Miaotofu01/Study-Mate

StudyMate – Your AI‑Powered Study Companion

What it is – StudyMate is a Node‑JS‑based, AI‑augmented learning workflow aimed at math and computer‑science subjects. It ships as a plugin for DeepSeek Harness (DSH), as a Google Antigravity multi‑agent plugin, and can be packaged for OpenAI Codex / ChatGPT Work. The system orchestrates five roles – material collection, diagram generation, curriculum design, teaching, and exercise evaluation – either within DSH or by delegating to sub‑agents.

Key features

• Learning mode preset for DSH (install via npx @yunmiao/studymate@latest install).
• Curriculum roadmap: knowledge‑point dependency graph, colour‑coded status, click‑to‑expand lesson cards.
• Lesson pages: Markdown‑based “textbook‑style” lessons with formulas, images, custom questions and project tasks, rendered live by DSH’s reading pane.
• Persistent memory: progress, preferences, and pain‑points are saved in a local workspace (~/StudyMate) and reused across subjects.
• Export: one‑click offline HTML export (npx @yunmiao/studymate export).
• Multi‑agent support via Google Antigravity and optional Codex/ChatGPT plugins.

How to get started

1. Install the DSH preset:

   npx -y @yunmiao/studymate@latest install
   

   – Desktop DSH ships the engine; CLI users install @deepseek-ai/dsh separately.
2. Open DSH, start a new learning mode session, and say something like “I want to learn calculus”.
3. The system creates a workspace, builds a syllabus, and presents the first lesson.
4. Use commands such as export, build‑antigravity, or the Codex/ChatGPT import zip to extend functionality.

Typical workflow

• Choosing a topic – Ask the assistant for suggestions or specify a subject directly.
• Curriculum generation – The agent asks about goals, level, and desired projects, then produces a dependency‑ordered outline.
• Lesson consumption – Lessons appear in the DSH reading pane; you can ask for clarification, request alternative explanations, or request practice problems.
• Assessment – The assistant can generate quizzes, run code‑based checks, and record results in the workspace.
• Progress persistence – All interactions are stored, so the next session picks up exactly where you left off.

Who it’s for

• Self‑learners of mathematics, algorithms, programming languages, machine‑learning theory, etc.
• Users who prefer a “learn‑by‑doing” approach with interactive coding or project tasks.
• Anyone who wants a locally‑run AI tutor without needing Python (the engine is pure TypeScript).

Project status

• Actively maintained (issues encouraged, MIT‑licensed).
• Core code lives in lib/tools, with extensive documentation in the docs/ folder.
• A separate web runtime (study‑mate‑web) exists on a dedicated branch and may be merged back once compatible with the latest engine.

---

All information above is taken directly from the repository’s README.