pasture

Keep more local AI models than fit on your disk.

pasture works with Ollama, the program that runs AI models on your own computer. You download and use models the way you always do. When they need more room than you give them, pasture moves the ones you have not used lately off the disk, and brings one back as soon as something asks for it.

New to pasture? Get started: install it, download a model, and use it, in about ten minutes.

Guides

Each guide walks you through one thing you might want to do.

• Get started
• Download, use, and remove models
• Choose how much space models use
• Keep models on an external drive
• Get ready to work offline
• Use your models with Cinchapi
• Use your models in other apps
• Share a computer's models with your team
• Update or remove pasture
• When something goes wrong

Reference

These pages describe each part of pasture in full, for when you want the details.

• How pasture works: where pasture runs, and what it promises about your models.
• Install pasture: build it yourself, the service on macOS and Linux, and how pasture finds or downloads Ollama.
• Choose models with pasture onboard: how onboard measures the machine and picks models.
• Keep a library bigger than your disk: the budget, the reserve, the status, and the archive folder.
• Pairing and the link to your account: how Cinchapi reaches a paired machine.
• Let web apps use your models: allowed origins, and what an allowed page can do.
• Serve HTTPS on a public host: HTTPS with API keys.
• Commands and configuration: every command, the Ollama settings, and the config file.
• Files: where pasture keeps its settings, state, and logs.
• Troubleshooting: what each error means.
• Third-party licenses: the open-source code inside pasture, and its licenses.