AI Controller

Installation

Install Qubrid AI Controller on Ubuntu primary and worker nodes

This guide covers installing the Qubrid AI Controller on Linux hosts. For an overview of the product and a direct download link, see AI Controller Overview.

Minimum system requirements

Ensure your system meets the following requirements to install AI Controller:

  • 64-bit kernel and CPU support for virtualization
  • At least 16 GB of RAM
  • Minimum 10 CPU cores
  • Ubuntu 22.04 LTS or Ubuntu 24.04 LTS
  • Python 3.10 and above

Supported platforms

AI Controller provides a .deb package for the following Linux distributions and architectures:

Platformx86_64 / amd64
Ubuntu22.04 LTS, 24.04 LTS

Supported GPUs

Data Center

  • L Series: L40, L40S, L4
  • H Series: H200 SXM, H100 PCIe, H100 NVL
  • A Series: A100, A30, A40, A10
  • RTX: A5000, A6000, A4000
  • T Series: T4
  • V Series: V100

GeForce

  • RTX 40 Series: RTX 4090, RTX 4090D, RTX 4080, RTX 4080 Super, RTX 4070, RTX 4070 Ti, RTX 4070 Super, RTX 4060
  • RTX 30 Series: RTX 3090, RTX 3090 Ti, RTX 3080, RTX 3080 Ti, RTX 3070, RTX 3070 Ti, RTX 3060, RTX 3060 Ti, RTX 3050

General system requirements

To install AI Controller successfully, ensure your Linux host meets the following requirements:

  • 64-bit kernel and CPU support for virtualization
  • At least 16 GB of RAM
  • Minimum 10 CPU cores
  • Ubuntu 22.04 LTS and Ubuntu 24.04 LTS
  • Python 3.10 and above

Ensure you meet the system requirements outlined earlier and follow the distro-specific prerequisites before installing.

  • Ensure the system has a proper internet connection during installation
  • Enable ports 3000, 4000, and 5000 on each node for system-level operations
  • Ensure ports 4000-5000 are opened on each node for users' workloads
  • Make sure all nodes have static IP addresses

Download and install

Download the installation package

Download the package from this documentation site or from the Qubrid AI Portal.

qubrid_packages.tar.bz2

Download

You can also fetch the package with wget (if your organization hosts the file at a signed URL from the portal):

wget https://qubridaic-packages.s3.us-east-1.amazonaws.com/v1.2.0-trail/qubrid_packages.tar.bz2

Install on Ubuntu

Prepare the host

Open your terminal and navigate to the directory where qubrid_packages.tar.bz2 is located.

Extract and install the primary node package

Select your OS version directory inside the archive (ubuntu22.04 or ubuntu24.04), then install:

sudo apt-get update
sudo apt-get install bzip2
tar xvf qubrid_packages.tar.bz2
cd qubrid_packages
cd ubuntu22.04
dpkg -i qubridAIcontroller-primary-node.deb

Replace ubuntu22.04 with ubuntu24.04 if you are on Ubuntu 24.04 LTS.

Choose the correct node package

Choose packages according to the node type (primary or worker node). Use the primary-node .deb on your first controller host; use worker-node packages on additional compute nodes.

Start the AI Controller on the primary node

After installation, launch Qubrid AI Controller by typing the following command in your terminal:

qubridAIcontroller

The installer checks firewall status, required ports, RAM, CPU, NVIDIA GPU, Python version, and internet connectivity. When setup completes, you will see a message similar to:

Qubrid AI Controller setup complete!
Access Qubrid AI Controller at: http://<primary-node-ip>:3000
All tasks completed successfully.

Access the AI Controller on the primary node

Open one of the following URLs in a browser on your local network:

  • http://localhost:3000 - if you are accessing from the same server's browser
  • http://<primary-node-ip>:6000 - if you are accessing from a different machine on the same internal network (as documented in the installation guide)

Click Admin sign up to create the first administrator account.

Password must be at least 8 characters long and include uppercase, lowercase, and a digit.

After sign-up, you can log in and see the nodes on the dashboard.

Adding a worker node

To expand your cluster:

  1. Install the worker node .deb package from qubrid_packages on the target server (same extract steps, worker package instead of primary).
  2. From the AI Controller dashboard, add the worker using:
    • Node Name of your choice
    • Static IP address
    • Hostname (pre-configured on the worker host)

See Compute Management for detailed add-node steps in the UI.

Next steps

Dashboard

Monitor nodes and install AI packages after the controller is running.

User Management

Create enterprise users and assign GPU, model, and Docker permissions.

PreviousOverviewNextDashboard