Changelog

0.2.1

validate(tree, options?): reports the mistakes that render silently wrong, with a path into the tree and a fix in each message. Checks fill and spacer placement, justify alongside a fill, table keys that no row has (with a suggestion), empty columns, unregistered fonts, long text directly in a row, and padding on a table. Never throws, never renders.
Documentation site at https://grandbusta.github.io/spyde/, with llms.txt and llms-full.txt for AI assistants and a page of rules for writing Spyde with one.
README shortened to install, one example, and a pointer to the documentation.
The raw PDFKit comparison example was removed from the repository.

Documents are no longer limited to one page.

Breaking

render returns Uint8Array instead of Buffer. Files, responses, and uploads accept it unchanged; Buffer.from(pdf) wraps it if a Node-only method is needed.
Tall content produces more pages. A root column now continues onto further pages instead of being clipped to one.

New

Pagination: columns and tables break across pages; keep, pageBreak; table headers repeat and are never orphaned.
image, spacer, divider, width, height, table.
align and justify on row and column; padding, background, margin as container options; keep as an option on column and table.
renderHtml and renderDisplayList; field on text styles; the live preview example.
Pagination is a single pass; text measurement is memoised. A 2000-row statement went from 888 ms to 84 ms.
A text with a line height tighter than the font's natural spacing now draws every line.

README updates.

First release: the single-page engine with text, padding, background, fill, row, column, and render.