← back to the arcade

Bring Your Own Model

XenoXanadu's AI games run on a model you host. Nothing is sent to us — the browser talks straight to a model on your own machine.

Heads up — you're on the public site.

Browsers block a public HTTPS page from reaching your localhost model (Private Network Access + CORS), so the AI opponents are available when you run XenoXanadu locally: cd public && python3 -m http.server 8000, then open http://localhost:8000/. The connection tester below only works from a local copy (or against a public HTTPS tunnel to your model).

Connect & test origin

This is shared across every AI game — set it once here.

API key (only for servers that require one)

Quick start with Ollama

  1. Install Ollama and pull a small, fast model: 
    ollama pull llama3.2:3b
  2. Let this site talk to it. Stop Ollama, then start it with this origin allowed: 
    OLLAMA_ORIGINS="…" ollama serve

    On macOS/Linux that's one line in a terminal. On Windows, set OLLAMA_ORIGINS as a user environment variable, then restart Ollama.

  3. Come back here and hit Test connection.

Other servers work too

Anything that speaks the OpenAI /v1/chat/completions API will connect — pick OpenAI-compatible (or leave it on Auto-detect):

If the test won't connect

The browser hides why a local request failed, so check these in order:

  1. Is the server actually running? Open the endpoint in a new tab — Ollama shows Ollama is running.
  2. CORS / origin: the server must allow this page's origin. For Ollama that's the OLLAMA_ORIGINS command above.
  3. HTTPS → localhost block:

Prefer zero browser hassle? Run the whole arcade locally: cd public && python3 -m http.server 8000 and open http://localhost:8000/ai-setup.html — then everything is plain http://localhost and the blocks above disappear.