封装minmax语音克隆接口

src/BowongModalFunctions/api.py

@@ -4,8 +4,9 @@ import sentry_sdk
 from sentry_sdk.integrations.loguru import LoguruIntegration, LoggingLevels
 from sentry_sdk.integrations.fastapi import FastApiIntegration
 from fastapi.middleware.cors import CORSMiddleware
+
 from .utils.KVCache import KVCache
-from .router import ffmpeg, cache, comfyui
+from .router import ffmpeg, cache, comfyui, text2speech
 from .config import WorkerConfig
 
 config = WorkerConfig()
@@ -82,3 +83,4 @@ async def scalar():
 web_app.include_router(ffmpeg.router)
 web_app.include_router(cache.router)
 web_app.include_router(comfyui.router)
+web_app.include_router(text2speech.router)  # 添加文本转语音路由

src/BowongModalFunctions/config.py

@@ -19,4 +19,39 @@ class WorkerConfig(BaseSettings):
     comfyui_s3_input: Optional[str] = Field(default="comfyui-input", description="ComfyUI input S3文件夹名")
     comfyui_s3_output: Optional[str] = Field(default="comfyui-output", description="ComfyUI output S3文件夹名")
 
-    modal_config: Any = SettingsConfigDict()
+    # MiniMax API 配置
+    # BaseSettings 会自动从环境变量 MINIMAX_GROUP_ID 读取值
+    minimax_group_id: str = Field(
+        default="",
+        description="MiniMax API 组ID"
+    )
+
+    # BaseSettings 会自动从环境变量 MINIMAX_API_KEY 读取值
+    minimax_api_key: str = Field(
+        default="",
+        description="MiniMax API 密钥"
+    )
+
+    # 语音克隆相关配置
+    voice_clone_max_file_size: int = Field(
+        default=20 * 1024 * 1024,  # 20MB
+        description="语音克隆最大文件大小限制"
+    )
+
+    voice_clone_allowed_formats: list = Field(
+        default=[".mp3", ".wav", ".m4a", ".aac", ".ogg"],
+        description="语音克隆支持的音频格式"
+    )
+
+    voice_clone_output_dir: str = Field(
+        default="voice_clone/minmax",
+        description="语音克隆输出目录（相对于S3挂载目录）"
+    )
+
+    model_config = SettingsConfigDict(
+        # 自动从环境变量加载，大小写不敏感
+        case_sensitive=False,
+        # 支持从 .env 文件加载（如果存在）
+        env_file=".env",
+        env_file_encoding="utf-8"
+    )

src/cluster/app.py

@@ -10,8 +10,8 @@ config = WorkerConfig()
 
 app = modal.App(config.modal_app_name,
                 include_source=False,
-                secrets=[modal.Secret.from_name("cf-kv-secret",
-                                                environment_name=config.modal_environment)])
+                secrets=[modal.Secret.from_name("cf-kv-secret", environment_name=config.modal_environment),
+                         modal.Secret.from_name("minimax-api", environment_name=config.modal_environment)])
 
 app.include(media_app)
 app.include(ffmpeg_app)

src/cluster/web.py

@@ -24,12 +24,12 @@ with fastapi_image.imports():
     @app.function(scaledown_window=60,
                   secrets=[
                       modal.Secret.from_name("cf-kv-secret", environment_name='dev'),
+                      modal.Secret.from_name("minimax-api", environment_name=config.modal_environment)
                   ],
                   volumes={
                       config.S3_mount_dir: modal.CloudBucketMount(
                           bucket_name=config.S3_bucket_name,
-                          secret=modal.Secret.from_name("aws-s3-secret",
-                                                        environment_name=config.modal_environment),
+                          secret=modal.Secret.from_name("aws-s3-secret", environment_name=config.modal_environment),
                       ),
                   },
                   )