Insights (/app/<workspace>/insight, sidebar) shows how your flows are doing on the engine you selected.
What it shows
- Period: the last 7, 30 or 90 days (
/workspace/:id/stat?days=), and Include test runs (real runs only by default). - Tiles: Runs, Success rate, Failed, Average run, AI (tokens and cost).
- Runs per day: a bar per UTC day; hover for
12 runs · 10 succeeded · 1 failed. - Per flow table: Flow, Runs, Success rate, Failed, Average run, Last run, AI.
- Per campaign table: the same per (its own runs, called flows' own runs and row actions), when any ran.
Success rate = succeeded ÷ runs. Average run counts finished runs only.
Where the numbers come from
The engine aggregates its Run records for your workspace (). AI cost uses the prices on your AI keys; keys without prices add tokens only.
Monitoring the engine itself
For operators: the engine's endpoint gives Prometheus gauges (browsers by state, active and waiting runs, runs and AI use in the last 24 hours).
FAQ
Why is the success rate low on an endless flow?
Endless flows end only when you stop them, so they count as Cancelled. Look at failures, not the rate.