Plate Recognizer VisionAlert On-premises

VisionAlert On-premises runs in your own environment so that camera snapshots never leave your network. It runs as a set of Docker Compose services: the web dashboard and API, a worker that processes uploads, the anomaly scheduler, a bridge that routes SFTP uploads, an SFTP server that cameras upload snapshots to, and a PostgreSQL database.

To use it, get a VisionAlert On-premises license from the VisionAlert page in your Plate Recognizer account, or contact us. An internet connection is required during installation and for periodic license validation.

note

The on-premises edition currently supports cameras that push snapshots over SFTP. Direct integrations with Cisco Meraki, Nx Witness, and OpenEye are only available in VisionAlert Cloud.

---

Install

1. Install Docker and the Docker Compose plugin on your server. See the system requirements.

   • Docker Engine install instructions and Docker Compose install instructions
   • Stuck? Check our FAQ for help.

2. Download the Docker Compose bundle. On the VisionAlert page, open the dropdown of your On-premises license and click Download Docker Compose. Unzip vision-alert.zip on the server.

3. Set the admin password. Open vision-alert/.env and fill in ADMIN_PASSWORD. The other required values are already filled in.

4. Set the server address. Open vision-alert/docker-compose.yml and add SITE_URL to the environment list of the web service, for example - SITE_URL=http://YOUR_SERVER_IP:8080. Webhook payloads use it to build image links. See SITE_URL.

5. Start VisionAlert.

   Start the VisionAlert services

   cd vision-alert
   docker compose up -d

6. Connect your cameras over SFTP and open the dashboard at http://YOUR_SERVER_IP:8080.

See the Manual Installation Steps for the full steps, including how to write the .env file yourself.

---

Upgrade

Your data is stored in Docker volumes, so upgrading replaces only the containers.

1. Go to the folder that contains docker-compose.yml.

2. Pull the latest images and recreate the services:

   Upgrade VisionAlert

   docker compose pull
   docker compose up -d

caution

Back up your data before upgrading so you can roll back if needed. See Data and Storage.

---

Stop and Uninstall

Run these commands from the folder that contains docker-compose.yml.

• To stop VisionAlert and keep your data, run docker compose down. Start it again with docker compose up -d.
• To uninstall and delete all data, run docker compose down -v.

danger

docker compose down -v deletes the database, all snapshots, and the SFTP server state. This cannot be undone.

---

System Requirements

Run VisionAlert on a 64-bit x86 (amd64) Linux server or virtual machine. No GPU is required. Camera health checks run on the CPU.

Minimum requirements:

• A 64-bit x86 Linux OS that supports Docker, with Docker Engine and the Docker Compose plugin installed. See the installation guide.
• 2 CPU cores and 4 GB of RAM for tens of cameras. For larger deployments, use 4 CPU cores and 8 GB of RAM.
• 10 GB of free disk space for the images and system files, plus about 0.3 MB per retained snapshot. For example, 100 cameras checked hourly produce roughly 20 GB of snapshots per month.

At steady state the application services use about 1.2 GB of RAM. CPU usage is close to idle between check cycles and rises only while a batch of snapshots is being analyzed.

Network:

Direction	Port / Destination	Purpose
Inbound	8080 (TCP)	Web dashboard and API.
Inbound	2022 (TCP)	SFTP server that cameras upload snapshots to.
Outbound	api.platerecognizer.com on ports 80 and 443	License activation and periodic license validation.
Outbound	Google Gemini API on port 443	Only if you enable the AI glare reviewer (GEMINI_API_KEY).

tip

If you are behind a firewall, whitelist the required IP addresses. Offline license activation is not available.