Merge origin/development: 整合 TTS（远端结构为主）+ memoir-cover，保留本地环境配置

Made-with: Cursor

api/app/features/conversation/ws/pipeline.py

@@ -26,6 +26,7 @@ from app.features.conversation.ws.profile_collector import (
     get_missing_profile_fields,
 )
 from app.features.user.models import User
+from app.core.config import settings
 from app.core.dependencies import get_asr_provider, get_tts_provider
 from app.features.memoir.state_service import get_or_create_state
 
@@ -47,7 +48,7 @@ async def _send_tts_audio(conversation_id: str, text: str) -> None:
             "conversation_id": conversation_id,
             "data": {
                 "audio_base64": base64.b64encode(audio_bytes).decode("utf-8"),
-                "format": "mp3",
+                "format": settings.tts_codec,
             },
             "timestamp": datetime.now(timezone.utc).isoformat(),
         })