Add code

app.py

@@ -54,8 +54,6 @@ OUT_CHUNK = 20 * 4096
 
     
 def response(audio: tuple[int, np.ndarray], conversation: list[dict], img: str | None):
-    conversation.append({"role": "user", "content": gr.Audio(audio)})
-    yield AdditionalOutputs(conversation)
 
     sampling_rate, audio_np = audio
     audio_np = audio_np.squeeze()
@@ -69,6 +67,7 @@ def response(audio: tuple[int, np.ndarray], conversation: list[dict], img: str |
     )
 
     segment.export(audio_buffer, format="wav")
+    conversation.append({"role": "user", "content": gr.Audio((sampling_rate, audio_np))})
     conversation.append({"role": "assistant", "content": ""})
 
     base64_encoded = str(base64.b64encode(audio_buffer.getvalue()), encoding="utf-8")