Targeted recovery
Re-run one job or only the failed jobs, cancel a queued or active job, and delete terminal runs when their record is no longer needed.
CI/CD
A pipeline is YAML checked into the repo. The server discovers it, schedules its jobs, and dispatches them to CI agents you run — then the results report back to the branch as commit statuses.
The pipeline
Jobs declare which agent they need, what they depend on, and an ordered list of steps. Matrixed jobs fan out across values, each step records its status, exit code, and timing, and a run moves from queued to running to a final outcome you can watch live.
ci/build, ci/testYour own agents
Jobs run on agents you register — nothing executes on hardware you don't own. An agent registers once with a one-time token, then pulls work it's eligible for. Free-text labels route each job to the right machine.
linux, gpu, or prod-readonly match a job's requirement to an online agent.Cross-repo builds
A job can declare that it needs an artifact from the registry before it dispatches — even one produced by a different repository. The coordinates resolve when the run is created, a version prefix like 6.0.* matches any patch, and the job waits, then dispatches once the last requirement is published.
requires:
- artifact: core-server
version: 6.0.*
- artifact: web-bundle
version: 6.0.*
# waits, then dispatches on publishOperate and release
CI operations stay at the level where the problem happened. Restart a failed job without replaying successful work, stop work that should no longer run, and move artifacts through release and promotion gates with the evidence attached.
Re-run one job or only the failed jobs, cancel a queued or active job, and delete terminal runs when their record is no longer needed.
Put a cron schedule beside the jobs in repository YAML and let Bosca create the run at the appointed time.
Publish native release artifacts, promote an existing build, and keep approval and verification gates in the pipeline that ships it.
Keep exploring