chore(design-system): re-sync to latest and drop the redundant next/font wiring - #8
Merged
Merged
Conversation
…artifacts react 2.0.1 -> 2.3.4, tokens 0.5.0 -> 0.12.0, ai-patterns 0.9.2 -> 0.20.0, eslint-config 0.3.0 -> 0.8.0. `ds-resync migrate` reports no token renames left in `src/` — every hit is inside the generated `.claude/skills/` trees, which `ds-resync artifacts` regenerates here. The new files are the palette/platform stylesheets and the self-hosted Geist and JetBrains Mono faces the brand skill now ships. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
… faces tokens 0.7.1 stopped reaching Google Fonts and started shipping Geist and JetBrains Mono as woff2 in the package, so importing tokens.css is now enough to load them. The `--ds-font-*-override` block and the two `next/font/google` loaders were pointing the family tokens at a second, build-time-fetched copy of the same typeface; the hooks still exist for a family the system does not ship. Verified in the browser: `--font-sans` resolves to the package's Geist and `--font-mono` to JetBrains Mono, both with a loaded face. The page renders identically in light and dark. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The random port in `3000..3999` could not be predicted by anything else in the repo: `.claude/launch.json`, `playwright.config.ts` and the E2E job in CI all name 3000, so tooling that reads them opened a port the dev server was not on and you had to read the startup log to find the real one. It was also broken outright under a pty. `console.log` colorizes a number when stdout is a TTY, so `next dev -p` was handed `\e[33m3891\e[39m` and rejected it as "not a non-negative number". Piped output hid that, which is why it passed review. The conflict the random port was avoiding — a second worktree or a concurrent session — is now handled explicitly: `PORT=3010 pnpm dev`. That keeps the port knowable when nobody sets it, which is the case the config files describe. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
EliRobinson
force-pushed
the
chore/design-system-resync
branch
from
August 26, 2026 08:43
8c4bcb8 to
4bfcda1
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Brings the four
@elirobinson/*packages to latest, re-syncs the generated agent artifacts, and drops the app-side font wiring the upgrade made redundant.@elirobinson/react@elirobinson/tokens@elirobinson/ai-patterns@elirobinson/eslint-configMigration surface
ds-resync migratefinds no token renames left insrc/— every hit it reports is inside the generated.claude/skills/trees, whichds-resync artifacts --writeregenerates from the installed packages. The app's own design system surface is small (Card,CardContent,Eyebrow, the.t-*classes) and none of it changed shape.The artifact sync also brought in the palette/platform stylesheets (
palettes.css,mobile.css) and the self-hosted Geist / JetBrains Mono faces the brand skill now ships.Fonts
@elirobinson/tokensstopped reaching Google Fonts at 0.7.1 and now ships Geist and JetBrains Mono as woff2 inside the package, so importingtokens.cssis enough to load them. The--ds-font-*-overrideblock inglobals.cssand the twonext/font/googleloaders inlayout.tsxwere pointing the family tokens at a second, build-time-fetched copy of the same typeface, so both are gone. The override hooks still exist upstream for a family the system does not ship —AGENTS.mdnow says that instead of describing the old wiring.Confirmed in the browser after the change:
--font-sansresolves to the package'sGeistand--font-monotoJetBrains Mono, both with a loaded face.Dev server port
pnpm devpicked a random port in3000..3999, which nothing else in the repo could predict —.claude/launch.json,playwright.config.tsand the E2E job in CI all name 3000, so tooling that reads them opened a port the server was not on and you had to read the startup log to find the real one. It was also broken outright under a pty:console.logcolorizes a number when stdout is a TTY, sonext dev -preceived\e[33m3891\e[39mand rejected it as "not a non-negative number". Piped output hid that, which is why it passed review.The script is now
next dev --turbopack -p ${PORT:-3000}. The conflict the random port was avoiding — a second worktree, a concurrent session — is handled explicitly withPORT=3010 pnpm dev, which keeps the port knowable when nobody sets it. Documented inREADME.mdandAGENTS.md.Verified both: the preview tool opens 3000 and finds the app there, and
PORT=3010 pnpm devbinds 3010.Verification
pnpm type-check— passpnpm lint— pass (oneslint-config0.8.0)pnpm format:check— passpnpm test— 2 passedpnpm build— passpnpm test:e2e— 16 passed, including the design system contracts (touch targets, visible focus, WCAG AA contrast) in light and dark across chromium, firefox, webkit and mobile-chromeScreenshots
Before/after at 1280×900, light and dark: the rendered page is pixel-identical in both themes. The home page only uses the sans family and neutral surfaces, both of which resolve to the same values before and after; the new status hues and tinted-badge colours in tokens 0.6.0–0.12.0 have no call site here yet.