Retrace Kit logo Retrace Kit
Sign up

Documentation

Incidents

An incident groups different issues that share a route, a release, or a session — so you triage one problem instead of a scattered issue list.

Issue vs incident

  • Issue — one fingerprint. The same exception message and stack pattern, counted many times. Example: TypeError: Cannot read 'id' of undefined on checkout.
  • Incident — several different issues on the same surface: a URL path, a release string, or a session id. Example: null cart item, failed payment fetch, and a currency error on /checkout in the same window.

Resolving an issue closes one crash type. Resolving an incident marks the whole cluster handled — linked issues stay attached for context.

How they form

On each new error, Retrace Kit runs three correlation rules. When enough distinct issues share a surface inside a rolling ~24 hour window, it creates or updates an incident and links those issues.

Correlation rules

Route

At least two different issues on the same normalized URL path (for example /checkout). Useful when one page fails in several ways at once.

Release

At least three different issues with the same init({ release }) value. Useful after a deploy that introduces a cluster of new failures. Set release in production if you want this rule.

Session

At least two different issues in the same browser sessionId. Useful when one visit hits a cascade of failures. Session incidents promote to Active as soon as they open.

Severity (low / medium / high) follows affected users and sessions. Route and release incidents start as Candidate and move to Active when impact crosses a threshold (more users, sessions, or linked issues).

Lifecycle

  • Candidate — cluster detected; watching impact.
  • Active — open work (auto-promoted by impact, or immediately for session).
  • Investigating — set in the app when you are working on it.
  • Resolved — cluster handled.
  • Reopened — related errors return within the reopen window (default 24 hours).

When new events stop for a while, the app can suggest resolve so quiet incidents do not sit open forever.

In the app

  • Incident list with status, severity, surface, and linked issue count
  • Detail view: linked issues, users / sessions affected, environments
  • AI explanation on the incident (title and summary of the cluster), plus AI summaries on each linked issue
  • Included on Free, Plus, and Pro — see pricing