ollama-helm

Why we included this project

Running Ollama on Kubernetes instead of a single workstation means you have to sort out deployment, storage, and networking yourself. This community Helm chart handles that work in a single install: it deploys Ollama with GPU support for NVIDIA and AMD, persistent volumes for model storage, and optional ingress so the REST API is reachable outside the cluster. At startup the chart manages models, pulling or creating the ones you specify and preloading them into memory. It can also render a Knative Service instead of a plain Deployment when you want scale-to-zero serving. Platform engineers who want reproducible LLM serving without hand-writing manifests will find the values file approach familiar, and the documented upgrade notes and examples cover the common GPU and ingress setups.