Get Key Usage
Everything the console’s Requests page draws, per key.
Four series, all read from counters that already exist rather than derived from an estimate:
days/month_to_date— the daily counter written byauth._check_rate_limiton the pipeline that already runs for every keyed request. Backs the 30-day view and the monthly quota.hours— the hourly request and error counters, for the traffic chart and its error strip.latency— six fixed buckets summed over the same hourly window. Percentiles are bucket UPPER BOUNDS, never interpolated; see :class:LatencySummary.recent— the per-key ring of the lastRECENT_MAXrequests, newest first. Route TEMPLATES only, so no request identifiers are retained.
Only the caller’s own keys. The prefixes are read from api_keys
filtered by user_id and never from user input, so there is no way to
address another tenant’s telemetry by asking for it.
Fails SOFT: a Redis outage yields zeroes and empty lists, not a 500. This is a dashboard panel — a broken chart must not take down the page that also shows someone their API key. A zero here is visibly a zero; it is not presented as a measured absence of traffic.
Authorizations
Issue from /v1/keys (or admin-issued for enterprise tier).
Headers
Response
Successful Response
Sum over the current UTC calendar month
Requests this key may make in one UTC calendar month, from its rate-limit tier. Reported so a caller can see the headroom before it matters; NOT enforced yet.
Approximate latency over the same window
Hourly buckets, oldest first, over the requested hours window
Most recent requests for this key, newest first