Restyle websites with typed catalog choices

01 · Role in the system

What Jev does here

Code first reads the page: real colours are clustered into roles (background, surface, text, muted, accent, border) and summarized in English, because Jev reads words better than hex codes. One request then asks everything at once - which palette, which font pair, how much spacing, which corner style, reading mode or not, and what kind of request this is (new look, adjust, fix images, a page element, undo) - with every answer constrained to a catalog of 40 palettes and 8 font pairs, so nothing invalid can come back. The compiler tags each element with the palette role its original colour maps to and applies one stylesheet; saved picks are recompiled on later visits rather than regenerated. Picking a single element works the same way: code lists landmarks and blocks in plain English and Jev chooses, with speculative fan-out on very large pages.

02 · Control boundary

Where Jev sits

Jev as the chooser in a catalog-constrained restyling pipeline: code reads the page and builds a facts summary, one typed call selects palette, fonts, spacing and intent from fixed options, and a compiler turns picks into role-stamped CSS.

Code owns the loop, permissions, thresholds, validation, and side effects. Jev owns only the bounded judgments described above.

03 · Known limits

What this evidence does not prove

• Restyling quality is bounded by the 40-palette and 8-font catalog; custom brands need new catalog entries.
• Role clustering of site colours is heuristic; unusual palettes can map to the wrong roles.
• Brand-new repo (created 2026-09-28, no stars yet); no external evaluation.

04 · Attribution

Public sources

This is a Community record: the project was published by a third-party community author.

• apneduniya ↗Community · github · public · checked 2026-09-29