Cross-filtering and column summaries

Open any dataset and you’ll see a row of column summaries above the table. Each summary is bound to its column and to a shared selection — a Mosaic primitive that sits between every component in the page. Mutate the selection from one component and every other component reading from it repaints.

Summary types

ST-1 picks a summary kind based on the column’s declared type:

• Numeric / temporal — histogram. Drag to select a range; release to apply.
• Boolean — two-bar count. Click either bar to filter; click again to clear.
• Low-cardinality categorical — top-K bars. Click a bar to filter to that value; click another to switch.

Combining filters

Filters across columns combine with AND. The clause-count badge at the top of the table tells you how many are active. Clear all filters empties the selection; clicking the × on a single summary clears just that column.

Promote to DBT

Filters reset on navigation by default — they live in the page, not the dataset. To keep a particular cross-filter, click Promote to DBT. ST-1 crystallizes the active predicate into authored SQL (SELECT * FROM <source> WHERE …) and saves it as a new derived table. From then on it’s a regular DBT — readable, editable, lineage-tracked.

Fan-out across charts

Every visualization on a dataset listens to the same selection. Filter the dataset and every chart that reads from it — or from any DBT downstream — repaints reactively. There’s no round-trip; the data lives in DuckDB-WASM in your browser.