Apps & data pipelines
jev-oncall routes production alerts with one typed Jev call per alert, then lets boring, auditable code decide what the probabilities are allowed to do. The model judges; a page threshold, a review band, and a fail-open default do the rest.
Only you can see your notes.
Recording unavailable. Open the experiment ↗
Someone clearly ran this on a real pager: jev-1.13.0 is pinned instead of an alias because aliases drift, every threshold is labeled a starting point, and a triage rotation catches alerts when Jev can't be asked. The replayed demo shows alert #5 hitting P(page) = 1.00 yet going to human review — the checkout incident it links to belongs to another team, and paging them twice helps nobody.
Try this tool's comparison mode if your app sends too many warnings. It records what it would do so you can compare that with your current rules. A developer can set it up. Start by checking whether several alerts come from one problem. The browser demo only shows decisions; it sends no notifications.