smart Transcriber, Summarizer & Tutorial Creator")

    with gr.Row():
        audio_input = gr.Audio(sources=["microphone", "upload"], type="filepath", label="🎙️ Record or Upload Audio")
        yt_input = gr.Textbox(label="🎥 Or paste YouTube link")

    trans_btn = gr.Button("🚀 Transcribe")
    transcript_output = gr.Textbox(label="📝 Transcription", lines=8)

    with gr.Row():
        lang_choice = gr.Radio(["English", "Urdu"], label="Select Summary Language", value="English")

    sum_btn = gr.Button("🧠 Generate Detailed Summary")
    summary_output = gr.Textbox(label="📋 Summary", lines=10)

    tut_btn = gr.Button("📘 Create Beginner Tutorial")
    tutorial_output = gr.Textbox(label="🎓 Tutorial", lines=12)

    # Button actions
    trans_btn.click(transcribe_audio, [audio_input, yt_input], transcript_output)
    sum_btn.click(summarize_text, [transcript_output, lang_choice], summary_output)
    tut_btn.click(generate_tutorial, [transcript_output, summary_output, lang_choice], tutorial_output)

demo.launch()