GitHub radar

DeepTutor: Open-Source AI Learning Workspace

An AI-native learning platform that turns your PDFs and YouTube videos into personalized courses with quizzes and a tutor that remembers your progress — runs locally with Docker.

01HKUDS/DeepTutor 40kPython

DeepTutor is an open-source AI learning workspace created by the HKUDS group at Hong Kong University, started in December 2025. Upload your PDFs, books, or YouTube links — the system builds an interactive course with quizzes, flashcards, and progress tracking, tutored by an AI that remembers what you already understood across sessions (three-layer memory: traces, summaries, synthesis). You can connect live AI agents — Claude Code, Codex, DeepSeek, Kimi, and others — and consult them directly within a chat turn. Supports multiple RAG engines (LlamaIndex, PageIndex, GraphRAG, LightRAG), a Markdown co-writer for notes, Obsidian vault import, and immersive YouTube playback with timestamp-grounded tutoring. Runs via a single Docker container (ghcr.io/hkuds/deeptutor:latest), or via pip install deeptutor. Actively maintained — v1.6.8 released September 14, 2026.

Why a vibe-coder should care

Vibe coders often need to pick up a new technology or domain fast — not by reading through docs page by page, but by asking questions and getting tested. DeepTutor turns your own materials into a structured course with a tutor that tracks where you are. One Docker command, no infrastructure required, Claude Code can be wired in as a study partner.

How to install

Copy this and send it to your agent — Claude Code, Codex, any of them:

Run DeepTutor from https://github.com/HKUDS/DeepTutor with: docker run --rm -p 3782:3782 ghcr.io/hkuds/deeptutor:latest — open http://localhost:3782 and help me connect an AI model. Ask me for keys if needed.

Runs on any laptop — only Docker required.

Open on GitHub