System Status
Real-time probes to 5 endpoints — measured from the edge, no mock data.
No probes recorded yet. Every refresh and every scheduled check is stored — this timeline fills with real data from now on. Nothing is fabricated.
Each refresh hits /api/status which fetches 5 real endpoints in parallel (SVG, PNG, search, categories, list) with no-store and 5s timeout. Latency and ok/fail are measured live — not mocked — and every run is appended to our database, so the 7D / 90D / 12M charts show only probes that actually happened.
→ { status, avgResponseTime, services: [{id, ms, ok}], checks }
Every probe run is written to our own Postgres database (schema hylab_status) — per-day, per-service buckets, a run log and an incident log. It survives restarts, redeploys and machine changes, so the history accumulates for months and years. A daily Vercel Cron (/api/cron/probe) records even with no visitors; opening this page probes every 30s on top.