IKEA Dirigera Hub Integration for Home Assistant

A Home Assistant integration for the IKEA Dirigera hub, built on the dirigera Python library (v1.2.6). Originally forked from sanjoyg/dirigera_platform and now under active development, with a strong focus on reliability : application-level WebSocket keepalive, automatic state re-sync on reconnect, and dynamic device discovery. Alongside broad device support and regular fixes. The most recent work lands in the changelog below; issues are actively triaged.

Contributions are welcome, feel free to open issues or submit pull requests. Device data dumps are especially helpful for adding support for new devices (how to dump).

Highlights

• Dynamic Device Discovery : devices added, removed, renamed, or moved to a different room in the IKEA Home app are automatically reflected in Home Assistant without a restart
• Split-Device Merging : newer IKEA devices (GRILLPLATS, TOFSMYGGA, TIMMERFLOTTE, MYGGSPRAY) expose as multiple API devices linked by relationId; the integration automatically merges them into single HA entities
• Real-Time Updates : full WebSocket event support for instant state changes across all device types
• Resilient Connection : application-level WebSocket keepalive reduces hub "inactivity" disconnects, and all device state is automatically re-synced on every reconnect, so entities never silently go stale after the hub drops the connection
• Reliable Startup : automatic retry on connection failure (ConfigEntryNotReady) instead of requiring manual reloads
• Multiple Hubs : run more than one Dirigera hub side by side. Add the integration once per hub; each hub keeps its own device registry and event connection, so reloading one never disturbs another

Upstream issues addressed in this fork

If you arrived here from an open issue on sanjoyg/dirigera_platform, the fix may already be in this fork. Switching to this repository in HACS should resolve the listed issues without further action.

Upstream issue	Title	Fixed in
#155	Integration failed setup on 2025.6.0	v0.2.7 (ConfigEntryNotReady retry + deprecated hass.loop migration)
#160	TRADFRI on/off switch — not properly integrated	v0.2.6 (multi-button scene-creation per controller half)
#165	VALLHORN motion sensor not providing illuminance	v0.2.1 (separate light sensor entities with lux conversion)
#168	Disconnected due to inactivity	v0.2.11/12 (application-level WebSocket keepalive)
#175	MotionSensor fails setup when batteryPercentage is missing	v0.2.1 (batteryPercentage made optional)
#177	Styrbar 4-button — no events	v0.2.6 (per-button scene creation)
#183	MYGGSPRAY motion sensor not visible	upstream 2.7.1 + v0.2.16 (battery-sensor dedup unique to this fork)
#184	BILRESA dual button — no activities	v0.2.6 (dual-button to controller map + scenes)
#195	Lights can't be used as automation triggers	v0.2.9 (merged upstream PR #197)
#198	ikea_bulb_device_set has no attribute 'entity'	v0.2.8 (merged upstream PR #196)
#152	IKEA Inspelning (plug with power sensor)	v0.2.5 (outlet + electricalSensor split-device merge)
#148	Energy Consumed at Last Reset not updating	v0.2.5 (electricalSensor events routed to outlet)

Not addressed (out of scope or different root cause)

• #143 — Power factor for Inspelning: the Dirigera API does not expose currentPowerFactor; would require a user-configurable correction factor + derived sensor, not a code fix.
• #150 — Duplicate Devices with Matter & HACS: caused by running both Matter and Dirigera integrations against the same device, not by this integration.
• #194 — TIMMERFLOTTE humidity sensor: TIMMERFLOTTE temperature works (split-device merging), humidity needs separate device support — not yet implemented.
• New device support requests (STARKVIND quirks, VINDYKSTRA air quality, Matter device coverage, etc.) are out of scope for this fork's reliability focus.

Supported Devices

Category	Devices	Notes
Lights	TRÅDFRI, FLOALT, and other IKEA lights	RGBWW with dynamic color mode switching (HS ↔ color temp)
Outlets	INSPELNING, GRILLPLATS, TOFSMYGGA	Energy monitoring (voltage, current, power, energy); split-device pairs auto-merged
Motion Sensors	VALLHORN, MYGGSPRAY (E2494)	Handles both motionSensor and occupancySensor device types
Light Sensors	MYGGSPRAY (E2494)	Illuminance in lux (Matter raw value conversion)
Open/Close Sensors	PARASOLL, MYGGBETT
Environment Sensors	VINDSTYRKA (PM2.5, VOC), ALPSTUGA (CO2), TIMMERFLOTTE	Split-device pairs auto-merged; state_class: measurement for Long Term Statistics
Blinds	FYRTUR, KADRILJ
Remotes	STYRBAR, RODRET, SOMRIG, BILRESA	remotePressEvent + shortcutController support; automation triggers for all buttons
Air Purifiers	STARKVIND	Native unit of measurement fix
Water Sensors	BADRING
Scenes	All Dirigera scenes	Exposed as entities with Activate button

Requirements

• Home Assistant 2026.1 or newer (v0.2.6 and earlier work with HA 2024.12+)
• IKEA Dirigera hub on your local network
• HACS installed

Installation

Via HACS (recommended)

Or manually:

1. In Home Assistant, go to HACS → Integrations → ⋮ (top right) → Custom repositories
2. Add https://github.com/nrbrt/dirigera_platform as an Integration
3. Search for "Dirigera" and install

Setup

1. Go to Settings → Devices & Services → Add Integration → search for Dirigera
2. Enter the IP address of your Dirigera hub (find it in your router's DHCP client list)
3. When prompted, press the action button on the Dirigera hub, then click Submit
4. Your devices will appear automatically

To test with mock devices, enter mock as the IP address.

Bug Reports & Feature Requests

When opening an issue, please include a device data dump, it helps enormously:

1. Go to Developer Tools → Services
2. Call dirigera_platform.dump_data (no parameters needed)
3. Copy the output from the Home Assistant log

See detailed instructions.

Changelog

See Releases for a full changelog.

Recent

• v0.3.14 (2026-07-11) Docs: clarify that the push throttle is a minimum interval, not an exact one (issue #40). Because a write only happens when a device push actually arrives, the effective interval rounds up to the first push after the limit elapses. With a sensor that pushes every ~30s and a 60s setting, the second push (~59s) is still blocked and the third (~89s) clears it, so the real interval is about 90s. This is correct throttle semantics (a floor), so rather than change the behaviour the option now documents it in its help text. Thanks to @m3gg3 for the precise measurement (89.2s observed vs 89.1s predicted) on real ALPSTUGA hardware.
• v0.3.13 (2026-07-11) Extend the push throttle to the environment sensors (issue #40). Following v0.3.12, the VINDSTYRKA/ALPSTUGA temperature, humidity, current PM2.5, VOC index and CO2 sensors now also set should_poll = False and honour the throttle interval, since these values arrive over the WebSocket push (they are in the environmentSensor push allowlist) rather than needing the poll. The min/max measured PM2.5 sensors are the exception: those values are not pushed and only refresh on the poll fetch, so they keep polling (they change rarely, so their recorder footprint is negligible). The "Power sensor push throttle" option now governs these sensors too. Thanks to @m3gg3 for confirming from debug logs that ALPSTUGA push updates arrive independently of the poll cycle.
• v0.3.12 (2026-07-11) Fix: the v0.3.10 power-sensor push throttle was silently bypassed by Home Assistant's own polling (issue #40, reopened). The current amps, active power and voltage sensors never overrode should_poll, so HA kept polling them on its default ~30s cycle and wrote the (already WebSocket-updated) value straight to the recorder, ignoring the configured throttle. Because these sensors' outlet device sets skip_update=True and they are driven entirely by the ~8s WebSocket push, that poll produced nothing but redundant recorder rows. They now set should_poll = False, so the throttled push is the only state-write path and the configured interval is actually honoured. Energy totals are unaffected. Thanks to @m3gg3 for catching the gap. Note: the environment sensors (VINDSTYRKA/ALPSTUGA) are poll-fetch driven rather than push driven, so they need a different approach, tracked in the issue discussion.
• v0.3.11 (2026-07-09) Fix: stop the WebSocket listener from pushing HA state during shutdown once the event loop is closed (issue #42). The listener runs on its own thread and keeps handling hub events for a few seconds after Home Assistant has already closed its asyncio loop; every event that reached schedule_update_ha_state in that window raised RuntimeError: Event loop is closed and was logged as a warning with a full traceback, once per incoming event. on_message now checks _request_to_stop/loop.is_closed() before scheduling the update, the same guard the keepalive path already used, so shutdown stays quiet. Normal operation is unchanged. Thanks to @m3gg3 for the precise root-cause analysis and the ready-made patch.
• v0.3.10 (2026-07-01) Throttle HA state pushes for outlet power sensors (issue #40). Smart plugs with power metering push currentActivePower/currentAmps over the WebSocket roughly every 8 seconds, so the current amps, active power and voltage sensors wrote about 10,800 recorder rows per day per entity. Each of those sensors now rate-limits its HA state push to a minimum interval (default 60s, cutting it to about 1,440 rows/day). The sensor's internal value is still updated on every push, so native_value keeps returning the live reading and only the recorder write is skipped. The interval is configurable in the integration options ("Power sensor push throttle"); set it to 0 to disable throttling. Energy totals (total_energy_consumed, TOTAL_INCREASING) are unaffected. Thanks to @m3gg3 for the detailed analysis and the ready-made patch.
• v0.3.9 (2026-06-30) Fix: guard the device state-update cascade against listeners that were never added to Home Assistant (issue #41). A device fans a hub push out to its sensor listeners via schedule_update_ha_state; a listener that was registered but never added to HA has hass=None, so HA dereferences self.hass.loop and raises AttributeError. That surfaced as a flood of "error processing hub event" warnings, typically on outlets and electrical sensors pushing power readings every few seconds. The cascade now skips listeners with no hass, which is behaviour-neutral since an entity not registered with HA cannot receive a push anyway.
• v0.3.8 (2026-06-24) Fix: per-hub device registry (issue #39). The device registry was a single shared dict and stop() cleared all of it, so with two hubs, unloading or reloading one config entry wiped the other hub's registrations too. After that, every event for the other hub's devices came in as an unknown device and its state was never applied, so those devices stopped updating entirely (not tied to one device type). The registry is now keyed per hub, so reloading one hub no longer affects another. The device registry was a single shared dict and stop() cleared all of it, so with two hubs, unloading or reloading one config entry wiped the other hub's registrations too. After that, every event for the other hub's devices came in as an unknown device and its state was never applied, so those devices stopped updating entirely (not tied to one device type). The registry is now keyed per hub, so reloading one hub no longer affects another.
• v0.3.7 (2026-06-24) Fix: re-sync all device state on WebSocket reconnect (issue #39). Dirigera hubs drop the event WebSocket fairly often; on reconnect the integration only waited for new events and never re-pulled current state, so any change made during the disconnect gap was lost and entities silently went stale over time, affecting every device type, not just one. (A config reload worked around it because a reload re-fetches all state.) On reconnect the integration now fetches /devices once and replays each device through the normal update path, with discovery suppressed during the replay. No more manual reloads.
• v0.3.6 (2026-06-24) Fix: stop the runtime re-discovery loop for devices without an entity (issue #39). State events for environment-sensor sub-channels that have no HA entity triggered a full hub lookup on every event with no caching, hundreds of redundant /devices/{id} calls and a wall of "discovery not yet implemented" log lines. Such devices are now remembered after the first failed attempt, so subsequent events short-circuit.
• v0.3.5 (2026-06-23) Fix: support multiple Dirigera hubs (issue #39). The device list and discovery coordinator were kept in a single shared slot, so with two hubs the second overwrote the first at startup (devices went missing, IDs collided). Both are now stored per config entry, so multiple hubs no longer step on each other.
• v0.3.4 (2026-06-23) Fix: prevent duplicate hub config entries (issue #39). The integration now refuses to add the same hub a second time and back-fills existing setups so a stray duplicate can't sneak in.
• v0.3.3 (2026-06-23) Perf: fetch /devices once during setup instead of once per device type (issue #38).
• v0.3.2 (2026-06-16) Fix: offload the blocking get_outlet_by_id() lookup to the executor so runtime outlet discovery no longer blocks the event loop (issue #37).
• v0.3.1 (2026-06-12) Fix: clamp power/amps readings to 0 while an outlet is off (issue #36).
• v0.3.0 (2026-06-12) Reliability and lifecycle release: single update listener, a real unload_ok and a per-entry event listener on unload; visible event-processing errors and an interruptible reconnect retry; hub-side scene updates now propagate; light fixes (stale color temperature, brightness clamp, HS dedupe); VOC index device class, controller push updates and air-purifier cleanups; energy-timestamp sensor repairs; cover/blind-level and fan-state crash fixes; serialized concurrent multi-entity device updates (issue #34).
• v0.2.18 to 0.2.21 (2026-05-25 to 2026-06-06) Incremental fixes; see Releases.
• v0.2.17 (2026-05-19) — Fix: GRILLPLATS / TOFSMYGGA energy data missing on runtime-added plugs (issue #31). When a smart plug is added to the Dirigera hub while the HA integration is already running, the ADD-event discovery path constructed the outlet directly via dict_to_outlet() and bypassed the energy-attribute merge from the linked electricalSensor device. Result: current_active_power, total_energy_consumed, voltage and amps were missing until an integration restart. The runtime-discovery path now uses get_outlet_by_id(), applying the same merge used at integration startup. Falls back to raw payload with a warning log if the lookup fails.
• v0.2.16 (2026-04-27) — Battery sensor fixes (PRs #27, #28, #29 by @ermitovski):
  • Fix: avoid duplicate Battery Percentage entity on MYGGSPRAY split-devices. The hub exposes MYGGSPRAY as occupancySensor + lightSensor sharing a relation_id, both reporting battery_percentage; the duplicate *_battery_percentage_2 entity is now suppressed by binding the battery diagnostic to the motion side only. Pre-existing _2 entities become orphaned (state unavailable) and can be deleted from the device page.
  • Fix: de-duplicate the controller battery diagnostic on multi-button controllers (BILRESA, SOMRIG, RODRET, STYRBAR, ...). Group controllers by relation_id and elect a single primary that becomes the HA entity; secondary halves rebind to the primary in the device registry so remotePressEvents still resolve to a registered entity. Multi-button device-trigger generation is preserved.
  • Feature: emit battery_percentage diagnostic when a sensor (waterSensor, motionSensor, occupancySensor, openCloseSensor) is paired with HA already running. The WebSocket-driven discovery path now mirrors the static startup path, so freshly paired BADRING / motion / open-close sensors show the battery diagnostic without a HA restart.
• v0.2.15 (2026-04-23) — Fix: guard _color_mode writes against unsupported modes on brightness-only lights (TRÅDFRI Driver). Prevents HomeAssistantError: "... set to unsupported color mode hs" after scene activation. (PR #26 by @charleslemaux)
• v0.2.14 (2026-04-20) — Fix: proper split-device entity naming using has_entity_name (PR #25 by @crowbarz)
• v0.2.13 (2026-04-20) — Fix: split-device entity naming — secondary entities (e.g. MYGGSPRAY illuminance) now inherit the user-configured name from the primary entity
• v0.2.11/12 (2026-04-17) — Fix: application-level WebSocket keepalive to prevent Dirigera hub "disconnected due to inactivity" (issue #12)
• v0.2.10 (2026-04-16) — Fix: honor 10s sleep between listener reconnects; add on_open/on_close diagnostics for WebSocket disconnect debugging
• v0.2.9 (2026-04-15) — Feature: light device triggers — turned_on/turned_off automations (upstream PR #197)
• v0.2.8 (2026-04-15) — Fix: ikea_bulb_device_set registry_entry registration (upstream PR #196)
• v0.2.7 (2026-04-13) — Fix: replace deprecated hass.loop with asyncio.get_event_loop() (HA 2026.1+ compatibility)
• v0.2.6 (2026-04-04) — BILRESA dual button support; consistent device naming for split-devices
• v0.2.5 (2026-04-04) — Split-device plug support (GRILLPLATS, TOFSMYGGA)
• v0.2.0 (2026-03-27) — Split-device merging framework; TIMMERFLOTTE support