Module download 

Resolving models from Hugging Face and Docker, through llama.cpp’s own cache.

This crate’s examples use the hf-hub crate, which keeps its own cache. These functions share the cache llama.cpp’s CLI tools use, so a model pulled by llama-cli is found here and vice versa.

Functions

list_cached_json

Every model in llama.cpp’s cache, as a JSON array of {"repo","tag","name"} objects.

remove_cached

Delete a cached model. Returns whether anything was removed.

resolve_docker

Resolve a Docker model reference to a local path.

resolve_hf

Resolve repo[:tag] to a local path, downloading if needed.

split_repo_tag

Split repo:tag into its parts.