change?

app.py

@@ -128,11 +128,14 @@ def generate_response(audio_path, message, chat_history=None):
                     conversation.append({"role": "assistant", "content": bot_msg})
 
         if audio_path:
-            # For files, we don't include an audio_url in the conversation
-            # Instead we just process the audio data directly
+            # Match the format that was working in the original code
+            # Create a placeholder audio URL to ensure the <|AUDIO|> token is generated
             conversation.append({
                 "role": "user",
-                "content": message
+                "content": [
+                    {"type": "audio", "audio_url": "placeholder_audio_url"},
+                    {"type": "text", "text": message}
+                ]
             })
 
             audio_data = process_audio_file(audio_path)