DeepTutor Learning Workspace

Core facts

Read by a script from the official API, not rewritten by a model.

Repository: https://github.com/HKUDS/DeepTutor
License: Apache-2.0
Languages: Python
Maintainer: HKUDS
Homepage: http://arxiv.org/abs/2604.26962
Verified: 2026-10-31

What it is

DeepTutor is an agent-native learning workspace that connects tutoring, problem solving, quiz generation, research, visualization, and mastery practice in one extensible system. Chat, Ask Questions, Quiz, Research, Visualize, Solve, Mastery Path, Immersive Reading, and Audio Overview share one capability runtime and session context while keeping purpose-built loops and pipelines. The project is licensed under the Apache License 2.0.

Main features

Task Board tracks study tasks in To do, In progress, and Done, with notes, drag-and-drop or keyboard-accessible move buttons, and an archive you can restore from.
Adding YouTube or Bilibili links to Reading gives native playback, transcript search, timestamp-grounded tutoring, saved notes, and resumable progress.
Knowledge is offered as versioned RAG libraries across LlamaIndex, PageIndex, GraphRAG, LightRAG, a remote LightRAG Server, a self-hosted WeKnora knowledge base, a Tencent IMA or MarginNote 4 library, a connected Kiwix ZIM archive, or a linked Obsidian vault, with pluggable document parsing.
Extensions include built-in tools, MCP servers, CLI apps, image / video / voice generation models, and installable community skills from EduHub.
In DeepTutor's personalization memory, L1 traces, L2 surface summaries, and L3 synthesis make personalization visible and editable.

Getting started

DeepTutor ships four installation paths.
Option 1, the PyPI installation of the full local Web app and CLI, needs Python 3.11–3.14 and a Node.js 20+ runtime on PATH, because the packaged Next.js standalone server is spawned by deeptutor start.
Skipping init is fine for a quick trial; the app boots with default ports and empty model settings, which can be configured later in Settings → Providers and Language models.

From the README · bash

docker run --rm --name deeptutor \
  -p 127.0.0.1:3782:3782 \
  -v deeptutor-data:/app/data \
  ghcr.io/hkuds/deeptutor:latest

From the README · bash

mkdir -p my-deeptutor && cd my-deeptutor
pip install -U "deeptutor[full]"
deeptutor init     # prompts for ports + LLM provider + optional embedding/search
deeptutor start    # starts backend + frontend; keep the terminal open

From the README · bash

git clone --branch dev https://github.com/HKUDS/DeepTutor.git
cd DeepTutor

# Create a venv (macOS/Linux). Windows PowerShell:
#   py -3.11 -m venv .venv ; .\.venv\Scripts\Activate.ps1
python3 -m venv .venv && source .venv/bin/activate
python -m pip install --upgrade pip

# Install backend + frontend deps
python -m pip install -r requirements/full.txt
python -m pip install -e . --no-deps
( cd web && npm ci --legacy-peer-deps )

deeptutor init
deeptutor start --dev

Who it is for

Authentication is off by default, and DeepTutor runs single-user. Option 4 is a CLI-only setup from a source checkout: when the Web UI is not needed, use the CLI from that checkout and skip Node/npm. Partners are persistent companions with their own soul, model policy, library, memory, and channels.

Partners are not a separate bot engine. Memory is a file-backed, three-layer system you can read, curate, and audit, deliberately not a hidden vector store. Each knowledge base is bound to one engine.

With the json format option added to a run command, each turn streams NDJSON, one event per line, with every line tagged with its session id. The DeepTutor team states that, so far, it has no paid online products of any form.

Related tools listed on this site.