Using llama.cpp directly has never felt more polished ...
The best part? This voice assistant runs entirely locally ...