KevinLiss/ApeAdmin

ApeAdmin – Open‑source admin platform built for modern AI applications

What it is – ApeAdmin is a full‑stack, plug‑in‑enabled admin framework designed to host AI‑driven services. It combines a FastAPI back‑end with a Vue 3 front‑end and adds a MCP‑SSE gateway that exposes management functions (RBAC, plugins, logs, etc.) as tools that AI agents can call via function‑calling. In short, it gives you a ready‑made “backend‑as‑a‑service” that can be extended with independent Python plugins and accessed both by human operators and by LLM agents.

---

Core capabilities

Category	What you get
Installation wizard	WordPress‑style three‑step setup (DB config → site & admin → finish). Auto‑creates MySQL databases, generates JWT secrets, and writes a lock file to prevent re‑runs.
RBAC	Five‑table model (users, roles, menus, departments, association). Four‑layer permission checks (no‑login → no‑auth → rule → data scope) and front‑end v‑permission directive for button‑level hiding.
Plugin system	Python‑package plugins discovered via importlib. Supports hot‑load/unload, Docker‑isolated deployment, ZIP upload, and an online plugin marketplace. Plugins can register routes, MCP tools, and event listeners.
MCP‑SSE gateway	Translates internal management APIs into Tools / Resources / Prompts that LLM agents can invoke. JSON‑Schema is auto‑derived from function signatures, RBAC‑filtered, and secured with one‑time tickets and 30 s timeout. All calls are audited.
AI chat assistant	Multi‑model support (OpenAI, DeepSeek, Alibaba Tongyi, Zhipu GLM, or any OpenAI‑compatible endpoint). Streaming SSE output, markdown rendering, and up to five rounds of tool‑calling loops. Model keys are stored encrypted with Fernet.
Audit & monitoring	Request‑ID middleware, operation logs, per‑API latency tracking, and searchable audit logs.
Database flexibility	SQLite for quick local dev, MySQL for production; automatic driver switch via DB_TYPE. Optional Redis cache with graceful fallback to in‑memory store.

---

Architecture at a glance

apeadmin/
  backend/          # FastAPI (async SQLAlchemy, Alembic migrations)
    src/
      api/          # REST endpoints (auth, users, roles, plugins, AI chat…)
      core/         # config, security (JWT + bcrypt), middleware, seeds
      crud/         # generic CRUD with RBAC hooks
      db/           # engine & session handling
      models/       # ORM definitions (rbac, plugins, AI, MCP logs)
      mcp/          # MCP manager, routes, built‑in tools
      plugins/      # plugin interface, manager, builtin examples
      ai/           # agent orchestration, function‑calling loop
      schemas/      # Pydantic request/response models
      main.py       # app entry point (lifespan, setup mode)
  frontend/         # Vue 3 + Vite + TypeScript + Element Plus UI
    src/
      api/          # Axios wrappers for all back‑end routes
      components/   # header, sidebar, etc.
      directives/   # `v-permission`
      router/       # dynamic routes generated from back‑end menu tree
      stores/       # Pinia stores for user, menu, permissions
      views/        # pages (login, system admin, plugin market, AI chat…)

The design principle is “install‑and‑run” – an un‑configured instance redirects to /setup, writes a .env file, creates tables, and then behaves as a normal multi‑tenant admin system.

---

Who might use it?

Scenario	Why ApeAdmin helps
AI SaaS platform	Expose internal tools (data retrieval, user management, custom business logic) to LLM agents without writing separate APIs.
Enterprise internal tools	Quickly spin up a permission‑aware admin UI for managing users, departments, and custom plugins while keeping a single source of truth.
Plugin ecosystem	Teams can develop independent Python plugins (e.g., data connectors, model wrappers) that are discovered and hot‑loaded, keeping the core stable.
Rapid prototyping	Use the SQLite dev mode and the three‑step wizard to get a functional admin panel running in minutes.

---

Getting started (quick‑start)

1. Back‑end

   cd backend
   python -m venv .venv && source .venv/bin/activate   # Windows: .venv\Scripts\activate
   pip install -e .
   uvicorn src.main:app --reload --host 0.0.0.0 --port 8000
   

2. Front‑end

   cd frontend
   npm install --legacy-peer-deps
   npm run dev   # serves at http://localhost:5173
   

3. Open the URL; the app will detect the missing configuration and launch the setup wizard where you can pick SQLite (zero‑config) or point it at a MySQL instance.
4. After installation, log in with the admin credentials you defined, then explore:
   • System → Users / Roles / Menus / Departments
   • Plugins → Marketplace, upload ZIP, enable/disable
   • MCP → Tools / Resources / Prompt templates
   • AI → Add model API keys and start a chat session that can call your registered tools.

For production you can follow the Docker/Nginx scripts in deploy/DEPLOY.md.

---

Technology stack

Layer	Tech
Back‑end	FastAPI, async SQLAlchemy 2.0, Alembic, Pydantic‑Settings, JWT, bcrypt
Front‑end	Vue 3.5, Vite 6, TypeScript 5.7, Element Plus 2.9, Pinia, ECharts
Database	MySQL (via aiomysql) or SQLite (aiosqlite)
Cache	Redis (optional)
AI integration	MCP (Model Context Protocol) over Server‑Sent Events, supports OpenAI‑compatible endpoints
Packaging	MIT license, pip‑editable install for the back‑end, npm for the front‑end

---

License & community

• License: MIT – free for commercial and non‑commercial use.
• Community: Official website, plugin marketplace, documentation portal, and a WeChat user group (QR code in the repo). Contributions follow a standard fork‑branch‑PR workflow with clear branch‑naming conventions.

---

TL;DR

ApeAdmin is a plug‑in‑first, AI‑aware admin framework that lets you manage users, permissions, and custom business logic while automatically exposing those capabilities as callable tools for LLM agents. It ships with a ready UI, an installation wizard, and a full plugin ecosystem, making it a solid foundation for any modern AI‑centric product.