Skip to content

v.5.122.0#

Changes

  • Support for sending monitoring data to ClickHouse has been added.

    All LUNA PLATFORM services now support sending monitoring data to ClickHouse, a high-performance, columnar database management system.

    Unlike InfluxDB, which divides data into tags and fields, ClickHouse uses a unified table structure with support for the JSON format. Each record is a table row, where:

    • The time field is the timestamp when the record was created;
    • The data field is a JSON object that combines all data (including what would otherwise be stored as tags or fields in InfluxDB).

    The main advantage over InfluxDB is speed: ClickHouse processes complex analytical queries and aggregations significantly faster, especially on large data volumes.

    Note: Currently, InfluxDB remains the default option. Future versions of LUNA PLATFORM may migrate to ClickHouse for monitoring.

    For more information, see the "Monitoring" section. Monitoring setup procedures are also described in the sections on installing and updating LUNA PLATFORM.

  • Support for the following services has been completely discontinued: Backport 3, Backport 4, UI 3, and UI 4. All related information has been removed from the LUNA PLATFORM documentation and Docker Compose script.

  • The ability to select a mode for evaluating deepfake has been deprecated and is no longer included in queries.

    The "estimation_mode" parameter has been removed from the LUNA_REMOTE_SDK_DEEPFAKE_ESTIMATOR_SETTINGS and LUNA_VIDEO_AGENT_DEEPFAKE_ESTIMATOR_SETTINGS parameter groups.

    For backward compatibility, the mode field is preserved in the response but always takes the default value of 2.

  • The OpenAPI specification has been updated.

    The resource "/videosdk" now displays the human_tracking analytics description by default instead of people_count.

    To use people_count for analytics, you must specify the corresponding value in the "analytics > name" field.

  • New settings for managing Video Agent service resources have been added to the "LUNA_VIDEO_AGENT_RUNTIME_SETTINGS" parameter group.

    These parameters are LUNA SDK runtime settings. For more information, see the SDK documentation.

    List of new settings:

    • PROGRAM_CACHE_SIZE — limits the number of cached elements, which speeds up work with a large number of tasks;
    • DEFAULT_GPU_DEVICE — allows you to select the default GPU device (relevant for systems with multiple GPUs);
    • CPU_HIGH_WATERMARK and GPU_HIGH_WATERMARK — set the maximum memory size for the CPU and GPU;
    • PIN_THREADS — enables/disables thread binding to processor cores.

  • The ability to manage access to lambdas has been added through the new shared parameter in requests.

    Previously, only the user who created a lambda could use it via the "/lambdas/{lambda_id}/proxy" resource. Now, when creating, updating, and importing, you can grant this permission to other users by configuring the shared parameter in the parameters section.

    This parameter accepts the following values:

    • 0 (default value): the lambda will be available only to the user who created it;
    • 1: the lambda becomes available for use by other users.

    Important: Restriction for Agent lambdas: if the agent is external (the external_agent parameter is set to 1), such a lambda cannot be publicly accessible. This is because agents in "external mode" interact with the platform only through the API (without access to core services).

    For more information, see the Lambda section of the developer guide.

  • A new example for using Agent-Lambda has been added to the Lambda service developer guide.

    Agent-lambda allows you to use not only standard analytics (people tracking and people counting), but also custom video analysis algorithms based on various neural network models.

    This example demonstrates the implementation of animal detection analytics in a video stream. The algorithm automatically starts tracking when animals appear in the frame and stops tracking when they leave the frame.

    For more information, see Agent lambda examples.