Add failure marker in SSH Docker panel for entries that failed to set up

[Copilot]

When a container is removed from the remote host, its config entry enters setup_error state and HA only surfaces this on the integrations dashboard — the SSH Docker sidebar panel shows nothing.

Changes

frontend/ssh-docker-panel.js

• _refreshFailedEntries(): async method that calls hass.callApi("GET", "config/config_entries/entry?domain=ssh_docker"), filters for state === "setup_error", and re-renders only on change. A simple concurrent-fetch guard prevents overlapping calls. Triggered only from connectedCallback(), visibilitychange, and pageshow handlers — not on every hass update.
• _getEntityByEntryId(): looks up hass.entities (entity registry) to build a map of config_entry_id → hass.states entity for all failed entries. Used to retrieve host info from the sensor's attributes and to identify which entities belong to failed entries.
• _getFailedEntityIds(): derives a Set of entity IDs from _getEntityByEntryId() — used to exclude those entities from the regular container grid so they no longer appear under "Unknown host".
• _renderFailedSection(): renders a red "⚠ Failed to set up" section above the container grid — one card per failed entry with the container name, host (when available from the entity registry), hint text, and an Open settings button that navigates to the HA integrations page.
• _getAllContainers(): excludes entities belonging to setup_error entries so they are shown only in the failed section, not also in the regular host-grouped grid.

strings.json / translations/en.json / translations/de.json

• Added setup_failed_badge, setup_failed_section, setup_failed_hint, btn_open_settings keys.

Result

Entries that fail to set up now appear prominently in the panel above the regular container grid and are no longer duplicated under "Unknown host":

⚠ Failed to set up
┌──────────────────────────────────────┐
│ My Container          ⚠ Setup failed │
│ Host: 192.168.1.100                  │
│ The container may have been removed  │
│ from the remote host.                │
│ [ Open settings ]                    │
└──────────────────────────────────────┘

---

⌨️ Start Copilot coding agent tasks without leaving your editor — available in VS Code, Visual Studio, JetBrains IDEs and Eclipse.

[gensyn]

@copilot I don't think a periodic refresh is useful for failed entries - remove that. I just realized that failed entries are shown on the panel but under "Unknown host" which is wrong, because the host is always known. It would suffice for me when the entires are prominently displayed on the panel as failed.

[Copilot]

@copilot I don't think a periodic refresh is useful for failed entries - remove that. I just realized that failed entries are shown on the panel but under "Unknown host" which is wro...

Done in 946859f:

Periodic refresh removed — _refreshFailedEntries() is no longer called from set hass() and has no throttle timer. It now only runs on panel connect, tab visibility restore, and page restore.

"Unknown host" fix — When a sensor entity for a setup_error entry still exists in hass.states (e.g. after a failed reload), it was slipping into the regular container grid with an empty host attribute and appearing under "Unknown host". The fix:

• _getEntityByEntryId(): single loop over hass.entities (entity registry) that maps config_entry_id → hass.states entity for all failed entries.
• _getFailedEntityIds(): derives its Set from the above — no duplicated loop.
• _getAllContainers(): filters out entities whose entity_id is in the failed set.
• _renderFailedSection(): shows Host: <value> (from the entity's host attribute) in the failed card whenever it's available.

[gensyn]

@copilot The way the failed entries are shown on the panel is great, but they are still also listed under "Unknown hosts" which is wrong and unncessary. Please show them only as failed.

[Copilot]

@copilot The way the failed entries are shown on the panel is great, but they are still also listed under "Unknown hosts" which is wrong and unncessary. Please show them only as fail...

Fixed in 29ea77d.

The root cause: the previous fix relied solely on hass.entities (entity registry) to identify which sensor entities belong to failed config entries. When hass.entities isn't available in the panel context (or doesn't expose config_entry_id), the exclusion silently did nothing — so the entity still showed under "Unknown host".

The new _findEntityForEntry() uses two strategies in order:

1. hass.entities registry — matched by config_entry_id (most precise when available)
2. Name-attribute match — state.attributes.name === entry.title (reliable fallback, since the sensor's name attribute always equals the config entry title)

_getFailedEntityIds() now uses this helper, so entities belonging to setup_error entries are excluded from _getAllContainers() regardless of whether hass.entities is available — they only appear in the "⚠ Failed to set up" section.

[gensyn]

@copilot The failed entries still show up under "Unknown host".

[Copilot]

@copilot The failed entries still show up under "Unknown host".

Fixed in 890f9c3. There were three bugs in the entity-matching chain that together allowed the entity to slip through:

1. Strategy 1 required a hass.states lookup — if the entity was in hass.entities (entity registry) with the right config_entry_id but its state hadn't been restored yet, the if (state) guard silently dropped it without ever adding the entity_id to the exclusion set.

2. Strategy 2 matched on state.attributes.name — restored/unavailable entities have empty attributes ({}), so state.attributes.name is undefined and the match never fired.

3. Both strategies required a state object — useless for the exact scenario that causes the bug.

The new _getFailedEntityIds() has three truly independent strategies:

• Strategy 1 (entity registry): adds entity_ids directly from hass.entities by config_entry_id — no hass.states lookup required.
• Strategy 2 (name-attribute): fallback for entities with populated attributes when hass.entities is unavailable.
• Strategy 3 (derived entity_id): sensor.ssh_docker_ + _slugify(entry.title) — covers the hardest case: hass.entities absent AND attributes empty. Mirrors HA's Python slugify() exactly.

Strategies 2 and 3 are combined into a single hass.states pass.