Skip to content
GitHub · 2.7.0

Get started

Docs screenshots

The docs site uses PNGs from this folder. Capture widgets and examples with the script below, or capture a full window interactively.

Widgets - automated via tools/widget_screenshot.py

For each registry entry, the script builds a minimal App with the widget or Hello EMG pair. It waits about 3 s, or 7 s for signal viewers, then captures the window via Quartz + screencapture, saves the PNG and exits. This capture script requires macOS, screen-capture permission and the dev extra for Quartz. Re-run after changes to widget styling, themes, fonts or ImGui, and when adding widgets.

Shell
uv run --extra dev python tools/widget_screenshot.py --all
# or one at a time
uv run --extra dev python tools/widget_screenshot.py signal_viewer
uv run --extra dev python tools/widget_screenshot.py hello_emg --out docs/images/hello-emg.png
# Check every isolated panel and the three start_here apps (local review captures).
uv run --extra dev python tools/widget_screenshot.py --examples
# Or run one unchanged example with the same timed capture helper.
uv run --extra dev python tools/widget_screenshot.py --example examples/start_here/myocontrol.py --out .superpowers/screens/app-font-myocontrol.png
# Reproduce the classification overview without a generator or hand process.
uv run --extra dev python tools/widget_screenshot.py --overview

The registry covers

File Shows
widgets/hello_emg.png signal + recording controls (use --out above for the homepage)
widgets/AppHeader.png Connect / Record / Train / Predict, Record selected
widgets/StatusBar.png stream, participant and data/P-014 folder
widgets/SimulatorControls.png persistent Rest / Fist preview
widgets/FollowRun.png guided recording controls
widgets/signal_viewer.png 4-channel synthetic signal
widgets/recording_controls.png 4-class button strip
widgets/session_manager.png 3 selected mock takes for P-014, with Rest / Fist durations
widgets/FilterControl.png one-euro defaults
widgets/process_launcher.png 2 dummy sleep 60 processes
widgets/TargetPanel.png connected, running and idle targets
widgets/pipeline_panel.png idle Pipeline
widgets/FeatureSelector.png RMS / MAV / WL / VAR / ZC
widgets/app_logo.png shipped wordmark, centred
widgets/prediction_label.png “Fist” @ 82%
widgets/VhiMovementPanel.png 12 movements, “Fist” highlighted

The session fixture creates three takes in P-014/recordings/, calls SessionManager.select_all("P-014"), then renders that participant. No Rescan click is needed. The newest day opens with all three takes visible.

Examples

--overview runs the classification tutorial with an in-process synthetic EMG source in a temporary data directory. It records Rest and Fist takes, trains the tutorial’s CatBoost model, starts prediction and captures gui-overview.png. The screenshot uses the real widgets and fitted model; no hardware is needed.

Fullscreen, manual

For the classification tutorial screenshot (gui-overview.png) and any future per-example shots, run the example fullscreen and capture the whole window. App.run(fullscreen=True) maximizes to the monitor work area, keeping the menu bar / dock visible.

Python
# Add temporarily to the example's main() or pass through __main__:
app.run(fullscreen=True)

Then start the example, launch EMG Generator in Connect, connect TestEMG1, and optionally connect the Virtual Hand. Run a Record → Train → Predict cycle so the signal viewer and prediction card show non-empty state, and capture:

Shell
# Interactive: click the window you want to capture
screencapture -i -W docs/images/gui-overview.png

For scripted macOS capture, obtain the Quartz window ID as _self_capture_after does in tools/widget_screenshot.py; a process ID is not a window ID.

On Linux, capture one window with gnome-screenshot -w or use ImageMagick import.

Style notes

  • Crop tightly for full-width display on phones.
  • PNG, < 1 MB each. Compress with pngquant --quality=70-90 *.png.
  • 16:10 is the safest aspect; the demo looks good at 1280×800. Fullscreen captures use native monitor resolution. Downsize Retina captures with sips -Z 1920 *.png before committing.
  • Use light or dark themes, whichever looks clearer in each screenshot. The docs support both. Mixing themes across screenshots is fine.

Once captured

Run cd website && npm run dev after capturing to copy the images into the Astro site. Reference paths in docs/widget-gallery.md, docs/api/widgets.md, the homepage and the tutorials are already in place. The site’s npm run check:links check fails if a referenced local image is missing.