# #!/usr/bin/env python3
# """
# Article Summarizer with Text-to-Speech
# Scrapes articles, summarizes with Qwen3-0.6B, and reads aloud with Kokoro TTS
# """

# import sys
# import torch
# import trafilatura
# import soundfile as sf
# import time
# from transformers import AutoModelForCausalLM, AutoTokenizer
# from kokoro import KPipeline

# # --- Part 1: Web Scraping Function ---

# def scrape_article_text(url: str) -> str | None:
#     """
#     Downloads a webpage and extracts the main article text, removing ads,
#     menus, and other boilerplate.
    
#     Args:
#         url: The URL of the article to scrape.

#     Returns:
#         The cleaned article text as a string, or None if it fails.
#     """
#     print(f"🌐 Scraping article from: {url}")
#     # fetch_url downloads the content of the URL
#     downloaded = trafilatura.fetch_url(url)
    
#     if downloaded is None:
#         print("❌ Error: Failed to download the article content.")
#         return None
        
#     # extract the main text, ignoring comments and tables for a cleaner summary
#     article_text = trafilatura.extract(downloaded, include_comments=False, include_tables=False)
    
#     if article_text:
#         print("✅ Successfully extracted article text.")
#         return article_text
#     else:
#         print("❌ Error: Could not find main article text on the page.")
#         return None

# # --- Part 2: Summarization Function ---

# def summarize_with_qwen(text: str, model, tokenizer) -> str:
#     """
#     Generates a summary for the given text using the Qwen3-0.6B model.

#     Args:
#         text: The article text to summarize.
#         model: The pre-loaded transformer model.
#         tokenizer: The pre-loaded tokenizer.

#     Returns:
#         The generated summary as a string.
#     """
#     print("🤖 Summarizing text with Qwen3-0.6B...")

#     # 1. Create a detailed prompt for the summarization task
#     prompt = f"""
#     Please provide a concise and clear summary of the following article. 
#     Focus on the main points, key findings, and conclusions. The summary should be 
#     easy to understand for someone who has not read the original text.

#     ARTICLE:
#     {text}
#     """
    
#     messages = [{"role": "user", "content": prompt}]

#     # 2. Apply the chat template. We set `enable_thinking=False` for direct summarization.
#     # This is more efficient than the default reasoning mode for this task.
#     text_input = tokenizer.apply_chat_template(
#         messages,
#         tokenize=False,
#         add_generation_prompt=True,
#         enable_thinking=False 
#     )

#     # 3. Tokenize the formatted prompt and move it to the correct device (CPU or MPS on Mac)
#     model_inputs = tokenizer([text_input], return_tensors="pt").to(model.device)

#     # 4. Generate the summary using parameters recommended for non-thinking mode
#     generated_ids = model.generate(
#         **model_inputs,
#         max_new_tokens=512,  # Limit summary length
#         temperature=0.7,
#         top_p=0.8,
#         top_k=20
#     )
    
#     # 5. Slice the output to remove the input prompt, leaving only the generated response
#     output_ids = generated_ids[0][len(model_inputs.input_ids[0]):]
    
#     # 6. Decode the token IDs back into a readable string
#     summary = tokenizer.decode(output_ids, skip_special_tokens=True).strip()
    
#     print("✅ Summary generated successfully.")
#     return summary

# # --- Part 3: Text-to-Speech Function ---

# def speak_summary_with_kokoro(summary: str, voice: str = "af_heart") -> str:
#     """
#     Converts the summary text to speech using Kokoro TTS and saves as audio file.

#     Args:
#         summary: The text summary to convert to speech.
#         voice: The voice to use (default: "af_heart").

#     Returns:
#         The filename of the generated audio file.
#     """
#     print("🎵 Converting summary to speech with Kokoro TTS...")
    
#     try:
#         # Initialize Kokoro TTS pipeline
#         pipeline = KPipeline(lang_code='a')  # 'a' for English
        
#         # Generate speech
#         generator = pipeline(summary, voice=voice)
        
#         # Process audio chunks
#         audio_chunks = []
#         total_duration = 0
        
#         for i, (gs, ps, audio) in enumerate(generator):
#             audio_chunks.append(audio)
#             chunk_duration = len(audio) / 24000
#             total_duration += chunk_duration
#             print(f"  📊 Generated chunk {i+1}: {chunk_duration:.2f}s")
        
#         # Combine all audio chunks
#         if len(audio_chunks) > 1:
#             combined_audio = torch.cat(audio_chunks, dim=0)
#         else:
#             combined_audio = audio_chunks[0]
        
#         # Generate filename with timestamp
#         timestamp = int(time.time())
#         filename = f"summary_audio_{timestamp}.wav"
        
#         # Save audio file
#         sf.write(filename, combined_audio.numpy(), 24000)
        
#         print(f"✅ Audio generated successfully!")
#         print(f"💾 Saved as: {filename}")
#         print(f"⏱️  Duration: {total_duration:.2f} seconds")
#         print(f"🎭 Voice used: {voice}")
        
#         return filename
        
#     except Exception as e:
#         print(f"❌ Error generating speech: {e}")
#         return None

# # --- Part 4: Voice Selection Function ---

# def select_voice() -> str:
#     """
#     Allows user to select from available voices or use default.
    
#     Returns:
#         Selected voice name.
#     """
#     available_voices = {
#         '1': ('af_heart', 'Female - Heart (Grade A, default) ❤️'),
#         '2': ('af_bella', 'Female - Bella (Grade A-) 🔥'),
#         '3': ('af_nicole', 'Female - Nicole (Grade B-) 🎧'),
#         '4': ('am_michael', 'Male - Michael (Grade C+)'),
#         '5': ('am_fenrir', 'Male - Fenrir (Grade C+)'),
#         '6': ('af_sarah', 'Female - Sarah (Grade C+)'),
#         '7': ('bf_emma', 'British Female - Emma (Grade B-)'),
#         '8': ('bm_george', 'British Male - George (Grade C)')
#     }
    
#     print("\n🎭 Available voices (sorted by quality):")
#     for key, (voice_id, description) in available_voices.items():
#         print(f"  {key}. {description}")
    
#     print("  Enter: Use default voice (af_heart)")
    
#     choice = input("\nSelect voice (1-8 or Enter): ").strip()
    
#     if choice in available_voices:
#         selected_voice, description = available_voices[choice]
#         print(f"🎵 Selected: {description}")
#         return selected_voice
#     else:
#         print("🎵 Using default voice: Female - Heart")
#         return 'af_heart'

# # --- Main Execution Block ---

# if __name__ == "__main__":
#     print("🚀 Article Summarizer with Text-to-Speech")
#     print("=" * 50)
    
#     # Check if a URL was provided as a command-line argument
#     if len(sys.argv) < 2:
#         print("Usage: python qwen_kokoro_summarizer.py <URL_OF_ARTICLE>")
#         print("Example: python qwen_kokoro_summarizer.py https://example.com/article")
#         sys.exit(1)
        
#     article_url = sys.argv[1]

#     # --- Load Qwen Model and Tokenizer ---
#     print("\n📚 Setting up the Qwen3-0.6B model...")
#     print("Note: The first run will download the model (~1.2 GB). Please be patient.")

#     model_name = "Qwen/Qwen3-0.6B"
    
#     try:
#         tokenizer = AutoTokenizer.from_pretrained(model_name)
#         model = AutoModelForCausalLM.from_pretrained(
#             model_name,
#             torch_dtype="auto",  # Automatically selects precision (e.g., float16)
#             device_map="auto"    # Automatically uses MPS (Mac GPU) if available
#         )
#     except Exception as e:
#         print(f"❌ Failed to load the Qwen model. Error: {e}")
#         print("Please ensure you have a stable internet connection and sufficient disk space.")
#         sys.exit(1)

#     # Inform the user which device is being used
#     device = next(model.parameters()).device
#     print(f"✅ Qwen model loaded successfully on device: {str(device).upper()}")
#     if "mps" in str(device):
#         print("   (Running on Apple Silicon GPU)")

#     # --- Run the Complete Process ---
    
#     # Step 1: Scrape the article
#     print(f"\n📰 Step 1: Scraping article")
#     article_content = scrape_article_text(article_url)
    
#     if not article_content:
#         print("❌ Failed to scrape article. Exiting.")
#         sys.exit(1)
    
#     # Step 2: Summarize the content
#     print(f"\n🤖 Step 2: Generating summary")
#     summary = summarize_with_qwen(article_content, model, tokenizer)
    
#     # Step 3: Display the summary
#     print("\n" + "="*60)
#     print("✨ GENERATED SUMMARY ✨")
#     print("="*60)
#     print(summary)
#     print("="*60)
    
#     # Step 4: Ask if user wants TTS
#     print(f"\n🎵 Step 3: Text-to-Speech")
#     tts_choice = input("Would you like to hear the summary read aloud? (y/N): ").strip().lower()
    
#     if tts_choice in ['y', 'yes']:
#         # Let user select voice
#         selected_voice = select_voice()
        
#         # Generate speech
#         audio_filename = speak_summary_with_kokoro(summary, voice=selected_voice)
        
#         if audio_filename:
#             print(f"\n🎧 Audio saved as: {audio_filename}")
#             print("🔊 You can now play this file to hear the summary!")
            
#             # Optional: Try to play the audio automatically (macOS)
#             try:
#                 import subprocess
#                 print("🎶 Attempting to play audio automatically...")
#                 subprocess.run(['afplay', audio_filename], check=True)
#                 print("✅ Audio playback completed!")
#             except (subprocess.CalledProcessError, FileNotFoundError):
#                 print("ℹ️  Auto-play not available. Please play the file manually.")
#         else:
#             print("❌ Failed to generate audio.")
#     else:
#         print("👍 Summary completed without audio generation.")
    
#     print(f"\n🎉 Process completed successfully!")
#     print(f"📝 Summary length: {len(summary)} characters")
#     print(f"📊 Original article length: {len(article_content)} characters")
#     print(f"📉 Compression ratio: {len(summary)/len(article_content)*100:.1f}%")