Usage
The /usage page charts your tenant's daily traffic: request counts and
billable units, sliced by one dimension at a time. Visible to every role.
Group dimensions
| Group by | Series | Use it for |
|---|---|---|
feature | ocr, face, fraud, liveness, liveness_challenge | The default: which features drive traffic. liveness_challenge is the challenge-token issuance, billed separately from the check. |
variant | OCR: the document type (cid, src, passport, work_permit). Face: verify, register, registrations_verify, search, dedupe. Fraud: detect. Liveness: passive, burst, challenge, challenge_issue. | Drill into one feature's sub-types. |
status | success | Only calls that completed are metered, so this dimension currently shows one series. Rejected requests (4xx, 5xx) are visible in the audit log, not here. |
Metric toggle
- Requests — one per completed API call, whatever it billed.
- Billable units — what the platform metered. It differs from the request count where a call does more work: a two-sided OCR scan bills two units, a burst liveness check two and a challenge check three. Requests rejected before inference (validation, auth and rate-limit errors) are not metered at all.
Use billable units for reconciliation, requests for capacity planning.
Time window and mode
7, 30 or 90 days, and the same live ⇄ test toggle as the rest of the dashboard. Test-mode usage is metered and shown separately; plan quotas and invoices are based on live usage.
Empty state
A new tenant, or one that has only used test mode, sees "No usage in the selected window" rather than a flat line. That is the empty state, not a fault.
Hard caps
If your plan carries a hard cap on a feature, crossing it returns
429 quota_hard_cap_exceeded for that feature until the next billing
window. Those rejections appear in the audit log individually.