DeepTutor
HKUDS/DeepTutorReviewed by AI
What it does
DeepTutor is a learning workspace you run on your own computer. You add documents, videos or web sources, and the tutor answers with citations to them, builds quizzes and practice, writes study books and narrated video lessons, keeps a notebook and tracks what you have learned. It brings no AI model of its own: you connect a provider key or a local model such as Ollama or LM Studio. There are preview installers for Windows and macOS that bundle Python and Node.js, plus Docker and pip installs for people comfortable with a terminal.
How easy it is to install
B · installer plus one simple step · Windows
- Windows B
- macOS B
- Linux D
Windows and macOS are a B: release v1.7.1 has a Setup .exe, a portable ZIP and a .dmg, and the README says the desktop build 'bundles Python, Node.js, and the full learning workspace, so no terminal or separate runtime installation is needed' — you then pick a provider and model in Settings. Linux is a D, because there is no desktop build and the documented routes are Docker or pip with commands such as 'deeptutor init' and 'deeptutor start'.
Numbers on GitHub
- Stars
- 52,820on Oct 31, 2026
- Downloads
- 204204 this week
- Growth
- +8,939stars in 5 days
Stars over time
Facts
- Runs on
- Windows, macOS, Linux
- You need to install
- Python, Docker
- To build from source
- Python, Docker
- Disk space
- 461 MB–1.3 GB · from the installer
- Accounts or keys
- A language model provider of your choice, set in Settings before use, OpenAI
- Cost
- Free, API costs extra · The software is free (Apache-2.0). You pay whichever model provider you connect, unless you point it at a local model.
- License
- Apache-2.0 (Permissive)
- Can replace
- NotebookLM, Quizlet, ChatGPT Plus
- Last update
- Oct 31, 2026
Risk signals (5)
- Attentiondesktop builds are previews and unsigned
README 'Desktop — macOS and Windows (Preview)': 'Desktop previews are available on dev; the v1.7.0 Release does not include desktop installers', with downloads taken from CI artifacts that 'expire after seven days', and 'Windows preview builds are unsigned and may show a SmartScreen warning.'
AI review
- Attentionai can run commands on your computer
The chat tool list includes 'exec', and the Dockerfile says 'the restricted-subprocess backend (which the office skills... rely on for exec) is enabled by default via the sandbox_allow_subprocess runtime setting (system.json, default on)'. docker-compose.yml adds that on 'bare-metal / macOS deployments where the runner is not started' the app 'degrades to bwrap (Linux, if installed) or... a restricted subprocess', so how well commands are boxed in depends on how you installed it.
AI review
- Attentioninstalls third party code while running
Dockerfile note: 'git is required to install CLI apps — most of the CLI-Anything catalog installs with pip install git+…'. The release notes list '101 CLI Apps the tutor can run' and 'Install community skills from ClawHub with deeptutor skill install behind a security gate'. Compose files also re-apply DEEPTUTOR_EXTRAS and DEEPTUTOR_APT_PACKAGES on every start.
AI review
- Attentiondocker files publish ports to the network
compose.yaml (podman) binds '127.0.0.1:${HOST_PORT_BACKEND:-8001}:8001', but docker-compose.yml and docker-compose.ghcr.yml publish '${DEEPTUTOR_DOCKER_BACKEND_PORT:-8001}:...' with no 127.0.0.1 prefix, so on those files the app answers on every network interface of the machine.
AI review
- Infokeys and accounts kept in the data folder
docker-compose.yml comments: the data tree holds 'accounts/grants/audit (data/system)' and 'never data/system, which holds auth state, provider API keys, and per-owner secrets' is called out as the part the sandbox must not see. Everything you configure lives in that folder on your disk.
AI review
Evidence is shown in its original language.
What Radar could not check
- Whether the preview desktop builds work and match the source in this repository
- What the model calls cost once you use books, quizzes and research for real
- Real disk use after indexing your own documents, on top of the 461 MB installer
- Whether the restricted sandbox keeps a generated command away from the rest of your files
- Whether the Anki, calendar and Word exports come out as described