adjust app.py

app.py

@@ -3,6 +3,8 @@ from contextlib import asynccontextmanager
 import uvicorn
 from whisper_live.server import TranscriptionServer
 import logging
+import json
+import numpy as np
 
 # Configure logging
 logging.basicConfig(level=logging.INFO)
@@ -39,27 +41,76 @@ async def root():
 @app.websocket("/ws")
 async def websocket_endpoint(websocket: WebSocket):
     await websocket.accept()
+    client_uid = None
+    
     try:
-        # Start the transcription server with the WebSocket connection
+        # Receive initial configuration
+        config = await websocket.receive_json()
+        client_uid = config.get("uid")
+        
+        if not client_uid:
+            await websocket.close(code=4000, reason="No client UID provided")
+            return
+            
+        # Send ready message
+        await websocket.send_json({
+            "uid": client_uid,
+            "message": "SERVER_READY",
+            "backend": "faster_whisper"
+        })
+        
+        # Start the transcription server
         transcription_server.run(
             host="0.0.0.0",
-            port=7860,  # Hugging Face Spaces uses port 7860
-            backend="faster_whisper",  # Using faster_whisper as the backend
-            single_model=True  # Use single model mode for better resource usage
+            port=7860,
+            backend="faster_whisper",
+            single_model=True,
+            language=config.get("language", "de"),
+            task=config.get("task", "transcribe"),
+            model=config.get("model", "tiny"),
+            use_vad=config.get("use_vad", True)
         )
         
-        # Handle WebSocket communication
+        # Handle incoming audio data
         while True:
             data = await websocket.receive()
-            if data.get("type") == "websocket.disconnect":
+            
+            if data["type"] == "websocket.disconnect":
                 break
-            # Process the received data through the transcription server
-            # Note: You might need to adapt this part based on your specific needs
-            await websocket.send_text(data)
+                
+            if data["type"] == "websocket.receive":
+                if isinstance(data.get("bytes"), bytes):
+                    # Process binary audio data
+                    audio_data = np.frombuffer(data["bytes"], dtype=np.float32)
+                    # Process audio data through transcription server
+                    segments = transcription_server.process_audio(audio_data)
+                    
+                    if segments:
+                        await websocket.send_json({
+                            "uid": client_uid,
+                            "segments": segments
+                        })
+                elif isinstance(data.get("text"), str):
+                    # Handle text messages (like END_OF_AUDIO)
+                    msg = data["text"]
+                    if msg == "END_OF_AUDIO":
+                        break
+                        
     except Exception as e:
         logger.error(f"WebSocket error: {e}")
+        if client_uid:
+            try:
+                await websocket.send_json({
+                    "uid": client_uid,
+                    "error": str(e)
+                })
+            except:
+                pass
     finally:
-        await websocket.close()
+        try:
+            await websocket.close()
+        except:
+            pass
 
 @app.get("/health")
 def health_check():