Free, private AI with Ollama

Ollama runs AI models on your own computer. It’s free, works offline, and nothing you read leaves your machine. It needs a reasonably recent computer; larger models need more memory.

1. Install Ollama and download a model. In Terminal: ollama pull mistral
2. Let browser extensions use it. On a Mac, in Terminal: launchctl setenv OLLAMA_ORIGINS "chrome-extension://*", then quit Ollama from the menu bar and open it again. On Windows or Linux, set the environment variable OLLAMA_ORIGINS to chrome-extension://* and restart Ollama.
3. In RapidReader, open Integrations → Ollama, enter http://localhost:11434, click 🧪 Test, and choose your model.

Reading more at once

Ollama Context Window lets a local model read more of a long document at once, from 8K to 128K, capped at what the model supports. A bigger window uses more memory: mistral measured about 5.6 GB at 8K and 8.9 GB at 32K.

“Ollama refused the request (403)”

Ollama hasn’t been told to trust browser extensions. Run the OLLAMA_ORIGINS step above, then quit and reopen Ollama.