Files
zeekayandHanzo Dev f5531da6e4 Borders you can see are not the same as borders that shout
The dark theme drew every emphasised edge at --neutral-500 (#737373) — a 4.43:1
mid-grey. One of those around a card on near-black is a wireframe, and that is
what "garish white on dark for outlines" was: hanzoai/id put it on .hanzo-id-input,
base/ui-react put it on .input, .btn--outline, a dashed .btn--add and an iframe
frame. They were following the file's own instruction to reach for --border-strong
whenever an edge had to be visible, because the only alternative was --border at
#1f1f1f — 1.27:1, invisible the moment it lands on anything lifted. Two rungs, an
octave apart, and nothing in between.

So there is a middle now, and the tiers split by DUTY rather than by degree.
--border (.10) and --border-strong (.16) are DECORATIVE: alpha, so they composite
correctly on the page, on a card and inside a popover, and they owe no ratio.
--border-control (--neutral-500) is PERCEIVABLE: it is the edge that IS the
affordance, it owes 3:1, and it is named for the duty so it cannot be spent on a
card by someone who just wanted "stronger". The gate follows the duty — it pins
--border-control and --ring, and lets the hairlines be as quiet as they look.
--input, which every shadcn `border-input` resolves through, was #1f1f1f at
1.27:1 and is now the control rung: that was a live WCAG 1.4.11 failure on every
field in the fleet, and nothing measured it because the gate watched a token the
inputs were not using.

The rest is the reference surface (hanzoai/extension sidebar.css) carried over:
#0a0a0a rather than #000000, a 0a/0f/17/26 lift ladder that actually steps,
#fafafa ink instead of #ffffff (21:1 on pure black is the maximum-contrast pair
that exists, and it halates), a graded text ramp at .78/.55/#a3a3a3, and
--glass/--glass-strong so a surface can rise by 5% of white instead of a block
fill. --border-hairline and --border-card were two more names for the one
hairline and are gone; the 46 components that used them now say --border.

MOBILE-FIRST, and free. Spacing is authored at phone values and scales up at md.
--gutter is max(1rem, safe-area-inset), so a notch is cleared with no per-app
rule. Under pointer:coarse every control clears a 44px --tap-target, checkbox and
radio hit areas expand to 44px via a pseudo-element while the 16px box is left
alone (WCAG 2.5.8 puts the floor at 24px; every one of them shipped at 16), and
fields render at --text-control = 16px so iOS stops zooming the viewport on
focus. --text-control has to be a token rather than a media query because every
component sets its size inline and an inline style outranks any stylesheet.

Light was quietly broken and is not a theme you can leave for later: --card,
--muted, --secondary and --accent were four names for #f5f5f5, so the surface
ladder was flat; ::selection had no light value; and the components painted from
the --white-* ladder, which does not invert — a tier2 badge was white-on-white
and the table header was a black bar with grey text on it. Components now paint
from --glass/--surface-*, and check 4 fails the build if any semantic token built
from a white rung has no .light counterpart.

Verified in Chromium at 390 / 768 / 1280 on the real components, reading computed
styles: no horizontal overflow at any width, fields 13px -> 16px on touch,
targets under 44px 16/19 -> 1/19 (the last is the 16px checkbox box, whose hit
area measures 44x44 and takes a tap 18px off-centre), --ring and --border-control
at 3.19:1 dark / 4.05:1 light worst case.

Co-authored-by: Hanzo Dev <[email protected]>
2026-08-04 09:47:04 -07:00

12 lines
617 B
React

import React from 'react'
export function Progress({ value = 0, max = 100, height = 6, style, ...rest }) {
const pct = Math.max(0, Math.min(100, (value / max) * 100))
return (
<div role="progressbar" aria-valuenow={value} aria-valuemin={0} aria-valuemax={max}
style={{position:'relative',width:'100%',height,borderRadius:'var(--radius-full)',background:'var(--glass-strong)',overflow:'hidden',...style}} {...rest}>
<div style={{height:'100%',width:pct+'%',background:'var(--primary)',borderRadius:'var(--radius-full)',transition:'width var(--duration-base) var(--ease-out)'}} />
</div>
)
}