KubeVision

What is KubeVision

KubeVision is an AI-native Kubernetes dashboard designed for teams that need fast root-cause analysis without handing AI unlimited access. It watches cluster state through Informer caches and pushes updates to browsers over WebSockets, so changes appear in under a second with zero polling. Users can browse Deployments, ReplicaSets, Pods, Services, and other resources, ask questions in natural language, and execute safe, auditable actions from a single web interface.

The backend uses a clean layered architecture: a middleware chain handles request IDs, logging, and authentication before passing requests to handlers, services, and a Kubernetes repository layer. Reads hit the Informer cache first for sub-millisecond response times, with automatic fallback to the API Server on cache misses. Eight core resource types are cached continuously, while more than eighteen additional types are fetched on demand. Writes flow through the API Server for validation, then the Informer detects the change, triggers an event listener, and broadcasts the update to all connected clients instantly.

Key features of the AI-native Kubernetes dashboard

• AI Operations Workspace – investigate live resources and logs with page context, then execute RBAC-controlled actions with explicit approval. AI tools inherit the user's permissions, so a read-only user cannot trigger destructive changes.
• Real-time sync – Informer Watch feeds a WebSocket hub, delivering sub-second updates to every browser session without repeated polling.
• Multi-cluster management – switch between cluster contexts instantly from one dashboard, with separate views for each environment.
• Terminal and logs – full Pod terminal with session recording and playback in asciinema format; real-time log streaming with search and filtering.
• Enterprise security – five-level RBAC, TOTP two-factor authentication with QR setup and recovery codes, audit logging, Secrets masking by default, and instant token revocation.
• Resource topology – a visual ownership graph shows relationships between Deployments, ReplicaSets, Pods, Services, and more.
• Dry-run diff preview – every change is validated by the API Server and displayed as a side-by-side YAML diff before applying, so operators can see exactly what will happen.
• kubectl hints – each UI action shows the equivalent kubectl command, helping users learn the underlying CLI while working in the dashboard.
• Extensibility – add a new resource type with one line of Go in the registry and one UI config block. Custom Resource Definitions (CRDs) are discovered automatically at runtime, and a plugin system supports integrations like Prometheus, Grafana, and ArgoCD.

How to use it

1. Install via Helm by adding the repository: helm repo add kubevision https://kubevision.github.io/charts, then run helm install kubevision gocronx/kubevision.
2. Alternatively, build with Docker: docker build -f deploy/Dockerfile -t kubevision:latest . and run it while mounting your kubeconfig file read-only on port 8080.
3. The default login for a fresh install is admin / admin123; change credentials immediately after first access.
4. Open the dashboard, choose a cluster context, and browse resources. Use the AI workspace to ask about failures or inspect logs and terminal output.
5. Before applying any change, review the dry-run diff. The UI shows the equivalent kubectl command and pauses for confirmation.
6. Administrators can enforce 2FA enrollment, replay recorded terminal sessions, and audit all mutation records.

Who it is for

KubeVision suits platform engineers, SREs, and DevOps teams that manage one or more Kubernetes clusters and need faster troubleshooting with guardrails. It is especially useful for organizations that want natural-language help but refuse to grant AI unrestricted cluster access. Multi-cluster operators get a single pane of glass with instant context switching. Teams with compliance requirements benefit from RBAC inheritance, two-factor authentication, audit logging, and secret masking. Developers new to Kubernetes can use the kubectl hints to learn the CLI while performing everyday tasks.

What to know before you use it

• KubeVision does not replace kubectl or official Kubernetes tools; it layers a visual and AI-assisted interface on top of the standard API.
• AI answers are only as accurate as the cluster state KubeVision can read. If a resource is not in the Informer cache, it falls back to the API Server, but there may be brief propagation delays during rapid changes.
• Mutations always pause for human confirmation and are re-authorized at execution time, but the AI still inherits the user's RBAC permissions. A user with elevated rights should understand what those rights allow.
• The tool does not perform automatic remediation. It proposes changes and shows previews, but the operator must decide to apply them.
• Terminal recording and Secrets masking reduce accidental exposure, but logs or terminal output may still contain sensitive data if not filtered. Regular audits are recommended.
• For authoritative cluster operations, security policies, or vendor-specific behavior, consult official Kubernetes documentation or your cloud provider's support channels.