refactor: Improve message handling in bot_response function by streamlining history updates and enhancing user interaction during streaming.

app.py

@@ -9,27 +9,24 @@ MODELS = [
 ]
 
 def bot_response(user_message, history, model_name, system_message, max_tokens, temperature, top_p, oauth_token: gr.OAuthToken | None, local_endpoint: str):
-    if not user_message:
+    if not user_message or user_message.strip() == "":
         yield history, ""
         return
         
     token = oauth_token.token if oauth_token else None
     
-    # --- Logic switch: local vLLM vs Hugging Face Hub ---
     if model_name.startswith("Local-"):
         client = InferenceClient(base_url=local_endpoint, token="vllm-token")
     else:
         client = InferenceClient(token=token, model=model_name)
     
-    # 1. Construct messages for API
-    api_messages = [{"role": "system", "content": system_message}] + history + [{"role": "user", "content": user_message}]
+    # 1. Append the new user message to history
+    history.append({"role": "user", "content": user_message})
+    # 2. Append an empty assistant message to be filled by the stream
+    history.append({"role": "assistant", "content": ""})
     
-    # 2. Update UI history (Gradio 5/6 format)
-    new_history = history + [
-        {"role": "user", "content": user_message},
-        {"role": "assistant", "content": ""}
-    ]
-    yield new_history, "" 
+    # Prep the API payload (history now includes the new user message)
+    api_messages = [{"role": "system", "content": system_message}] + history[:-1]
 
     try:
         response_text = ""
@@ -44,12 +41,17 @@ def bot_response(user_message, history, model_name, system_message, max_tokens,
         for chunk in stream:
             token_content = chunk.choices[0].delta.content or ""
             response_text += token_content
-            new_history[-1]["content"] = response_text
-            yield new_history, ""
+            # Update the last message in history (the assistant's content)
+            history[-1]["content"] = response_text
+            # Yield history and keep the input box text as is during streaming
+            yield history, gr.update(interactive=False) 
+
+        # Final yield to clear the input box and make it interactive again
+        yield history, ""
             
     except Exception as e:
-        new_history[-1]["content"] = f"**Error:** {str(e)}"
-        yield new_history, ""
+        history[-1]["content"] = f"**Error:** {str(e)}"
+        yield history, ""
 
 with gr.Blocks() as demo:
     with gr.Sidebar():