Update app.py

app.py

@@ -117,22 +117,22 @@ def transcribe(audio_path):
 
     return text
 
-def model(text, web_search):
+def model(text, web_search, max_tokens, temperature):
     if web_search is True:
         """Performs a web search, feeds the results to a language model, and returns the answer."""
         web_results = search(text)
         web2 = ' '.join([f"Link: {res['link']}\nText: {res['text']}\n\n" for res in web_results])
         formatted_prompt = system_instructions1 + text + "[WEB]" + str(web2) + "[OpenGPT 4o]"
-        stream = client1.text_generation(formatted_prompt, max_new_tokens=300, stream=True, details=True, return_full_text=False)
+        stream = client1.text_generation(formatted_prompt, max_new_tokens=max_tokens, temperature=temperature, stream=True, details=True, return_full_text=False)
         return "".join([response.token.text for response in stream if response.token.text != "</s>"])
     else:
         formatted_prompt = system_instructions1 + text + "[OpenGPT 4o]"
-        stream = client1.text_generation(formatted_prompt, max_new_tokens=300, stream=True, details=True, return_full_text=False)
+        stream = client1.text_generation(formatted_prompt, max_new_tokens=max_tokens, temperature=temperature, stream=True, details=True, return_full_text=False)
         return "".join([response.token.text for response in stream if response.token.text != "</s>"])
 
-async def respond(audio, web_search, voice):
+async def respond(audio, web_search, voice, max_tokens, temperature):
     user = transcribe(audio)
-    reply = model(user, web_search)
+    reply = model(user, web_search, max_tokens, temperature)
     communicate = edge_tts.Communicate(reply, voice=voice)
     with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp_file:
         tmp_path = tmp_file.name
@@ -151,7 +151,8 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
                     audio_input = gr.Audio(label="🎤 Speak or Upload Audio", sources="microphone", type="filepath")
                     web_search = gr.Checkbox(label="🔍 Enable Web Search", value=False)
                     voice = gr.Dropdown(label="🎤 Choose Voice", choices=voices, value="en-US-JennyNeural")
-                    submit_btn = gr.Button("🚀 Submit")
+                    max_tokens = gr.Slider(minimum=50, maximum=500, value=300, label="Max Tokens")
+                    temperature = gr.Slider(minimum=0.1, maximum=1.0, value=0.7, label="Temperature")
                 with gr.Column():
                     audio_output = gr.Audio(label="🤖 AI Response", autoplay=True)
                     user_text = gr.Textbox(label="👤 You Said", interactive=False)
@@ -162,8 +163,6 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
         
         with gr.TabItem("Settings"):
             gr.Markdown("### ⚙️ Settings")
-            max_tokens = gr.Slider(minimum=50, maximum=500, value=300, label="Max Tokens")
-            temperature = gr.Slider(minimum=0.1, maximum=1.0, value=0.7, label="Temperature")
             gr.Markdown("Adjust the parameters to customize the AI's behavior.")
 
     # Store conversation history
@@ -173,9 +172,10 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
         conversation_history.append([user_input, ai_response])
         return conversation_history
 
-    submit_btn.click(
+    # Automatically submit when audio is detected
+    audio_input.change(
         fn=respond,
-        inputs=[audio_input, web_search, voice],
+        inputs=[audio_input, web_search, voice, max_tokens, temperature],
         outputs=[audio_output, user_text, ai_text]
     ).then(
         fn=update_history,