Helm releases
Open Helm releases from a cluster to see releases by namespace. Install or upgrade takes a release name, a packaged chart (.tgz, up to 8 MB) and values as JSON. PS-pocketEdge runs a dry-run first and shows a review before applying. Using an existing release name upgrades it.
Each release has History and rollback, and Uninstall. No helm command-line tool is needed. Operations can take up to five minutes.
Run an AI model with Ollama
Deploy Ollama creates an Ollama deployment, an internal service and a persistent volume for the model. Enter a service name, the model (for example llama3.2:1b), CPU, RAM, storage and an optional GPU. The model downloads when the service starts and it becomes ready once the model is available. The service address inside the cluster is shown after deployment.
The cluster needs a storage class that can provision a ReadWriteOnce volume. Deleting the deployment keeps the service and the model volume.