Removed intermediate message object

api/audio.py

@@ -31,10 +31,8 @@ class STTManager:
         self.config = config
         self.streaming = os.getenv("STREAMING", False)
 
-    def speech_to_text(self, audio, convert_to_bytes=True):
-        if convert_to_bytes:
-            audio = numpy_audio_to_bytes(audio[1])
-
+    def speech_to_text(self, audio, chat_display):
+        audio = numpy_audio_to_bytes(audio[1])
         try:
             if self.config.stt.type == "OPENAI_API":
                 data = ("temp.wav", audio, "audio/wav")
@@ -54,7 +52,8 @@ class STTManager:
         except Exception as e:
             raise APIError(f"STT Error: Unexpected error: {e}")
 
-        return transcription
+        chat_display.append([transcription, None])
+        return chat_display
 
 
 class TTSManager:

api/llm.py

@@ -122,15 +122,16 @@ class LLMManager:
         messages = self.get_problem_prepare_messages(requirements, difficulty, topic)
         yield from self.get_text_stream(messages)
 
-    def update_chat_history(self, code, previous_code, message, chat_history):
+    def update_chat_history(self, code, previous_code, chat_history, chat_display):
+        message = chat_display[-1][0]
         if code != previous_code:
             chat_history.append({"role": "user", "content": f"My latest code:\n{code}"})
         chat_history.append({"role": "user", "content": message})
 
         return chat_history
 
-    def send_request_full(self, code, previous_code, message, chat_history, chat_display):
-        chat_history = self.update_chat_history(code, previous_code, message, chat_history)
+    def send_request_full(self, code, previous_code, chat_history, chat_display):
+        chat_history = self.update_chat_history(code, previous_code, chat_history, chat_display)
 
         reply = self.get_text(chat_history)
         chat_display.append([None, reply])
@@ -138,8 +139,8 @@ class LLMManager:
 
         return chat_history, chat_display, code
 
-    def send_request_stream(self, code, previous_code, message, chat_history, chat_display):
-        chat_history = self.update_chat_history(code, previous_code, message, chat_history)
+    def send_request_stream(self, code, previous_code, chat_history, chat_display):
+        chat_history = self.update_chat_history(code, previous_code, chat_history, chat_display)
 
         chat_display.append([None, ""])
         chat_history.append({"role": "assistant", "content": ""})

app.py

@@ -133,7 +133,7 @@ with gr.Blocks(title="AI Interviewer") as demo:
                     end_btn = gr.Button("Finish the interview", interactive=False)
                     chat = gr.Chatbot(label="Chat", show_label=False, show_share_button=False)
                     audio_input = gr.Audio(interactive=False, **default_audio_params)
-                    message = gr.Textbox(label="Message", lines=3, visible=False)
+                    # message = gr.Textbox(label="Message", lines=3, visible=False)
 
         with gr.Accordion("Feedback", open=True) as feedback_acc:
             feedback = gr.Markdown()
@@ -166,16 +166,14 @@ with gr.Blocks(title="AI Interviewer") as demo:
         fn=llm.end_interview, inputs=[description, chat_history], outputs=[feedback]
     )
 
-    audio_input.stop_recording(fn=stt.speech_to_text, inputs=[audio_input], outputs=[message]).then(
+    audio_input.stop_recording(fn=stt.speech_to_text, inputs=[audio_input, chat], outputs=[chat]).then(
         fn=lambda: None, outputs=[audio_input]
-    ).then(fn=add_candidate_message, inputs=[message, chat], outputs=[chat]).then(
+    ).then(
         fn=llm.send_request,
-        inputs=[code, previous_code, message, chat_history, chat],
+        inputs=[code, previous_code, chat_history, chat],
         outputs=[chat_history, chat, previous_code],
     ).then(
         fn=tts.read_last_message, inputs=[chat], outputs=[audio_output]
-    ).then(
-        fn=lambda: "", outputs=[message]
     )
 
 demo.launch(show_api=False)