Home

Real-time biosignal experiment GUI builder. A compact Python framework that turns a short script into a live experiment - signal viewers, recording, training, prediction - without classes, registries, or config files.

from myogestic import App, Stream
from myogestic.sources import LSLSource
from myogestic.widgets import recording_controls, signal_viewer

app = App("Hello EMG")
app.streams(Stream("emg", source=LSLSource("EMG"), window_seconds=1.0))


@app.ui
def ui(ctx):
    signal_viewer(ctx, "emg")
    recording_controls(
        ctx, ["Rest", "Fist"], on_record=app.start_recording, on_stop=app.stop_recording
    )


app.run()

That's the whole loop. Add a Pipeline, decorate extract / train / predict, and you have a closed-loop experiment.

Live signal viewer (right) plus process launchers, recording controls, train/predict, output filter, and session manager (left). One script, one window.

Top-level data flow

Sources push samples into a Stream ring buffer. The pipeline's @extract, @train, @predict decorators read from that buffer and emit predictions to outputs - LSL, UDP, Serial, or the Virtual Hand Interface over LSL+gRPC. The render thread runs @app.ui at 60 fps independently of the predict thread; see Concepts → Threading.

Try it in your browser

A live MyoGestic app running entirely in your browser via Pyodide. Pick a gesture, record a few seconds, train an LDA, watch the prediction flip on every click. No install, no Python on your machine, just a tab.

Pyodide + imgui-bundle WASM + scikit-learn, ~30 s first-load, cached after.

Open the Playground

Where to go next

• Getting Started - install, run the synthetic-EMG demo
• Anatomy of an app - walk through one complete script in the order you write it
• Guides - tutorials (line-by-line walkthroughs) + how-to recipes
• Concepts - streams, pipeline, threading, recording, design principles
• Troubleshooting - symptom-first reference for the things that go wrong
• Reference - auto-generated API + cheatsheet + glossary

Design principles

• No base classes, no inheritance, no registration, no config files.
• Public API fits on one page. Each widget is a single public function; implementation may split into private _<widget>_*.py helpers when it grows.
• Errors tell you what to write, not what went wrong.
• Plain functions with typed arguments and typed returns.
• Immediate-mode rendering - widgets are stateless function calls.

See Design principles for the full list.