llama.cpp

d983516f - `tool-call`: let the tool call handler expand chat template, moving builtin_tools down as extra_context