feat: Update generate_speech endpoint to enforce MP3 format, modify request and response models for clarity, and enhance documentation to reflect language code usage and format restrictions.

models/mcp_models.py

@@ -98,12 +98,12 @@ class GenerateSpeechResponse(BaseModel):
     model_config = ConfigDict(
         json_schema_extra={
             "example": {
-                "audio_path": "temp_audio/tts_e9b78164.wav"
+                "audio_path": "temp_audio/tts_e9b78164.mp3"
             }
         }
     )
     
-    audio_path: str = Field(..., description="Path to generated audio file")
+    audio_path: str = Field(..., description="Path to generated MP3 audio file")
 
 class ExtractTopicResponse(BaseModel):
     """Response model for topic extraction"""

routes/mcp_routes.py

@@ -65,17 +65,19 @@ class MatchKeypointRequest(BaseModel):
 
 
 class GenerateSpeechRequest(BaseModel):
-    """Request pour générer de la parole"""
+    """Request pour générer de la parole (TTS gTTS, MP3 seulement)"""
     text: str = Field(..., description="Texte à convertir en parole")
-    voice: str = Field(default="Aaliyah-PlayAI", description="Voix à utiliser")
-    format: str = Field(default="wav", description="Format audio (wav, mp3, etc.)")
-    
+    # gTTS attend un code langue (ex: 'en', 'fr')
+    voice: str = Field(default="en", description="Code langue à utiliser (ex: 'en', 'fr')")
+    # Le backend gratuit ne supporte que le format MP3
+    format: str = Field(default="mp3", description="Format audio (uniquement 'mp3')")
+
     class Config:
         json_schema_extra = {
             "example": {
                 "text": "Hello, this is a test",
-                "voice": "Aaliyah-PlayAI",
-                "format": "wav"
+                "voice": "en",
+                "format": "mp3",
             }
         }
 
@@ -174,16 +176,27 @@ async def list_mcp_tools():
             ),
             ToolInfo(
                 name="generate_speech",
-                description="Convertit du texte en fichier audio",
+                description="Convertit du texte en fichier audio (TTS gTTS, MP3 seulement)",
                 input_schema={
                     "type": "object",
                     "properties": {
-                        "text": {"type": "string", "description": "Texte à convertir en parole"},
-                        "voice": {"type": "string", "description": "Voix à utiliser", "default": "Aaliyah-PlayAI"},
-                        "format": {"type": "string", "description": "Format audio", "default": "wav"}
+                        "text": {
+                            "type": "string",
+                            "description": "Texte à convertir en parole",
+                        },
+                        "voice": {
+                            "type": "string",
+                            "description": "Code langue à utiliser (ex: 'en', 'fr')",
+                            "default": "en",
+                        },
+                        "format": {
+                            "type": "string",
+                            "description": "Format audio (uniquement 'mp3')",
+                            "default": "mp3",
+                        },
                     },
-                    "required": ["text"]
-                }
+                    "required": ["text"],
+                },
             ),
             ToolInfo(
                 name="generate_argument",
@@ -286,14 +299,14 @@ async def call_mcp_tool(request: ToolCallRequest):
     }
     ```
     
-    5. **generate_speech** - Générer de la parole:
+    5. **generate_speech** - Générer de la parole (MP3 seulement):
     ```json
     {
       "tool_name": "generate_speech",
       "arguments": {
         "text": "Hello, this is a test",
-        "voice": "Aaliyah-PlayAI",
-        "format": "wav"
+        "voice": "en",
+        "format": "mp3"
       }
     }
     ```
@@ -488,10 +501,17 @@ async def mcp_transcribe_audio(file: UploadFile = File(...)):
         if os.path.exists(temp_path):
             os.unlink(temp_path)
 
-@router.post("/tools/generate-speech", summary="Générer de la parole à partir de texte")
+@router.post("/tools/generate-speech", summary="Générer de la parole à partir de texte (MP3 uniquement)")
 async def mcp_generate_speech(request: GenerateSpeechRequest):
-    """Convertit du texte en fichier audio (téléchargeable)"""
+    """Convertit du texte en fichier audio MP3 (téléchargeable) via l'outil MCP generate_speech"""
     try:
+        # Validation explicite pour être cohérent avec le backend gTTS
+        if request.format != "mp3":
+            raise HTTPException(
+                status_code=400,
+                detail="Only 'mp3' format is supported by the TTS backend",
+            )
+
         result = await mcp_server.call_tool("generate_speech", {
             "text": request.text,
             "voice": request.voice,
@@ -520,7 +540,7 @@ async def mcp_generate_speech(request: GenerateSpeechRequest):
         # Nettoyer le chemin si c'est une représentation string d'objet
         if audio_path and isinstance(audio_path, str):
             # Si c'est une représentation d'objet TextContent, extraire le chemin
-            if "text='" in audio_path and ".wav" in audio_path:
+            if "text='" in audio_path:
                 import re
                 match = re.search(r"text='([^']+)'", audio_path)
                 if match:
@@ -533,8 +553,8 @@ async def mcp_generate_speech(request: GenerateSpeechRequest):
         if not Path(audio_path).exists():
             raise HTTPException(status_code=500, detail=f"Audio file not found: {audio_path}")
         
-        # Déterminer le type MIME
-        media_type = "audio/wav" if request.format == "wav" else "audio/mpeg"
+        # Déterminer le type MIME (MP3 uniquement)
+        media_type = "audio/mpeg"
         
         # Retourner le fichier pour téléchargement
         logger.info(f"Speech generated: {audio_path}")