forked from wylab/llama-swap
980a4e6baa
- Login to git.wylab.me instead of ghcr.io - Use Gitea-hosted llama.cpp-rocm base image instead of ghcr.io - Rewrite fetch_llama_tag to use anonymous OCI registry API - Add LS_UPSTREAM for release binary fetches on forks - Add REGISTRY and BASE_TAG overrides for self-hosted builds - Only build rocm platform Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>