@jupyter-kit/executor-pyodide

In-browser Python executor for @jupyter-kit/core. Pyodide runs in a dedicated Web Worker so the main thread stays responsive during execution.

Install

pnpm add @jupyter-kit/executor-pyodide @jupyter-kit/core

Usage

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

const executor = createPyodideExecutor({
  packages: ['numpy', 'pandas'],   // preload
  autoloadImports: true,           // auto-install from `import` statements
  figureFormats: ['svg'],          // matplotlib output formats
  onStatus: (s) => console.log('pyodide:', s),
});

createRenderer({
  executor,
  plugins: [createEditorPlugin({...})],
});

What you get

• stdout / stderr captured and streamed as Jupyter stream outputs.
• Last-expression value rendered via IPython-style _repr_*_ methods (_repr_html_, _repr_latex_, _repr_png_, _repr_svg_, _repr_markdown_, _repr_json_). Pandas DataFrames, sympy matrices, etc. work out of the box.
• matplotlib figures auto-captured at end of cell as display_data (SVG by default; configurable via figureFormats: ['svg' | 'png']).
• Errors with IPython-style ANSI-coloured tracebacks (helper frames filtered out).
• IPython magics (%matplotlib, %autosave, !pip ...) are commented out before execution rather than raising SyntaxError.
• Sympy tuples / lists render as combined LaTeX when sympy is loaded.

Options

| Option | Default | Notes | |---|---|---| | src | jsdelivr full build | URL of pyodide.js | | indexURL | derived from src | Directory for .wasm / .whl | | version | '0.26.2' | Pyodide version pin | | packages | [] | Preload list | | autoloadImports | false | Scan source for imports | | figureFormats | ['svg'] | ['svg'], ['png'], or both | | onStatus | — | 'idle' / 'loading' / 'installing' / 'running' / 'ready' / 'error' |

Bundle size

~3KB gz (script + worker code). Pyodide itself (~10MB compressed) is fetched from the CDN at first use.

License

Apache-2.0. Pyodide itself is MPL-2.0 and served from the CDN.