Run
A [run](../glossary.md#run) is one [task](../glossary.md#task) going through the pipeline, in a copy of your repo.
A run is one task going through the pipeline, in a copy of your repo.
What it is
A run is one task going through the pipeline. Adding the task file queues a run. Serving the factory is what moves that run along. The run is not your checkout: it works in a worktree, a separate copy of each repo the task named, taken from the base branch. Your working tree stays as you left it.
The same task can have more than one run. The first add is run 1. If you requeue, you get run 2 of that task, on a new branch, with your note as feedback. A run also has tries inside it: a step may be attempted again, and a loop may repeat its group as try 2 of 3. Those tries are still the same run. A new run is a new pass through the pipeline.
When the run needs a person, it waits. When the work is done, delivery is a local commit on a branch the factory created. You merge. You push. The copy of the repo is removed when the run no longer needs you; the branch stays.
What you see
States, in words you will see on the Dashboard and in status:
- Queued — waiting to start
- Running — a step is in progress, including wrapping up
- Waiting for you — a gate
- Needs attention — the run stopped for a person (out of tries, a rejected gate, a partial delivery)
- Delivered — the pipeline succeeded
- Failed — a step failed and the run ended
- Cancelled — stopped or rejected
The runs list is those states for a factory. Open a run and you see the task title, the path through the pipeline, the current try, cost and time, and the evidence. Waiting time at a gate does not count toward the run's time limit.
The worktree is a copy. You can open it while the run needs attention. Do not use it as your daily checkout. Take over if you want that copy left in place and the run stopped so you can finish by hand.
Attempts show up as retries on a single step and as tries on a loop. A resume of a partial delivery is the same run, failed repos only. A restart is the same run from the base commits, with the previous tips kept aside. Neither of those is a new task.
What you decide
You decide when to add the task, when to pause or stop, and what to do with a run that needs attention: requeue with a note, resume the failed repos, restart from the base, dismiss, or take over.
You decide when a waiting run should continue, by approving or rejecting the gate, and whether a delivered branch is work you want to merge. Do not confuse the copy with your checkout.
You do not have to watch a running step. kusaw serve runs the pipeline; the Dashboard is how you inspect and recover.
Where to go next
Meet a run as a row on the Dashboard, as kusaw add then status in the CLI, and as a branch named for the task. The worktree is the copy.