Public service status (auth, inference, billing, per-model)
Public, unauthenticated status surface. Aggregates authentication
and inference health into a top-level
status (ok | degraded | down), plus per-component detail and an
active-only catalog of per-model health badges. incidents are
the manually-published current/recent incidents. health_data_age_seconds
is the age (in seconds) of the most recent health probe — clients
can show a “stale data” warning when this exceeds ~5 minutes.
Cached for 10 seconds and served with
Cache-Control: public, max-age=10, stale-while-revalidate=60,
so frequent polling is inexpensive.
Response
Successful response
Worst-status wins. unknown on any component
(e.g. auth probe couldn't run) demotes to degraded
rather than masquerading as ok.
ok, degraded, down Per-model badges, active catalog only.
false when incidents.json is missing, malformed,
or contains a row that fails validation. The page
still renders other components but should surface
"incidents temporarily unavailable" instead of
claiming "no current incidents".