Local and Custom Models

Configure a supported cloud provider or OpenAI-compatible endpoint.

Run tars setup to choose Google, OpenAI, Anthropic, a local endpoint, or another OpenAI-compatible service.

For a local or custom endpoint, configure the base URL and the exact model ID exposed by that server. A typical local base URL is:

http://127.0.0.1:8080/v1

Use the endpoint’s actual context limit in setup. Oversized context settings do not increase model capacity and can cause failed requests or poor compression timing.

Privacy considerations

A loopback endpoint can keep inference traffic on the host, but local inference alone does not make the full system private:

• enabled extensions may access the network;
• web tools contact public services;
• Discord messages pass through Discord;
• dashboard exposure and OS-level access remain operator responsibilities.

Bind local inference servers to loopback unless remote access is intentionally protected. If the server requires an API key, store it with tars secret set rather than in config.json.