A self-hosted, on-premise AI stack you can install on a single Linux machine in under an hour.
Tronexus is an opinionated, production-grade AI infrastructure stack built for developers, tinkerers, and small teams who want to run their own AI services — without depending on cloud providers, paying per token, or compromising on privacy.
Named after the foundational computing architecture of the Federation, Tronexus is the nexus your self-hosted AI stack connects through.
Alpha release v0.1
Please report any issues or bugs, I will be more than happy to help you.
A curated, pre-integrated collection of open source services that work together out of the box:
- Local LLM inference via Ollama — run models on your own GPU or CPU
- AI chat interface via Open WebUI — full-featured, multi-model, multi-user
- LLM routing and proxy via LiteLLM — unified API across local and remote models
- Workflow automation via n8n — AI-powered automations without writing code
- Authentication via Tronexus Auth — Google OAuth, JWT tokens, API keys, role management
- Reverse proxy and TLS via Caddy — automatic HTTPS, zero config
- Monitoring and alerts via a built-in monitor script — Telegram notifications, container health, disk, security
- Hardened Ubuntu base — fail2ban, UFW, SSH hardening, automatic security updates, and audited service configuration out of the box
- Remote inference support — connect a separate GPU machine as your inference backend
Everything runs in Docker. Everything is configured through a single .env file.
| Component | Minimum |
|---|---|
| OS | Ubuntu 24.04 LTS |
| RAM | 15 GB |
| GPU VRAM | 8 GB (NVIDIA) |
| Storage | 256 GB SSD |
| Network | Any broadband connection |
No cloud account required. No API keys required to get started. DNS/DDNS is optional.
A second machine can be used as a dedicated inference server — Tronexus supports remote Ollama endpoints out of the box.
Full installer coming soon. For now, run this command on your Ubuntu Server
curl -fsSL https://tronexus.dev/install.sh -o install.sh && sudo bash install.shOr clone and configure manually.
git clone https://github.com/Tronexus/tronexus-stack
cd tronexus-stack
cp .env.example .env
# Edit .env with your settings
docker compose up -dDetailed setup guide: tronexus.dev
┌─────────────────────┐
│ Caddy (TLS) │
└────────┬────────────┘
│
┌──────────────────────┼──────────────────────┐
│ │ │
┌──────▼──────┐ ┌───────▼──────┐ ┌───────▼──────┐
│ Open WebUI │ │ n8n │ │ Tronexus Auth│
└──────┬──────┘ └───────┬──────┘ └───────┬──────┘
│ │ │
┌──────▼──────────────────────▼──────┐ ┌───────▼──────┐
│ LiteLLM │ │ Postgres │
└──────────────────┬────────────────┘ └──────────────┘
│
┌───────────▼───────────┐
│ Ollama (local GPU) │
│ or Remote inference │
└───────────────────────┘
- Bootstrap installer script (single
curl | bashcommand) - Bootable ISO installer for fresh Ubuntu 24.04 installs
- Web-based management UI for container control and config
- Built-in documentation site (self-hosted, runs in the stack)
- MCP server integration for AI tool access
- Multi-node support (dedicated inference server setup)
- Community module system for extending the stack
Local first. Every service runs on your hardware. Your data does not leave your machine unless you explicitly configure it to.
Opinionated but extensible. Tronexus makes choices so you do not have to spend weeks integrating disparate tools. Every choice is documented and reversible.
Minimal hardware floor. Designed to run on a single mid-range machine with a consumer GPU. Cloud-grade hardware is not required.
Production-grade from day one. TLS, authentication, monitoring, and automated container recovery are included — not afterthoughts.
Security is not optional. Tronexus ships with a hardened Ubuntu 24.04 base — fail2ban, UFW, SSH hardening, automatic security updates, and audited service configuration are part of the default install. The same security standards applied to professional infrastructure are applied here.
Tronexus is in early development. The best way to contribute right now is to try it, break it, and open an issue.
MIT — use it, fork it, build on it.
"The computer is the most creative tool we have ever built. It is time we ran our own."
— tronexus.dev