Skip to main content
Proactive detection is on our roadmap, coming within the Growth tier. Once available, it will watch your telemetry continuously and open low-confidence incidents when it spots statistical anomalies — often minutes before your alert thresholds would fire. By the time your real alert arrives, sureops would already have context from the early detection. This page describes the intended behavior so you know what to expect. Nothing here is available in the product today — there is no Settings toggle, no sensitivity tuning UI, and no filter-bar option for it yet.

​
Why it’ll be useful

Most incident detection today is reactive: something breaks, an alert threshold trips, and only then does the response begin. Proactive detection is designed to close that gap by comparing live telemetry against each service’s own historical baseline, so a real anomaly can start getting diagnosed before it becomes customer-visible. The plan is for detection to run continuously, compare current metric values against a rolling baseline for the same service, and open a distinct, clearly-marked low-confidence incident when the deviation is significant enough — without replacing or delaying your existing alerting.

​
What we’re planning to monitor

The default metric set we’re planning to watch is small and latency/error focused:
  • User-facing request latency (p95)
  • Overall request volume/error rate
  • Inter-service (RPC) latency (p95)
Tuning (if shipped) is planned to be organization-wide, not per-service — a single sensitivity setting per org rather than per-service knobs.

​
How you’ll be able to respond

The design calls for three responses when a proactive incident appears: confirm it’s real, dismiss it as a false alarm, or take no action and let it resolve on its own if no real alert follows. Confirming or dismissing is intended to feed back into the confidence model so detection gets more accurate for your environment over time.

​
Current status

Proactive detection does not have a shipped, customer-facing surface yet. Internally, the statistical core (baseline comparison + anomaly scoring) exists as an early foundation, but it isn’t wired to live telemetry and isn’t reachable from any UI. There’s nothing to enable, configure, or suppress in your .sureops/services.yaml for this feature today — any instructions suggesting otherwise are out of date. We’ll update this page with real setup instructions once proactive detection ships.