Merge branch 'main' into tasks/#47_tts_handlers_repositories

src/assistant/lib/api/v1/handlers/__init__.py

@@ -1,3 +1,7 @@
 from .health import basic_router
+from .voice_responce_handler import VoiceResponseHandler
 
-__all__ = ["basic_router"]
+__all__ = [
+    "VoiceResponseHandler",
+    "basic_router",
+]

src/assistant/lib/api/v1/handlers/voice_responce_handler.py

@@ -0,0 +1,45 @@
+import http
+import io
+
+import fastapi
+
+import lib.stt.services as stt_services
+
+# import lib.tts.services as tts_service
+# import lib.models as models
+
+
+class VoiceResponseHandler:
+    def __init__(
+        self,
+        stt: stt_services.SpeechService,
+        # tts: tts_service.TTSService,
+    ):
+        self.stt = stt
+        # self.tts = tts
+        self.router = fastapi.APIRouter()
+        self.router.add_api_route(
+            "/",
+            self.voice_response,
+            methods=["POST"],
+            summary="Ответ голосового помощника",
+            description="Маршрут возвращает потоковый ответ аудио",
+        )
+
+    async def voice_response(
+        self,
+        voice: bytes = fastapi.File(...),
+    ) -> fastapi.responses.StreamingResponse:
+        voice_text: str = await self.stt.recognize(voice)
+        if voice_text == "":
+            raise fastapi.HTTPException(status_code=http.HTTPStatus.BAD_REQUEST, detail="Speech recognition failed")
+        # TODO: Добавить обработку текста через клиента openai
+        # TODO: Добавить синтез речи через клиента tts
+        # TODO: Заменить заглушку на реальный ответ
+        # response = await self.tts.get_audio_as_bytes(
+        #     models.TTSCreateRequestModel(
+        #         text=voice_text,
+        #     )
+        # )
+        # return fastapi.responses.StreamingResponse(io.BytesIO(response.audio_content), media_type="audio/ogg")
+        return fastapi.responses.StreamingResponse(io.BytesIO(voice), media_type="audio/ogg")

src/assistant/lib/api/v1/schemas/__init__.py

@@ -1,3 +1,5 @@
 from .base import HealthResponse
 
-__all__ = ["HealthResponse"]
+__all__ = [
+    "HealthResponse",
+]

src/assistant/lib/app/app.py

@@ -115,11 +115,18 @@ class Application:
                 models.VoiceModelProvidersEnum.ELEVEN_LABS: tts_eleven_labs_repository,
             },
         )
+        
         # Handlers
 
         logger.info("Initializing handlers")
         liveness_probe_handler = api_v1_handlers.basic_router
 
+        # TODO: объявить сервисы tts и openai и добавить их в voice_response_handler
+        voice_response_handler = api_v1_handlers.VoiceResponseHandler(
+            stt=stt_service,
+            # tts=tts_service,  # TODO
+        ).router
+
         logger.info("Creating application")
 
         fastapi_app = fastapi.FastAPI(
@@ -132,6 +139,7 @@ class Application:
 
         # Routes
         fastapi_app.include_router(liveness_probe_handler, prefix="/api/v1/health", tags=["health"])
+        fastapi_app.include_router(voice_response_handler, prefix="/api/v1/voice", tags=["voice"])
 
         application = Application(
             settings=settings,

src/assistant/lib/app/settings.py

@@ -1,30 +1,16 @@
-import pydantic
 import pydantic_settings
 
 import lib.app.split_settings as app_split_settings
 
 
 class Settings(pydantic_settings.BaseSettings):
-    api: app_split_settings.ApiSettings = pydantic.Field(default_factory=lambda: app_split_settings.ApiSettings())
-    app: app_split_settings.AppSettings = pydantic.Field(default_factory=lambda: app_split_settings.AppSettings())
-    postgres: app_split_settings.PostgresSettings = pydantic.Field(
-        default_factory=lambda: app_split_settings.PostgresSettings()
-    )
-    logger: app_split_settings.LoggingSettings = pydantic.Field(
-        default_factory=lambda: app_split_settings.LoggingSettings()
-    )
-    openai: app_split_settings.OpenaiSettings = pydantic.Field(
-        default_factory=lambda: app_split_settings.OpenaiSettings()
-    )
-    project: app_split_settings.ProjectSettings = pydantic.Field(
-        default_factory=lambda: app_split_settings.ProjectSettings()
-    )
-
-    proxy: app_split_settings.ProxySettings = pydantic.Field(default_factory=lambda: app_split_settings.ProxySettings())
-    voice: app_split_settings.VoiceSettings = pydantic.Field(default_factory=lambda: app_split_settings.VoiceSettings())
-    tts_yandex: app_split_settings.TTSYandexSettings = pydantic.Field(
-        default_factory=lambda: app_split_settings.TTSYandexSettings()
-    )
-    tts_eleven_labs: app_split_settings.TTSElevenLabsSettings = pydantic.Field(
-        default_factory=lambda: app_split_settings.TTSElevenLabsSettings()
-    )
+    api: app_split_settings.ApiSettings = app_split_settings.ApiSettings()
+    app: app_split_settings.AppSettings = app_split_settings.AppSettings()
+    postgres: app_split_settings.PostgresSettings = app_split_settings.PostgresSettings()
+    logger: app_split_settings.LoggingSettings = app_split_settings.LoggingSettings()
+    openai: app_split_settings.OpenaiSettings = app_split_settings.OpenaiSettings()
+    project: app_split_settings.ProjectSettings = app_split_settings.ProjectSettings()
+    proxy: app_split_settings.ProxySettings = app_split_settings.ProxySettings()
+    voice: app_split_settings.VoiceSettings = app_split_settings.VoiceSettings()
+    tts_yandex: app_split_settings.TTSYandexSettings = app_split_settings.TTSYandexSettings()
+    tts_eleven_labs: app_split_settings.TTSElevenLabsSettings = app_split_settings.TTSElevenLabsSettings()

src/assistant/lib/stt/repositories.py

@@ -1,8 +1,11 @@
+import http
 import mimetypes
 import tempfile
 
+import fastapi
 import magic
 import openai
+import pydantic
 
 import lib.app.settings as app_settings
 import lib.stt as stt
@@ -24,15 +27,24 @@ class OpenaiSpeechRepository:
 
     async def speech_to_text(self, audio: bytes) -> str:
         file_extension = self.__get_file_extension_from_bytes(audio)
-        if not file_extension:
-            raise ValueError("File extension is not supported")
-
-        voice: stt.models.SttVoice = stt.models.SttVoice(
-            audio_size=len(audio) // 1024,  # audio size in MB,
-            audio_format=file_extension,
-            audio_data=audio,
-            voice_settings=self.settings.voice,
-        )
+        if not file_extension or file_extension not in self.settings.voice.available_formats:
+            raise fastapi.HTTPException(
+                status_code=http.HTTPStatus.UNSUPPORTED_MEDIA_TYPE,
+                detail=f"File extension is not supported. "
+                f"Available extensions: {self.settings.voice.available_formats}",
+            )
+        try:
+            voice: stt.models.SttVoice = stt.models.SttVoice(
+                audio_size=len(audio) // 1024,  # audio size in MB,
+                audio_format=file_extension,
+                audio_data=audio,
+                voice_settings=self.settings.voice,
+            )
+        except (pydantic.ValidationError, ValueError) as e:
+            raise fastapi.HTTPException(
+                status_code=http.HTTPStatus.BAD_REQUEST,
+                detail=f"Voice validation error: {e}",
+            )
 
         try:
             with tempfile.NamedTemporaryFile(suffix=f".{file_extension}") as temp_file:
@@ -40,8 +52,14 @@ class OpenaiSpeechRepository:
                 temp_file.seek(0)
                 transcript = openai.Audio.transcribe(self.settings.openai.stt_model, temp_file)  # type: ignore
         except openai.error.InvalidRequestError as e:  # type: ignore[reportGeneralTypeIssues]
-            raise ValueError(f"OpenAI API error: {e}")
+            raise fastapi.HTTPException(
+                status_code=http.HTTPStatus.BAD_REQUEST,
+                detail=f"OpenAI request error: {e}",
+            )
         except openai.error.OpenAIError as e:  # type: ignore[reportGeneralTypeIssues]
-            raise ValueError(f"OpenAI API error: {e}")
+            raise fastapi.HTTPException(
+                status_code=http.HTTPStatus.BAD_REQUEST,
+                detail=f"OpenAI API error: {e}",
+            )
 
         return transcript.text  # type: ignore[reportUnknownVariableType]