ezra-environment/the-nexus/deepdive/bin/phase4_generate_audio.py

#!/usr/bin/env python3
"""
Deep Dive Phase 4: Audio Generation
Converts text briefing to spoken audio podcast.

Usage:
    python phase4_generate_audio.py [--date YYYY-MM-DD] [--output-dir DIR] [--tts TTS_PROVIDER]

Issue: the-nexus#830
"""

import argparse
import os
import re
import subprocess
from datetime import datetime
from pathlib import Path
from typing import Optional


class AudioGenerator:
    """Generate audio from briefing text using TTS."""
    
    # TTS providers in order of preference
    TTS_PROVIDERS = ['edge-tts', 'openai', 'elevenlabs', 'local-tts']
    
    def __init__(self, output_dir: Path, date: str, tts_provider: str = 'edge-tts'):
        self.output_dir = output_dir
        self.date = date
        self.tts_provider = tts_provider
        self.briefings_dir = output_dir / "briefings"
        self.audio_dir = output_dir / "audio"
        self.audio_dir.mkdir(parents=True, exist_ok=True)
    
    def load_briefing(self) -> str:
        """Load briefing markdown from Phase 3."""
        briefing_file = self.briefings_dir / f"{self.date}.md"
        if not briefing_file.exists():
            raise FileNotFoundError(f"Phase 3 output not found: {briefing_file}")
        
        with open(briefing_file) as f:
            content = f.read()
        
        # Remove YAML frontmatter if present
        if content.startswith('---'):
            parts = content.split('---', 2)
            if len(parts) >= 3:
                content = parts[2]
        
        return content
    
    def clean_text_for_tts(self, text: str) -> str:
        """Clean markdown for TTS consumption."""
        # Remove markdown syntax
        text = re.sub(r'\*\*', '', text)  # Bold
        text = re.sub(r'\*', '', text)    # Italic
        text = re.sub(r'`[^`]*`', 'code', text)  # Inline code
        text = re.sub(r'\[([^\]]+)\]\([^)]+\)', r'\1', text)  # Links
        text = re.sub(r'#{1,6}\s*', '', text)  # Headers
        text = re.sub(r'---', '', text)  # Horizontal rules
        
        # Remove URLs (keep domain for context)
        text = re.sub(r'https?://[^\s]+', ' [link] ', text)
        
        # Clean up whitespace
        text = re.sub(r'\n\s*\n', '\n\n', text)
        text = text.strip()
        
        return text
    
    def add_podcast_intro(self, text: str) -> str:
        """Add standard podcast intro/outro."""
        date_str = datetime.strptime(self.date, '%Y-%m-%d').strftime('%B %d, %Y')
        
        intro = f"""Welcome to Deep Dive, your daily AI intelligence briefing for {date_str}. This is Hermes, delivering the most relevant research and developments in artificial intelligence, filtered for the Timmy organization and agent systems development. Let's begin.

"""
        
        outro = """

That concludes today's Deep Dive briefing. Sources and full show notes are available in the Hermes knowledge base. This briefing was automatically generated and will be delivered daily at 6 AM. For on-demand briefings, message the bot with /deepdive. Stay sovereign.
"""
        
        return intro + text + outro
    
    def generate_edge_tts(self, text: str, output_file: Path) -> bool:
        """Generate audio using edge-tts (free, Microsoft Edge voices)."""
        try:
            import edge_tts
            import asyncio
            
            async def generate():
                communicate = edge_tts.Communicate(text, voice="en-US-AndrewNeural")
                await communicate.save(str(output_file))
            
            asyncio.run(generate())
            print(f"[Phase 4] Generated audio via edge-tts: {output_file}")
            return True
        except Exception as e:
            print(f"[WARN] edge-tts failed: {e}")
            return False
    
    def generate_openai_tts(self, text: str, output_file: Path) -> bool:
        """Generate audio using OpenAI TTS API."""
        try:
            from openai import OpenAI
            client = OpenAI(api_key=os.environ.get('OPENAI_API_KEY'))
            
            response = client.audio.speech.create(
                model="tts-1",
                voice="alloy",
                input=text[:4000]  # OpenAI limit
            )
            
            response.stream_to_file(str(output_file))
            print(f"[Phase 4] Generated audio via OpenAI TTS: {output_file}")
            return True
        except Exception as e:
            print(f"[WARN] OpenAI TTS failed: {e}")
            return False
    
    def generate_elevenlabs_tts(self, text: str, output_file: Path) -> bool:
        """Generate audio using ElevenLabs API."""
        try:
            from elevenlabs import generate, save
            
            audio = generate(
                api_key=os.environ.get('ELEVENLABS_API_KEY'),
                text=text[:5000],  # ElevenLabs limit
                voice="Bella",
                model="eleven_monolingual_v1"
            )
            
            save(audio, str(output_file))
            print(f"[Phase 4] Generated audio via ElevenLabs: {output_file}")
            return True
        except Exception as e:
            print(f"[WARN] ElevenLabs failed: {e}")
            return False
    
    def generate_local_tts(self, text: str, output_file: Path) -> bool:
        """Generate audio using local TTS (XTTS via llama-server or similar)."""
        print("[WARN] Local TTS not yet implemented")
        return False
    
    def generate_audio(self, text: str) -> Optional[Path]:
        """Generate audio using configured or available TTS."""
        output_file = self.audio_dir / f"{self.date}.mp3"
        
        # If provider specified, try it first
        if self.tts_provider == 'edge-tts':
            if self.generate_edge_tts(text, output_file):
                return output_file
        elif self.tts_provider == 'openai':
            if self.generate_openai_tts(text, output_file):
                return output_file
        elif self.tts_provider == 'elevenlabs':
            if self.generate_elevenlabs_tts(text, output_file):
                return output_file
        
        # Auto-fallback chain
        print("[Phase 4] Trying fallback TTS providers...")
        
        # Try edge-tts first (free, no API key)
        if self.generate_edge_tts(text, output_file):
            return output_file
        
        # Try OpenAI if key available
        if os.environ.get('OPENAI_API_KEY'):
            if self.generate_openai_tts(text, output_file):
                return output_file
        
        # Try ElevenLabs if key available
        if os.environ.get('ELEVENLABS_API_KEY'):
            if self.generate_elevenlabs_tts(text, output_file):
                return output_file
        
        print("[ERROR] All TTS providers failed")
        return None
    
    def run(self) -> Optional[Path]:
        """Run full audio generation pipeline."""
        print(f"[Phase 4] Generating audio for {self.date}")
        
        briefing = self.load_briefing()
        print(f"[Phase 4] Loaded briefing: {len(briefing)} characters")
        
        clean_text = self.clean_text_for_tts(briefing)
        podcast_text = self.add_podcast_intro(clean_text)
        
        # Truncate if too long for most TTS (target: 10-15 min audio)
        max_chars = 12000  # ~15 min at normal speech
        if len(podcast_text) > max_chars:
            print(f"[Phase 4] Truncating from {len(podcast_text)} to {max_chars} characters")
            podcast_text = podcast_text[:max_chars].rsplit('.', 1)[0] + '.'
        
        output_file = self.generate_audio(podcast_text)
        
        if output_file and output_file.exists():
            size_mb = output_file.stat().st_size / (1024 * 1024)
            print(f"[Phase 4] Audio generated: {output_file} ({size_mb:.1f} MB)")
        
        return output_file


def main():
    parser = argparse.ArgumentParser(description='Deep Dive Phase 4: Audio Generation')
    parser.add_argument('--date', default=datetime.now().strftime('%Y-%m-%d'),
                       help='Target date (YYYY-MM-DD)')
    parser.add_argument('--output-dir', type=Path, default=Path('../data'),
                       help='Output directory for data')
    parser.add_argument('--tts', default='edge-tts',
                       choices=['edge-tts', 'openai', 'elevenlabs', 'local-tts'],
                       help='TTS provider')
    args = parser.parse_args()
    
    generator = AudioGenerator(args.output_dir, args.date, args.tts)
    result = generator.run()
    
    if result:
        print(f"[DONE] Audio file: {result}")
    else:
        print("[FAIL] Audio generation failed")
        exit(1)


if __name__ == '__main__':
    main()