Skip to content

Ollama

Ollama runs AI models on your own computer. Nothing you type leaves it, and there is no key and no daily limit. You need a computer that can run a model well; small models make more mistakes with rare words.

  1. Install Ollama and download a model, for example ollama pull llama3.1.

  2. Ollama only answers web pages and extensions it trusts. Start it with Kotiko allowed:

    Terminal window
    OLLAMA_ORIGINS=chrome-extension://*,moz-extension://* ollama serve

    If Ollama runs as a background service, set OLLAMA_ORIGINS in that service's settings instead (Ollama's FAQ explains how on each system), then restart it.

  3. In Kotiko's Settings, Word lookups, choose Ollama under Look words up with. The address is http://localhost:11434/v1.

  4. Choose your model, then select Test.

If Test can't reach Ollama, it is either not running or not started with OLLAMA_ORIGINS as above.