KubeVision

An AI-native Kubernetes dashboard for understanding, troubleshooting, and operating clusters with confidence.

English · 简体中文

Pod Terminal & Logs	Resource Topology
Global Search	Terminal Audit

KubeVision is built around an AI assistant that can use live cluster context, not just answer generic Kubernetes questions. Ask it to inspect resources, summarize Pod logs, explain unhealthy workloads, query Prometheus, or prepare a resource change in natural language.

AI actions stay inside the dashboard's operational controls:

• Context-aware: Understands the selected cluster, namespace, page, and resource
• Tool-enabled: Reads resources, logs, cluster health, and Prometheus data instead of guessing
• Permission-aware: Every tool call is checked against the current user's RBAC role
• Human-approved: Create, update, patch, and delete operations pause for explicit confirmation
• Auditable: Approved AI mutations are recorded with user, target, tool, and outcome
• Model-flexible: Connects to OpenAI-compatible providers such as OpenAI, OpenRouter, DeepSeek, and Qwen

KubeVision is designed for human-in-the-loop operations. The assistant helps investigate and execute, but it does not bypass RBAC or make unapproved cluster changes.

Full documentation is available at: kubevision-docs

• 🚀 Quick Start - Installation, quick start, configuration
• 🤖 AI Operations - Context-aware investigation and guarded actions
• 🏛️ Architecture - System design, data flow, component interactions
• 📘 User Guide - Features walkthrough and usage guides
• 🔌 API Reference - REST & WebSocket API documentation

• AI Operations Workspace: Context-aware investigation and RBAC-controlled cluster actions from natural language
• Real-time Sync: Sub-second updates with zero polling via Informer to WebSocket Push architecture
• Multi-cluster: Manage all your clusters efficiently from a single, unified dashboard
• Pod Terminal & Logs: Integrated terminal emulator with fully recorded and replayable sessions
• Live Pod Metrics: Current CPU and memory usage with per-container requests and limits from Metrics Server
• Resource Topology: Visual ownership graph displaying real-time relationships for workloads
• Global Search: Cmd+K fuzzy search across your entire cluster ecosystem
• Deployment Ops: Seamlessly scale, restart, rollback, and preview diffs before applying changes
• Helm Workspace: Quickly deploy PostgreSQL, Redis, or gocron, browse other charts, and run guarded installs, upgrades, rollbacks, and removals as recoverable background operations
• Security & Access Control: 5 built-in RBAC roles, TOTP 2FA, and asynchronous audit logging
• Fully Extensible: 26+ built-in resources with CRDs that are auto-discovered at runtime
• Modern UI: Responsive desktop and mobile workflows with full i18n localization and native dark mode support

The fastest way to deploy KubeVision in your cluster is via Helm:

# 1. Install KubeVision from the public OCI registry
helm install kubevision oci://ghcr.io/gocronx/charts/kubevision \
  -n kubevision \
  --create-namespace

# 2. Temporarily access the dashboard from your local machine
kubectl port-forward --namespace kubevision svc/kubevision 8080:8080

# 3. Open the web interface
# http://localhost:8080

Default login credentials: admin / admin123

kubectl port-forward is intended for local evaluation and troubleshooting. For production, expose KubeVision through an HTTPS Ingress/Gateway or a LoadBalancer Service. See the Installation Guide.

For other deployment methods (Docker, Binary, Development), please refer to the Installation Guide.

The kubevision binary doubles as a powerful administrative CLI. You can manage accounts natively or chat with your cluster from the shell:

# Account management
kubevision reset-password --username admin
kubevision create-user --username dev --role editor --email dev@example.com

# Terminal AI assistant (uses the AI provider configured in Settings)
kubevision ai "why is the web pod crashing in default?"   # one-shot query
kubevision ai                                              # interactive REPL

Contributions are welcome! Please open an issue first to discuss what you would like to change.

One thing to note: commit messages are validated by a git hook (commitlint), so use the interactive commit tool instead of git commit:

pnpm install      # first-time setup (installs git hooks in the web directory)
pnpm run commit   # create a properly formatted commit

This project is licensed under the MIT License. See the LICENSE file for details.