Llama.cpp guide – Running LLMs locally on any hardware, from scratch