Add voice-based model auto-detection and voice discovery endpoint

Features:
- Optional model parameter in /v1/audio/speech - auto-detects from voice name
- Voice-to-model cache loaded at startup for fast O(1) lookups
- First-match strategy for duplicate voice names across models
- New /v1/voices endpoint with extended voice info (engine, sample_rate, voice count)
- /v1/models kept OpenAI-compatible (minimal fields)

Implementation:
- speech.py:274: Made model parameter Optional[str] = None
- speech.py:253-260: Added detect_model_from_voice() using cached mapping
- speech.py:42: Added voice_to_model_cache global dict
- speech.py:723-732: Cache initialization at startup (227 voices)
- speech.py:336-383: New /v1/voices endpoint with voice lists and metadata
- speech.py:401-408: Auto-detection logic when model is None

Tested:
- bm_george auto-detected to tts-1-kokoro (unique voice)
- alloy auto-detected to tts-1 (first match of duplicate)
- /v1/models returns OpenAI-compatible minimal format
- /v1/voices returns extended info for all 4 models

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

speech.py

@@ -16,6 +16,7 @@ from fastapi.responses import StreamingResponse
 from loguru import logger
 from openedai import OpenAIStub, BadRequestError, ServiceUnavailableError
 from pydantic import BaseModel
+from typing import Optional
 import uvicorn
 
 @contextlib.asynccontextmanager
@@ -38,6 +39,9 @@ kokoro_pipeline = None
 kokoro_lang = None
 args = None
 
+# Voice-to-model lookup cache (loaded at startup)
+voice_to_model_cache = {}
+
 def unload_model():
     import torch, gc
     global xtts
@@ -246,6 +250,15 @@ def preprocess(raw_input):
     #logger.debug(f"preprocess: after: {[raw_input]}")
     return raw_input
 
+# Auto-detect which model a voice belongs to (uses cached mapping)
+def detect_model_from_voice(voice: str) -> str:
+    """Find which model supports a given voice name.
+    Returns the first model that has this voice, or None if not found.
+    Uses voice_to_model_cache populated at startup for fast lookups.
+    """
+    global voice_to_model_cache
+    return voice_to_model_cache.get(voice, None)
+
 # Read voice map on demand so it can be changed without restarting the server
 def map_voice_to_speaker(voice: str, model: str):
     default_exists('config/voice_to_speaker.yaml')
@@ -258,7 +271,7 @@ def map_voice_to_speaker(voice: str, model: str):
             raise BadRequestError(f"Error loading voice: {voice}, KeyError: {e}", param='voice')
 
 class GenerateSpeechRequest(BaseModel):
-    model: str = "tts-1" # or "tts-1-hd"
+    model: Optional[str] = None  # Auto-detected from voice if not provided
     input: str
     voice: str = "alloy"  # alloy, echo, fable, onyx, nova, and shimmer
     response_format: str = "mp3" # mp3, opus, aac, flac
@@ -288,7 +301,41 @@ def build_ffmpeg_args(response_format, input_format, sample_rate):
 
 @app.get("/v1/models")
 async def list_models():
-    """List all available TTS models and their supported voices"""
+    """List all available TTS models (OpenAI-compatible format)"""
+    # Return minimal OpenAI-compatible model list (no extra fields)
+    return {
+        "object": "list",
+        "data": [
+            {
+                "id": "tts-1",
+                "object": "model",
+                "created": 1700000000,
+                "owned_by": "uncloseai"
+            },
+            {
+                "id": "tts-1-hd",
+                "object": "model",
+                "created": 1700000000,
+                "owned_by": "uncloseai"
+            },
+            {
+                "id": "tts-1-silero",
+                "object": "model",
+                "created": 1700000000,
+                "owned_by": "uncloseai"
+            },
+            {
+                "id": "tts-1-kokoro",
+                "object": "model",
+                "created": 1700000000,
+                "owned_by": "uncloseai"
+            }
+        ]
+    }
+
+@app.get("/v1/voices")
+async def list_voices():
+    """List all available voices with model mapping and metadata (extended endpoint)"""
     default_exists('config/voice_to_speaker.yaml')
 
     with open('config/voice_to_speaker.yaml', 'r', encoding='utf8') as file:
@@ -300,11 +347,11 @@ async def list_models():
         if isinstance(voices, dict):
             voice_list = list(voices.keys())
 
-            # Add model metadata
+            # Add model metadata with extended info
             model_info = {
                 "id": model_id,
                 "object": "model",
-                "created": 1700000000,  # Static timestamp
+                "created": 1700000000,
                 "owned_by": "uncloseai",
                 "voices": voice_list,
                 "voice_count": len(voice_list)
@@ -351,6 +398,15 @@ async def generate_speech(request: GenerateSpeechRequest):
     response_format = request.response_format.lower()
     speed = request.speed
 
+    # Auto-detect model from voice if model not provided
+    if model is None:
+        detected_model = detect_model_from_voice(voice)
+        if detected_model:
+            logger.info(f"Auto-detected model '{detected_model}' for voice '{voice}'")
+            model = detected_model
+        else:
+            raise BadRequestError(f"Voice '{voice}' not found in any model. Please specify a model.", param='voice')
+
     # Set the Content-Type header based on the requested format
     if response_format == "mp3":
         media_type = "audio/mpeg"
@@ -665,6 +721,17 @@ if __name__ == "__main__":
     default_exists('config/pre_process_map.yaml')
     default_exists('config/voice_to_speaker.yaml')
 
+    # Build voice-to-model cache for fast lookups
+    with open('config/voice_to_speaker.yaml', 'r', encoding='utf8') as file:
+        voice_map = yaml.safe_load(file)
+        for model_id, voices in voice_map.items():
+            if isinstance(voices, dict):
+                for voice_name in voices.keys():
+                    # First match wins (for duplicate voice names across models)
+                    if voice_name not in voice_to_model_cache:
+                        voice_to_model_cache[voice_name] = model_id
+    print(f"Voice-to-model cache initialized with {len(voice_to_model_cache)} voices")
+
     logger.remove()
     logger.add(sink=sys.stderr, level=args.log_level)