Glossary

Account (tenant). The top-level iLive customer entity. Collections, API keys, and usage counters all belong to an account. Each account is fully isolated from every other account.

API key. An opaque bearer secret (iksk_...) issued to a tenant and sent on the Authorization header of every request. See API keys.

Collection. A named, per-tenant bucket of indexed reference faces. Search and dedupe queries run against one collection at a time.

External face ID. Your own identifier stored alongside an indexed face — typically your internal user ID. Returned in search hits so you can correlate back to your system.

External image ID. Optional identifier attached to an image when indexing, useful when you're storing multiple faces per user.

Face ID. The UUID iLive assigns to an indexed face. Use it to fetch, delete, or run a same-collection search starting from that face.

ICAO photo. A passport-standard, 480×600, front-facing JPEG produced at the end of a successful liveness session. See ICAO-compliant photo.

Liveness verdict. The final outcome of a managed session, one of pass, fail, or retry. See Liveness overview.

Managed session. A server-created liveness session where iLive hosts the capture UI and calls you back with the verdict. Contrast with a direct session, which is created client-side.

Passive mode. A liveness session configured without interactive challenges — the user holds still for a few seconds. Contrast with active mode, which uses per-session challenges like a blink or a head turn.

Probe photo. The image you want to check against a reference — for example, the selfie a returning user just took.

Reference photo. The trusted image a probe is compared against — for example, the ICAO photo stored at first onboarding.

Similarity score. A value between 0.0 and 1.0 returned by compare and search calls. Higher means more likely to be the same person.

Threshold. The similarity cutoff at which your application decides "same person." Thresholds are a business decision; typical starting points are 0.65 for one-to-one match and 0.5 for one-to-many dedupe.