Portal UX
Frontend UX direction
This is the TASK-017 design direction for the fixture portal. It was written before implementation so the redesign starts from reviewer jobs rather than CSS tweaks.
- Source
docs/frontend-ux-direction.md- Mode
- rendered markdown
- State boundary
- authored documentation; not live state
Frontend UX direction
This is the TASK-017 design direction for the fixture portal. It was written before implementation so the redesign starts from reviewer jobs rather than CSS tweaks.
Diagnosis
/issues/replaces the old portal roadmap and should not become another local/confluence/uses a runtime-cached tree on the left and a validated page/evidence/lists where screenshots live but does not render the screenshots./mockups/is useful as a design comparison artifact, not as the product UI.- The shared navigation wrapped into a tall pill stack on mobile, pushing useful
backlog. It should explain the Forgejo issue workflow, label taxonomy, and agent trigger contract while sending live planning back to Forgejo.
workspace on the right. Empty and refresh-failure states are first-class.
Evidence review needs direct visual proof first and paths second.
It should move out of primary attention once a durable direction is chosen.
work below the first screen. It also gave every route equal weight, including temporary design mockups.
Useful Existing Portal Patterns
The existing RWS architecture portal demonstrates useful patterns to carry over:
- A real shared shell with active navigation instead of route-by-route chrome.
- A Confluence-like workbench with a tree, selected content, and an inspector.
- Progressive disclosure for large hierarchies and generated model data.
- Stronger RWS-blue and yellow accents that feel institutional rather than like a
generic dashboard.
Do not blindly copy its theme switcher, mobile header height, or heavier issue workflow visualizations. This workbench needs to feel more operational: source fidelity, evidence proof, task state, and review handoff are more important than broad portal exploration.
Information Architecture
Status: triage entry point and route discovery.Source: primary architecture reviewer anchor; tree, page, metadata, findings,Review: findings, compliance, and preview review should eventually group intoIssues: workflow contract, not a local backlog. Show Forgejo entry points,Evidence: visual proof gallery with manifest/readiness metadata nearby.Mockups: internal design-decision route; remove from top-level emphasis after
and source evidence.
one review area or section navigation instead of three equal top-level routes.
label semantics, implementation trigger, safety gates, and evidence flow.
the visual direction is selected.
Page Hierarchy And Navigation
- First viewport should answer: where am I, what needs review, what is blocked,
- Use compact page headers on task routes. Large hero panels are reserved for
- Use row-based lists and split panes for workbench data. Cards are summaries only.
- Keep breadcrumbs as route context, but avoid repeated explanatory paragraphs
- On mobile, global navigation should stay below roughly one compact band plus two
what proof exists, and where do I go next.
discovery pages, not operational queues.
once a page has moved into operational mode.
rows of route links; the active work/source/evidence content must start quickly.
Density, Spacing, And Typography Rules
- Use an architecture operations ledger visual language: compact rows, restrained
- Cap task page headings around
2remto4rem, not the previous oversized - Prefer
8pxto16pxinternal spacing for task rows and10pxto14px - Metadata, paths, and immutable IDs stay visible but visually subordinate to
- Replace pill spam with structured rows, short status labels, and evidence tags.
borders, smaller radii, RWS-blue/yellow accents, and minimal decorative effects.
5rem treatment.
radii for work surfaces.
human-readable titles, state, owner, review state, dependencies, and proof.
Remove Or De-emphasize
- Duplicate local planning registers on
/issues/. - Giant action panels and hero blocks on operational routes.
- Decorative gradients and large rounded cards that do not improve scanning.
- Fixture prefixes in source display names.
- Evidence pages that only list screenshot paths.
- Mockup route prominence after a direction is selected.
Acceptance Criteria
- Desktop
/issues/at 1440px shows Forgejo issue entry points, trigger label, - Mobile
/issues/at 390px keeps navigation compact and exposes the issue /issues/does not render a local backlog or duplicate issue state.- Issue labels expose richer work semantics: type, status, priority, automation
- Desktop and mobile source pages keep tree context visible while runtime cache
- Evidence renders screenshots inline in the next implementation slice, with
- All inspected routes avoid page-level horizontal overflow and keep visible focus.
delivery mode, and safety boundary in the first viewport.
workflow before the label reference.
intent, gate, and area.
state remains explicit.
route, viewport, commit or generation time, and manifest path available.
First Redesign Slice
The first proof slice is /issues/ plus shared navigation. It implements the issue workflow contract, reduces mobile nav height, documents the label model, and makes Forgejo the live planning surface. Evidence screenshots and source display-name cleanup remain the next coherent slices.