It seems that this is possibly not necessary, since LLaMA.cpp already integrates Jinja with CPP implementation (through minja)