code

app.py

@@ -26,7 +26,7 @@ else:
 
 
 
-def transcribe(audio: tuple[int, np.ndarray], conversation: list[dict]):
+def transcribe(audio: tuple[int, np.ndarray], conversation: list[dict], gradio_convo: list[dict]):
    
     output = pipe({"audio": audio[1], "turns": conversation, "sampling_rate": audio[0]},
                   max_new_tokens=512,
@@ -35,8 +35,10 @@ def transcribe(audio: tuple[int, np.ndarray], conversation: list[dict]):
 
     conversation.append({"role": "user", "content": output["transcription"]})
     conversation.append({"role": "assistant", "content": output["reply"]})
+    gradio_convo.append({"role": "user", "content": output["transcription"]})
+    gradio_convo.append({"role": "assistant", "content": output["reply"]})
 
-    yield AdditionalOutputs(conversation)
+    yield AdditionalOutputs(conversation, gradio_convo)
 
 
 with gr.Blocks() as demo: