What you’re about to do
Bring up one edge box at one site. It’ll start streaming camera frames + posting health, and you’ll see it light up green in the Edge tab.Prerequisites
- A Linux host (Ubuntu 22.04+ recommended) reachable by ssh
- Docker installed (the install script will install it if missing, but only on Ubuntu/Debian)
- Outbound network to your Oxy server (typically port 443 if Oxy is behind TLS, or via Tailscale)
- At least one RTSP camera reachable from that host
Step 1 — Add the device in the UI
- Open your Oxy workspace → Edge → List → Edge boxes tab.
- Click Add device.
- Pick the site this box belongs to, give it a hardware label (e.g. “kitchen-jetson-1”), and submit.
- Copy the one-line install command the wizard shows. It looks like:
Step 2 — Run the install command on the box
SSH into the edge host and paste the command. It will:- Install Docker if missing
- Drop the device identity file at
/var/lib/oxy/device.json - Download the latest signed
docker-compose.yml - Install a
oxy-edge.servicesystemd unit systemctl enable --now oxy-edge.service
Step 3 — Verify in the UI
Within 60 seconds of the worker booting, you should see:- Edge → Topology — the new box appears under its site with a green ring.
- Edge → List → Edge boxes — the row shows status
active,auth_modechip showingjwt, and anedge v…version line. - JWT cutover banner — if it was showing earlier because of this box, it should clear.
What just happened (high level)
/var/lib/oxy/device.json. The update agent watches for new image tags the operator sets via the Rollouts tab and applies them with cosign signature verification + watchdog rollback.
Supported hardware
Officially tested:- NVIDIA Jetson Orin Nano (8GB) — recommended for kitchens with 2–4 cameras
- Generic x86 box (Intel N100 or better, 8GB RAM) — recommended for offices / retail
- Raspberry Pi 5 (8GB) — OK for 1 camera; YOLO inference is the bottleneck
- Older Jetson Nano (4GB) — too tight on RAM for reliable inference
Adding cameras
Once the box is up, add cameras to it from the Cameras tab. You’ll need:- The RTSP URL (e.g.
rtsp://10.0.1.50:554/Streaming/Channels/101) - Optional: substream URL for the lower-resolution preview
- Optional: credentials reference (workspace secret name) if the camera is auth-gated