Skip to content

LUNA PLATFORM v.5.1.3#

Changes

  • Accept header was added for the "/sdk" resource. It enables you to specify the request content-type: JSON or MessagePack. MessagePack packages data more efficiently than JSON. It is not a binary format, so it does not require decoding from BASE64. See "sdk" > "sdk resource".

  • Glasses estimation was added for the "/sdk" resource. See "sdk" > "sdk resource".

  • Filters list_id__gte and list_id__lt were added for resources "/lists" and "/lists/count" (method "GET"). Received lists are ordered by "list_id".

    Using these filters, you can perform pagination that is:

    • faster than the pagination by "page" and "page_size" parameters. It does not slow down with a large number of lists
    • more stable than pagination by "page" and "page_size" parameters. When lists are changed during the pagination process, it does not cause the lists to be lost or duplicated in the response.

    See "lists" > "get lists"

    See "lists" > "get lists count"

  • The "/lists/deletions" resource was added to the Faces service for tracking of the deleted lists. See "administration" > "get lists deletions".

    The request for removing deleted lists log was added. See "administration" > "clear lists deletions log".

  • The possibility to extract basic attributes for faces without basic attributes was added to the "/additional_extract" resource. Biometric samples should be available for the samples. See "tasks processing" > "additional extract task".

  • The "wait_events_saving" query parameter was added for method POST on the "/events" resource. See "events" > "create new events".

    When this parameter is enabled, the system will wait for the event to be saved to the database before returning a response (this default behavior was used earlier).

    When this option is disabled, the system does not wait for events to be saved to the database. It sends a response immediately after validating the event and adding it to the buffer.

  • The memory consumption when caching a large number of small lists was reduced for the Python Matcher service.