v0.3.12-beta.1

Beta — testing before stable 0.3.12.

Added

• Bus events on transitions — the coordinator now fires entity_availability_offline and entity_availability_recovered on the Home Assistant event bus when a monitored entity crosses offline/recovered (after cooldown, outside the startup grace period). Event data: entity_id, group, plus offline_since / downtime_seconds. Use these as native automation triggers instead of watching sensor attributes with templates.
• Reliability sensors (MTBF + MTTR) — two new diagnostic sensors per group that flag devices which keep flaking out. Mean Time Between Failures shows how often devices break (in hours); Mean Time To Recovery shows how long each outage lasts (in minutes). Together they distinguish a genuinely flaky device from one that had a single long outage at the same uptime %. Both are EntityCategory.DIAGNOSTIC (grouped under the device's Diagnostic section, kept off the main dashboard) with device_class: duration. Each exposes total_offline_events and its own per_device breakdown as attributes (the MTBF sensor lists mtbf_hours + offline_events; the MTTR sensor lists mttr_minutes + offline_events). Derived from all-time event counters (no extra storage growth, no state_class so they do not generate long-term statistics). Values populate once an entity has completed at least one offline→recovery cycle. Combined-group counterparts are not included in this release.
• reset_statistics action — new service to clear availability history and reliability counters (MTBF/MTTR, offline-event count) for an entity or an entire group. Useful after planned maintenance so a known outage does not skew the numbers. Accepts optional entity_id and/or group, matching the suppress actions.

Changed

• Translatable sensor names — all sensors now use translation_key instead of hardcoded English names, so their display names localize with the Home Assistant UI language. Names are provided for all bundled languages (de, es, fr, it, nl, da, nb, pl, pt, sv) with English fallback. Entity IDs and unique IDs are unchanged, so existing dashboards, cards, and automations are unaffected.

Fixed

• Statistics bloat: group summary sensors — Group Summary and Combined Summary sensors no longer declare state_class: measurement. Their value is an entity count that only changes when a group is edited, so the recorder was writing ~288 identical rows per day per group to statistics_short_term indefinitely. They remain valid state sensors (current count still shown); only long-term statistics generation stops. The availability-% sensors are unaffected — those are legitimate time-series statistics.
  • Existing rows: this stops new statistics from being generated but does not delete the rows already accumulated. Home Assistant surfaces a Settings → System → Repairs issue for the affected sensor.*_group_summary / sensor.*_combined_summary entities ("... no longer has a state class"); resolve it to purge the orphaned statistics, or use Developer Tools → Statistics and delete them manually.
• Card: empty filtered list — when entity_filter is offline or online and no entities match (e.g. all healthy), the card no longer renders the dangling Entity / State / Bat. column legend. The section header still shows the filtered count (e.g. Problem Entities (0)) so the active filter stays visible.
• Combined group creation: no group pre-selected — the "groups to include" multi-select on the Create Combined Group form no longer auto-selects the first group. The Required field lacked a default, so Home Assistant's frontend pre-checked the first option; it now defaults to an empty selection (the ≥2 groups validation is unchanged). The edit/options flow already behaved correctly.