Doc no. SK-404DatasheetRev 2026.11Sheet 2.22 of 4
Study Copilot
macOS menu-bar study helper: select a screen region or clipboard text and get a step-by-step explanation from a hosted model with your own API key, or an offline mock.
Active
Source on GitHub
$ ./run.sh
Block diagram
Characteristics
| UI shell | Native macOS menu-bar app built with rumps and PyObjC (app.py), keeping inference on background worker threads |
|---|---|
| Inputs | Captures interactive screen regions (screencapture -i), full screen, or current macOS clipboard text |
| Vision backends | Supports offline mock mode (no network/API key), Gemini (GEMINI_API_KEY) and Claude (ANTHROPIC_API_KEY) |
| Practice sandbox | Includes an isolated local test quiz server (http://127.0.0.1:8001) for verifying capture and explanation flows safely |
| Requires | macOS with uv (automatically provisions .venv) and Screen Recording permission for capture |
Overview & build notes
Sits quietly in the macOS menu bar so you can drag a selection box around a dense derivation or copy a paragraph to the clipboard and inspect a structured explanation without switching browser tabs.
explainer_engine.py decouples image/text prompting behind swappable backends — defaulting to Gemini vision when GEMINI_API_KEY is exported, falling back to Anthropic Claude or running completely offline via a deterministic mock backend.
A self-contained local quiz page (mock_practice_server.py on :8001) makes it easy to test OCR/vision capture end to end offline.
Files & artifacts
| Kind | File | Details | Size |
|---|---|---|---|
| Source | app.py | macOS menu-bar lifecycle, hotkeys and background worker loop on GitHub | GitHub |
| Source | explainer_engine.py | Prompt wrapper across offline mock, Gemini and Claude vision APIs on GitHub | GitHub |
| Source | mock_practice_server.py | Local 127.0.0.1:8001 practice sandbox server on GitHub | GitHub |
| Setup | run.sh | Auto-provisions .venv with uv and launches the menu bar app | GitHub |