Hey, some llama.cpp folders exists under /models folder. I didn't find a cause why this exists...
You can load a local gguf model and use it for NPC responses in free dialogue mode instead of the markov chain algorithm
