AI systems
React Native · Expo · TypeScript · DNS
Open source · on the App Store
A Bun monorepo pairing an Expo client with a Cloudflare Workers API, hardened around idempotent writes and honest health reporting.
$ trace "ai-calories-tracker"
A food log is a write-heavy mobile app talking to an edge API, where a retried request must not create a duplicate entry and a broken maintenance job must not look healthy.
Scheduled maintenance tasks have to fail independently, and the health endpoint has to report degradation explicitly instead of returning a green 200.
Independent checks for media garbage collection and idempotency, a health endpoint that answers 503 when either fails, per-task decoupled scheduling with structured logs, and environment parsing with safe fallbacks.
Failure of one maintenance task no longer blocks the other, and the regression suite covers the degraded paths rather than only the happy one. Still in progress and without a published README.
MN://WORK