Kubernetes and Argo CD manifests for running an Ollama backend together with the Open WebUI frontend. The manifests default to deploying the stack into an ai namespace and exposing the UI through an ingress.

• ai/ – Kustomize base that provisions the Ollama DaemonSet, Open WebUI deployment, service, ingress, and the ai namespace.
• ingress-lb/ – Kustomize base for a cluster LoadBalancer service that fronts an existing NGINX ingress controller (handy on MicroK8s).
• app-ai-stack.yaml / app-ingress-lb.yaml – Argo CD Application resources that point at the ai and ingress-lb folders in this repository.

kubectl apply -k ai
kubectl apply -k ingress-lb

Update hostnames or other defaults in the manifests before applying if they differ from your environment (for example the ai/home.lan host in ai/openwebui-ingress.yaml).

1. Commit this repository to your own Git remote and update the spec.source.repoURL values in both app-*.yaml files.
2. Apply the Argo CD applications: kubectl apply -f app-ai-stack.yaml and kubectl apply -f app-ingress-lb.yaml.
3. Sync the applications in the Argo CD UI/CLI; automated sync and self-heal are already enabled.

The Ollama DaemonSet uses a hostPath volume at /var/lib/ollama to persist downloaded models on each node. Ensure that path exists (or change it) on every node where the DaemonSet should run.

• Adjust resource requests/limits in ai/ollama-daemonset.yaml to match your cluster capacity.
• Configure TLS or external DNS for the Open WebUI ingress if exposing it beyond a trusted network.