feat(tts): implement CosyVoice3 engine with HTTP client and dynamic p…

app/engines/tts/cosyvoice3/__init__.py

@@ -0,0 +1,5 @@
+# CosyVoice3 TTS Engine (HTTP Client)
+from app.engines.tts.cosyvoice3.config import CosyVoice3Config
+from app.engines.tts.cosyvoice3.engine import CosyVoice3Engine
+
+__all__ = ["CosyVoice3Config", "CosyVoice3Engine"]

app/engines/tts/cosyvoice3/config.py

@@ -0,0 +1,84 @@
+"""
+CosyVoice3 TTS Engine Configuration
+
+Configuration for CosyVoice3 HTTP client engine.
+Connects to an external CosyVoice3 FastAPI server for TTS synthesis.
+"""
+
+from pydantic import BaseModel, Field
+
+from app.models.engine import EngineConfig
+
+
+class VoiceConfig(BaseModel):
+    """Configuration for a single voice (zero-shot voice cloning)."""
+
+    prompt_wav_path: str = Field(
+        ..., description="Path to the prompt WAV file for voice cloning"
+    )
+    prompt_text: str = Field(..., description="Transcript of the prompt WAV file")
+    description: str = Field(
+        default="", description="Human-readable description of the voice"
+    )
+
+
+class CosyVoice3Config(EngineConfig):
+    """
+    Configuration for CosyVoice3 TTS Engine (HTTP Client)
+
+    Attributes:
+        model_name: Model identifier (for display/logging only)
+        service_url: URL of the CosyVoice3 FastAPI server
+        sample_rate: Output audio sample rate (CosyVoice3 default: 22050)
+        voices: Mapping of voice names to VoiceConfig (prompt_wav_path + prompt_text)
+        default_voice: Default voice name when none specified in request
+        speed: Default speech speed multiplier
+        system_prompt: System prompt prefix for CosyVoice3 prompt_text
+        connect_timeout: HTTP connection timeout in seconds
+        read_timeout: HTTP read timeout in seconds (None = no timeout for streaming)
+    """
+
+    service_url: str = Field(
+        default="http://localhost:50000",
+        description="URL of the CosyVoice3 FastAPI server",
+    )
+    sample_rate: int = Field(
+        default=22050,
+        ge=8000,
+        le=48000,
+        description="Output audio sample rate in Hz",
+    )
+    voices: dict[str, VoiceConfig] = Field(
+        default_factory=dict,
+        description="Voice name to VoiceConfig mapping for zero-shot cloning",
+    )
+    default_voice: str | None = Field(
+        default=None,
+        description="Default voice name when none specified",
+    )
+    speed: float = Field(
+        default=1.0,
+        gt=0.0,
+        le=3.0,
+        description="Default speech speed multiplier",
+    )
+    system_prompt: str = Field(
+        default="You are a helpful assistant.",
+        description="System prompt prefix for CosyVoice3",
+    )
+    connect_timeout: float = Field(
+        default=10.0,
+        ge=1.0,
+        description="HTTP connection timeout in seconds",
+    )
+    read_timeout: float | None = Field(
+        default=None,
+        description="HTTP read timeout (None = no timeout for long synthesis)",
+    )
+    allowed_voice_dirs: list[str] = Field(
+        default_factory=list,
+        description=(
+            "Allowed directories for dynamic prompt_wav_path from API kwargs. "
+            "If empty, dynamic paths are rejected (only configured voices allowed)."
+        ),
+    )