Skip to main content
Upstream source: Current installation guide. This CPU-only image stores downloaded models under /root/.ollama. Ollama’s API has no built-in authentication, so the example stays on the server’s private application network. Model size and speed depend on available RAM and CPU.

Towbar manifest

.towbar/services/ollama.service.yml

Configure

  1. Prepare the example server, connect the repository, and map production to the branch containing these manifests. Replace the example server IP. Commit the manifests, then sync the repository and inspect the resolved configuration.
  2. Keep the Service’s named volumes attached across deployments. Towbar’s Datastore backup policy does not back up Service volumes, so include them in your own recovery plan.
  3. This manifest has no public domain. Use a trusted connection to the server for the first check. Add a public route only after configuring the product’s authentication or an access gateway.
  4. Deploy the Service, then perform the checks below before enabling auto-deploy.

Verify

From a trusted shell on the server, pull a small model in the running container and query /api/tags. Redeploy and confirm the model is still listed. For field constraints, see Service manifest.
Last modified on September 27, 2026