Status
How to check whether the API is up, from your own monitoring.
Health endpoint
curl https://api.cognitivers.com/health
# {"status":"ok"}Public, no credentials, safe to call from a monitor every thirty seconds. It answers 200 with a
JSON body when the API is serving, and a non-200 when it is not. Treat anything other than 200
with a parseable body as unhealthy rather than reading the body's wording.
A monitor that does not cost anything
A health check that requires a key will eventually page you when a key expires rather than when the service breaks. Use the health endpoint for availability, and a real one-token request with a dedicated key at a lower frequency if you want to prove that end-to-end inference works, not just that the front door is open.
Two checks, two questions:
| Check | Question it answers | Frequency |
|---|---|---|
GET /health | Is the API reachable and serving? | Every 30 seconds |
| One small chat request | Does a request reach a model and come back? | Every few minutes |
When something is wrong
GET /healthto separate a network problem on your side from an outage on ours.- A minimal request with the same key, to separate a key problem from a service problem.
- If both are healthy, the failure is in your client's configuration: check the base URL includes
/v1, then check the model id.
Most reports that reach us as "the API is down" are a key that was rotated, a base URL missing /v1,
or a client timeout shorter than the time a long prompt takes to read.
Planned work
Changes that could affect a client are announced before they are deployed: Changelog.