Skip to content

v0.5.0 — The Extraction Depth Release

Choose a tag to compare

View all tags
@zk-xyz zk-xyz released this 14 Apr 22:52
v0.5.0
bb1d77d

Make your brand machine-readable from any source. Not just websites anymore.

New Tools

brand_extract_pdf — Ingest brand guidelines from a PDF. Colors, typography, spacing, and rules extracted and merged into core-identity.yaml as source: "guidelines". Guidelines outrank web extraction by default because they're authoritative.

brand_resolve_conflicts — When you've extracted from multiple sources (website + PDF + Figma), this tool shows where they disagree and lets you resolve it. Uses the source_priority field in brand.config.yaml: guidelines > figma > visual > web > manual.

Visual Token Extraction

Rendered-page extraction now captures what dembrandt does:

  • Spacing scale — base unit detection, common padding/margin values
  • Border radius — dominant style (rounded, sharp, pill), value distribution
  • Shadows — elevation levels with usage context (cards, elevated, subtle)
  • Component variants — buttons, inputs, badges with actual computed styles

All wired into design-synthesis.json, DESIGN.md, and tokens.json.

Source Catalog

Every extraction now records which source contributed each value in .brand/source-catalog.json. This is what makes conflict resolution possible. When web says #00749a and guidelines say #00A3E0, the catalog knows both and the agent can resolve it.

Numbers

36 tools. 337 tests. 0 vulnerabilities. 2,666 npm downloads in 10 days.

Assets 2
Loading