Release Notes

v0.3.1

This update continues the recording-session reliability work from v0.3.0 — fewer false session endings on transient download failures, cleaner schedule-end transitions, and quieter repeat checks for out-of-schedule streamers — and adds GPU health monitoring to the System Health page. The /api/health endpoint is also significantly faster and lighter, so opening the System Health page stays instant even on busy systems.

Recording reliability

• Fewer false session endings during temporary download failures

  Some failed downloads could be treated as a stronger "stream is offline" signal than they really were. rust-srec now keeps better track of which download engine and stream type reported the failure, so temporary network trouble is less likely to cut a recording into separate sessions.

• Recording schedules now end sessions more cleanly

  When a stream leaves its allowed recording schedule, rust-srec now closes the active session through the normal session flow. This keeps the dashboard, notifications, post-processing, and the session timeline aligned with the schedule decision.

• Repeated out-of-schedule checks are quieter

  If a streamer is already out of schedule and has no active recording, repeated checks no longer keep rewriting the same session state in the background.

• Session timeline data is more consistent

  Session timeline events are now handled the same way as the rest of the session detail data. This does not change how the timeline looks, but it makes the page more reliable when older or malformed event details are present.

• Removed an inactive session timing setting

  The old session gap time setting has been removed because it no longer affected recordings. Stream ending is controlled by Offline Check Delay and Offline Detection Count. Backups that still contain the old value continue to import successfully, and the value is ignored.

• Queue freshness settings now save correctly

  Changes to the queue freshness threshold are now stored reliably, so rust-srec keeps your chosen re-check timing after settings updates.

System health

• GPU health is now tracked on the System Health page

  If your container loses GPU access (a known issue with the NVIDIA Container Toolkit on cgroup v2 hosts), you'll get a notification right away instead of finding out from the next failed remux job. The probe interval is configurable from the global settings page.

• /api/health is faster and lighter on resources

  The dashboard's health endpoint now reads a cached snapshot instead of re-running every check on each poll. Health components refresh in the background on per-component cadences (cheap atomic checks every 5 s, disk capacity every 30 s), so opening the System Health page is instant and background CPU stays low even on busy systems.

Compatibility

• One new database migration runs automatically on startup (adds the GPU health probe interval to global config).
• The old session gap time setting is removed from the configuration UI and ignored on import. Backups from earlier versions continue to import successfully — no manual cleanup is needed.