llama-cpp-python

fix(chat_format): parse Gemma 4 native tool-call tokens into tool_calls (#2227)

#2232

Open