A catalogue is a database wearing typography

The Data sources panel imports CSV, TSV, JSON (an array or newline-delimited), Parquet and Excel .xlsx — each as a table in DuckDB, running as WebAssembly in your browser. A remote CSV, TSV, JSON or Parquet URL can be added too; it is fetched only from an origin you have consented to.

The Data query panel takes SQL with DuckDB’s own diagnostics, offers filter, sort and group builders that write the SQL for you, and previews the first 50 rows. Every query passes a guard that admits exactly one read-only statement over your sources, and each source has a refresh policy: manual, on open, or every N seconds for a remote one.

A binding connects a query to the page through an expression in paged.data’s own binding language: 42 functions, arithmetic, comparison and concatenation, with errors as values. Number, currency, percent and date formats follow a locale — nine are built in, from en and de to fr, it, es and nl — and a field can carry its own.

Bindings write native content: a variable becomes a tagged placeholder field, a table binding a native table, an image binding a placed image, a visibility binding shows or hides an element, and a style rule applies the document’s own paragraph, character or cell styles to the records it fires on — with a preview of which records those are.

EAN-13, UPC-A, Code-128 and QR are encoded by the engine and drawn as one closed vector path per dark module, scaled to their frame — never pixels, so they print at any size.

When the data changes, nothing is overwritten behind your back. Each binding shows its state — synced, stale, pinned, overridden or error — with Pin, Unpin and Accept source. “What changed?” lists the rows added, removed and changed since the document was last written from the data, each changed cell before and after, and which bindings each change reaches. Refresh fields then rewrites only the fields whose value changed, in one undo step.