Doorboard

An open-source, self-hosted smart-door and smart-room system built around a removable door-mounted appliance. A Raspberry Pi 5 with a Hailo AI accelerator runs the door-side camera and visitor experience, an ESP32 handles immediate physical I/O, and an indoor NUC provides the wider control plane. The system also ties together an ambient wallboard, visitor sessions, local face recognition, birds, aircraft, satellites, 3D-printer status, food recommendations, a photo booth, and a NAS archive.

Make a feature-rich smart door react immediately even when the NUC, NAS, cloud services, Home Assistant, background jobs, or a fresh AI inference are unavailable.

• Physical door interactions must never depend on the network or control plane
• Face recognition and visitor media need a privacy-aware, local-first data model
• Services span an ESP32, door Pi, NUC, BirdNET Pi, ADS-B Pi, and NAS
• Hardware bring-up must preserve the behavior already accepted in simulation

• Split the system into a real-time door plane and a wider control plane
• Give the ESP32 ownership of physical feedback while the Pi 5 + Hailo own local vision and media
• Use typed event contracts and offline-tolerant queues between services
• Keep the project self-hosted and open source under the MIT license

1. 01

   Defined the system architecture, trust boundaries, latency budgets, event catalog, and wire protocols

2. 02

   Implemented and verified the eight software milestones from real-time I/O through security and performance hardening

3. 03

   Built simulator, integration, end-to-end, and performance suites so the distributed system can be accepted without the final hardware online

4. 04

   Prepared hardware bring-up, deployment, monitoring, backup, and recovery runbooks

• The cleanest way to guarantee immediate physical feedback is to make the fast path independent by architecture, not by hoping every dependency stays online.
• Simulation is most useful when it exercises the same contracts, state machines, and failure modes as the eventual hardware.