Overview job counts cover the whole period
The dashboard overview now counts every job in the period you pick, instead of stopping at your 50 most recent jobs.
The cards on the dashboard overview were computed from your 50 most recent jobs, so a workspace with more than 50 jobs saw the same totals for 30 days, 3 months, 12 months and All Time. Jobs, Success Rate, Avg Duration and Active Jobs now come from a server-side count over the whole period you pick. Time in Queue already worked this way.
Avg Duration now averages completed jobs only, as it did before for those 50. A failed job can end in under a second or run until its timeout, and neither tells you how long a working job takes.
The jobs table also stays live now. A new job’s Created time counts up every second instead of jumping every 30 seconds, and a running job shows how long it has been running instead of a dash.
GET /jobs/stats takes three more values for window: 90d, 365d and all. With all, the window starts at your first job. Windows longer than 90 days return weekly activity buckets. In the SDK, client.jobs.stats({ window: "all" }) works the same way.