API Sentinel Fleet — Prometheus Classic

Inspiration

APIs fail silently — dashboards show up/down, drift kills first. Built a fleet that spots drift via prediction-market math, Prometheus-ready.

What it does

API Sentinel Fleet — Prometheus Classic polls N endpoints every 8s, scores via 3-model ensemble (Phi + momentum + staleness HEALTHY→WARN→FAILOVER). Generalized from predictrader-ai.

Live: https://api-sentinel-fleet.vercel.app · GitHub: https://github.com/binasalama12/api-sentinel-fleet

How we built it

• Stack: Next.js 14, TypeScript, Node fetch, Docker (Cloud Run), OpenAPI 3.1
• Infra: Vercel + Cloud Run — Prometheus metrics ready

Challenges

Generalizing trading drift to latency — z-score recalibration.

Accomplishments

One base → 5 sentinels, all live 200 — reuse >90%.

What we learned

Trading infra is API infra.

What's next

Wire Prometheus alerts on FAILOVER, add fleet metrics endpoint.

Built With

• cloud-run
• docker
• next.js
• openapi
• typescript
• vercel