Odysseus: a private AI workspace on my own box

Everything else on this site has been plumbing - engines, memory, gateways, agents. Useful, but it all lives in a terminal. Odysseus is the friendly face I put on top of it: a self-hosted AI workspace, the local-first version of the ChatGPT or Claude window you already know - running on my own hardware, on my own data.

This is the piece that turns “I run models in a terminal” into “I have a private AI I can actually use, from my phone.”

What it is

Odysseus is a single web app that sits in front of my whole stack. It is open-source, it runs on my box, and everything in it stays on my box. Out of the box it gives me:

• Chat with any of my local models - it speaks to vLLM, Ollama, and others, so the same models I serve behind my gateway show up here to talk to.
• An agent that can take a task and use tools (web, files, shell, memory) to finish it.
• Deep research - multi-step runs that gather sources, read them, and write up a tidy report.
• Documents, notes, tasks, calendar, and email triage - the everyday things, with AI as an assistant sitting beside them, not in charge of them.
• A “cookbook” that scans my hardware and recommends models that will actually fit - built on llmfit, the same fit-checker from the four gates.
• It works on a phone. It installs like an app and is built for a small screen, not just a desktop.

How I run it

I run it as a small always-on service on the box, reachable only over my private network - the same tailnet-only approach I use for everything I do not want on the public internet. It connects to the models I already serve, so I am not running anything twice.

I will keep the sensitive details out of this post on purpose - the login, the email account settings, the keys all live in a local environment file that never leaves the machine, and never goes near a git repo. That is the one rule that does not bend: the workspace can be as rich as you like, but its secrets stay home.

Why it matters

For a long time my local AI was powerful but awkward - great if you were happy in a terminal, invisible to everyone else. Odysseus closes that gap. It is the difference between owning an engine and owning a car you can actually drive. My models, my email, my notes, my research - all in one private place, usable by a normal human on a normal phone, and all of it staying on my own hardware.

That is the whole sovereign-stack idea finally made comfortable: the AI runs on your own machine, and using it feels as easy as the thing you were trying to replace.

You do not need my hardware to want this. A self-hosted workspace like Odysseus runs on modest machines too, and points at whatever models you have - even a single small one:

• It gives your local models a proper home - one calm window instead of a terminal.
• It keeps your data yours - email, notes, chats, all local-first.
• It makes the whole thing shareable - a private AI the rest of your household can actually use.

The plumbing is what makes local AI possible. Something like this is what makes it something you will actually reach for.