Local AI on Omarchy: A Private Chatbot on the Leanest Desktop You Own

Omarchy already treats AI as a first-class citizen — ten cloud coding agents ship preconfigured. But the same machine is also the best local AI box most people will ever own: a lean Arch system with no bloat competing for RAM, and Ollama one command away. Here’s the private-chatbot setup.

This post joins two of our series: the Omarchy productivity series and the local AI series. If you’re on Windows or Mac, the local AI series stands alone; if you’re on Omarchy, this is where they meet.

Why Omarchy is a great local AI machine

Local models live and die by free memory, and a fresh Omarchy install idles dramatically lighter than Windows 11 or macOS — no background bloat, no mystery services. The RAM your OS doesn’t eat is RAM your model gets, which on a 16GB machine is the difference between a comfortable 8B model and a cramped one. Omarchy’s official manual even lists local models alongside the cloud agents in its AI story: the point was never one vendor.

The setup: one package, one command

1. Install Ollama — from the Omarchy menu (Super + Space > Install > Package) or the terminal: yay -S ollama. (Why Ollama and not LM Studio? On a keyboard-first Linux box it’s not close — our comparison explains.)
2. Start it and pull a model matched to your RAM: ollama run llama3.2:3b on 8GB, an 8B Llama or Qwen on 16GB. Our full picks by machine size are in the small-models guide.
3. Chat. You’re offline, private, and unmetered.

NVIDIA GPU? The Arch package ecosystem has CUDA-enabled builds that make responses dramatically faster — check the manual’s NVIDIA notes. AMD and CPU-only both work fine out of the box.

Wiring it into the Omarchy workflow

This is where the tiling desktop earns its keep:

• Give the model a room. A terminal running ollama run on a dedicated workspace — say, Super + 6 — makes your private AI a one-keystroke glance from anywhere, exactly like the cloud-agent room in our developers guide.
• Feed it with the capture keys. Super + Ctrl + PrintScreen OCRs any text on screen to your clipboard; Super + Ctrl + V holds your clipboard history. Grab text from anywhere, paste it to the local model, nothing ever leaves the machine — a fully private summarize-anything pipeline.
• Point other tools at it. Ollama serves an API on localhost, and a growing pile of editors, note apps, and Obsidian plugins will happily use it as their AI backend instead of a cloud key.

Cloud agent or local model? Both.

The honest division of labor: Claude Code and its peers are senior engineers — use them for real work on real codebases. The local model is the utility layer — private questions, offline drafting, summarizing sensitive documents, and every task where “free and instant” beats “brilliant.” Omarchy is unusual in making both first-class: Super + Shift + Ctrl + A for your cloud agent, Super + 6 for the model that answers to no one.

From the Omarchy productivity series and the local AI series (start here). New to Omarchy? The explainer and install guides for Windows PCs and Intel Macs.