Ollama

Ollama is a simple way to run models locally with one command. Yttri supports it separately from the generic OpenAI-compatible adapter: for Ollama, the model list can be pulled straight from its own API, not only from the standard OpenAI endpoint.

Connecting

1. Install Ollama and pull a model: ollama pull llama3.1 (or any other — see ollama.com/library).
2. Ollama starts a local server on http://localhost:11434 by itself.
3. In Yttri: Settings → Assistant → Models → Add provider → the Ollama preset → “Test connection”.

Caution

Ollama’s address has no /v1 at the end, unlike most OpenAI-compatible servers: http://localhost:11434, not http://localhost:11434/v1. A manually added /v1 is a common reason the connection fails.

No API key needed, the “Local provider” flag is set automatically.

If it won’t connect

• Make sure Ollama is running: ollama serve (it usually starts on its own after installation), or ollama list to see what’s pulled.
• If the model list in Yttri is empty — check that at least one model is pulled (ollama pull ...); an empty Ollama library shows an empty list.