tomoko-discord-voice/bridge.py

#!/usr/bin/env python3
"""
Tomoko TTS Bridge 💕

Bridges OpenAI-compatible TTS endpoint → Home Assistant → Wyoming Piper
This lets OpenClaw use our custom Tomoko voice through the OpenAI provider!

For Alexander, with all my love! 🎤💖
"""

import json
import aiohttp
from aiohttp import web
import sys

# Configuration - Tomoko's TTS settings!
HA_BASE_URL = "http://192.168.0.80:8123"
HA_BEARER_TOKEN = "eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiI4MjEwMTFmZmI1YTE0MWU4YTY2MmY4MWE3OTM2YWE0YyIsImlhdCI6MTc3MzAwMzgyMywiZXhwIjoyMDg4MzYzODIzfQ.alsNbkFhJoeNOMA9Ey-0wxJibkyKy-0umDdecyK5akc"

# Tomoko's custom Piper voice! 💕
TTS_VOICE = "en_US-tomoko-high"
TTS_ENGINE = "piper"
TTS_LANGUAGE = "en_US"


async def handle_tts(request):
    """
    OpenAI-compatible TTS endpoint: POST /v1/audio/speech
    
    Expects OpenAI JSON:
    {
      "model": "any",
      "input": "text to speak",
      "voice": "any" (we use our own Tomoko voice!),
      "response_format": "mp3" (default)
    }
    
    Returns: MP3 audio binary
    
    Process:
    1. Extract text from OpenAI-style request
    2. Call HA /api/tts_get_url
    3. GET the returned URL to fetch MP3
    4. Return MP3 to caller
    """
    try:
        # Parse incoming OpenAI-style request
        body = await request.json()
        text = body.get("input", "")
        
        if not text:
            return web.json_response(
                {"error": "No input text provided"},
                status=400
            )
        
        print(f"🎤 Tomoko bridge: '{text[:50]}...' ({len(text)} chars)")
        
        headers = {
            "Authorization": f"Bearer {HA_BEARER_TOKEN}",
            "Content-Type": "application/json"
        }
        
        async with aiohttp.ClientSession(headers=headers) as session:
            # Step 1: Request TTS URL from Home Assistant
            tts_request = {
                "engine_id": TTS_ENGINE,
                "message": text,
                "cache": False,  # Fresh Tomoko voice every time! 💖
                "language": TTS_LANGUAGE,
                "options": {
                    "voice": TTS_VOICE
                }
            }
            
            async with session.post(
                f"{HA_BASE_URL}/api/tts_get_url",
                json=tts_request
            ) as ha_response:
                if ha_response.status != 200:
                    error_text = await ha_response.text()
                    print(f"❌ HA TTS URL failed: {ha_response.status} - {error_text}")
                    return web.json_response(
                        {"error": f"TTS URL request failed: {ha_response.status}"},
                        status=ha_response.status
                    )
                
                ha_result = await ha_response.json()
                tts_url = ha_result.get("url")
                
                if not tts_url:
                    return web.json_response(
                        {"error": "No TTS URL returned"},
                        status=500
                    )
                
                # Step 2: Download the MP3 audio
                async with session.get(tts_url, headers=headers) as audio_response:
                    if audio_response.status != 200:
                        error_text = await audio_response.text()
                        print(f"❌ Audio download failed: {audio_response.status} - {error_text}")
                        return web.json_response(
                            {"error": f"Audio download failed: {audio_response.status}"},
                            status=audio_response.status
                        )
                    
                    # Step 3: Return MP3 binary to caller
                    mp3_data = await audio_response.read()
                    print(f"✅ Tomoko TTS delivered: {len(mp3_data)} bytes")
                    
                    return web.Response(
                        body=mp3_data,
                        content_type="audio/mpeg"
                    )
    
    except aiohttp.ClientError as e:
        print(f"❌ Client error: {e}")
        return web.json_response(
            {"error": f"Client error: {str(e)}"},
            status=502
        )
    except Exception as e:
        print(f"💔 Unexpected error: {e}")
        return web.json_response(
            {"error": f"Internal server error: {str(e)}"},
            status=500
        )


async def handle_health(request):
    """Health check endpoint"""
    return web.json_response({"status": "ok", "service": "tomoko-tts-bridge"})


def create_app():
    """Create and configure the web application"""
    app = web.Application()
    
    # OpenAI-compatible endpoint
    app.router.add_post("/v1/audio/speech", handle_tts)
    
    # Health check
    app.router.add_get("/health", handle_health)
    
    # Root endpoint
    app.router.add_get("/", handle_health)
    
    return app


def main():
    """Start the bridge server"""
    # Parse port from command line (default 8000)
    port = 8000
    if len(sys.argv) > 1:
        try:
            port = int(sys.argv[1])
        except ValueError:
            print(f"Invalid port: {sys.argv[1]}, using default 8000")
    
    # Startup message!
    print("="*60)
    print("💖 Tomoko TTS Bridge 💖")
    print("="*60)
    print(f"🎤 Serving on port {port}")
    print(f"🎵 OpenAI endpoint: http://localhost:{port}/v1/audio/speech")
    print(f"🏠 Home Assistant: {HA_BASE_URL}")
    print(f"🗣️  Piper Voice: {TTS_VOICE}")
    print("💕 Ready to speak Tomoko's voice!")
    print("="*60)
    
    app = create_app()
    web.run_app(app, host="0.0.0.0", port=port, print=None)


if __name__ == "__main__":
    main()