home and cliproxy were reverted to their original designs; before/after
screenshots would show identical content. Keep only health migration
screenshots and refocus the HTML report on the design-system docs
restructure (three identity-strip patterns).
User feedback caught a Phase 2 mistake: forcing home and cliproxy into
a one-size-fits-all PageShell + PageHeader chrome regressed density on
both pages.
- home: original 1-row hero (logo + title + version + 4 inline stats)
was split into PageHeader + KpiRow, doubling vertical footprint
- cliproxy: original rail-anchored brand was duplicated by a top
PageHeader, stealing ~80px from the 3-pane body
Resolution: revert both pages to dev-branch state, then restructure
ui/docs/design-system.md to extract THREE identity-strip patterns from
the canonical references rather than imposing one:
1. HeroBar (1-row dense) -> home pattern
2. Rail-anchored (no top chrome) -> cliproxy pattern
3. PageHeader (title + description) -> health pattern
Health stays migrated -- the Monitor archetype + PageHeader is a real
improvement there because the gauge + KPI row + group cards don't fit
the HeroBar pattern.
Future page migrations adapt to whichever identity-strip pattern fits
the page's content shape, not the other way around. design-decisions.md
gets a v1.1 revision row capturing the rationale.
Self-contained HTML report at ui/docs/design-system-phase2-preview.html
with embedded base64 before/after screenshots so reviewers can scrub
the cliproxy / home / health migrations side-by-side without leaving
GitHub. Captured via Playwright with Privacy mode on; data anonymized.
Self-contained HTML report at ui/docs/design-system-preview.html with
embedded base64 screenshots so the PR review and the report itself remain
viewable without external hosting. Companion to the live /_styleguide
route.
Screenshots captured with Privacy mode ON via Playwright; all data
anonymized (Provider A/B/C, fake metrics) -- no real account names,
emails, or OAuth identifiers.