Skip to main content

Health check

Langbly has two health check endpoints: a public liveness check and an internal, authenticated dependency check.

GET /health​

This public, unauthenticated endpoint returns a simple liveness response. It does not call the database, Redis, or perform any translation. It responds within milliseconds and is guaranteed not to fail due to downstream dependency issues.

{
  "ok": true,
  "service": "langbly-workers",
  "region": "global",
  "timestamp": "2025-04-02T12:00:00Z"
}
  • The region field is "global" on the global endpoint (api.langbly.com) and "eu" on the EU endpoint (eu.langbly.com).
  • The timestamp is an ISO 8601 date string.

GET /health/deep (internal, requires admin authentication)​

A deeper health check that tests the database, Redis cache, and performs a real translation to confirm end-to-end functionality.

  • Requires an Authorization: Bearer <ADMIN_API_KEY> header. Not intended for regular API users, only for internal or operational monitoring.
  • Each dependency returns "ok", "fail", or "skip" (if not configured).

Frequently asked questions​

Why doesn't /health check the database?​

The previous implementation included database and Redis calls in the public health endpoint. This caused load balancers to mark the service as unreachable when a slow or failing dependency occurred, even though the worker process itself was healthy. /health now only confirms the process is alive and responding, isolating dependency issues to separate monitoring paths.

Is there a public deep health check?​

No. The GET /health/deep endpoint requires admin authentication and is intended solely for internal operational monitoring. Regular API users should use GET /health for liveness checks and rely on other endpoints (like translation API calls) to verify end-to-end functionality.

How does the region differ between endpoints?​

When calling the global endpoint (api.langbly.com), the region field is "global". When using the EU endpoint (eu.langbly.com), it returns "eu". No other regions are currently supported.

Can I use /health to monitor my application's connectivity to Langbly?​

Yes. Because /health is fast and unauthenticated, it is ideal for uptime monitoring and load balancer health checks. A successful response ("ok": true) confirms that the Langbly workers are running and reachable. However, it does not verify the availability of dependent services like the database or cache. Use a test translation request for full end-to-end verification.