llama.cpp

53d0a126 - server : allow specifying reasoning_format in HTTP request (#15238)