I’ve published a custom integration for the Somfy Connect UAI+ gateway, which drives Somfy SDN (RS-485) motors. It talks to the gateway locally over its telnet JSON-RPC interface — no cloud, and no external Python dependencies.
Why another one
There are already a few UAI+ integrations, and they all assume every motor reports its position. That assumption breaks Irismo drapery motors.
Irismo has no SDN NodeType. It reaches the bus only through Somfy’s 1811129 bridge module, which does open, close, stop and group addressing — and nothing else. The gateway reports those nodes as type SDN Module; the word “Irismo” never appears on the wire, so a classifier looking for it matches nothing. Integrations that advertise SET_POSITION for every node leave those blinds with a dead slider and a permanently unknown state.
This one asks the gateway what each node is, verifies the answer with a position probe, and builds each entity from what that node can actually do. Unrecognised hardware degrades to working open/close/stop rather than a broken slider.
What you get
- Automatic discovery of motors and groups — no hand-typed node addresses
- One cover per group (enabled), one per motor (disabled by default, so a large install doesn’t add state-change traffic you didn’t ask for)
- Position and a working slider for Sonesse; open/close/stop for Irismo
- Optionally add the gateway’s web password (separate from the telnet one) and Irismo motors report open/closed state too
- Per-motor capability override for hardware it doesn’t recognise, adjustable poll interval, and diagnostics with credentials redacted
Install
HACS → ⋮ → Custom repositories → add the repo as an Integration, then install Somfy SDN and restart. A PR to add it to the HACS default list is open, so eventually it should be searchable directly.
Scope, honestly
It’s been developed and validated against a single UAI+ (HW 04.04 / FW 02.03.11) running a mix of Sonesse and Irismo motors across a couple of dozen groups. The capability model is designed so other firmware and motor families degrade gracefully rather than break, but I can’t claim they’re tested. Issues and reports very welcome — particularly from anyone running Glydea, or a different firmware revision.