Update app.py

app.py

@@ -13,7 +13,7 @@ print("Zephyr is getting ready to charm! 🌟")
 model = None
 tokenizer = None
 
-@spaces.GPU(duration=120)
+@spaces.GPU
 def load_model():
     global model, tokenizer
     if model is None:
@@ -61,8 +61,10 @@ def chat_with_zephyr(message, history):
     # Extract Zephyr's response
     zephyr_response = response.split("Zephyr:")[-1].strip()
     
-    # Return the new message pair
-    return "", zephyr_response
+    # Append the new exchange to the history
+    history.append((message, zephyr_response))
+    
+    return history
 
 css = """
 body {
@@ -80,18 +82,12 @@ with gr.Blocks(css=css) as iface:
     msg = gr.Textbox(placeholder="Tell Zephyr what's on your mind...", label="Your message")
     clear = gr.Button("Clear Chat")
 
-    def user(user_message, history):
-        return "", history + [[user_message, None]]
+    def respond(message, chat_history):
+        bot_message = chat_with_zephyr(message, chat_history)
+        chat_history.append((message, bot_message[-1][1]))
+        return "", chat_history
 
-    def bot(history):
-        user_message = history[-1][0]
-        bot_response = chat_with_zephyr(user_message, history[:-1])
-        history[-1][1] = bot_response
-        return history
-
-    msg.submit(user, [msg, chatbot], [msg, chatbot], queue=False).then(
-        bot, chatbot, chatbot
-    )
+    msg.submit(respond, [msg, chatbot], [msg, chatbot])
     clear.click(lambda: None, None, chatbot, queue=False)
 
     gr.Markdown("""