Skip to content

v.5.140.0#

Changes

  • The Python version has been updated to 3.14 in the API, Events, Faces, Image Store, Handlers, Python Matcher, and Remote SDK service containers.

    Support for older Python versions has been discontinued.

  • Automatic restart support has been added for video streams with decoding errors.

    Now, if decoding errors occur, the system automatically attempts to restore stream processing on the agent side (not supported for the videofile type). For this purpose, a new setting, decode_error_retry, has been added to the behavior_parameters stream behavior parameters. This setting includes two parameters (similar to eof_retry):

    • attempt_count — number of reconnection attempts (0 - disable auto-restart, default: 10);
    • delay — delay between attempts in milliseconds (default: 100 ms).

    If stream processing is not restored after all attempts, it will be marked as failed.

    See requests create stream, put stream, get stream, get streams.

  • A new callback type, kafka, has been added to the handler's "callbacks" policy.

    It allows notifications of events generated by handlers to be sent to Apache Kafka. This expands integration capabilities with systems using Kafka.

    The following parameters are used to configure the kafka callback:

    • servers - one or more Kafka server addresses
    • topic - topic for sending messages
    • protocol - security protocol
    • username, password - username and password for SASL authentication

    See the create handler request and the "Send events to a third-party service" documentation section.

  • The Lambda service monitoring system has updated the structure of the Lambda logs series.

    Events are now logged using operation and error codes instead of text messages. This change simplifies data analysis and allows for more efficient monitoring of lambda operations.

    The message and type fields have been replaced with the following:

    • operation_code — operation type code (e.g., 100 — lambda image deletion)
    • error_code — error code (0 — operation completed successfully, 10000 — image deletion error)

    For more information, see the "Monitoring" section of the developer guide.

  • Logging detail has been enhanced for callback sending.

    All callback dispatch logs now contain the stream identifier stream_id. This applies to dispatch start messages (log level - debug), successful completion messages (info), and errors (error).

    This change simplifies tracking and analyzing events for specific streams.

  • The Agent Interaction section in the Developer guide has been updated.

    This section now provides a detailed, practical guide with a clear structure and a demo example of agent implementation. The corresponding section in the Administrator guide has also been updated to reflect the full interaction cycle.

Fixed errors

  • Fixed an issue related to frame loss when decoding RTSP streams.

    Previously, if an RTSP stream was explicitly created with an fps parameter that matched the source video's frame rate, significant frame loss could occur during decoding. Decoding now works reliably regardless of the fps setting method.

  • Fixed an error that caused the Python Matcher service to hang during cache initialization.

    Previously, if an error occurred during this step (for example, if the database connection was lost), the service container would remain running, but the port would not be opened, making Python Matcher unavailable. Now, in this case, the service will terminate with an error message.