The hub is a small Ubuntu Server box running a Django/Python stack (PostgreSQL, Redis, Supervisor, etc.). It sits on-prem by design, on a UPS with the router and PoE switch. Nightly LVM/Borg backups go to removable media; recovery is a bare-metal restore, not a cloud dance.
The backbone is Ethernet/PoE. Panels are DIN-rail “Game Changer” boards with 16 universal I/O plus digital interface bays for buses like DALI or I²C. You snap in relay/dimmer/input modules for lights, valves, sensors, blinds, gates, etc. Critical loops (e.g. wall switch ↔ light) run directly on the panel, so basic behavior continues even if the hub is down. Rooms are handled by Sentinel units: PoE ceiling devices with presence sensing, environment sensors, AI voice assistant, and a 100 dB siren, all talking to the local hub over LAN.
Some concrete details that may interest HN:
- Wire-first, security-first architecture: PoE spine, no radios in the backbone, UPS-aware power handling, and offline-first behavior for core functions. - Pro-grade permissions and roles: hub masters, instance superusers, and per-component ACLs so electricians, clients, residents, and guests each see and control exactly what they should. - Developer-facing design: Django Admin and key-only remote SSH into the hub, Python virtualenv for integrations, structured logs under /var/log/simo, and a documented extension model for third-party systems. - Automation model: components (lights, valves, sensors, scenes, alarm groups, global state, etc.) plus Script components running plain Python. The app can generate an initial Python script from a natural-language description; you then edit it in Admin. - Mobile-first commissioning: an installer can create an instance, pair panels, map IO, build scenes/presence lighting/alarm groups, and hand over access via QR code from a phone, then drop into Admin/SSH only when deeper work is needed.
The goal is not a pile of Wi-Fi gadgets, but something an electrician or integrator can deploy as infrastructure: repeatable panels, an auditable hub, predictable behavior under failure, and an open-source stack that developers can actually reason about.
I’d appreciate blunt feedback on the architecture, security model, and developer ergonomics, especially from people who run or deploy whole-house systems today.
Website: https://simo.io Docs: https://simo.io/wiki/ GitHub: https://github.com/SIMO-io