让每个孩子拥有平等的学习机会 · Equal Learning Opportunities for Every Child
中文 | English
EduCore（光合啟途）是一个专为乡村教育设计的 AI 驱动自适应学习平台。
我们相信：教育公平，不是给每个孩子一样的东西，而是给每个孩子最适合他们的东西。
EduCore is an AI-powered adaptive learning platform designed specifically for rural education in China.
We believe: Educational equity means giving every child what they need, not the same thing for everyone.
| 🔬 科学自适应 | Science-based Adaptive Learning |
|---|---|
| 基于 BKT / IRT / SM-2 算法，动态调整题目难度和出现频率 | Dynamically adjusts question difficulty based on BKT/IRT/SM-2 algorithms |
| ❤️ 温暖不评判 | Warm & Non-judgmental |
|---|---|
| 永远不说"你错了"，只说"让我们一起再想想看" — 零压力学习环境 | Never says "you're wrong," only "let's think together" — zero-pressure learning |
| 📶 离线优先 | Offline-first |
|---|---|
| PWA + 本地大模型，即使没有互联网也能完整使用 | Works fully offline with PWA + on-device LLMs |
| 🏫 全员协同 | Everyone together |
|---|---|
| 学生 ↔ 家长 ↔ 老师 ↔ 志愿者，一个都不能少 | Students ↔ Parents ↔ Teachers ↔ Volunteers, all connected |
| 痛点 | Pain Point | 我们的解法 | Our Solution |
|---|---|---|---|
| 师资匮乏 | One teacher covers 3-5 grades | AI 老师 24 小时在线，个性化辅导每个孩子 | AI teacher 24/7, personalized for every student |
| 资源贫瘠 | No internet, no textbooks | 树莓派 + 本地大模型，完全离线运行 | Raspberry Pi + local LLM, fully offline |
| 信息断层 | Migrant parents disconnected | 微信小程序推送每周学习报告 | Weekly reports via WeChat Mini Program |
本地 Ollama (树莓派) → 云端 Ark API → 规则引擎
Local Ollama (Pi) → Cloud Ark API → Rule Engine
- 有网时：云端大模型提供高质量讲解
- 弱网时：自动切换到本地小模型（Qwen 2.5 1.8B）
- 断网时：规则引擎兜底，保证学习不中断
| 层级 | Layer | 技术 | Technology |
|---|---|---|---|
| 后端 | Backend | Fastify + TypeScript + MongoDB + Redis | |
| 前端 | Frontend | React 19 + Vite + TailwindCSS 4 + Zustand | |
| 算法 | Algorithms | TypeScript (BKT, IRT, SM-2) | |
| AI 模型 | AI Models | Qwen 2.5, Ollama, Ark API | |
| 单体仓库 | Monorepo | Turborepo + pnpm workspaces |
| 阶段 | Phase | 状态 | Status | 完成度 | Progress |
|---|---|---|---|---|---|
| Phase 0 · 核心引擎 | Core Engine | ✅ 完成 | Done | 100% | |
| Phase 1 · PWA + 多语言 | PWA + i18n | ✅ 完成 | Done | 100% | |
| Phase 2 · 多科目 + 捐赠 | Multi-subject + Donation | ✅ 完成 | Done | 100% | |
| Phase 3 · 离线终端 | Offline Terminal | 🚧 进行中 | In Progress | ~60% | |
| Phase 4 · 家校协同 | Family-School Sync | ⏳ 待启动 | Planned | 0% |
当前里程碑：模型抽象层 + 本地大模型集成完成 ✅ Current Milestone: Model abstraction layer + local LLM integration complete ✅
- 光合 (Light Synthesis) → 像阳光一样，公平地洒在每个孩子身上
- 啟途 (Path Opening) → 开启一条通向未来的道路
我们不做"慈善"，我们做"赋能"。 我们不施舍，我们建设基础设施。
We don't do "charity" — we build infrastructure.
| 方式 | Way | 你可以做的 | What you can do |
|---|---|---|---|
| 💻 代码贡献 | Code | 前端 / 后端 / AI 算法 / 科目内容 | Frontend / Backend / AI Algorithms / Content |
| 📚 内容创作 | Content | 编写练习题、知识点讲解 | Create exercises and explanations |
| 💰 设备捐赠 | Donation | 赞助一套树莓派基站 | Sponsor a Raspberry Pi station |
| 🗣️ 传播推广 | Outreach | 告诉更多有需要的乡村学校 | Tell rural schools that need it |
| 🧪 测试反馈 | Testing | 帮我们找到 Bug，优化体验 | Find bugs, help improve UX |
See CONTRIBUTING.md for details.
| 文档 | Document | 说明 | Description |
|---|---|---|---|
| ARCHITECTURE.md | 架构文档 | 系统设计、技术选型、模块划分 | Architecture, tech stack, modules |
| ROADMAP.md | 开发路线图 | Phase 0-5 完整计划 | Phase 0-5 full plan |
| CONTRIBUTING.md | 贡献指南 | 如何参与项目 | How to contribute |
| CODE_OF_CONDUCT.md | 行为准则 | 社区参与规则 | Community rules |
| docs/ | 全部文档 | All documentation |
这不是一个商业项目，这是一个关于可能性的实验。
我们想证明：技术和善意结合，真的可以改变一些事情。
This is not a commercial project. This is an experiment in possibility.
We want to prove: Technology + Goodwill really can change things.
🌱 让每个孩子，都能站在同一条起跑线上。
🌱 May every child stand on the same starting line.
Equal Learning Opportunities for Every Child
EduCore is an open-source, AI-powered adaptive learning platform designed specifically for rural education in China. It combines cutting-edge edge AI (Raspberry Pi + local LLMs) with proven learning science (BKT/IRT/SM-2 algorithms) to deliver personalized education to students who need it most.
- Offline-first: Works completely without internet using local LLMs on Raspberry Pi
- Adaptive learning: Dynamically adjusts difficulty based on student performance
- Warm, non-judgmental: Designed to encourage, not criticize
- Multi-stakeholder: Connects students, parents, teachers, and volunteers
- Open source: 100% open source, community-driven
- Backend: Fastify + TypeScript + MongoDB + Redis
- Frontend: React 19 + Vite + TailwindCSS 4 + Zustand
- AI: Qwen 2.5 family + Ollama + Model Provider abstraction layer
- Hardware: Raspberry Pi 5 / RK3566 with local LLM inference
This project is licensed under the GNU Affero General Public License v3.0 — see the LICENSE file for details.
Why AGPLv3? To ensure any improvements to this project remain open and accessible to everyone, even when run as a cloud service.
Thank you for your interest in making education more equitable for every child.