v2.3.1

Bug Fixes

• Fix incorrect Prometheus metric descriptions for on_disconnect, code_project_calls, plate_recognizer_calls, and plate_recognizer_errors counters
• Add null guard on thread pool executor in MQTT on_message callback to prevent crash if a message arrives before initialization completes
• Log unhandled exceptions in main loop to the log file before re-raising, so crashes appear in both docker logs and the log file (closes #55, closes #82)

v2.3.0

What's Changed

• Fix hanging tests: start_healthcheck_server() now returns (server, thread) so tests can properly shut down HTTP servers. Added a pytest fixture for automatic cleanup.
• Fix test_run_mqtt_client: Updated to match the current loop-based shutdown pattern instead of the removed loop_forever() API.
• MQTT retain flag: Plate messages are now published with retain=True so the last known plate is available to new MQTT subscribers.
• pytest-timeout: Added a 30-second timeout safety net to prevent future test hangs.

Release 2.2.1

Merge branch 'master' of github.com:ljmerza/frigate_plate_recognizer

Release 2.2.0

Merge branch 'master' of github.com:ljmerza/frigate_plate_recognizer

Release 2.1.1

fix string format

Release 2.0.0

min version frigate 0.15

Release 1.8.14

Merge branch 'master' of github.com:ljmerza/frigate_plate_recognizer

Release 1.8.13

remove init conversion on box dimensions

Release 1.8.12

fix missing return value

Release 1.8.11

mqtt auth clean up