Update response perf.

app.py

@@ -66,7 +66,6 @@ def get_chat_response(chatbot, task_history, model, tokenizer):
             history_filter.append((pre, a))
             pre = ""
     history, message = history_filter[:-1], history_filter[-1][0]
-    
     response, history = model.chat(tokenizer, message, history=history)
     image = tokenizer.draw_bbox_on_latest_picture(response, history)
     if image is not None:
@@ -83,8 +82,10 @@ def get_chat_response(chatbot, task_history, model, tokenizer):
         if chat_response != "":
             chatbot.append((None, chat_response))
     else:
-        chatbot[-1] = (format_text(chat_query), response)  # Hier verwenden wir format_text statt _parse_text
-    full_response = format_text(response)  # Hier verwenden wir format_text statt _parse_text
+        chatbot[-1] = (format_text(chat_query), response)
+    full_response = format_text(response)
+    task_history[-1] = (query, full_response)
+    return chatbot
 
     task_history.append((message, full_response))
     chatbot.append((format_text(message), format_text(full_response)))  # Hier verwenden wir format_text statt _parse_text