Bring your own AI

What stays the same

The game engine still decides every dice roll, wound, and consequence — server-side, impartially, exactly as before. BYO only changes who generates the narration. Your local model or cloud provider writes the prose; the engine keeps the world honest. All the same guardrails apply.

Requirements

BYO AI is a benefit of the Platform subscription ($10.99/month). Free and Premium tiers use our hosted AI and cannot connect external providers. To get started, subscribe to Platform in Settings > Billing.

Option 1: Connect a cloud API key

If you already have an API key from Google, OpenAI, or Anthropic, you can connect it directly. Your turns run on your key; your Platform credits pause and stay as a fallback.

Supported providers

• Google Gemini — Get a key from Google AI Studio. Recommended model: gemini-flash-latest.
• OpenAI — Get a key from your OpenAI dashboard. Recommended model: gpt-4o-mini.
• Anthropic (Claude) — Get a key from the Anthropic Console. Recommended model: claude-haiku-4-5.
• Self-hosted / OpenAI-compatible — Any endpoint that speaks the OpenAI API (vLLM, LocalAI, text-generation-webui with OpenAI extension, etc.).

How to connect

1. Go to Settings > AI.
2. Expand "Connect a cloud provider".
3. Select your provider and enter the model name.
4. Paste your API key.
5. Click Connect.

We validate the key once to confirm it works, then store it encrypted. You can disconnect at any time from the same page.

Option 2: Run locally with the connector

For completely free, unlimited play on your own hardware, run a local model with Ollama and our connector app. No API key needed — your machine does all the inference.

Step 1: Install Ollama and pull a model

Download Ollama from ollama.com, then pull a model. For a good balance of speed and quality:

ollama pull llama3.2:3b

Larger models (like llama3:8b or llama3:70b) give richer narration but need more RAM and a capable GPU. Start small and upgrade if your hardware handles it.

Step 2: Get your connection token

1. Go to Settings > AI.
2. Expand "Run on your own machine".
3. Click Generate Token.
4. Copy the token — you will paste it into the connector.

Step 3: Download and run the connector

Download the connector from Settings > AI — Windows and Linux are available now, macOS is coming. Run it, paste your token, and leave it running while you play.

Prefer the terminal, or installing on a headless server? Each snippet below downloads the connector, installs it, and starts it against your token. Replace YOUR_TOKEN with the token from Step 2 — or copy the ready-made command from Settings > AI, which splices your token in for you.

Windows (PowerShell)

iwr https://worldofthaedoria.com/downloads/connector/windows -OutFile wot-connector-setup.exe
.\wot-connector-setup.exe /S
& "$env:LOCALAPPDATA\wot-connector\wot-connector.exe" --token YOUR_TOKEN

/S installs silently. Drop the last line to just install it and launch the app from the Start menu instead.

Linux (Debian / Ubuntu)

curl -fL https://worldofthaedoria.com/downloads/connector/linux-deb -o wot-connector.deb
sudo apt install -y ./wot-connector.deb
wot-connector --token YOUR_TOKEN

Linux (AppImage — any distro, no install)

curl -fL https://worldofthaedoria.com/downloads/connector/linux-appimage -o wot-connector.AppImage
chmod +x wot-connector.AppImage
./wot-connector.AppImage --token YOUR_TOKEN

On a headless box without FUSE, prefix it with APPIMAGE_EXTRACT_AND_RUN=1.

Passing --token runs the connector headlessly — no window — which is what you want on a server. Run wot-connector --help for every flag, or --list-models to check which models your endpoint exposes.

The connector opens an outbound connection to our server — no firewall changes or port forwarding needed. While it is running, your turns route to your local model. When you close it, you automatically fall back to your Platform credits.

Connector environment variables

For non-default setups, you can configure the connector with environment variables:

• WOT_LOCAL_URL — Your local endpoint (default: http://localhost:11434/v1)
• WOT_LOCAL_KEY — Bearer token if your server needs one (default: ollama)

Troubleshooting

My turns are slow

Local inference speed depends on your hardware. If you are running on CPU only, expect longer waits. Try a smaller model (the 3B variants are much faster than 8B or larger), or consider a cloud provider if speed matters more than running locally.

The connector disconnects

The connector auto-reconnects if the connection drops. If it keeps failing, check that Ollama is running (ollama serve) and that the model is pulled. You can test locally with curl http://localhost:11434/v1/models.

I see an error about my subscription

BYO AI requires an active Platform subscription. If your subscription lapsed, you will fall back to the free daily allowance until you resubscribe. Check your status in Settings > Billing.

Can I use my ChatGPT Plus subscription?

ChatGPT Plus is a consumer product and does not include API access. You need a separate OpenAI API account with its own billing. However, if you are already paying for API access through another project, you can reuse that key here.

Security and privacy

Your API keys are stored encrypted at rest with AES-256-GCM. We validate them once when you connect, then never display them again. You can revoke access at any time by disconnecting in Settings.

The local connector never exposes your machine to inbound connections — it dials out to our server. Your model, your keys, and your compute stay entirely on your hardware.