Kiac v0.13.0 adds a native Metal service for Apple silicon and makes local GPU workloads easier to inspect.
- Native Metal alpha. A Linux Pod on an ordinary Apple/container node can use a Kubernetes DRA claim to request fixed Metal compute or bounded llama.cpp inference from a service on the Mac. The service runs on macOS and grants one exclusive Kiac lease. Inference requires a Metal-enabled
llama-serverand a local GGUF model. The Venus/Vulkan GPU backend remains in beta. #93 - GPU dashboard and inference lab. See Mac GPU and Metal capabilities, worker health, and DRA reservations in
kiac ui. The inference preset gives the worker more CPU and memory. The lab verifies inference again after restarting the GPU worker. #91 · #92 - Standalone kubeconfig export.
kiac get kubeconfig --name devprints a config for a single cluster without modifying your existing kubeconfig. It preserves IPv6 endpoints for IPv6-only clusters. Thanks to @vishalkotak for the contribution. #95
Install with brew install --cask saiyam1814/tap/kiac. For an existing installation, run brew update && brew upgrade --cask kiac.