Add LG Styler ST_B_E4H01Y_APL by tunarabuuu · Pull Request #115 · anszom/rethink

tunarabuuu · 2026-08-02T21:14:32Z

Third of the per-device-class splits from #111.

ST_B_E4H01Y_APL — an AA..BB ThinQ2 Styler / garment-care unit (deviceType 203). The byte layout was recovered by injecting single-byte-changed state frames through the management API and reading the LG cloud's own decode back, one offset per observation (three of the bytes are bitfields, split bit by bit).

Addressing your review notes from #111:

processAABB — reads now go through the base class's processAABB instead of a custom processData; the frame arrives with its leading AA/FF and trailing CRC/BB stripped, so the record offsets are simply two less than the whole-frame positions.
logging utility — the stray console.warn calls now use log(...).
header comment — trimmed to what's needed to read the code.

All labels are English. npm test passes; tsc clean.

Port from upstream Node PR #115 (anszom/rethink#115). AABB class 0x31 with 40-byte records (last record is current). Field map and F024/F026 command frames from 2026-07-28 LG-cloud oracle and convert/control captures, including Standard/Quick start blocks.

anszom · 2026-08-22T20:36:05Z

+    17: '8000000000000000000000000000001e780000000000000000000000000000000000000000', // Timed Dry 30
+    18: '8000000000000000000000000000003c780000000000000000000000000000000000000000', // Timed Dry 60
+    19: '8000000000000000000000000000005a780000000000000000000000000000000000000000', // Timed Dry 90
+    20: '8000000000000000000000000000003c780000000000000000000000000000000000000000', // Timed Dry 120


This is identical to Timed Dry 60. Is this correct?

anszom · 2026-08-22T20:36:54Z

+                    // Replaced by the two settable halves above. Removing a component from the
+                    // config is not enough to retire the entity Home Assistant already made —
+                    // a config carrying nothing but its platform is what says "this is gone",
+                    // and it has to carry that, or HA rejects the whole device payload.
+                    night_care_window: { platform: 'sensor' } as ComponentInfo,


That entity was never exposed by a released version of rethink, so we don't need to include code to withdraw it.

anszom · 2026-08-22T20:40:50Z

+                    remaining: minutes('remaining', 'Time remaining', { icon: 'mdi:timer-sand' }),
+                    total: minutes('total', 'Total time', { icon: 'mdi:timer-outline' }),
+                    reserved_at: sensor('reserved_at', 'Reserved time', { icon: 'mdi:calendar-clock' }),


Please follow the naming conventions of other devices (remaining_time, reserve_time)

anszom · 2026-08-22T20:40:58Z

+                ...HADevice.config(meta, { name: 'LG Styler' }),
+                components: {
+                    power: toggle('power', 'Power', { icon: 'mdi:power' }),
+                    status: sensor('status', 'State', { icon: 'mdi:hanger' }),


Other devices use the 'Status' label

anszom · 2026-08-22T20:52:19Z

+import HADevice from './base'
+import AABBDevice from './aabb_device'
+
+/*


This header is overly verbose. Keep only the most important protocol details, definitely drop mentions of older revisions that we have never seen.

anszom · 2026-08-22T21:06:53Z

+describe(MODEL_ID, () => {
+    /*
+     * Writable exactly where a command frame was captured off LG's own capability API and seen to
+     * take (CL-0000). Deliberately absent: `remoteControlStatus setRemoteControlType`, which


The fields claimed to be absent do appear to be present...

anszom · 2026-08-22T21:07:36Z

+        const { ha } = makeDevice()
+        const components = ha.devices[DEVICE_ID].config!.components as Record<string, Record<string, unknown>>
+        for (const [name, comp] of Object.entries(components)) {
+            if (comp.unique_id === undefined) continue // a removal, not an entity


This line should not be needed if we remove the night_care_window reference entirely.

anszom · 2026-08-22T21:07:48Z

+    test('the retired night-care summary is published as a removal', () => {
+        const { ha } = makeDevice()
+        const components = ha.devices[DEVICE_ID].config!.components as Record<string, Record<string, unknown>>
+        // Its two settable halves replace it, but dropping it from the config would leave the old
+        // entity behind, frozen at whatever it last read.
+        assert.deepEqual(components.night_care_window, { platform: 'sensor' })
+    })


This should not be needed if we remove the night_care_window reference entirely.

anszom · 2026-08-22T21:09:55Z

+                    energy: sensor('energy', 'Energy monitoring', {
+                        icon: 'mdi:flash',
+                        entity_category: 'diagnostic',
+                    }),


Is this in a useful real-world unit?

anszom · 2026-08-22T21:12:27Z

+                    internal_light: sensor('internal_light', 'Interior light setting', {
+                        icon: 'mdi:lightbulb-outline',
+                        entity_category: 'diagnostic',
+                        // LG's snapshot exposes styler.internalLightingTime but this model's
+                        // modelJSON does not declare its value list, so the raw code is published
+                        // rather than an invented label. 0 read back NO_SETTING, 1 read LIGHTING_TIME_0.
+                    }),


What is the meaning of this entity? If it has only two values, would a binary_sensor be a better choice?

anszom · 2026-08-22T21:20:12Z

Also please add a README entry just as for the other devices.

An AA..BB ThinQ2 garment-care unit (deviceType 203). The byte layout was recovered by injecting single-byte-changed state frames through the management API and reading the LG cloud's decode back, one offset per observation. Reads go through the base class's processAABB — the frame arrives with its leading AA/FF and trailing CRC/BB already stripped, so the record offsets are two less than the whole-frame positions the probes recorded. Stray warnings go through the logging utility rather than console. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

tunarabuuu · 2026-08-22T21:47:25Z

All addressed, one commit, rebased on master.

Header too verbose (:9) — cut from 105 lines to about 70, and everything that narrated this driver's own history is gone ("the earlier note said control was refused", "which the first pass got wrong", the retelling of what an older revision published). What is left is the read map, the frame/record shape, the write vocabulary and the course-block rule — the parts you need to read the code. The same treatment went to the styler.state / styler.error comment blocks and to one comment in the test file.
Outdated Korean references (:264) — gone. Where the source of a label matters the comment now says it is an English translation of LG's ko-KR pack, which is what the tables actually hold.
Timed Dry 120 identical to Timed Dry 60 (:218) — it is correct, and it surprised me too, so I checked it against the model file rather than my transcription. LG's modelJSON gives course 20 a Drying1_Time default of 60, exactly as course 18 does; I regenerated the whole table from Course[id].function with the documented rule and every one of the 21 entries came back byte for byte identical to what is in the PR. The run length evidently comes from the course id, not from this block. There is now a comment saying so.
Naming conventions (:528) — remaining → remaining_time, total → initial_time, reserved_at → reserve_time, matching F_V8_Y___W.B_2QEUK and friends. reserve_time is now a duration in minutes rather than a clock string: LG declares Reserve_Time_H as a 3..19 hour range, so it is a delay before the course starts, and 0 means no reservation.
'Status' label (:515) — done.
night_care_window withdrawal (:573, tests :124 and :582) — you are right, that entity never shipped. The tombstone component, the test for it, and the unique_id === undefined skip in the writability test are all gone.
Energy monitoring in a useful unit? (:587) — no, and I could not make it one. The two probes only establish that 34..35 is a big-endian 16-bit value; LG's EnergyMonitoring block for this model is a powertable keyed by level, which does not say what the counter counts. Rather than publish a number with no unit I dropped the entity (and its probe fixture). The offset stays documented in the header.
internal_light meaning (:598) — dropped for the same reason. styler.internalLightingTime has no value list in this model's modelJSON and nothing in its language pack, and the two codes I saw (0 → NO_SETTING, 1 → LIGHTING_TIME_0) do not settle whether "time 0" means the light is off or on — so a binary sensor would have a coin-flip polarity. If a capture ever pins it down it is a two-line addition.
Test :99 — the comment was stale: it claimed remoteControlStatus and course start/pause/resume were deliberately absent, while all four are in the writable set. It now says what the rule actually is — writable where a frame the appliance accepts was established, either captured from LG's converter or (for remote-maintain ON, which the converter never emits) built from the same shape, acknowledged, and read back on the next state frame.

442 tests passing (409 on master + 33 here), tsc --noEmit and prettier --check clean. README entry added as S5BBP (ST_B_E4H01Y_APL).

tunarabuuu mentioned this pull request Aug 2, 2026

Add nine Korean LG appliances, an ac_common base, and persistent driver stores #111

Closed

tunarabuuu force-pushed the dev-styler branch 2 times, most recently from 95b0228 to 250d3a1 Compare August 2, 2026 22:13

anszom reviewed Aug 22, 2026

View reviewed changes

tunarabuuu force-pushed the dev-styler branch from 250d3a1 to 080a8b4 Compare August 22, 2026 21:46

Navigation Menu

Add LG Styler ST_B_E4H01Y_APL - #115

Add LG Styler ST_B_E4H01Y_APL#115
tunarabuuu wants to merge 1 commit into
anszom:masterfrom
tunarabuuu:dev-styler

tunarabuuu commented Aug 2, 2026

Uh oh!

anszom Aug 22, 2026

Uh oh!

anszom Aug 22, 2026

Uh oh!

anszom Aug 22, 2026

Uh oh!

anszom Aug 22, 2026

Uh oh!

anszom Aug 22, 2026

Uh oh!

anszom Aug 22, 2026

Uh oh!

anszom Aug 22, 2026

Uh oh!

anszom Aug 22, 2026

Uh oh!

anszom Aug 22, 2026

Uh oh!

anszom Aug 22, 2026

Uh oh!

anszom commented Aug 22, 2026

Uh oh!

tunarabuuu commented Aug 22, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

tunarabuuu commented Aug 2, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

anszom commented Aug 22, 2026

Uh oh!

tunarabuuu commented Aug 22, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants