Aligns the project guides with the branching/CI/environment changes landed in the previous commits: - CLAUDE.md: per-stage CI gate now closes against gitea.lan; describes the main/development/feature/* flow and the workflow surface - docs/ARCHITECTURE.md: new section 18 "CI and Environments" covering branches, workflows, and the local-dev / dev-deploy / local-ci triad; section numbering shifted accordingly - tools/local-ci/README.md: marked as fallback (offline / runner isolation only) - tools/local-dev/README.md and ui/README.md: cross-link to tools/dev-deploy/ for production-shaped testing Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Local Gitea CI (fallback)
Status: fallback / opt-in. The primary CI target is now
gitea.lanwith its host-modeact_runner. The per-stage CI gate closes againstgitea.lan, not against this stack. Use this directory when you want to validate.gitea/workflows/*without reachinggitea.lan— for example, when iterating on a workflow file from a flight without LAN access — or when isolating a runner issue from production-shaped infrastructure.
Self-contained Gitea + Actions runner for verifying
.gitea/workflows/* honestly before pushing to gitea.lan. Runs
natively on arm64 (Apple Silicon) — every image below has an arm64
variant, so Docker pulls the right architecture and the runner
executes workflow steps without QEMU emulation.
Prerequisites
- Docker (Colima or Docker Desktop)
python3,curl,bash— all built into macOS
First time
make -C tools/local-ci up
This:
- brings up the Gitea container;
- creates an admin user (
galaxy/galaxy-dev); - creates the
galaxy/galaxyrepo; - fetches a runner registration token from the Gitea API;
- brings up the runner with that token (the runner persists its credentials in a Docker volume and ignores the token on subsequent restarts).
The script is idempotent — re-running it is safe.
Pushing a branch
make -C tools/local-ci push
This adds a local-gitea remote on the first run and then pushes the
current HEAD. Equivalent manual flow:
git remote add local-gitea \
http://galaxy:galaxy-dev@localhost:3000/galaxy/galaxy.git
git push local-gitea HEAD
The Tier 1 workflow fires on push to any branch and the Tier 2
workflow fires on tags matching v*. Watch runs at:
http://localhost:3000/galaxy/galaxy/actions
Operational targets
| Target | What it does |
|---|---|
make up |
Bring up Gitea + runner (idempotent) |
make down |
Stop both containers (state preserved) |
make logs |
Tail logs from both containers |
make status |
Show container status |
make push |
Push current HEAD to local Gitea |
make clean |
Stop and wipe all local state (full reset) |
What's in the box
| Component | Image | Role |
|---|---|---|
| Gitea | gitea/gitea:1.23 |
Server with SQLite backend |
| act_runner | gitea/act_runner:0.6.1 |
Single-capacity runner registered on boot |
| Workflow | catthehacker/ubuntu:act-latest |
Image spawned per job (multi-arch) |
The runner mounts the host Docker socket and spawns workflow
containers on the same Docker network as Gitea, so
actions/checkout reaches the server at http://gitea:3000 from
inside spawned containers.
Caveats
- Gitea's
ROOT_URLis set tohttp://gitea:3000/so spawned workflow containers reach the server through the compose network. The web UI works athttp://localhost:3000via port mapping, but copy-paste URLs in the UI may showgitea:3000instead oflocalhost:3000. Harmless for local dev; switch the host part by hand when copying. - The runner is single-capacity (
runner.capacity: 1inconfig.yaml). Concurrent jobs queue. Bump if you need parallel jobs. - First push from a fresh checkout uploads the full repo history (~tens of MB). Subsequent pushes are deltas.
actions/upload-artifact@v4requires Gitea ≥ 1.21 — we pin1.23to stay above the cutoff.- Workflow steps run as
rootinside the spawned container; this matches the upstream catthehacker behaviour. Keep that in mind if you add steps that touch host-mounted directories. - On Apple Silicon the runner image and its catthehacker child run
natively as arm64. Some pre-built tools that ship in the image are
amd64-only and would fall back to QEMU;
setup-go,setup-node, andpnpm/action-setupall download arm64 binaries themselves, so the workflow steps we care about stay native.