Quick Start

Get OKO OS running with AI in five steps. This guide applies to all supported devices.

1

Download & Install

Download the installer ISO for your device from the device page, write it to a USB drive, and boot from it. The TUI installer walks you through disk selection, hostname, admin user, and timezone.

2

SSH In

ssh admin@<server-ip>
3

Get a Model Config

Download a tested model configuration from the OKO-configs repository:

curl -o ~/oko-ai.yaml https://raw.githubusercontent.com/satisfanly/OKO-configs/main/coder-qwen38-DavidAU.yaml
curl -o ~/oko-ai-embed.yaml https://raw.githubusercontent.com/satisfanly/OKO-configs/main/embed-gemma.yaml
4

Download Models

This reads both config files and downloads all referenced GGUF models to ~/models/ and ~/models-embed/:

oko-dl --auto
5

Restart the Proxies

The router, stream, and OpenAI-compat services are already running and will pick up the backends automatically:

sudo systemctl restart oko-ai-proxy
sudo systemctl restart oko-ai-proxy-embed
6

Test

You should see your model listed. The Ollama-compatible API is on port 11434, the OpenAI-compatible API on port 8000.

curl http://localhost:11434/api/tags
7

Use with ollama run

From any machine on your network, point the Ollama client at your OKO server:

export OLLAMA_HOST=http://<oko-server-ip>:11434
ollama run <model-name>

Replace <oko-server-ip> with your OKO server's IP address and <model-name> with the model name from your config (e.g. qwen38-DavidAU).