Skip to main content

Zhiheng AI Teaching Assistant

AI Teaching Assistant for Morality & Rule of Law (Grades 7–9)


Overview​

Zhiheng is an AI-powered teaching assistant platform for middle school Morality & Rule of Law education, designed for border schools in Guangxi, China.

The system provides:

⚠️ Status: These are planned features. The current system is at Phase 1 (scaffolding + infrastructure). See Roadmap for implementation status.

  • AI lesson plan generation — RAG-driven from digitized textbooks and school knowledge base
  • AI PPT generation — auto-create slides from lesson plans
  • Voice input & AI completion — minimal-input workflow with smart autocomplete
  • Mind map & blackboard design — auto-generate knowledge graphs and handwritten-style blackboard layouts
  • Exam point mapping — auto-connect knowledge points to curriculum standards and exam syllabi
  • Material archiving — export Word/PDF/ZIP with version control
  • Knowledge base — digitized 2026 edition textbooks, chunked and semantically tagged, with knowledge graph and RAG engine
  • Content audit — full logging, compliance tracking, and role-based permissions
  • Current affairs push — curated political and news content relevant to each lesson topic

Target Users​

Approximately 37 border schools in Guangxi (China–Vietnam border region), covering grades 7–9 with around 350 class hours of Morality & Rule of Law curriculum.

Tech Stack​

  • Backend: Python (FastAPI)
  • Frontend: Vue 3.5 + Vite (Docusaurus for docs)
  • LLM: External vLLM endpoint (Qwen3.6-35B-A3B-FP8)
  • RAG: LightRAG (PostgreSQL + pgvector + AGE graph), KohakuRAG (local)
  • Embedding: Qwen3-Embedding-8B via OpenRouter (qwen/qwen3-embedding-8b)
  • Chat Agent: DeerFlow (SSE streaming, PPT generation)
  • Auth: JWT + bcrypt, role-based access control + Cloudflare Worker session cookie auth
  • Cache: Redis 7 (rate limiting, chat threads, token store)
  • Design System: DESIGN.md (Mintlify-inspired, Tailwind exported)

Project Structure​

zhiheng/
├── backend/ # FastAPI backend
│ ├── models.py # SQLAlchemy models (User, Lesson)
│ ├── routers/ # auth.py, lessons.py
│ └── ...
├── frontend/ # Vue 3.5 + Vite frontend
│ ├── src/pages/ # Dashboard, Editor, Materials, Login, Register
│ ├── src/components/ # Sidebar, TopBar, Layout, LessonForm
│ ├── src/locales/ # zh.json, en.json (react-i18next)
│ └── ...
├── nginx/ # Reverse proxy + Basic Auth
├── rag/ # RAG pipeline (scaffold)
├── docs/ # This docs site (Docusaurus) + project docs
├── DESIGN.md # Design system specification
└── DESIGN.tailwind.json # Tailwind token export

Team​

TeamLead
LLM & Agent DevChi Yupeng, Tang Ba Minh
Curriculum & ContentYe Haibo, Wang Dong, Huang Linyi, Zheng Siwen
AIGC Micro-DramasYang Yixin, Zeng Rou
VR/ARShenzhen Huiying

Quick Start​

Backend​

cd backend
pip install -r requirements.txt
uvicorn main:app --reload

Frontend​

cd frontend
npm install
npm run dev

Docs Site (this site)​

cd docs
npm install
npm run start