Optimize the display of thinking messages (#2747)

Author: openvino-dev-samples

notebooks/deepseek-r1/gradio_helper.py

@@ -4,6 +4,7 @@
 from threading import Event, Thread
 import queue
 import sys
+import re
 
 max_new_tokens = 256
 
@@ -206,6 +207,8 @@ def default_partial_text_processor(partial_text: str, new_text: str):
         updated text string
 
         """
+        new_text = re.sub(r'^<think>', '<em><small>I am thinking...', new_text)
+        new_text = re.sub('</think>', 'I think I know the answer</small></em>', new_text)
         partial_text += new_text
         return partial_text