Skip to main content

​
Deployment

The engine is a single static-ish binary with no runtime service dependencies. It runs the same way on Linux, in containers, and on Windows.

​
Release Binaries

Tagged releases publish standalone, statically linked binaries for Linux, macOS, and Windows (using musl on Linux and static CRT on Windows). For binary downloads, checksum verification, and desktop installer guides, see Installation and Quick Start.

​
Container Deployment

The container image contains only blockscope with no external runtime dependencies or assets. Client jars, resource packs, and job directories must be mounted at runtime. For container build commands, volume mounting, worker execution, and supervisor setup, see the Docker Guide and Service Mode. Always verify container sizing using blockscope capacity against the host cgroup limits before deploying.

​
Linux via WSL

WSL2 with Docker Desktop, or a plain Ubuntu distro, is enough for a local Linux check:
To verify cgroup-aware sizing without Docker, use a transient systemd scope:
Expected output: cores: 1, cpuQuotaMilli: 1000, memoryLimitBytes: 1073741824, suggestedWorkers: 1, suggestedThreads: 1, default tier 1024x1024 @ ss1.

​
CI

.github/workflows/ci.yml runs two jobs:
  • test on ubuntu-latest, windows-latest, and macos-latest: fmt, clippy with -D warnings, and the full workspace test suite.
  • container on ubuntu-latest: builds the image, then runs capacity and a protocol smoke test inside --cpus=1 --memory=1g so cgroup parsing is exercised on every push.
.github/workflows/release.yml runs on a v* tag: it checks that the tag matches [workspace.package] version, creates the GitHub Release from CHANGELOG.md, and uploads the five archives with sha256 files. Running the workflow manually builds and packages the same targets with dry-run, so the pipeline can be verified before the first tag exists.

​
Capacity reference

blockscope capacity also reports suggestedMaxQuads: 80 percent of available memory divided by roughly 400 bytes per quad. The default mesh cap is the lower of 5,000,000 quads and that budget, so a 1 GiB container rejects oversized schematics earlier than a 32 GiB host. --force-limits bypasses the cap; --lod auto uses the budget to pick a factor. These values come from the formulas in blockscope capacity; override them with --workers, --threads, or --memory-budget when the automatic value is wrong for a deployment.