A local server
Ollama, LM Studio, or any OpenAI-compatible endpoint.
In short
If you run a model on your own machine, Thesivor can use it through an OpenAI-compatible endpoint. Presets exist for Ollama and LM Studio, and no API key is needed for a local server.
1Start your server
For Ollama the default address is http://localhost:11434/v1; for LM Studio it is http://localhost:1234/v1.
2Add it in Settings
Settings → AI engines → OpenAI-compatible: enter the base URL and, if your server needs one, a key. Press Test.
For technical users
Wire: /chat/completions over SSE with an optional Bearer key, and /models for the list (local models are shown as local/free). Errors use the shared mapper and name the engine, so a message reads "Could not reach Ollama…" rather than a generic failure. This is the only way to keep every prompt on your machine while still using the model-backed features. Larger local models are slower; the same visible retries and Stop apply.