← Back to archive

Repo of the Day

GitHub - HKUDS/DeepTutor: "DeepTutor: Agent-Native Personalized Learning Assistant"

Published: Jul 21, 2026

Open repository ↗

"DeepTutor: Agent-Native Personalized Learning Assistant" - HKUDS/DeepTutor

Summary

DeepTutor (HKUDS/DeepTutor) is an Apache-2.0, Python 3.11+ learning platform that pairs a Next.js 16 frontend with a multi-agent backend for AI-assisted study and research. The repository ships a CLI, an interactive web UI, and plug-in integrations for retrieval, code assistants, and chat platforms.

What it is useful for

  • RAG over personal knowledge bases using selectable engines: LlamaIndex, LightRAG, GraphRAG, and PageIndex (the v1.4.6 release notes call out GraphRAG / PageIndex / LightRAG / linked-KB / Obsidian support).
  • Document-grounded study: "Immersive Reading" opens a PDF/EPUB beside the chat with page-anchored citations (v1.5.14), and a Book Engine streams a "living book" from source material (v1.2.0).
  • Deep research workflows that combine the configured web-search providers — including Doubao, Bocha, Zhipu, Firecrawl, Qianfan, Aliyun IQS, and Serply (v1.5.12) — with retrieval over local files.
  • Multi-agent "Partners" that can be wired to local coding CLIs (Claude Code, Codex, Gemini, Kimi, opencode, MiMo) and given private memory and skills (v1.4.7, v1.5.3).
  • Question banks, quizzes, and guided learning sessions that record mastery state per learner.
  • Optional multi-user deployments with isolated user workspaces, admin grants, and auth (v1.3.8), plus IM integrations on Discord, Feishu, WeChat, Mattermost, Zulip, and Matrix.

How engineers can use it

  1. Confirm Python 3.11+ and clone HKUDS/DeepTutor; the README badges mark Python 3.11+ and Next.js 16 as the runtime stack.
  2. Follow the "Get Started" anchor in the README for environment prep — the supplied excerpt is truncated and the full walkthrough is not present here, so consult the upstream docs at deeptutor.info for any steps not visible in this excerpt.
  3. Launch the app from the command line with deeptutor start (documented in v1.3.9). Run deeptutor doctor to surface environment issues (v1.5.17).
  4. Install community skills with deeptutor skill install from ClawHub (v1.4.4), then point the tutor at a folder to use it as a content workspace (v1.6.5).
  5. Wire a local coding agent under /agents (v1.4.7), and add a knowledge base through the Knowledge Center to enable RAG-backed chat.
  6. Reference the arXiv paper 2604.26962 for the architectural details behind the v1.4.0 rewrite.

Documented limitations: the supplied README excerpt is truncated and the "Get Started" / installation section is not fully present here; full setup commands, Docker deployment, model-provider configuration, and MCP wiring should be verified against the current README and the docs at deeptutor.info before running. The current release as of the excerpt is v1.6.13.