Issues and grouping

1How the fingerprint is made

If the SDK sends a fingerprint array, that wins. Otherwise Blipit hashes the exception type plus the last five in-app stack frames (file name and function). When there are no in-app frames it uses the exception type and the message with numbers, hex addresses and UUIDs replaced, so 'order 4821 not found' and 'order 9 not found' group together. Events without an exception group by the normalised message.

2Change grouping with rules

Open the project page and add a grouping rule. Rules run in order and the first match wins. A rule matches on exception type equals, message regex, stack frame file name or function contains, or tag equals. Group puts every matching event into one issue; Ignore drops the event before it is stored. Rules take up to a minute to apply. Regexes are limited to 200 characters, must compile, and may not use back-references, lookarounds or nested repeats such as (a+)+.

3Resolve, ignore, reopen

An issue is unresolved, resolved or ignored. Resolve it when you ship a fix: if the error comes back, the issue reopens as a regression and you are alerted. The issue says whether it came back in the same release (the fix never shipped) or in a later one (something reintroduced it). Ignore keeps counting events but stops alerts. Reopen sets it back to unresolved. The Issues list has bulk actions, and alert emails carry Resolve and Ignore links that ask for confirmation and work once.

4Snooze

From the issue page, snooze for 1 hour, 24 hours, 7 days, or the next 100 or 1,000 occurrences. Events are still recorded and counted; new-issue and regression alerts are skipped until the snooze ends. Security alerts are not affected.

Good to know

• Blipit keeps the full payload for an issue's first event, every regression, every security event and the first 20 events in each UTC hour. Other events store a summary (id, time, platform, level, release, environment, tags, user id, SDK). The issue page says 'Showing full details for N of M events'; the stack trace, breadcrumbs and AI fix read the latest full payload.
• Past your plan's monthly event cap, 1 event in 10 is kept and the rest are counted as sampled out; nothing is billed. The Usage page shows the sampled count.
• Alerts are budgeted at 30 new-issue or regression alerts per project per hour, so an error storm cannot flood your inbox.