headlamp - SKILL.md Agent Skill

name: headlamp description: > Kubernetes Web UI with AI Plugins: Dynamic Official and Custom Plugins, in-cluster admin authentication, HolmesGPT integration, and Gateway API routing. license: MIT compatibility: - opencode metadata: author: dotfiles tags: [kubernetes, headlamp, UI, holmesgpt, litellm, cluster-admin, plugins]

Headlamp Kubernetes UI Skill

Headlamp is deployed in-cluster in the headlamp namespace and acts as the premium, unified control plane UI for both Kubernetes administrators and AI assistants.

Deployment Details

Namespace: headlamp
Port: 80 (ClusterIP mapped via Gateway HTTPRoute to headlamp.cluster.home)
Gateway Domain: http://headlamp.cluster.home and https://headlamp.cluster.home (routes work on both HTTP and HTTPS listeners).

Authentication

Headlamp is deployed with a pre-configured cluster-admin ServiceAccount. To log in:

Retrieve the permanent login token:

kubectl get secret headlamp-admin-token -n headlamp -o jsonpath='{.data.token}' | base64 -d

Paste this token into the Headlamp Login screen.

AI Plugins Integration

Headlamp dynamically loads two AI-focused plugins in its pod lifecycle using a custom initContainer pipeline:

Official ai-assistant Plugin:
- Downloads from Headlamp’s GitHub releases.
- Automatically patched during deployment (sed in initContainer) to target the ai namespace (where HolmesGPT runs) instead of default.
Custom headlamp-holmes Plugin:
- Compiled from React/TypeScript.
- Mounted as a ConfigMap into the pod.
- Configured dynamically via Ansible variables to point to our in-cluster LiteLLM proxy and route to local models (e.g. rk1-npu-local or fallback tiers).

Useful Commands

# Deploy Headlamp stack
make headlamp

# Verify resources
kubectl get deploy,pods,svc,httproute -n headlamp

# Fetch setup logs (plugin downloads, config patching)
kubectl logs -n headlamp deploy/headlamp -c setup-plugins

# Fetch server logs
kubectl logs -n headlamp deploy/headlamp -c headlamp