The trial is widening. Subjects from outside the lab’s local population are getting the wrong reading on their chart, and the lab director just walked in with a stack of complaint forms — she wants the audit log to tell her, after the fact, exactly which vision_state the lab recorded for which subject. Your job: populate all three OpenFeature evaluation-context layers (global, transaction, invocation) on the Spring Boot lab, register a custom audit Hook, and prove the lab can target the right reading per cohort — of course without setup headache again.
What You’ll Learn
- OpenFeature’s three evaluation-context layers — global, transaction, invocation — and the precedence rule between them (invocation > transaction > global)
- Spring
HandlerInterceptorfor per-request context, including why aThreadLocalTransactionContextPropagatoris the right primitive for Servlet-based apps - Authoring your own OpenFeature
Hook— cross-cutting behaviour (audit logging today, OpenTelemetry tracing tomorrow) without modifying every flag-evaluation call site - How a fixed PII-safe attribute allowlist keeps audit logs honest
Deadline
Tuesday, 26 May 2026 at 23:59 CET
You can still complete the challenge after this date, but points will only be awarded for submissions before the deadline.
Get Started
Discussion
Share your solutions, your questions, and the cohort attribute you forgot to clear in afterCompletion (it’s always one of them). Looking forward to seeing how you wire the contexts together.
Other levels
Beginner: Stand up the lab — already live
Expert: Phase 3 — read the chart (coming soon)
