Platform Health & Public Status: Green Means Something

Health checks are not glamorous. That is their charm. They tell platform staff whether the app, database, auth, storage, mail, the notification queue, cron, webhooks, the public API, billing, the external secret store, the activity audit and observability are behaving well enough for normal work. Thirteen components, one question each: is this thing doing its job right now?

/api/health

The technical health endpoint returns app readiness with component checks, latency and sanitized status. It is built for uptime monitoring and deploy smoke checks, not for leaking database details to the internet.

Platform cockpit

/admin/platform/health is the incident console for platform owners and staff; platform viewers do not see the sidebar item and are redirected if they open the address, and organization admins cannot open it at all. It shows the global status, release readiness, a support summary, likely causes, status per dependency with its category — platform, data, external, tenant integration, billing, queue or evidence — the number of affected tenants where that is safe to derive, the last check, last success and last error, and recent job runs. Cron routes and background jobs write bounded job-run records so the platform can answer whether a scheduled task ran, failed, was blocked, ran partially or quietly took a nap.

Words, not codes

The console speaks support language. Internal severity codes are replaced by impact labels — urgent outage, high impact, incident follow-up needed, low-priority review — and two states exist precisely so nobody panics at the wrong card: “not yet configured” means a dependency is deliberately absent or has no direct monitor, which is not an outage, and “could not be checked” means the page could not read or verify the source data. Billing may report that subscription billing is not configured while local package enforcement keeps working. Each component card links to the most relevant admin, settings or health page. When support opens a ticket during an incident, a snapshot of this context can be attached to the ticket automatically.

Public status

/api/status/public exposes a sanitized component feed for a status page or external monitor. It can say that file storage or scheduled jobs are degraded without revealing tenant names, payloads, secrets or stack traces. Transparency is good; oversharing is still oversharing.

Why it matters

When something is wrong, support needs a shared source of truth. Health turns “is it just me?” into “storage is degraded, uploads are affected, everything else is fine.” That sentence saves a surprising amount of oxygen.