ollama wraps llama.cpp into a docker container, correct? Besides that it seeks like a go server for chat?