html-artifact-redesign

Scrape every route and embedded data payload of the original site first, then rebuild it as one self-contained HTML file applying the chosen design language. Verify the result with node --check on extracted scripts and headless Chrome DOM assertions.

Curl the Refero style URL and concatenate the self.__next_f.push chunks, then decode with unicode_escape and html.unescape to recover the DESIGN.md prose, color tokens, type scale, and component list. The authored description prose serves as the build blueprint.

Yes. Use Chrome headless --dump-dom to grep for headings, asset paths, and links, or set up a Py-Playwright venv to assert runtime facts like canvas pixels, element counts, and horizontal overflow at multiple viewport widths.

A single unbalanced brace or stray closing token silently kills the entire inline script, leaving the page rendered but non-functional. Extract the script block to a temp file and run node --check before trusting any render.

Yes, a React/Vite/TypeScript branch is supported where npm run build (tsc -b && vite build) is the authoritative gate when browser automation is blocked. The @/* path alias must be set in both vite.config.ts and tsconfig.json.

The --dump-dom check is static only; it proves the file parses and all markup, links, and images are present, but it runs no JavaScript and cannot judge visual layout or motion. Aesthetic sign-off remains with the user.