Get started

Lumi Virtual Lab runs from a clone of the code. One command sets it up and opens the lab in your browser.

What you need

Python 3.12 or later
The start script creates its own environment, so nothing else is installed globally.
Node.js 22 or later
Only the first time, to build the interface.
A model key
A key from one of nine model sources. OpenRouter is one, and gives you many models with a single key. You pay the provider directly for what the lab uses.
Docker, if you want a sandbox
It runs generated code away from your files. The smallest sandbox image is about 3 GB, and the fuller ones are much larger. Without Docker, you can leave code off.

Install and start

To see what the script would do before it does it, run python3 start.py --plan.

  1. Clone the repository

    git clone https://github.com/AdviktheGreat/Lumi_Virtual_Lab.git
    cd Lumi_Virtual_Lab
  2. Start the lab

    python3 start.py

    The first start takes a few minutes. It creates a Python environment, installs the lab, builds the interface, and starts a server that only your computer can reach. It prints a link and opens it.

  3. Add a model key

    The first page tells you when no key is set and opens the key settings. The lab keeps keys in your operating system’s keychain. A .env file in the folder works too.

  4. Ask something you can check

    Start with a question you already know the answer to, so you can judge the work. You choose the spending limit, and the screen shows what each chat has spent.

Other things to know

You can also download the code as a ZIP file or browse it on GitHub.

Lumi Virtual Lab is early software, and there is no hosted version. If something breaks, open an issue.

Prefer to have it set up with you? Request a pilot.