Vega-Lite is a high-level declarative language for specifying interactive, multi-view visualizations. It compiles a concise JSON specification into a full Vega visualization, automatically inferring scales, axes, and legends from encoding declarations. The grammar-of-graphics encoding maps data fields to visual channels such as position, color, size, and shape, while a multi-view composition grammar enables layered, faceted, concatenated, and repeated layouts. Reactive parameter binding links named parameters to input widgets, selections, and expressions for dynamic updates.
The project supports interactive data exploration through selections, tooltips, and pan/zoom interactions. Statistical visualizations include histograms, box plots, density estimates, regression lines, and loess smoothing. Geographic mapping renders GeoJSON shapes using configurable cartographic projections for choropleths and point maps. A data transformation pipeline provides aggregation, binning, filtering, window calculations, and reshaping operations like fold and pivot.
Vega-Lite supports a wide variety of mark types, from bars, lines, and areas to points, text, images, and tick marks. Global configuration and theming allow uniform styling across axes, legends, and marks. Data can be loaded from inline JSON, remote URLs, runtime injection, or generated as numeric sequences, with shared datasets and per-field parsing rules.
The specification compiles to a low-level Vega description, which can be embedded in web pages and exported via the included viewer.