# app.py
import os
import subprocess
import glob
import re
import traceback
import gradio as gr
from openai import OpenAI

# Load OpenAI key from environment (Hugging Face Spaces -> Settings -> Secrets)
openai_api_key = os.getenv("OPENAI_API_KEY")
openai = OpenAI(api_key=openai_api_key)

# def download_audio(youtube_url):
#     try:
#         output_template = "/tmp/downloaded_audio.%(ext)s"

#         # Remove any old files
#         for f in glob.glob("/tmp/downloaded_audio.*"):
#             os.remove(f)

#         command = [
#             "yt-dlp", "-f", "bestaudio",
#             "--extract-audio", "--audio-format", "mp3",
#             "--audio-quality", "0",
#             "-o", output_template,
#             youtube_url
#         ]

#         result = subprocess.run(command, capture_output=True, text=True)
#         print("stdout:\n", result.stdout)
#         print("stderr:\n", result.stderr)

#         if result.returncode != 0:
#             raise RuntimeError(f"yt-dlp failed: {result.stderr}")

#         files = glob.glob("/tmp/downloaded_audio.*")
#         if not files:
#             raise FileNotFoundError("No audio file downloaded.")

#         return files[0]
#     except Exception as e:
#         raise RuntimeError(f"Download error: {e}")

from pytube import YouTube

def clean_youtube_url(url):
    match = re.search(r"(?:v=|shorts/)([a-zA-Z0-9_-]{11})", url)
    video_id = match.group(1) if match else None
    return f"https://www.youtube.com/watch?v={video_id}" if video_id else None

def download_audio(youtube_url):
    try:
        print(f"▶️ Original URL: {youtube_url}")
        output_template = "/tmp/downloaded_audio.%(ext)s"

        # Cleanup old files
        for f in glob.glob("/tmp/downloaded_audio.*"):
            os.remove(f)

        # ✅ Try yt-dlp first
        command = [
            "yt-dlp",
            "-f", "bestaudio",
            "--extract-audio", "--audio-format", "mp3",
            "--audio-quality", "0",
            "-o", output_template,
            youtube_url
        ]
        print("📡 Running yt-dlp...")
        result = subprocess.run(command, capture_output=True, text=True)
        print("📜 yt-dlp stdout:", result.stdout)
        print("🐞 yt-dlp stderr:", result.stderr)

        if result.returncode == 0:
            files = glob.glob("/tmp/downloaded_audio.*")
            if files:
                print("✅ yt-dlp success.")
                return files[0]

        # 🔁 Fallback: try pytube with cleaned URL
        print("🔁 yt-dlp failed. Trying pytube...")
        clean_url = clean_youtube_url(youtube_url)
        if not clean_url:
            raise ValueError("Unable to extract video ID for pytube fallback.")

        print(f"🧽 Cleaned URL for pytube: {clean_url}")
        yt = YouTube(clean_url)
        stream = yt.streams.filter(only_audio=True).first()
        if not stream:
            raise ValueError("No audio stream found via pytube.")

        output_path = "/tmp/fallback_audio.mp4"
        stream.download(filename=output_path)
        print("✅ pytube download success.")
        return output_path

    except Exception as e:
        print("❌ Final Download error:", e)
        raise RuntimeError(f"Download error: {e}")

def transcribe_audio(file_path):
    try:
        with open(file_path, "rb") as f:
            result = openai.audio.transcriptions.create(
                model="whisper-1",
                file=f,
                response_format="verbose_json"
            )
        return result["text"], result["language"]
    except Exception as e:
        raise RuntimeError(f"Transcription error: {e}")

def summarize_text(text, lang):
    lang = lang.lower()
    if lang.startswith("zh") or "chinese" in lang:
        prompt = "你是一位聰明的助手，能夠用繁體中文清楚且完整地摘要影片內容。"
    elif lang.startswith("ja") or "japanese" in lang:
        prompt = "あなたは日本語で要点を簡潔かつ分かりやすく要約する有能なアシスタントです。"
    else:
        prompt = "You are a helpful assistant that summarizes transcripts clearly and concisely."

    response = openai.chat.completions.create(
        model="gpt-3.5-turbo",
        messages=[
            {"role": "system", "content": prompt},
            {"role": "user", "content": f"Summarize the following transcript:\n\n{text}"}
        ]
    )

    summary = response.choices[0].message.content
    debug_info = f"🌐 Detected Language: {lang}\n🧠 Prompt Used: {prompt}"
    return summary, debug_info

def extract_video_id(url):
    match = re.search(r"(?:v=|shorts/)([a-zA-Z0-9_-]{11})", url)
    return match.group(1) if match else None

def full_process(youtube_url):
    try:
        video_id = extract_video_id(youtube_url)
        thumbnail_url = f"https://img.youtube.com/vi/{video_id}/maxresdefault.jpg" if video_id else None
        audio_path = download_audio(youtube_url)
        transcript, lang = transcribe_audio(audio_path)
        summary, debug = summarize_text(transcript, lang)
        return summary, debug, thumbnail_url
    except Exception as e:
        return f"❌ Error: {str(e)}", "", None

with gr.Blocks() as demo:
    gr.Markdown("## 🧠 YouTube AI Summarizer\nEasily extract summaries from YouTube videos using Whisper + GPT. Supports English/Japanese/Chinese.")

    with gr.Row():
        youtube_input = gr.Textbox(label="🎥 Enter YouTube Video Link")
        submit_btn = gr.Button("🔍 Summarize")

    summary_output = gr.Textbox(label="📝 AI Video Summary", lines=6)
    info_output = gr.Textbox(label="📄 Language & Model Info", lines=4)
    thumbnail_output = gr.Image(label="🎞️ Video Thumbnail", visible=True)

    submit_btn.click(fn=full_process, inputs=youtube_input, outputs=[summary_output, info_output, thumbnail_output])

if __name__ == "__main__":
    demo.launch()