Update main.py

main.py

@@ -1,29 +1,50 @@
-from fastapi import FastAPI, HTTPException  
-from fastapi.responses import StreamingResponse  
-from pydantic import BaseModel  
-from kokoro import KPipeline  
-import soundfile as sf  
-import io  
-  
-app = FastAPI()  
-  
-# Initialiser le pipeline au démarrage  
-pipeline = KPipeline(lang_code='a', device='cpu')  
-  
-class TTSRequest(BaseModel):  
-    text: str  
-    voice: str = 'af_heart'  
-    speed: float = 1.0  
-  
-@app.post("/tts/stream")  
-async def stream_speech(request: TTSRequest):  
-    def generate():  
-        # Le pipeline retourne un générateur qui produit des Result  
-        for result in pipeline(request.text, voice=request.voice, speed=request.speed):  
-            # Chaque result.audio contient un chunk d'audio  
-            buffer = io.BytesIO()  
-            sf.write(buffer, result.audio.numpy(), 24000, format='WAV')  
-            buffer.seek(0)  
-            yield buffer.read()  
-      
-    return StreamingResponse(generate(), media_type="audio/wav")
+from fastapi import FastAPI, HTTPException
+from fastapi.responses import StreamingResponse
+from pydantic import BaseModel
+from kokoro import KPipeline
+import soundfile as sf
+import io
+import logging
+import time
+
+# --- Configurer les logs ---
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger("kokoro-stream")
+
+app = FastAPI()
+
+# Initialiser le pipeline au démarrage
+logger.info("🔍 Initialisation du pipeline Kokoro...")
+pipeline = KPipeline(lang_code='a', device='cpu')
+logger.info("✅ Pipeline Kokoro initialisé")
+
+class TTSRequest(BaseModel):
+    text: str
+    voice: str = 'af_heart'
+    speed: float = 1.0
+
+@app.post("/tts/stream")
+async def stream_speech(request: TTSRequest):
+    logger.info(f"🚀 Streaming demandé pour le texte: '{request.text}'")
+    start_time = time.time()
+
+    def generate():
+        chunk_count = 0
+        for result in pipeline(request.text, voice=request.voice, speed=request.speed):
+            chunk_count += 1
+            if hasattr(result.audio, "numpy"):
+                audio_data = result.audio.numpy()
+            else:
+                audio_data = result.audio
+
+            logger.info(f"✅ Chunk {chunk_count} généré, taille={len(audio_data)} samples")
+
+            buffer = io.BytesIO()
+            sf.write(buffer, audio_data, 24000, format='WAV')
+            buffer.seek(0)
+            yield buffer.read()
+
+        elapsed = time.time() - start_time
+        logger.info(f"🏁 Streaming terminé, {chunk_count} chunks envoyés en {elapsed:.2f}s")
+
+    return StreamingResponse(generate(), media_type="audio/wav")