Health endpoints

MIRA exposes two unauthenticated HTTP probes for supervisors, load balancers, and uptime monitors. For guidance on which to point a given tool at, see Health checks & monitoring.

GET /livez — liveness

Is the process answering? Does no dependency I/O (no provider, DB, or subsystem), so it stays responsive even while a dependency is wedged.

Auth	None (public)
200	Body ok (text/plain). Always, whenever the server is serving.
Other	The endpoint has no failure branch — a connection refused / timeout (the server isn’t up) is the only “down” signal.

$ curl -i http://127.0.0.1:8087/livez
HTTP/1.1 200 OK
content-type: text/plain; charset=utf-8

ok

GET /readyz — readiness

Is MIRA fit to serve? Reflects whether the active model provider is reachable. The verdict is cached and refreshed out of band (see below).

Auth	None (public)
200	JSON {"status":"ready","cached_age_secs":<n>} — provider reachable.
503	JSON {"status":"unavailable","detail":"provider unreachable","cached_age_secs":<n>} — provider not reachable.

cached_age_secs is the age, in seconds, of the cached check the response was served from (0 on a freshly computed verdict).

$ curl -i http://127.0.0.1:8087/readyz
HTTP/1.1 200 OK
content-type: application/json

{"status":"ready","cached_age_secs":12}

Readiness cache

/readyz’s verdict is cached for server.readiness_cache_ttl_secs (default 30; 0 disables caching and checks inline on every probe). Within the TTL, probes are served the cached value with no provider round-trip; once stale, the cached value is still served immediately while a single background refresh runs. So however many callers probe, at most one provider round-trip is in flight per window and no probe blocks on the provider.

/health is not a probe

/health is the admin System Health page in the web UI, served by the SPA — it is deliberately not a server route. Machine probes are /livez and /readyz.

Changed in 0.339.0. /health was previously the machine health probe (a single overloaded endpoint that also ran a provider round-trip). It was split into /livez (liveness) and /readyz (readiness), and /health was freed for the System Health page. A sentinel guardian.process.probe_url ending in /health is auto-migrated to /livez on startup; update any external monitors or container HEALTHCHECKs you configured yourself.