Skip to main content

Ollama (Local / Self-Hosted)

Ollama is typically used for local or self-hosted inference. In most cases it does not require third-party cloud API keys, but it does require a reachable service endpoint.

1. Deployment Modes​

  • Local mode: install and run Ollama on your own device.
  • Remote mode: deploy Ollama on your own server and expose reachable networking.

2. Pre-checks​

  • Target model has been pulled on the server.
  • Endpoint address, port, and network policy are configured correctly.

3. Configure in Mask​

  1. Select Ollama as provider.
  2. Fill in local or remote service endpoint.
  3. Choose an installed model and run a test.

4. Common Issues​

  • Connection failed: endpoint unreachable, firewall, or closed ports.
  • Model not found: model not downloaded on server side.
  • Slow response: insufficient local compute or excessive concurrency.

5. Privacy Note​

  • Local Ollama is often the easiest way to keep data on-device.
  • If using remote self-hosted nodes, data is sent to your server; you are responsible for network and storage security.