Speed up chat display

app.py

@@ -2,6 +2,7 @@ import os
 
 import gradio as gr
 
+from config import LLM_NAME, STT_NAME, TTS_NAME
 from llm import end_interview, get_problem, read_last_message, send_request, speech_to_text, test_connection, text_to_speech
 from options import fixed_messages, topics_list
 
@@ -33,6 +34,11 @@ def add_interviewer_message(message):
     return f
 
 
+def add_candidate_message(message, chat):
+    chat.append((message, None))
+    return chat
+
+
 def hide_solution():
     solution_acc = gr.Accordion("Solution", open=False)
     end_btn = gr.Button("Finish the interview", interactive=False)
@@ -76,22 +82,22 @@ with gr.Blocks() as demo:
             with gr.Column(scale=1):
                 try:
                     audio_test = text_to_speech("Handshake")
-                    gr.Markdown("TTS status: 🟢")
+                    gr.Markdown(f"TTS status: 🟢.   Model: {TTS_NAME}")
                 except:
-                    gr.Markdown("TTS status: 🔴")
+                    gr.Markdown(f"TTS status: 🔴.   Model: {TTS_NAME}")
+
                 try:
                     text_test = speech_to_text(audio_test, False)
-                    gr.Markdown("STT status: 🟢")
+                    gr.Markdown(f"STT status: 🟢.   Model: {STT_NAME}")
                 except:
-                    gr.Markdown("STT status: 🔴")
+                    gr.Markdown(f"STT status: 🔴.   Model: {STT_NAME}")
 
                 try:
                     test_connection()
-                    gr.Markdown("LLM status: 🟢")
+                    gr.Markdown(f"LLM status: 🟢.   Model: {LLM_NAME}")
                 except:
-                    gr.Markdown("LLM status: 🔴")
+                    gr.Markdown(f"LLM status: 🔴.   Model: {LLM_NAME}")
 
-        pass
     with gr.Tab("Coding") as coding_tab:
         chat_history = gr.State([])
         previous_code = gr.State("")
@@ -164,7 +170,7 @@ with gr.Blocks() as demo:
 
     audio_input.stop_recording(fn=speech_to_text, inputs=[audio_input], outputs=[message]).then(
         fn=lambda: None, inputs=None, outputs=[audio_input]
-    ).then(
+    ).then(fn=add_candidate_message, inputs=[message, chat], outputs=[chat]).then(
         fn=send_request,
         inputs=[code, previous_code, message, chat_history, chat],
         outputs=[chat_history, chat, message, previous_code],

llm.py

@@ -89,7 +89,7 @@ def send_request(code, previous_code, message, chat_history, chat_display, clien
     reply = response.choices[0].message.content.strip()
 
     chat_history.append({"role": "assistant", "content": reply})
-    chat_display.append([message, reply])
+    chat_display[-1][1] = reply
 
     return chat_history, chat_display, "", code
 
@@ -117,7 +117,7 @@ def text_to_speech(text):
 
     if TTS_TYPE == "OPENAI_API":
         client = OpenAI(base_url=TTS_URL, api_key=os.getenv(f"{TTS_TYPE}_KEY"))
-        response = client.audio.speech.create(model=TTS_NAME, voice="alloy", input=text)
+        response = client.audio.speech.create(model=TTS_NAME, voice="alloy", response_format="opus", input=text)
     elif TTS_TYPE == "HF_API":
         headers = {"Authorization": "Bearer " + os.getenv(f"{STT_TYPE}_KEY")}
         response = requests.post(TTS_URL, headers=headers)
@@ -127,8 +127,9 @@ def text_to_speech(text):
 
 
 def read_last_message(chat_display):
-    last_message = chat_display[-1][1]
-    if last_message is not None:
-        audio = text_to_speech(last_message)
-        return audio
+    if len(chat_display) > 0:
+        last_message = chat_display[-1][1]
+        if last_message is not None:
+            audio = text_to_speech(last_message)
+            return audio
     return None

options.py

@@ -25,7 +25,7 @@ fixed_messages = {
     "intro": "Welcome to the coding interview! I am your AI interview assistant. For the start select the difficulty and topic of the problem you would like to solve. Then click on the 'Generate a problem' button. Good luck!",
     "start": (
         "Please take a moment to read the problem statement. Then you can share you initial thoughts and ask any questions you may have. "
-        "Please use the record button to communicate with me. Try to think out loud and record all you comments and questions as if it is the real interview."
+        # "Please use the record button to communicate with me. Try to think out loud and record all you comments and questions as if it is the real interview."
     ),
     "end": "The interview has concluded. Thank you for your participation! In a moment I will provide a detailed feedback on your performance.",
     "error": "An error occurred. Please try again.",