Integrations
jevtraces brings the same annotation idea to spans: an alpha OpenTelemetry processor asks Jev whether an operation looks worth diagnosing, whether it's business-critical, and whether to keep it, then caches the verdicts and stamps them on later matching spans. Every span is retained regardless.
Screenshot unavailable. Open the experiment ↗
The limits are stated more clearly than most projects manage: it judges operation metadata, never whole traces, and the README explicitly warns that a low score is not permission to discard anything. An opt-in experiment combines the annotations with tail sampling while keeping a full archive branch, and the eval set deliberately includes reworded operations and an unknown call.
A developer can adapt this project to label your application's performance events. The tool sends details about specific actions, like a database search, to Jev. Jev is an outside AI service. It guesses if the action is critical or helpful for fixing errors. The software then attaches these answers to the event records.
To start testing, a developer provides a configuration file. They also need an access key that connects the software to Jev. The tool protects privacy by sending only specifically approved details. It drops unlisted information like web addresses unless a developer adds them. The software always keeps every event, regardless of the service's opinion.