TeacherPro is a desktop app for planning lessons. It combines a lesson-plan editor, a weekly calendar, note-taking, a method bank and mindmaps in one window, and keeps everything in a folder you choose on your own disk — no account, no cloud, no subscription.
It's built with Tauri (Rust) and React, and the interface is available in English and German. I made it for my own teaching work and use it daily; it's shared here in case it's useful to someone else.
Everything lives in a vault — a normal folder you pick on first launch.
Lessons and mindmaps are JSON files, materials stay where you put them, and the
trash is just a Trash folder inside the vault. Because it's all plain files,
you can back the vault up, sync it with Dropbox or put it on a USB stick.
Avoid editing the same vault from two machines at the same time: the app has
no merge or conflict detection, and simultaneous autosaves can overwrite each
other.
Lessons. A rich-text editor with headings, lists, checklists, quotes,
links, tables, colours and highlights. There are a dozen bundled fonts so
documents look the same on every machine. Lesson tables are set up for the
typical pedagogy columns (time, phase, activity, social form, media) and can
grow rows or columns as needed. Lessons autosave; Ctrl/Cmd+S saves on demand.
Notes. Every lesson has a notes drawer that saves with it. Notes stay out of PDF export unless you add them to the lesson itself.
Method bank. Comes with a built-in set of phases, social forms and
methods, searchable and insertable into lesson tables from a side panel or via
/ inside a table cell.
Mindmaps. A node-based canvas for mapping out a unit or a topic. Nodes can be colour-coded, renamed, connected, and materials attached from the sidebar.
Calendar. A week view of all planned lessons. Click a card to open the lesson, drag it to another day to reschedule it.
Search and trash. The sidebar search field filters lessons and mindmaps by name and folder. Materials and trash are browsed separately. Deleted items go to the vault's trash where they can be restored or deleted permanently.
Printing and PDF. Lessons and mindmaps print through the native print
dialog and export as PDF into an Exports folder. The exported page matches
what's on screen, including table borders and colours.
Shortcuts. Ctrl/Cmd+S save, Esc close panels, Ctrl/Cmd+B/I/U basic
formatting, Ctrl/Cmd+E/L/R/J alignment, Ctrl/Cmd+K link, Ctrl/Cmd+P
print, Ctrl/Cmd+Shift+V paste as plain text.
AI is optional and runs entirely on your machine. TeacherPro starts its own
llama-server process when needed and shuts it down with the app — there is no
Ollama install, no runtime to set up and nothing sent to the internet.
- Models are downloaded in Settings → AI Features as GGUF files from Hugging Face and verified with SHA-256 when downloaded. The catalog runs from a 2B model for older laptops up to 31B, plus a 35B-A3B mixture-of-experts model for machines with a large GPU. A hardware probe recommends a model for the detected GPU and memory.
- Chat streams token by token, supports thinking models, and can be stopped mid-answer. The rewrite/translate actions in the editor use the same engine.
- Knowledge bases. You can import PDFs into subject databases. They're chunked and embedded locally (IBM Granite Embedding 311M Multilingual R2, CPU only) and stored in a single SQLite index per vault. Search combines vector similarity with keyword search, and answers cite the source file and page number; you can scope the search to a subject, class, topic or a single PDF.
- Lesson drafts. The sparkle button in the chat panel can generate a full lesson table from a topic, duration and the selected knowledge-base content.
- CUDA (optional). On NVIDIA machines a separate acceleration pack can be installed from Settings to run the chat model on the GPU. Without it the app uses the bundled Vulkan build; either way it falls back cleanly if something fails.
- Benchmark. Settings can measure prompt and decode speed for the selected model on demand; it runs only when clicked.
AI settings live in Settings → AI Features. Nothing AI-related is enabled by default, and disabling it leaves the rest of the app untouched.
Ready-to-run installers for Windows, macOS and Linux are on the
Releases page. On Linux both
.deb and .rpm packages plus an AppImage are provided.
The macOS build is not signed with a paid Apple Developer certificate, so Gatekeeper sometimes refuses to open it. If that happens:
-
Move
TeacherPro.appfrom the.dmginto/Applications. -
Open Terminal and run:
xattr -cr /Applications/TeacherPro.app
-
Open the app normally.
Without signing there isn't a nicer workaround; the command only removes the "downloaded from the internet" flag.
You need Node.js 20.19+ and a recent stable Rust toolchain. On Linux you also need the usual Tauri system packages (webkit2gtk, gtk3, etc.) — see the Tauri prerequisites.
git clone https://github.com/Panolix/TeacherPro.git
cd TeacherPro
npm install
# Fetches the pinned llama-server build for your platform into
# src-tauri/resources/engines/ (required before building; not committed)
npm run fetch:engines
npm run tauri dev # development with hot reload
npm run tauri build # installer in src-tauri/target/release/bundle/The engine pin lives in scripts/engine-config.json. To bump it, change the
pin there and regenerate the checksum with
node scripts/fetch-engines.mjs --print-hashes <asset>.
Releases are tag-driven. Write the notes in
release-notes/RELEASE_NOTES_vX.Y.Z.md, then push a matching tag:
git tag v2.10.1
git push origin v2.10.1The Release workflow builds the Windows, macOS and Linux installers and attaches them to a draft release. The "Build Installers" workflow can rebuild an existing version from the Actions tab.
- docs/FEATURE_TRACKER.md — implementation notes for maintainers
- release-notes/ — notes for every release
Model weights are not bundled with the app. They're downloaded at runtime from their official Hugging Face repositories and stay under their respective licenses.
- Chat and draft models: various GGUF builds (Qwen, Gemma, Ministral, …); each model's license is listed in the app's model catalog.
- Embedding: IBM Granite Embedding 311M Multilingual R2 (Apache 2.0). Its tokenizer derives from the Gemma 3 tokenizer and is subject to the Gemma Terms of Use.
- Engine: llama.cpp (MIT).
See THIRD_PARTY_LICENSES.md for the details.
TeacherPro is licensed under CC BY-NC-SA 4.0.
License text: human-readable summary · full legal code · SPDX entry · LICENSE
Source-available, not open source. You may use, fork, modify and share it for non-commercial purposes, provided you credit Panagiotis Smponias, link back to this repository, state what you changed, and license your version under the same terms.
Commercial use requires a paid license — see COMMERCIAL.md. Attribution, share-alike and trademark terms: NOTICE.



