Update app.py

app.py

@@ -87,7 +87,7 @@ def generate_response(messages, apikey):
     )
     if chat_completion.choices and hasattr(chat_completion.choices[0].message, "content"):
         return chat_completion.choices[0].message.content
-    return chat_completion.choices[0].message.content
+    return ""
 
 
 @gpu_decorator
@@ -96,7 +96,7 @@ def process_audio_input(audio_path, text, apikey, history, conv_state):
     Process audio and/or text input from the user:
       - If an audio file is provided, its transcript is obtained.
       - The conversation state and history are updated.
-      
+    
     Updated to construct the chat history as a list of dictionaries.
     """
     if not audio_path and not text.strip():
@@ -109,7 +109,7 @@ def process_audio_input(audio_path, text, apikey, history, conv_state):
     if not text.strip():
         return history, conv_state, ""
 
-    # Construct user message as a dict.
+    # Wrap the user input in a dict.
     user_msg = {"role": "user", "content": text}
     conv_state.append(user_msg)
     history.append(user_msg)
@@ -244,19 +244,19 @@ Have a conversation with an AI using your reference voice!
     def generate_audio_response(history, ref_audio, ref_text, remove_silence):
         """
         Generate an audio response from the last AI message in the conversation.
-        Updated to search for the last assistant message in dictionary format.
+        Returns the generated audio, the (possibly updated) reference text, and the unchanged chat history.
         """
         if not history or not ref_audio:
-            return None, ref_text
+            return None, ref_text, history
 
-        # Find the last message from the assistant.
+        # Find the last assistant message in the history.
         last_assistant = None
         for message in reversed(history):
             if message.get("role") == "assistant":
                 last_assistant = message
                 break
         if last_assistant is None or not last_assistant.get("content", "").strip():
-            return None, ref_text
+            return None, ref_text, history
 
         audio_result, _, ref_text_out = infer(
             ref_audio,
@@ -267,7 +267,7 @@ Have a conversation with an AI using your reference voice!
             speed=1.0,
             show_info=print,
         )
-        return audio_result, ref_text_out
+        return audio_result, ref_text_out, history
 
     def clear_conversation():
         """
@@ -299,7 +299,7 @@ Have a conversation with an AI using your reference voice!
     ).then(
         generate_audio_response,
         inputs=[chatbot_interface, ref_audio_chat, ref_text_chat, remove_silence_chat],
-        outputs=[audio_output_chat, ref_text_chat],
+        outputs=[audio_output_chat, ref_text_chat, chatbot_interface],
     ).then(lambda: None, None, audio_input_chat)
 
     text_input_chat.submit(
@@ -309,7 +309,7 @@ Have a conversation with an AI using your reference voice!
     ).then(
         generate_audio_response,
         inputs=[chatbot_interface, ref_audio_chat, ref_text_chat, remove_silence_chat],
-        outputs=[audio_output_chat, ref_text_chat],
+        outputs=[audio_output_chat, ref_text_chat, chatbot_interface],
     ).then(lambda: None, None, text_input_chat)
 
     send_btn_chat.click(
@@ -319,7 +319,7 @@ Have a conversation with an AI using your reference voice!
     ).then(
         generate_audio_response,
         inputs=[chatbot_interface, ref_audio_chat, ref_text_chat, remove_silence_chat],
-        outputs=[audio_output_chat, ref_text_chat],
+        outputs=[audio_output_chat, ref_text_chat, chatbot_interface],
     ).then(lambda: None, None, text_input_chat)
 
     clear_btn_chat.click(clear_conversation, outputs=[chatbot_interface, conversation_state])