Skip to content

LUNA PLATFORM v.5.72.1#

LP changes

  • The SDK has been updated to version 5.22.1.

    In this version of LUNA PLATFORM:

    • Support for the 116th neural network model for extracting body descriptors has been added.
    • The 107th neural network model for extracting body descriptors has been removed from the Remote SDK container. Support for this model is not discontinued.

    Important: If the 107th neural network model (parameter "DEFAULT_HUMAN_DESCRIPTOR_VERSION") is used at the time of updating, the Remote SDK service will not start without additional actions.

    You need to perform one of the following additional actions:

    • Request the missing model from VisionLabs and add it to the Remote SDK container yourself.
    • Perform the Additional extraction task before starting the update, and then specify the new version in the "DEFAULT_HUMAN_DESCRIPTOR_VERSION" parameter before starting the Remote SDK service, thus preserving the ability to use old descriptors.
    • Specify the new version in the "DEFAULT_HUMAN_DESCRIPTOR_VERSION" setting before starting the Remote SDK service, thereby ceasing to use old descriptors.

  • A new type of callback, luna-ws-notification, has been added to the "storage_policy", allowing to receive generated events when subscribing to the Sender service via the WebSocket mechanism.

    The new type of callback replaces the "notification_policy", which is now considered Deprecated. Simultaneous use of a callback and "notification_policy" is not provided (two notifications will be sent).

    The main advantage of the callback mechanism over the obsolete "notification_policy" is the ability to specify multiple callbacks with different filters in the handler creation request, resulting in only one event being sent.

  • For bucket creation scripts ("lis_bucket_create.py", "s3_bucket_create", and "monitoring_db_create.py") and database preparation ("db_create.py"), a -v argument has been added, setting the logging level to "DEBUG".

    This allows displaying additional information that can help in case of errors.

  • Matching by face list has been accelerated when caching is enabled ("cache_enabled" parameter in the Python Matcher service).

  • A new section "LUNA_LAMBDA_BUILD_LIMITS" has been added to the Lambda service, allowing to configure limitations on the use of memory and CPU allocated for Docker image building tasks.

  • A new verify token permission has been added, allowing to specify rights to perform a "perform verification" request that performs descriptor and image verification.

    Additionally, similar to the emit_events permission, verifier IDs can be placed in black or white lists. If "verifier_id" IDs are in the black list, their use will be prohibited. If the IDs are in the white list, only their use will be allowed. The maximum number of IDs in the lists is 100.

    If the "verify" permission is granted, all necessary objects will be created during verification regardless of the permissions set in the token. For example, a "sample" type permission does not affect the creation of a sample during verification. Thus, when using a verifier with the "store_sample" parameter and the absence of the "creation" right for "samples", a sample will still be created.

  • The migrate_ttl_settings.py script has been updated to perform TTL settings migration for objects in S3 faster.

    See the "Add TTL for S3 buckets" section in the upgrade manual.

  • The SDK version for Handlers-lambda has been updated from version 5.19.0 to version 5.21.0.

  • The ability to set TTL in S3 for archives with Lambda has been added.

    TTL is set in the "archive_ttl" parameter in the request to create/update lambda.

    See detailed information on setting TTL in S3 in the "Migration to apply TTL to objects in S3" section in the administrator manual.

  • Support for "origin_bounding_box" fields in Handlers-lambda has been added, allowing to specify bbox coordinates with face or body in the coordinate system of the original image.

  • Matching by multiple lists in a single request has been accelerated.

  • The monitoring point for matching in the Python Matcher service has been changed.

    Previously, monitoring points contained the list_id field necessary for face matching by list. Now such points contain the list_batch_size field, indicating the number of cached lists.

    See detailed information in the "Monitoring" section of the Python Matcher service developer manual.

  • Documentation for migrating from LUNA PLATFORM 3 to Backport 3 and from LUNA PLATFORM 4 to Backport 4 have been removed from the distribution package and online documentation.

    These documentation are now available upon request to VisionLabs.

LP fixes

  • Fixed error which caused token permissions to be ignored in the "detect faces" request, allowing, for example, the creation of a sample even if the "creation" permission is not present in the token.

  • Removed unused OPTIONS method from the OpenAPI specification of the API service for the "ws handshake" request.

  • Fixed startup error of services with environment variables specified for overriding settings (prefix VL_SETTINGS).

  • Fixed error where "ws handshake" requests with token authentication could be executed without considering the rights to view the event.

  • Fixed environment variable passing error for the database preparation script db_create.py.

  • Fixed error where GET requests without any authentication data (e.g., "get handlers" request) did not consider the "account_id" query parameter as a filter.

  • Fixed error where it was impossible to create a token without setting "visibility_area" for a "user" account type.

    Due to this error, a "user" type user did not have access to the API service user interface.

    Now for tokens created within a "user" account, the default "visibility_area" value is "account".

  • Improved validation of requests for the "/liveness" and "/tasks/schedules" resources.

    Previously, an "Internal server error" with status code 500 could occur due to an invalid request body. Now correct messages are returned.

  • Fixed error where the luna-prepare script with the --dump-file argument of the Storages utility terminated with an error No module named 'configurator_db_tools'.