Skip to content

llama-cpp

Version 2026.185.1930
Repo superproject

llama.cpp prebuilt binaries and GGUF conversion tools The llama-cpp candy downloads the latest llama.cpp release into ~/llama.cpp: the llama-quantize and llama-cli binaries with their shared libraries, plus the convert_hf_to_gguf.py script and the gguf-py package for converting HuggingFace models to GGUF. Every artifact lands at a fixed path under ~/llama.cpp, so its presence and executability are directly checkable.

This candy’s plan: — the runnable spec charly check executes against a live deployment. check: steps are idempotent probes; run: steps change state.

Intent Step
run command=install llama.cpp prebuilt ubuntu-x64 binaries + matching source GGUF tools into ~/llama.cpp (newest release that actually ships a binary)
check the llama-quantize binary is present under ~/llama.cpp
check the llama-cli binary is present and executable
check the HuggingFace-to-GGUF conversion script is present
check the gguf-py conversion package directory is present
agent-check running “llama-cli –version” prints a llama.cpp version string, confirming the prebuilt ubuntu-x64 binary executes on this host
check file=${HOME}/llama.cpp/llama-cli
check command=test -x ${HOME}/llama.cpp/llama-cli && test -x ${HOME}/llama.cpp/llama-quantize && echo ok
check command=ls ${HOME}/llama.cpp/lib*.so* >/dev/null 2>&1 && echo ok