Agent runs
DevEx › Agent runs (?tab=runs) is one history of every autonomous agent run, whatever started
it. The agents themselves (what each does, its model, its schedule and arming) are on
DataEx › Agents; the tab strip links there.
| Sub-tab | ?tab= | Reads |
|---|---|---|
| Runs | runs | GET /api/runs |
| Coding sessions | none | /api/session-queue |
| Doctor | doctor | GET /api/doctor |
| CI runners | none | GET /api/ci-runners |
Runs
The page answers two questions, in this order: what did TIDE’s agents do (a rollup by agent, then the timeline), and did it land. Delivered (did a person get something) and lineage (did the commit ship) are kept apart; a green run whose delivery is unknown says so, never “success”.
| Kind | Source |
|---|---|
| workflow agents | the console’s run store, topped up from GitHub Actions |
| pipelines | the run store’s pipeline runs |
| CLI sessions | tide session reports, and your own command queue |
| CronJobs | the ops doctor and the platform’s maintenance CronJobs |
| app agents | the apps’ latest agent-health reports |
Coding sessions
The queue of commands for tide session runs: you hand a command to a machine running
tide session serve, and its runs appear on Runs. See Coding sessions.
Doctor
The ops doctor (doctor ops) runs as a Kubernetes CronJob rather than a workflow, so it keeps
working when the runner pool it watches is down. The tab shows the CronJob’s state (present,
suspended, schedule, last runs), its recent Jobs and its tracking issue. A repair it proposes, such
as deleting a wedged runner pod, waits for an agent admin’s Approve repair here.
CI runners
TIDE’s GitHub Actions runner pools as its own controller manages them: pools, replicas, minimum and
maximum, busy and idle runners, the queue per runs-on label set with its oldest wait, the last
scaling decisions with their reasons, and the recycled runners. These are CI runners, not the
data-access runners on Integrations › Runners.