Update app.py

app.py

@@ -21,6 +21,7 @@ def respond(
     """
     client = InferenceClient(model=available_models[model_name])
 
+    # Prepare the conversation history for the API call
     messages = [{"role": "system", "content": system_message}]
     for user_input, assistant_response in history:
         messages.append({"role": "user", "content": user_input})
@@ -28,7 +29,9 @@ def respond(
     messages.append({"role": "user", "content": message})
 
     streamed_response = ""
+
     try:
+        # Generate a response from the model with streaming
         for response in client.chat_completion(
             messages=messages,
             max_tokens=max_tokens,
@@ -39,19 +42,24 @@ def respond(
             chunk = response.choices[0].delta.get("content", "")
             streamed_response += chunk
             yield streamed_response
+
     except Exception as e:
         yield f"**Error:** {str(e)}"
 
-# Latest updates
+# Latest updates (this is just placeholder text)
 latest_updates = """
 **Chatbot - Latest Updates:**
 
-* **Multiple Model Support:** You can now choose from different models like Zephyr 7B.
-* **Improved Error Handling:** The chatbot now provides clearer error messages if something goes wrong.
+* Multiple Model Support
+* Improved Error Handling
+* Enhanced System Message Input
 """
 
 def show_updates_and_respond(history, system_message, max_tokens, temperature, top_p, model_name):
-    history.append(("User: ", "Show me the latest updates"))
+    """
+    Shows the latest updates and generates a response from the model based on the updates.
+    """
+    history.append(("User", "Show me the latest updates"))
     yield from respond(
         message="Show me the latest updates",
         history=history,
@@ -61,8 +69,8 @@ def show_updates_and_respond(history, system_message, max_tokens, temperature, t
         top_p=top_p,
         model_name=model_name,
     )
-    history[-1] = ("User: ", "Show me the latest updates")
-    history.append(("Assistant:", latest_updates))
+    history[-1] = ("User", "Show me the latest updates")
+    history.append(("Assistant", latest_updates))
     yield from respond(
         message="What are the latest updates?",
         history=history,
@@ -73,7 +81,7 @@ def show_updates_and_respond(history, system_message, max_tokens, temperature, t
         model_name=model_name,
     )
 
-# Define the Gradio interface
+# Define the Gradio interface with the Blocks context
 with gr.Blocks(css=".gradio-container {border: none;}") as demo:
     chat_history = gr.State([])  # Initialize an empty chat history state
     system_message = gr.Textbox(
@@ -81,41 +89,24 @@ with gr.Blocks(css=".gradio-container {border: none;}") as demo:
         label="System message",
         lines=2
     )
-    max_tokens_slider = gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens")
-    temperature_slider = gr.Slider(minimum=0.1, maximum=2.0, value=0.7, step=0.1, label="Temperature")
-    top_p_slider = gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)")
-    model_dropdown = gr.Dropdown(
-        choices=list(available_models.keys()),
-        value="Zephyr 7B Beta",
-        label="Select Model",
-    )
-
+    max_tokens = gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens")
+    temperature = gr.Slider(minimum=0.1, maximum=2.0, value=0.7, step=0.1, label="Temperature")
+    top_p = gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)")
+    model_dropdown = gr.Dropdown(choices=list(available_models.keys()), value="Zephyr 7B Beta", label="Select Model")
+    
+    # Define the chat interface
     chatbot = gr.Chatbot()
 
-    def user_interaction(message, history, system_message, max_tokens, temperature, top_p, model_name):
-        history = history or []
-        history.append(("User", message))
-        bot_response = next(respond(message, history, system_message, max_tokens, temperature, top_p, model_name))
-        history.append(("Assistant", bot_response))
-        return history, history
-
-    message_input = gr.Textbox(label="Your message")
-
-    message_input.submit(
-        fn=user_interaction,
-        inputs=[message_input, chat_history, system_message, max_tokens_slider, temperature_slider, top_p_slider, model_dropdown],
-        outputs=[chatbot, chat_history],
-    )
-
-    # Add the "Show Updates" button and output area
+    # Add the "Show Updates" button
     updates_button = gr.Button("Show Latest Updates")
 
+    # Button click event
     updates_button.click(
         fn=show_updates_and_respond,
-        inputs=[chat_history, system_message, max_tokens_slider, temperature_slider, top_p_slider, model_dropdown],
-        outputs=[chatbot],
+        inputs=[chat_history, system_message, max_tokens, temperature, top_p, model_dropdown],
+        outputs=chat_history
     )
 
-# Launch the Gradio interface in full screen
+# Launch the Gradio interface
 if __name__ == "__main__":
-    demo.launch(share=True, fullscreen=True)
+    demo.launch(share=True)