Update app.py

app.py

@@ -9,32 +9,43 @@ model_list = {
 }
 
 def respond(message, history, system_message, max_tokens, temperature, top_p, selected_model):
-    # Create an InferenceClient for the selected model
-    client = InferenceClient(model_list.get(selected_model, "HuggingFaceH4/zephyr-7b-beta"))
-    
-    # Build conversation messages for the client
-    messages = [{"role": "system", "content": system_message}]
-    for user_msg, assistant_msg in history:
-        if user_msg:  # Only add non-empty messages
-            messages.append({"role": "user", "content": user_msg})
-        if assistant_msg:  # Only add non-empty messages
-            messages.append({"role": "assistant", "content": assistant_msg})
-    messages.append({"role": "user", "content": message})
-    
-    response = ""
-    
-    # Stream the response from the client
-    for token_message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = token_message.choices[0].delta.content
-        if token is not None:  # Handle potential None values
-            response += token
-            yield response
+    try:
+        # Create an InferenceClient for the selected model
+        client = InferenceClient(model_list.get(selected_model, "HuggingFaceH4/zephyr-7b-beta"))
+        
+        # Build conversation messages for the client
+        messages = [{"role": "system", "content": system_message}]
+        for user_msg, assistant_msg in history:
+            if user_msg:  # Only add non-empty messages
+                messages.append({"role": "user", "content": user_msg})
+            if assistant_msg:  # Only add non-empty messages
+                messages.append({"role": "assistant", "content": assistant_msg})
+        messages.append({"role": "user", "content": message})
+        
+        response = ""
+        
+        # Stream the response from the client
+        for token_message in client.chat_completion(
+            messages,
+            max_tokens=max_tokens,
+            stream=True,
+            temperature=temperature,
+            top_p=top_p,
+        ):
+            # Safe extraction of token with error handling
+            try:
+                token = token_message.choices[0].delta.content
+                if token is not None:  # Handle potential None values
+                    response += token
+                    yield response
+            except (AttributeError, IndexError) as e:
+                # Handle cases where token structure might be different
+                print(f"Error extracting token: {e}")
+                continue
+    except Exception as e:
+        # Return error message if the model call fails
+        print(f"Error calling model API: {e}")
+        yield f"Sorry, there was an error: {str(e)}"
 
 # Custom CSS for branding with blue, teal and gold accents
 css = """
@@ -174,7 +185,8 @@ with gr.Blocks(css=css) as demo:
             chatbot = gr.Chatbot(
                 label="Conversation", 
                 show_label=True,
-                avatar_images=("👤", "🔒"),
+                # Removed avatar_images which may not be supported in some Gradio versions
+            elem_id="chatbot",
                 height=500
             )
             with gr.Row():
@@ -206,7 +218,9 @@ with gr.Blocks(css=css) as demo:
         return "", history + [[user_message, None]]
     
     def bot(history, system_message, max_tokens, temperature, top_p, selected_model):
-        # Get the last user message from history
+                    # Get the last user message from history (with error checking)
+        if not history or len(history) == 0:
+            return history
         user_message = history[-1][0]
         # Call respond function with the message
         response_generator = respond(
@@ -223,7 +237,7 @@ with gr.Blocks(css=css) as demo:
             history[-1][1] = response
             yield history
     
-    # Wire up the event chain
+    # Wire up the event chain - use queue=True for the bot responses
     user_input.submit(
         user,
         [user_input, chatbot],
@@ -232,7 +246,8 @@ with gr.Blocks(css=css) as demo:
     ).then(
         bot,
         [chatbot, system_message, max_tokens_slider, temperature_slider, top_p_slider, model_dropdown],
-        [chatbot]
+        [chatbot],
+        queue=True
     )
     
     send_button.click(
@@ -243,11 +258,15 @@ with gr.Blocks(css=css) as demo:
     ).then(
         bot,
         [chatbot, system_message, max_tokens_slider, temperature_slider, top_p_slider, model_dropdown],
-        [chatbot]
+        [chatbot],
+        queue=True
     )
     
-    # Clear the chat history
-    clear_button.click(lambda: None, None, chatbot, queue=False)
+    # Clear the chat history - use a proper function instead of lambda
+    def clear_history():
+        return []
+    
+    clear_button.click(clear_history, None, chatbot, queue=False)
 
 if __name__ == "__main__":
     demo.launch()