Tuya local lane: SmartLife Wi-Fi devices as kit devices over the LAN protocol

Anchor row. Tuya Zigbee devices already work through zigbee2mqtt (ZY-ZTH02 bound to canopy since 2026-08-30). Tuya Wi-Fi (SmartLife) devices had no lane. Operator chose the local Tuya protocol route (tinytuya over TCP 6668, per-device local key from a one-time wizard run, brain never contacts Tuya). PASS T1 LANDED 2026-09-07 on feat/dashboard-v2 (uncommitted, not hotpatched to the Pi): brain tuya_local.py (per-device worker, persistent socket, DPS->datapoint normalisation via tuya_catalog.py archetypes smart_plug / smart_switch / water_tester, honesty states LIVE/STALE/OFFLINE/KEY CHANGED and PENDING/SYNCED/DIFFERS/FAILED), shared device_bindings.py, role rows merged into zigbee_by_role via a provider hook, entities exported as dsc_tuya_, tuya_switch rule action, lane-tagged GET /settings/devices/actuatable, routes under /settings/tuya/*, 15 tests against a fake device (377 brain tests green). SPA: TuyaLocalCard on Settings > Devices (two-step add drawer: key export guide + paste devices.json, per-device IP/version/Probe/type; shared bind row + honesty line; edit drawer with DPS map/scales; delete confirm), rule editor Tuya plug action. Operator guide docs/ops/TUYA-LOCAL-SETUP.md. Verified in the local dev pane end to end against a local brain (import, probe failure hint, type save, bind -> BOUND, actuatable target, rule action). STILL TO VERIFY: a real SmartLife plug and the operator’s water tester on the Pi (DPS defaults for the tester are the common Tuya profile and must be confirmed by Probe). Auxiliary plug* roles and sensors only: Tuya plugs keep last state if the brain dies. T2 (hygrometer, power strips, energy history) and T3 (LAN discovery helper, key-rotation detection) remain.