GT AI OS is a self-hosted enterprise AI platform for RKE2 (Kubernetes). A Control Panel handles administration; a tenant app gives users agents, chat, and document/RAG workflows. Data and inference stay in your environment.
This repository has releases and install documentation for v3.0.2 and later. Install from GitHub Releases; images are on ghcr.io/gt-edge-ai.
Some capabilities require an Enterprise license from GT Edge AI. See Enterprise license below.
Choose your platform for detailed runbooks in the wiki, then use one install entry point below.
| Platform | Guide |
|---|---|
| Ubuntu 24.04 (x86_64) | Installation — Ubuntu |
| NVIDIA DGX OS 7 (ARM64) | Installation — DGX |
| Method | Use when |
|---|---|
| Operator script | Interactive or scripted install on a fresh RKE2 host |
Quick Installer .deb |
You want gt-ai-os-operator installed via apt |
gt-ai-os-admin CLI |
Fully non-interactive install from release tarballs |
Start here: Installation (wiki) · Gen 3 Getting Started
v3.0.2+ uses registry-backed images only (no per-platform image bundle tarballs). While this repository is private, use gh auth or a release-read token; after the repository and GHCR are public, token-free install is supported.
Operator (public repo):
curl -fsSL https://github.com/GT-Edge-AI/GT-AI-OS/releases/latest/download/gt-ai-os-operator.sh | sudo bashUbuntu / DGX (operator menu):
sudo gt-ai-os-admin update --yes --namespace <namespace-name> --to latestValidate after upgrade:
sudo gt-ai-os-operator --validate --namespace <namespace> --lan-ip <your-lan-ip>For troubleshooting, see Updating and Gen 3 Admin — Updates.
Local-network installs (typical lab / LAN HTTPS):
| App | URL | Default login (first install) |
|---|---|---|
| Control Panel | https://<host>:3001/login |
gtadmin@test.com / Test@123 |
| Tenant App | https://<host>:3002/login |
gtadmin@test.com / Test@123 |
Replace <host> with the fixed LAN IP you set at install time. Change default passwords before production use. Accept the self-signed certificate warning in the browser.
| Platform | Architecture | Minimum resources |
|---|---|---|
| Ubuntu 24.04 | x86_64 | 8+ CPU cores, 16 GB RAM, 100 GB disk (see wiki) |
| DGX OS 7 | ARM64 (Grace) | See DGX installation guide |
| Production HA | Multi-node RKE2 | Per release manifest and wiki |
Clusters must reach ghcr.io/gt-edge-ai (or your approved mirror). Databases require a StorageClass suitable for CloudNativePG (for example Longhorn with immediate volume binding). Details are in the wiki.
| Topic | Detail |
|---|---|
| Registry | ghcr.io/gt-edge-ai/gt-ai-os-* tagged to match the release (for example v3.0.2) |
| Helm / manifest | gt-ai-os-v<version>.tgz and release-manifest.json on each Release |
| Image bundles | Not published for v3.0.2+ |
- Agent builder — Custom agents, templates, role-based access, and guardrails
- Datasets and RAG — Document-centric chat with retrieval-augmented generation
- GT API — OpenAI-compatible API for integrated applications (see wiki)
- Teams and sharing — Workgroup access to agents and datasets
- Observability — Usage dashboards, conversation review, and operational metrics
- Enterprise controls — SSO, compliance mode, billing, and GT API (license required; see below)
You can install and use GT AI OS without a license for evaluation, with low user limits. An Enterprise license raises seat limits and unlocks the integrations below. Licenses are not included in this repository—you request one from GT Edge AI and activate it on the Control Panel License page.
Contact GT Edge AI to request a license for your deployment: gtedge.ai/contact-us.
| Area | Requires Enterprise license |
|---|---|
| Control Panel — SSO | Identity providers and SCIM provisioning |
| Control Panel — Compliance mode | Compliance configuration |
| Control Panel — Financial controls | Billing policy, allocations, and related broadcasts |
| Tenant app — GT API | API keys and external application integrations (also enable GT API for the tenant) |
| Tenant app — Billing | Usage billing views and allocations (tenant owner) |
Included without a license (subject to seat caps): install and updates, agents, chat, RAG, datasets, models, teams, backup/restore, and most observability (billing tabs stay hidden until licensed).
After you receive a license file, open Control Panel → License, upload it, and confirm the status shows active before using the features above.
Full guides are in the wiki:
| Topic | Description |
|---|---|
| Gen 3 Getting Started | Primary guide for RKE2 deployments |
| Gen 3 Admin Getting Started | Control Panel administration |
| Gen 3 Chat | Tenant chat and conversations |
| Gen 3 Agents | Agent configuration and templates |
| Gen 3 GT API Overview | API keys and compatibility |
| Installation | Platform install runbooks |
| Updating | Upgrade an existing deployment |
| Troubleshooting | Common issues |
Legacy Gen 2 wiki articles remain for reference during migration; prefer Gen 3 articles for new installs.
gt-ai-os-operator # Install / update menu (use sudo -E when needed)
gt-ai-os-admin report --namespace <ns> # URLs and status
gt-ai-os-admin validate --namespace <ns> # Post-install checks
sudo gt-ai-os-operator --nuke --yes # Remove install (lab reset)| Version | Release host | GHCR owner |
|---|---|---|
| ≤ v3.0.1 | GT-Edge-AI-Internal/gt-ai-os-release |
gt-edge-ai-internal |
| ≥ v3.0.2 | This repository | gt-edge-ai |
v3.0.2 does not dual-publish to the Internal release repository. Use the supported upgrade path to rewire release repo, GHCR owner, and registry pulls.
┌────────────────────────────────────────────────────────────────┐
│ GT AI OS │
├──────────────────┬──────────────────────┬──────────────────────┤
│ Control Panel │ Tenant App │ Resource Cluster │
│ (Admin UI) │ (User UI) │ (AI inference routing)│
├──────────────────┴──────────────────────┴──────────────────────┤
│ PostgreSQL │
│ Control DB │ Tenant DB │
└────────────────────────────────────────────────────────────────┘
- Runbook issues: GitHub Issues
- In-app: Tenant Instructions and Contact support (when configured by your operator)
- Security: SECURITY.md and contact GT Edge AI
- Licensing: Request an Enterprise license from GT Edge AI; activate on Control Panel → License
Apache License 2.0 — see LICENSE.
GT AI OS · GT Edge AI