utils/unified-design-system/README.md
Krzysztof kuhy Rudnicki b9e2312779 unified-design-system: split the three docs — repo now fully under the cap
tokens.md        260 -> 219 + operability.md 69
  nielsen-audit.md 258 -> 154 + shipped-record.md 122
  README.md        257 -> 172 + operability-patterns.md 95

Worked the requested order rather than cutting straight to a split.

(a) Automate: the categorical ramp's four constraints were restated in prose
    with the measured margins hand-copied (ΔE 34.6/26.0/25.6/27.4, etc).
    ramp_check.py prints every one of those numbers and fails if any
    regresses, so the prose now points at the command instead of duplicating
    its output. The two consequences the numbers do not state — why the ramp
    is mid-toned, and why lightness falls monotonically — stay in prose.

(b) Dedup: measured, not assumed. Zero paragraphs are shared between the
    three files; README already delegates the rules to tokens.md and shows
    only per-stack code. Nothing to remove.

(c/d) Split on meaning: tokens.md is "frozen values", so the behavioural
    requirements that *use* those tokens moved to operability.md, with the
    per-stack code for them in operability-patterns.md. nielsen-audit.md is a
    live grid, so completed phases moved to shipped-record.md — following the
    existing phase3-record.md convention, not a new one.

The focus-ring token row had to stay in tokens.md: palette_check.py parses
that file, and moving the row made the check fail with "MISSING md:focus-ring"
— a token it cannot see is a token nothing compares across the four stacks.
The rationale lives in operability.md and the row links to it.

palette_check and ramp_check both at 0 failing checks; every relative link
between the docs resolves.

`bash scripts/check_file_length.sh --all` now exits 0.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CFSGeWxHPY5rHUGrPqhcmG
2026-08-22 15:42:27 +02:00

7.9 KiB

unified-design-system

The concrete, frozen design tokens (colors, spacing, type, radius, shadow policy) that implement safe-design-rules' 28 abstract rules identically across every one of kuhy's repos — Flutter, web/React, and Python/Tkinter. One palette everywhere (including Flutter launcher-icon glyphs), not just shared rules. Use whenever building or fixing UI in any of kuhy's repos — this is what a design audit's violations should be fixed against, and what new UI should be built from instead of inventing new values. Full frozen value table in tokens.md; annotated component references in components.html (9 components: Button/Card/Input/Chip/ Dialog/Switch/Divider/Checkbox/SnackBar) and button.html (the earlier single-component exemplar that set the doc format).

Read tokens.md for the full frozen value table before touching any theme/CSS/style file.

Why one palette, not per-app branding

Apps are visually unified down to the exact hex values, not just the same rules with different accents. This also means the Flutter launcher-icon family (the app-icon skill/generator) uses the same accent for every app's glyph — apps are distinguished by glyph shape, not color.

Per-stack implementation pattern

Flutter

Build ThemeData from explicit ColorScheme.dark(...)/.light(...) using the exact hex values in tokens.md — not ColorScheme.fromSeed, which algorithmically derives tones and won't reproduce hand-picked values. Add:

  • A ThemeExtension for roles M3 doesn't have (success/warning — error is already danger), e.g. AppStatusColors(success:, warning:). Only add this if the app actually has success/warning status indicators — grep for existing Colors.green/Colors.orange-style usage first; don't add it speculatively (it becomes untested dead code otherwise).
  • AppSpacing / AppRadius / AppTextSize static const classes mirroring the scales in tokens.md.
  • Wire every screen through Theme.of(context) — the dominant violation across every Flutter audit was raw Colors.* literals bypassing an existing (or newly added) theme. Grep for Colors\. / Color(0x after wiring the theme and route every hit through a theme role.
ThemeData buildTheme() => ThemeData(
  useMaterial3: true,
  colorScheme: const ColorScheme.dark(
    surface: Color(0xFF211D1B),        // ink
    surfaceContainerHighest: Color(0xFF38312E), // ink-raised-2
    onSurface: Color(0xFFECEAE9),      // text-on-dark
    onSurfaceVariant: Color(0xFFAAA09A), // muted-on-dark
    outline: Color(0xFF463E3A),        // line-dark
    primary: Color(0xFFB8862E),        // accent
    onPrimary: Color(0xFF211D1B),      // on-fill — NOT text-on-dark, see below
    error: Color(0xFFE2585F),          // danger
    onError: Color(0xFF211D1B),        // on-fill
  ),
  extensions: const [AppStatusColors(
    success: Color(0xFF8A9A3C),
    onSuccess: Color(0xFF211D1B),      // on-fill
    warning: Color(0xFFE0A63C),
    onWarning: Color(0xFF211D1B),      // on-fill
  )],
);

onPrimary/onError/onSuccess/onWarning are all on-fill (#211D1B, dark), never onSurface's near-white — see tokens.md's on-fill entry: every filled semantic color under-contrasts with near-white text.

Web (CSS custom properties)

One :root block (plus a light/dark override block as needed) with these variable names — kept identical across repos so the vocabulary transfers:

:root {
  --bg: #211d1b; --surface-1: #2b2624; --surface-2: #38312e;
  --border: #463e3a; --text: #eceae9; --text-muted: #aaa09a;
  --accent: #b8862e; --success: #8a9a3c; --warning: #e0a63c; --danger: #e2585f;
  --on-fill: #211d1b;   /* text/icons ON a filled accent/success/warning/danger surface */
  --on-scrim: #eceae9;  /* text/icons on a dark translucent overlay/backdrop — fixed, not theme-tied */
  --radius-sm: 8px; --radius-md: 12px; --radius-lg: 16px;
  --space-1: 4px; --space-2: 8px; --space-3: 16px; --space-4: 24px;
  --space-5: 32px; --space-6: 48px;
  --font-sans: -apple-system, "Segoe UI", Roboto, system-ui, sans-serif;
  --font-mono: ui-monospace, "Cascadia Code", monospace;
}
@media (prefers-color-scheme: light) {
  :root {
    --bg: #f6f4f3; --surface-1: #fcfbfb; --surface-2: #fcfbfb;
    --border: #e0dad7; --text: #211d1b; --text-muted: #70625b;
    /* --on-fill / --on-scrim stay fixed — the fills and the scrim don't change with page theme */
  }
}
body { font-size: 16px; font-family: var(--font-sans); }
button.primary { background: var(--accent); color: var(--on-fill); }

Don't reach for a single "on-accent" variable for both cases (an earlier dufs-cloud draft did, and it broke the moment the accent went from blue to gold): a filled button needs --on-fill (dark), a caption over a photo-tile scrim needs --on-scrim (near-white) — same-looking near-white token, wrong one silently fails contrast on fills.

Every hardcoded color/spacing/radius/shadow value in component CSS should route through one of these variables. No #fff/#000/rgba(0,0,0,…) literals — use var(--text)/var(--bg) or an ink-tinted rgba.

Python/Tkinter

No ttk.Style() in any of these apps currently — introduce (or, for the screen-locker/wake_alarm/diet_guard family, extend) a shared constants module/dataclass with the same token names, and have every tk.Label/ tk.Button/tk.Frame read from it instead of inline hex literals:

@dataclass(frozen=True)
class Tokens:
    bg: str = "#211D1B"
    fg: str = "#ECEAE9"
    muted: str = "#AAA09A"
    field_bg: str = "#2B2624"
    accent: str = "#B8862E"
    success: str = "#8A9A3C"
    warning: str = "#E0A63C"
    danger: str = "#E2585F"
    on_fill: str = "#211D1B"  # text on an accent/success/warning/danger bg — NOT fg
    font_family: str = "Arial"

A shared _button(bg=...) helper that always sets fg=self._colors.fg is the exact bug this caught: fg (near-white) is correct when bg is field_bg/bg (a dark neutral surface), but any button whose bg is accent/success/warning/danger needs fg=self._colors.on_fill instead — pick fg from bg, don't hardcode it per call site.

For screen-locker/wake_alarm/diet_guard specifically: this lives on gatelock.LockConfig (the shared token source already vendored in this same ~/utils monorepo, see ../gatelock/) rather than a per-repo duplicate, since all three already depend on it. Extending LockConfig requires production verification (/usr/bin/python3 -c "import gatelock") since it backs 3 live systemd services — a dev-venv-only check is not sufficient.

Pointer-free + small-screen

The rules are in operability.md; the per-stack code that satisfies them — Flutter, web and Python/Tkinter — is in operability-patterns.md.

Do NOT

  • Don't use ColorScheme.fromSeed for the unified palette — it can't reproduce hand-picked hex values exactly.
  • Don't pass a positive font size to Tk — that's points, ~37% oversized.
  • Don't make long-press, hover, or drag the only route to an action.
  • Don't set highlightthickness=0 on anything focusable, or set highlightbackground when you meant highlightcolor.
  • Don't centre-anchor content that can overflow — it clips symmetrically, taking the header and the submit button with it.
  • Don't tune a layout against a 1080p screen, and don't test responsiveness only at phone-portrait sizes; neither exercises 1366x768 or 1024x600.
  • Don't add shadows to dark surfaces, or mix shadow + border depth on the same element (see tokens.md's shadow policy).
  • Don't invent a new spacing/radius/font value outside the scales in tokens.md — round to the nearest step instead.
  • Don't add a ThemeExtension/status-color role speculatively — check the app actually uses that status (success/warning/etc.) before adding it.
  • Don't touch ../gatelock without the production-path verification above; a broken import there takes down 3 services at once.