Run completed stays always off (Washing Machine)

[eldi0s]

Describe the bug
Sensor run completed does not change from "off" to "on" when completed or only changes sporadic and not everytime

Expected behavior
Should reliably change to "on" when run is completed. It did work reliably in the past as I used it to notify
myself via the companion app on the phone, which stopped working now for most of the time.

Environment details:

• Environment (HASSIO, Raspbian, etc):
  Docker
• Home Assistant version installed:
  2026.6.4
• Component version installed:
  0.43.0
• Last know working version:
  maybe 1 or 2 months ago
• LG device type and model with issue:
  LG Washing Machine
  WTWN3-FL (WASHER)
• LG devices connected (list):
  Only this device

Output of HA logs
Nothing in the log

Additional context
Other sensor do also not update frequently (300 seconds) but only sporadic or random (between 5-20 minutes sometimes never)

If the washing machine is switched off completely and HA has to restart, all states become unavailable. When the washing machine comes back online as it got switched on, I have to manually reload the integration otherwise there will be no data polled from the cloud.

[blackscreener]

@eldi0s

I ran into these issues too and made some fixes that have been working well for me. Here's what I found:

washerDryer.py — localized state strings cause silent comparison failures

The LG ThinQ API returns localized strings for run states — for example, my Polish HA setup gets "Cykl został zakończony." instead of the English "STATE_END", and even English setups get "Power Off" instead of "STATE_POWER_OFF". Every comparison against the hardcoded English constants ends up false, breaking is_on, is_run_completed, save_run_states, and _get_runstate_key.

I added a _get_translated_labels() helper that cross-references the device's language packs (_model_lang_pack, _product_lang_pack, local_lang_pack) and expands each English constant into a set that includes both the key and its translated value. All comparisons now use any(state in value for state in labels) instead of value == "STATE...".

I also fixed WMStatus.update_status() to clear cached fields (_run_state, _pre_state, _process_state, _error) on every update — without that, stale values persist across state transitions even in English-only setups.

init.py — two fixes for devices getting stuck in unavailable

Fix 1 — recovery loop: When the coordinator catches MonitorUnavailableError, it sets _available = False and returns without calling poll() again. On subsequent cycles the code sees _available = False and bails out immediately — so the device stays unavailable forever, even after the LG API recovers. The fix: when _available is already false and we hit MonitorUnavailableError again, we attempt one more poll() call. If it succeeds, we restore _available = True and update the state. If it still fails, we stay in recovery mode and the coordinator's normal scan_interval will trigger another attempt.

Fix 2 — no retry on startup failure: If the LG API is temporarily unreachable at boot (washer offline, UseOfficialAPIError, etc.), both create_client_from_token and lge_devices_setup fail silently — the entry stays loaded (so the scan_interval number entity works) but no devices are ever created and nothing retries.

The fix schedules a periodic retry at 60s intervals (up to 12 attempts, ~12 minutes total). The retry count is stored in hass.data under a key that survives entry reloads, so the limit can't be bypassed by cycling reloads. When the limit is reached, a persistent notification appears in the GUI telling the user to reload or restart.

Two retry paths:

• Client creation failed (create_client_from_token throws): reloads the whole entry, starting setup from scratch with a fresh client.
• Device setup failed (lge_devices_setup fails or returns no devices): retries just lge_devices_setup with the existing client (lighter, no reload). On success, reloads the entry to complete setup.

During retries a persistent notification shows the current attempt count ("Connecting... 3/12"). It auto-dismisses on success or changes to "Setup Failed" when the limit is reached.

That's what I've got — hope it helps.

init.py
washerDryer.py

[eldi0s]

@blackscreener
Sounds great, will give it a shot when I find the time.
Thanks

[blackscreener]

@eldi0s
Quick follow-up — after more thorough testing the is_run_completed binary sensor was still flaky even with _get_translated_labels. The root cause is that the PreState display value (e.g. @WM_UX35_CYCLE_FINISHED_W) doesn't contain "STATE_END" as a substring, so the any(state in pre_state for state in end_labels) check always fails after the washer finishes.

The fix is dead simple — _get_pre_state() already stores the raw enum key in self._internal_pre_state (e.g. "END"). I added a fallback that checks whether self._internal_pre_state is a substring of any label in end_labels. Since "END" in "STATE_END" is True, the condition now catches completion regardless of the display value format.

washerDryer.py