mirror of
https://github.com/kevinveenbirkenbach/computer-playbook.git
synced 2026-11-10 04:36:40 +00:00
kevinveenbirkenbachandClaude Opus 5
d7533be83c
lmstudio/llmster-preview publishes amd64 only, so the role was pinned to x86_64 and arm64 runners could not deploy it. LM Studio does ship a linux-arm64 llmster release, and it carries its own CPU llama.cpp engine, so the role builds its own image instead of pulling the vendor one. install.sh picks the release for the build machine's architecture, verifies its sha512 and trims the bundle to the CPU layout by dropping the nvidia and vulkan backends; the build fails when no llama.cpp engine survives. daemon-run.sh starts llmster, waits for its ready marker and then the lms server. On amd64 the result matches the upstream image in user, workdir, entrypoint, port and all 82 entries under /app, and answers /lmstudio-greeting on the container address rather than loopback. The pin moves from the floating `cpu` tag to app_version, which ends the silent re-push that once made the daemon bind 127.0.0.1. The healthcheck moves out of the image onto the platform's curl flavor, and both architecture restrictions are gone, which retires the x86 pin test. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Integration Tests 🔗
See integration.md for the SPOT: framework, requirements, directory structure, and make test-integration usage.