Widget gallery¶
Visual contact-sheet of every public widget. Each card shows the widget rendered in isolation.
Widgets are composable parts; a typical app combines four or five of them. See Anatomy of an app and Getting started for end-to-end examples that wire several widgets together.
Captured by tools/widget_screenshot.py (re-run when widget styling changes):
Display¶
-

Renders a stream's ring buffer as a min/max envelope decimated for 60 fps. Per-channel toggles, optional display filters (rectify, DC removal, RMS envelope), pause/rescale. Keyed by
stream_name.
Recording and sessions¶
-

One button per class plus a Record toggle. Clicking a class button writes a
LabelEventand fires youron_gesturecallback. Status pill shows IDLE / RECORDING. -

Browses recorded sessions (folders or
.session.zip), lets the user tick which to include, and returns aTrainingDatafor@pipeline.train. Per-row class buttons select active classes.
Process management¶
-

Launch / stop external subprocesses (synthetic generator, Virtual Hand, custom acquisition tools) from the GUI. Shows live state per entry; framework adopts children for clean exit.
ML pipeline¶
-

Train / Predict button row plus state indicator. Buttons grey out automatically based on
pipeline.state(no Train while Predicting). Individual buttons:train_button,predict_button.
Output post-processing¶
-

Live-tunable post-prediction smoother (Identity / Gaussian / One Euro). Sliders tune parameters in place; Reset clears smoothing history. Pair with
output_filter(pose, t=time.monotonic())inside@pipeline.predict.
Feature engineering¶
-

Live tickbox panel for choosing which feature transforms feed the model. Construct with
{name: callable}; the selector concatenates active features along axis 0. Useselector.n_activeto size architecture hyperparams.
Branding¶
-

The MyoGestic wordmark fit-to-cell with aspect preserved. Drop into a grid cell as a branding header - pairs with the square OS icon
core.pywires into the dock / taskbar / title bar.
ML readout¶
-

Big, centred class-name readout of the current classifier output. Reads
pipeline.predictions["class"], colour-codes via the shared palette, optionally renders the predicted class's probability as a coloured progress bar.
Virtual Hand integration¶
-

Compact VHI control palette - auto-refreshes the cached movement list from the gRPC plane (off-thread, throttled) and dispatches button clicks to
VhiControlClient.set_movement. Highlights the current movement; greys out while disconnected.
Other widgets¶
These need richer fixtures (live data, trained models, recorded trials) than the screenshot script currently produces. They're documented in the API reference: