Make a Service Recover Like Production Should

Fix a health check and crash-recovery setup that both lie.

A advanced AI Engineering challenge worth 20 points. Solve it hands-on in a real AI Engineering environment in your browser - no local setup, no fake shells.

The Challenge

Scenario: A FastAPI service reports itself "ready" even when its dependency is down, and its supervisor never restarts it after a crash - so nobody finds out until a real user does.

Make the readiness endpoint honestly check its dependency and report HTTP 503 when it's down, then add bounded-delay automatic restart so a crashed process comes back with a new process ID. Run check.