A lightweight desktop multi-model AI programming scheduling center built with Tauri + Vue + antv/x6.
- Multi-Provider Support: Claude 4.7, Codex 5.3, MiniMax 2.5, GLM-5
- Visual Workflow: antv/x6 based collaboration flow visualization
- Harness Strategy Engine: Leader/Coder/Checker workflow orchestration
- ACP Protocol: Simplified Agent Communication Protocol for model routing
- Terminal-like Input: Command-style input for AI prompts
- Code Output: Syntax highlighting, copy, and save functionality
- Hot Reload Development: Instant UI updates during development
- Observable Collaboration: Real-time workflow visualization and logging
- Frontend: Tauri 2.x + Vue 3 + TypeScript + antv/x6 + Pinia
- Backend: Rust (Tauri) + Node.js core logic
- Build: Vite
- Node.js 18+
- Rust 1.70+
- npm or yarn
# Install dependencies
npm installThe application supports two development modes:
npm run tauri devThis starts the Tauri development server with hot module replacement (HMR) enabled. The frontend will reload automatically when you make changes.
For faster iteration, you can run the frontend and backend separately:
# Terminal 1: Start Vite dev server with HMR
npm run dev
# Terminal 2: Start Tauri in development mode
npm run tauri devThe Vite dev server runs on port 1420 with HMR enabled. Changes to Vue components, TypeScript files, and CSS will reflect instantly without rebuilding the Tauri application.
Before using the application, you need to configure your AI provider API keys:
- Open the application
- Navigate to Settings (gear icon in the header)
- Select the provider tab (Claude, Codex, MiniMax, or GLM)
- Enter your API key and configure the endpoint
Supported providers and their default endpoints:
| Provider | Model | Default Endpoint |
|---|---|---|
| Claude 4.7 | claude-4-20250514 | https://api.anthropic.com/v1 |
| Codex 5.3 | codex-5-3 | https://api.openai.com/v1 |
| MiniMax 2.5 | abab6.5s-chat | https://api.minimax.chat/v1/text/chatcompletion_v2 |
| GLM-5 | glm-5 | https://open.bigmodel.cn/api/paas/v4/chat/completions |
# Build for production
npm run tauri buildConfigure your AI model API keys through the application UI or by editing the config file directly.
- Click the Settings icon in the header
- Select the provider you want to configure
- Enter your API key (must be obtained from the respective provider)
- The endpoint is pre-configured but can be customized if needed
- Click Save to store your configuration
You can also set API keys via environment variables for development:
# Linux/macOS
export CLAUDE_API_KEY="your-claude-key"
export MINIMAX_API_KEY="your-minimax-key"
# Windows (PowerShell)
$env:CLAUDE_API_KEY="your-claude-key"
$env:MINIMAX_API_KEY="your-minimax-key"API keys are encrypted using a local encryption scheme before being stored. For production use, keys are stored in the OS keychain (Windows Credential Manager / macOS Keychain / Linux Secret Service).
CodeLink/
├── src/
│ ├── frontend/src/ # Vue frontend
│ │ ├── components/ # Vue components
│ │ │ ├── ModelSelector.vue
│ │ │ ├── FlowVisualization.vue
│ │ │ ├── CollaborationLog.vue
│ │ │ ├── TerminalInput.vue
│ │ │ ├── CodeOutput.vue
│ │ │ └── SettingsPanel.vue
│ │ ├── stores/ # Pinia stores
│ │ └── App.vue # Main app
│ └── core/ # Node.js core logic
│ ├── models/ # Model adapters
│ │ ├── types.ts # Type definitions
│ │ └── adapters.ts # Provider implementations
│ ├── harness/ # Harness engine
│ ├── protocol/ # ACP protocol
│ └── config/ # Config storage
├── src-tauri/ # Tauri backend (Rust)
├── docs/ # Documentation
└── package.json
- Configure Providers: Set up your AI provider API keys in Settings
- Select Models: Choose AI models from the header panel
- Enter Prompts: Type your request in the terminal input
- Watch the Workflow: Observe the flow visualization as AI processes your request
- Review Results: Check the collaboration log for detailed progress
- Copy Output: Use the code output panel to copy or save generated code
The application uses a Leader/Coder/Checker workflow:
- Leader: Analyzes user request, decomposes into subtasks
- Coder: Generates code based on subtasks
- Checker: Verifies code quality and suggests improvements
- Output: Final result with generated code
┌─────────┐ ┌─────────┐ ┌─────────┐
│ Leader │───▶│ Coder │───▶│ Checker │
└─────────┘ └─────────┘ └─────────┘
│ │ │
│ Decompose │ Generate │ Validate
│ Task │ Code │ Output
▼ ▼ ▼
CodeLink supports instant hot module replacement (HMR) during development:
- Frontend Changes: Vue components, TypeScript files, and styles update instantly
- No Full Rebuild: Unlike traditional Tauri development, you don't need to restart the entire application
- Separate Server: The Vite dev server runs on port 1420 with WebSocket-based HMR
To enable hot reload:
- Set the
TAURI_DEV_HOSTenvironment variable - Run
npm run devto start the Vite server - Run
npm run tauri devin another terminal
Example:
# Linux/macOS
TAURI_DEV_HOST=localhost npm run dev
# Windows (PowerShell)
$env:TAURI_DEV_HOST="localhost"; npm run dev# Run unit and integration tests
npm test
# Run tests in watch mode
npm run test:watchThe test suite covers:
- Unit Tests: Model adapters, Harness workflow engine, ACP protocol, config storage
- Integration Tests: Frontend-backend Tauri command communication
| Model | Provider | Capabilities |
|---|---|---|
| Claude 4.7 | Anthropic | Reasoning, code generation, analysis |
| Codex 5.3 | OpenAI | Code generation, debugging |
| MiniMax 2.5 | MiniMax | General purpose, code generation |
| GLM-5 | Zhipu AI | General purpose, local deployment |
We welcome contributions! Please see our Contributing Guide for details.
- Fork the repository
- Create a feature branch
- Make your changes
- Submit a pull request
MIT