Connect the camera
Complete installation first so you have a connection file at ~/.config/anava-local/camera-connection.json.
1. Set the camera to HTTPS-only
Before you import the connection file, the camera's administrator access must be HTTPS-only. Open the camera over HTTPS and set System > Network > HTTP and HTTPS > Admin to HTTPS only.
Local import is rejected on a camera that still allows plain HTTP admin access. This protects the shared secret and private CA certificate in the connection file from ever crossing an unencrypted connection.
2. Select Local inference mode
In the camera's Anava settings, open the AI inference section and choose Local as the mode. The Anava portal's setup dialog shows the exact verified gateway version and download links for your installation.
If your entitlement is not a paid Hosted EDGE license, this option is disabled and the camera shows: "Local inference requires an active paid Marketplace entitlement."
3. Import the connection file
Upload camera-connection.json through the camera's Local settings panel.
If the camera is not HTTPS-only, the import is refused with: "Private connection profiles require HTTPS. Open this camera over HTTPS, then set System > Network > HTTP and HTTPS > Admin to HTTPS only."
Importing a new connection file clears any prior connection state. The camera stays disarmed until you test the connection and run a sample again.
4. Test the connection
Run the connection check. This confirms the camera can reach the gateway, the certificate is trusted, and the shared secret is accepted. It does not confirm the model works.
If this fails, the camera shows: "Local inference connection failed. Check the endpoint, certificate, secret, and worker status." See Troubleshooting.
5. Analyze a real sample
Run a sample detection against a real scene. This is a separate step from the connection check and is required before you can arm detection.
Run this step again after any restart of the gateway or Ollama. The first call after a restart is a cold call and takes longer; see Requirements.
6. Choose second-stage verification, optional
Second-stage verification adds a confirmation call using the same local model. It is off by default. Leave it off if you want the fastest single-shot result; turn it on if you want an extra confirmation pass and can accept the added time.
7. Arm detection
Once the sample result looks correct, arm detection.
If your entitlement lapses after arming, the camera disarms and shows "entitlement inactive."
Next step
Related: anava.ai/edge/local