A desktop AI agent platform with a local runtime and an extensible Gin cloud service.
由 Tauri 2 + Vue 3 + NestJS + SQLite 桌面端和 Go + Gin + PostgreSQL + Redis 云端服务组成的 AI 智能体平台。
Xingyao Agent Platform is a desktop tool for managing AI agents, skills, and models in one place. It is designed for users who want a local, structured, and extensible workspace for building agent workflows.
星曜 Agent Platform 把 Agent、Skill、Model 和本地工具运行时统一在桌面应用中,并通过独立的 Gin 云端服务承载账号、积分、订单、Skill 市场和软件分发。
Key ideas / 核心理念:
-
Turn each agent into a “team member” with its own model and skill set
-
Make skills reusable and searchable, so they stay manageable even when the library grows to thousands
-
Keep the desktop app lightweight while the backend runs as a Tauri Sidecar
-
将每个 Agent 视作“团队成员”,为它绑定独立模型和技能
-
让技能可复用、可搜索、可分页,即使技能库增长到几千条也能稳定管理
-
用 Tauri Sidecar 承载后端服务,保持桌面应用轻量化
-
本地 NestJS 只负责 Agent Runtime,远程 Gin 服务负责平台业务,避免执行权限和云端账户耦合
-
Agent management / Agent 管理
- Card-based layout for browsing and editing agents
- Bind a model and multiple skills for each agent
-
Skill library / 技能库
- Supports
prompt,tool, andmixedskill types - Paginated search, filtering, and sorting for large libraries
- Preset skills can be imported with one click
- Supports
-
Model management / 模型管理
- Built-in provider presets for fast setup
- Custom model configuration with provider, base URL, and API key
-
Auth and profile / 认证与个人信息
- JWT login on the backend
- Default admin account is initialized automatically on first run
-
Desktop experience / 桌面体验
- System tray
- Single instance
- Close-to-tray behavior
- Backend starts automatically through Tauri Sidecar
-
Cloud service / 云端服务
- Gin modular monolith with PostgreSQL and Redis
- Registration, login, points ledger, orders, Skill versions, permissions, devices, releases, audit and outbox foundations
Smooth gradient background and floating animation.
渐变背景加轻量动效的登录页。
Card-style agent management with model and skill binding.
卡片式 Agent 管理,支持模型和技能绑定。
Optimized for large skill libraries with search, filters, and pagination.
面向大规模技能库的优化界面,支持搜索、筛选和分页。
Provider presets plus custom model setup.
支持厂商预设和自定义模型配置。
System and runtime controls.
系统与运行状态相关设置。
| Layer / 层级 | Stack / 技术 |
|---|---|
| Frontend / 前端 | Vue 3, Vite 6, Element Plus, Pinia, TypeScript |
| Backend / 后端 | NestJS 10, Prisma 6, SQLite |
| Desktop / 桌面端 | Tauri 2, Rust |
- Node.js 18+
- Rust stable
- Visual Studio Build Tools on Windows
npm install
cd frontend && npm install
cd nestjs && npm install && npm run prisma:generateRun backend and frontend separately:
分别启动后端和前端:
cd nestjs && npm run start:dev
cd frontend && npm run devRun the full desktop app with Tauri:
使用 Tauri 启动完整桌面应用:
cd nestjs && npm run build
npm run tauri devcd nestjs && npm run build
npm run tauri buildThe desktop bundle is generated under src-tauri/target/release/bundle/.
桌面安装包会输出到 src-tauri/target/release/bundle/。
The backend initializes a default admin account on first launch:
后端会在首次启动时自动初始化默认管理员账号:
- Username / 用户名:
admin - Password / 密码:
123456
D:\AIAgentPlatform\
├─ frontend/ Frontend app / 前端应用
│ ├─ src/
│ │ ├─ api/ Axios wrapper and API definitions / 接口封装
│ │ ├─ router/ Routing and guards / 路由与守卫
│ │ ├─ stores/ Pinia stores / 状态管理
│ │ └─ views/ Pages / 页面
│ └─ vite.config.ts
├─ nestjs/ Backend service / 后端服务
│ ├─ src/
│ │ ├─ auth/ Authentication / 认证
│ │ ├─ agent/ Agent module / Agent 模块
│ │ ├─ skill/ Skill module / Skill 模块
│ │ ├─ model/ Model module / Model 模块
│ │ └─ prisma/ Prisma service / Prisma 服务
│ └─ prisma/schema.prisma
├─ gin-server/ Cloud API / 云端 API
│ ├─ cmd/ API and migration entries / API 与迁移入口
│ ├─ internal/ Domain modules and infrastructure / 领域模块与基础设施
│ ├─ migrations/ PostgreSQL migrations / 数据库迁移
│ └─ compose.yaml Local cloud stack / 本地云端环境
├─ src-tauri/ Tauri shell / Tauri 壳
│ ├─ src/
│ │ ├─ lib.rs
│ │ ├─ main.rs
│ │ ├─ sidecar.rs
│ │ └─ tray.rs
│ └─ tauri.conf.json
├─ screenshots/ Screenshots / 截图
├─ PROJECT.md Detailed project docs / 详细项目文档
└─ README.md
Detailed technical notes are in PROJECT.md, including:
详细技术说明请看 PROJECT.md,里面包括:
- API routes / API 接口
- Data model / 数据模型
- Tauri sidecar flow / Tauri Sidecar 流程
- Skill pagination details / 技能库分页细节
- Development notes / 开发注意事项
- Gin cloud architecture and database foundation / Gin 云端架构与数据库基础
Cloud service setup and extension rules are documented in gin-server/README.md.
云端服务的启动方式与模块扩展规范见 gin-server/README.md。
The Skill page has already been optimized for large datasets:
Skill 页面已经针对大规模数据做了优化:
- Server-side pagination / 服务端分页
- Keyword search across name, description, and prompt / 按名称、描述、提示词搜索
- Type filtering / 类型筛选
- Sort controls / 排序控制
- Page-size limits to keep requests stable / 限制页大小,避免一次加载过多
This makes the library practical even when the number of skills grows to several thousand.
这让技能库即使增长到几千条,也依然可以稳定管理。
The app includes preset model providers for fast setup, such as:
应用内置了常见厂商预设,便于快速配置,例如:
- DeepSeek
- OpenAI
- Anthropic (Claude)
- Google (Gemini)
- Groq
- xAI (Grok)
- Mistral AI
- OpenRouter
You can also create custom model entries with your own base URL and API key.
你也可以使用自定义模型配置,填写自己的 Base URL 和 API Key。




