feat: Add OpenAI-compatible TTS endpoint and engines

- Implements POST /v1/audio/speech endpoint (OpenAI API compatible).
- Integrates Kokoro and XTTS engines (including dependencies and implementations).
- Updates main application to register new engines and router.
- Adds unit tests for OpenAI compatibility.
- Updates requirements.txt for new engines.

tests/test_openai_api.py

@@ -0,0 +1,74 @@
+import pytest
+from fastapi.testclient import TestClient
+from unittest.mock import AsyncMock, MagicMock
+from app.routers.openai_compatible import OpenAISpeechRequest
+
+def test_openai_speech_endpoint_success(app_client, app_instance, tmp_path):
+    # 1. Setup Mock Engine
+    mock_engine = MagicMock()
+    # Mock synthesis to return a dummy file path
+    dummy_file = tmp_path / "test_output.wav"
+    dummy_file.write_bytes(b"fake audio data")
+    mock_engine.synthesize = AsyncMock(return_value=str(dummy_file))
+    
+    # Inject mock engine into registry
+    app_instance.ENGINE_REGISTRY["mock-engine"] = mock_engine
+    
+    # 2. Make Request
+    payload = {
+        "model": "mock-engine",
+        "input": "Hello OpenAI",
+        "voice": "default",
+        "response_format": "wav"
+    }
+    response = app_client.post("/v1/audio/speech", json=payload)
+    
+    # 3. Assertions
+    assert response.status_code == 200
+    assert response.headers["content-type"] == "audio/wav"
+    assert response.content == b"fake audio data"
+    
+    # Verify engine call
+    mock_engine.synthesize.assert_called_once_with(
+        text="Hello OpenAI",
+        speaker="default",
+        model=None,
+        fmt="wav"
+    )
+
+def test_openai_speech_endpoint_engine_not_found(app_client):
+    payload = {
+        "model": "non-existent-engine",
+        "input": "Test",
+        "voice": "default"
+    }
+    response = app_client.post("/v1/audio/speech", json=payload)
+    assert response.status_code == 404
+    assert "not found" in response.json()["detail"]
+
+def test_openai_speech_endpoint_tts_1_mapping(app_client, app_instance, tmp_path):
+    # Setup Mock Engine for 'kokoro' (simulating it's active)
+    mock_engine = MagicMock()
+    dummy_file = tmp_path / "tts1_output.mp3"
+    dummy_file.write_bytes(b"mp3 data")
+    mock_engine.synthesize = AsyncMock(return_value=str(dummy_file))
+    
+    # Inject into registry
+    app_instance.ENGINE_REGISTRY["kokoro"] = mock_engine
+    
+    # Request 'tts-1' -> should map to 'kokoro'
+    payload = {
+        "model": "tts-1",
+        "input": "Mapping test",
+        "voice": "alloy",
+        "response_format": "mp3"
+    }
+    response = app_client.post("/v1/audio/speech", json=payload)
+    
+    assert response.status_code == 200
+    assert response.headers["content-type"] == "audio/mpeg"
+    
+    mock_engine.synthesize.assert_called_once()
+    # Check that speaker passed through
+    args, kwargs = mock_engine.synthesize.call_args
+    assert kwargs["speaker"] == "alloy"