🇺🇸 English | 🇨🇳 简体中文

• Introduction
• Features
• Tech Stack
• Project Structure
• Quick Start
• Configuration
• API Reference
• Deployment
• Contributing
• License

Introduction

Kubernetes AI Dashboard is an operations-oriented Kubernetes management platform with an integrated AI assistant. Built on the Fabric8 Kubernetes Client for full-resource cluster operations, Spring AI for tool-calling and chat, and a built-in A2A (Agent-to-Agent) protocol framework.

[!NOTE] Designed for ops scenarios — unified K8s cluster resource management with AI-assisted operations.

Features

• ☸️ Full K8s Resource Management — CRUD for Namespace / Pod / Deployment / StatefulSet / DaemonSet / Service / Ingress / PVC / PersistentVolume / StorageClass / ConfigMap / Secret / Event / Job / CRD
• ⚙️ Ops-Focused K8s Views — Node management (read-only), HPA autoscaling (view + min/max tuning), NetworkPolicy, RBAC (ServiceAccount / Role / ClusterRole / RoleBinding / ClusterRoleBinding), ResourceQuota (with usage-vs-limit view and CRUD)
• 🌍 Multi-Cluster Management — upload multiple kubeconfigs via the UI (persisted in MySQL), activate one cluster at a time, hot-switch without app restart
• 📊 Cluster Overview — node health and resource usage statistics
• 📁 Object Storage Browser — MinIO buckets, folder navigation, drag-and-drop upload, image/text preview
• 🐳 Image Registry Browser — Docker Registry v2 catalog, tags, manifest detail, image path copy
• 🔒 Resource Ownership Tracking — platform-created resources are labeled and audited per user; users can only delete what they created
• 🤖 AI Chat — Spring AI integration with tool-calling, streaming responses, and JDBC-backed conversation memory
• 🔌 MCP Toolchain — 7 built-in MCP tool services (K8s / System Info / Command Line / File / SSH / OSS / Docker Registry)
• 🤝 A2A Protocol Framework — Agent-to-Agent JSON-RPC 2.0 endpoint with a reserved SolverProvider extension point
• 🔐 Auth — Sa-Token, 32-char random token, Redis-backed sessions; zh-CN/en-US UI language switching

Tech Stack

Category	Technology	Version
Language	Java	21
Framework	Spring Boot	3.5.8
AI Framework	Spring AI	1.1.6
K8s Client	Fabric8 Kubernetes Client	7.3.1
ORM	MyBatis Plus	3.5.14
Auth	Sa-Token	1.44.0
Object Storage	MinIO	8.6.0
Database	MySQL	8.0+
Cache	Redis	7.0+
Frontend Framework	Vue	3.5
Frontend Build	Vite	8
UI Components	Element Plus	2.13

Project Structure

graph TD
    A[kubernetes-ai-dashboard] --> B[dashboard backend app]
    A --> C[dashboard-fe frontend]
    A --> D[common shared module]
    A --> E[ai-base AI core]
    A --> F[oss object storage]
    A --> G[user auth]
    A --> H[search full-text search]
    A --> I[message site messages]

kubernetes-ai-dashboard/
├── dashboard/              # Backend main app (cc.shiyi.dashboard)
├── dashboard-fe/           # Frontend app (Vue 3)
├── common/                 # Shared module (AOP annotations, utils)
├── ai-base/                # AI core (ChatClient pool, chat memory)
├── oss/                    # Object storage (MinIO)
├── user/                   # Auth (Sa-Token)
├── search/                 # Full-text search (Lucene)
├── message/                # Site messages
└── pom.xml

Quick Start

Prerequisites

• JDK 21+
• Maven 3.9+
• Node.js 18+
• Docker & Docker Compose
• Kubernetes cluster v1.24+

Start Dependencies

cd docker
docker-compose up -d mysql redis minio

Backend

mvn clean install -DskipTests
cd dashboard
mvn spring-boot:run

Backend runs at http://localhost:8080

Frontend

cd dashboard-fe
npm install
npm run dev

Frontend runs at http://localhost:3000

Configuration

Config file: dashboard/src/main/resources/application.yml. All values support ${ENV_VAR:default} overrides.

Env Var	Description	Default
SPRING_DATASOURCE_URL	MySQL JDBC URL	jdbc:mysql://127.0.0.1:3306/dashboard
SPRING_DATA_REDIS_HOST	Redis host	localhost
MINIO_ENDPOINT	MinIO URL	http://localhost:9000
SPRING_AI_OPENAI_BASE_URL	AI API URL	-
SPRING_AI_OPENAI_API_KEY	AI API Key	<YOUR_API_KEY>
KUBERNETES_CLIENT_TRUST_CERTS	Trust K8s API certs	true
DOCKER_REGISTRY_URL	Docker Registry v2 address	-
DOCKER_REGISTRY_USERNAME	Docker Registry username	-
DOCKER_REGISTRY_PASSWORD	Docker Registry password	-
A2A_API_KEY	A2A protocol API Key	<YOUR_API_KEY>

[!IMPORTANT] There is no kubeconfig entry point in config files. Upload kubeconfigs on the Cluster Config page of the web UI — they are stored in MySQL, several clusters may coexist, but exactly one is active and all K8s operations target the active cluster. Switching clusters takes effect immediately (no restart).

API Reference

Internal API (/api/v1/*, Sa-Token login required)

Method	Path	Description
POST	/api/v1/ai/chat	AI chat (streaming/non-streaming)
GET/POST/PUT/DELETE	/api/v1/k8s/*	K8s resource management (namespaced + cluster-scoped)
POST/GET/DELETE	/api/v1/k8s/cluster-config*	kubeconfig upload / list / activate / delete / test
GET	/api/v1/k8s/cluster/stats	Cluster statistics
GET	/api/v1/resources/ownership	Resource ownership query
GET	/api/v1/oss/*	Object storage (buckets / files / folders)
GET	/api/v1/registry/*	Image registry (repositories / tags / manifests)
POST/GET	/api/v1/user/*	User management

A2A Protocol (/a2a, API Key auth)

Method	Path	Description
GET	/a2a/card	Get AgentCard
GET	/.well-known/agent-card.json	Standard discovery endpoint
POST	/a2a	JSON-RPC 2.0 (message/send, tasks/get, tasks/cancel)

MCP Tools (7)

Tool Service	Capability
KubernetesMcpService	Namespace/Pod/Deployment ops, cluster stats
SystemInfoMcpService	CPU/memory/disk/uptime
CommandLineMcpService	Command execution (dangerous-command blacklist)
FileMcpService	File read/write (sandboxed)
SshMcpService	Remote SSH commands
OssMcpService	MinIO upload/list
DockerRegistryMcpService	Docker registry query

Deployment

For Docker deployment see DOCKER_README.md.

# Build backend image
docker build -f dashboard/Dockerfile -t <REGISTRY>/kubernetes-ai-dashboard-backend:latest .

# Build frontend image
docker build -f dashboard-fe/Dockerfile -t <REGISTRY>/kubernetes-ai-dashboard-frontend:latest ./dashboard-fe

Contributing

• Improve cluster monitoring alerts
• More K8s resource types

Issues and Pull Requests are welcome.

License

This project uses an Internal License and is not open-sourced.