@jupyter-kit/widgets

ipywidgets plugin for @jupyter-kit. Renders application/vnd.jupyter.widget-view+json outputs via @jupyter-widgets/html-manager.

Two modes

| | What you get | Requires | |---|---|---| | Static | Hydrates metadata.widgets["application/vnd.jupyter.widget-state+json"] from the saved notebook. Sliders show their last value but don't fire callbacks. | nothing extra | | Live | Sliders drive Python callbacks and outputs re-render on every change. | an executor with commProvider (e.g. executor-pyodide, executor-jupyter) |

The plugin auto-detects the executor's commProvider — no extra wiring.

Usage

Static (read-only viewer)

import { Notebook } from '@jupyter-kit/react';
import { createWidgetsPlugin } from '@jupyter-kit/widgets';

<Notebook
  plugins={[createWidgetsPlugin()]}
  ipynb={notebook}
/>;

Live (interactive, with Pyodide)

import { createPyodideExecutor } from '@jupyter-kit/executor-pyodide';
import { createWidgetsPlugin } from '@jupyter-kit/widgets';

<Notebook
  executor={createPyodideExecutor({ packages: ['ipywidgets'] })}
  plugins={[createWidgetsPlugin()]}
  ipynb={notebook}
/>;

Live (with a remote Jupyter kernel)

import { createJupyterExecutor } from '@jupyter-kit/executor-jupyter';
import { createWidgetsPlugin } from '@jupyter-kit/widgets';

<Notebook
  executor={createJupyterExecutor({ baseUrl, token })}
  plugins={[createWidgetsPlugin()]}
  ipynb={notebook}
/>;

Options

| Option | Type | Notes | |---|---|---| | loader | (moduleName: string, moduleVersion: string) => Promise<unknown> | Resolver for custom widget modules outside @jupyter-widgets/{base,controls,output}. Return the module exports containing view/model classes. |

Bundle cost

Pulls in @jupyter-widgets/{base,base-manager,controls,html-manager,output} — ~600 KB – 1 MB minified. The heavy deps are dynamic-imported on first setup() so the initial render path stays light. The plugin lives in a separate package so notebooks without widgets don't pay this cost.

Limitations

• Interactivity requires an executor with commProvider. Without it the plugin still renders saved widget state, but sliders are inert.
• widget-view outputs whose model_id wasn't saved show [widget — run the cell to render] until you re-execute.
• Output widget + matplotlib auto-capture works for typical @interact(...) patterns but exotic uses (multi-widget display, clear_output(wait=True) interleaving) may diverge from classic Jupyter.

More

Full reference: https://jupyter-kit.walkframe.com/reference/widgets/