corr_vars_widget

A collection of AnyWidgets for the corr_vars package.

Widgets

ObsWidget

Displays a polars.DataFrame in a Quak Widget with additional infos about the shape and the observation level.

ObsmWidget

Displays a list of polars.DataFrame in Quak Widgets within collapsible accordions with additional infos about the shape of the respective polars.DataFrame.

JsonWidget

Displays a searchable JSON viewer (filtering the keys of the first level). The JSON object is displayed in a collapsible view.

JsonmWidget

Displays a dict of JSON objects in collapsible accordions. A shared search bar filters the first-level keys across all JSON views.

TimeseriesWidget

Displays per-ID clinical timeseries over a shared, synchronized time axis, backed by DuckDB. It combines three lane types in one view: interval lanes (Gantt-style bars), event lanes (point-in-time dots), and value charts (line/area). Pick an ID with the searchable combobox, pan/zoom via the overview strip, and toggle labels from the settings popover. Provide a color_col (any CSS color or shadcn CSS variable) to colour rows explicitly.

A fluent Timeseries builder is available as the primary API:

(
    Timeseries(id_col="stay_id", start_col="start", end_col="end", value_col="value")
    .interval("Device", devices)
    .event("Sedation", sedation)
    .value("PaO₂", pao2)
)

Development

Development requires uv and pnpm.

uv venv
uv pip install -e . --group dev
pnpm dev # start a development server
uv run jupyter lab notebooks/example.ipynb