An agent-native intelligent learning companion with multi-agent collaboration and RAG.
python3 -m venv .venv
source .venv/bin/activate
pip install -e ".[cli]"
deeptutor init --clideeptutor chat
deeptutor run deep_research "Transformer 最新进展" -l zh
deeptutor run meta_analysis "LLM tutoring effect" -l zhA prepackaged DeepTutor.app is built under dist/ by PyInstaller.
It is signed ad-hoc with the local PhotoToCodex identity, which avoids the
macOS "damaged" warning, but Gatekeeper will still show "unverified developer"
on first open; right-click the app → Open to bypass.
For distribution outside your own Mac, you need a Developer ID Application
certificate + notarization.
The bundled app includes the full Next.js frontend, the new cat-face app icon,
and the frozen-app launcher fixes so start --detach works from the .app.
On first launch the app seeds its runtime home
(~/Library/Application Support/DeepTutor) with the LLM pool config from
packaging/default-settings/, so it comes up with a working provider
instead of an empty catalog. Only placeholder keys ship in the bundle;
real credentials stay in your local provider service.
Apache-2.0. See LICENSE.