Skip to content

Gen3 Self Hosted Installation

GT AI OS Release edited this page Jun 3, 2026 · 8 revisions

Self-Hosted installation

Choose one install runbook below. Each runbook takes you from a fresh Ubuntu or DGX host through the interactive operator wizard to your first Control Panel login.

After install, continue to Self-Hosted Control Panel setup. To upgrade an existing host, see Self-Hosted updating.


Install runbooks

Scenario Runbook When to use
Local LAN Install — Local LAN Operators reach Control Panel and the tenant app on the node LAN IP (https://<ip>:3001 and :3002).
Local NAT and DNS Install — Local NAT and DNS Internet users use public DNS names on port 443; LAN users can still use :3001 / :3002.
Cloudflare Install — Cloudflare Internet users reach both apps through Cloudflare Tunnel on public hostnames.

Shared prerequisites

Every scenario needs:

  • Ubuntu 24.04 (x86_64) or DGX OS 7 (ARM64): 8+ CPU cores, 16 GB RAM, 100 GB disk recommended
  • Outbound HTTPS to GitHub Releases and ghcr.io/gt-edge-ai
  • StorageClass suitable for CloudNativePG (for example Longhorn with immediate volume binding)

Public releases on GT-Edge-AI/GT-AI-OS do not require a GitHub PAT when release assets and GHCR are publicly reachable.

Each scenario runbook lists any extra prerequisites (LAN IP, DNS/NAT, or Cloudflare).


Troubleshooting

Symptom What to check
ImagePullBackOff Cluster must pull ghcr.io/gt-edge-ai/gt-ai-os-*:<tag>; GHCR packages must be public
Release download 401/403 Outbound HTTPS to github.com; proxy or firewall blocking anonymous release access
Database pods pending StorageClass and volume binding for CloudNativePG
Wrong URLs after install gt-ai-os-admin report --namespace <your-namespace>

Report runbook gaps on GitHub Issues.

Clone this wiki locally