UI/UX Improvement PRD¶
Source files:
Architechture & Research/RapidDraft/UX & UI/UIUX Improvement for RapidDraft.mdLast synthesized: March 2026
Executive Summary¶
RapidDraft's primary UX blockers are navigation clarity, overloaded primary actions, panel inconsistency, and fragmented review + issue workflows. The core fix is simple: reframe RapidDraft around "Reviews + Findings/Issues" as the north-star workflow, making every tool panel (DFM, Vision, Fusion, Report, DraftLint) feel like part of a single review pipeline rather than a collection of utilities. This document details the current problems, recommended fixes, and a phased implementation roadmap.
Current State: What RapidDraft Does Well¶
RapidDraft already has the foundation of a power tool:
- Viewer-first layout (large 3D viewport; minimal chrome) supports inspection tasks well
- Right-side task panels (Simple DFM Review, Vision Analysis, Fusion Analysis, Report Template Builder) are conceptually clean: "inputs → controls → generate"
- DraftLint's empty state is straightforward ("Ready for scan… Select a drawing file… then run scan")
- The breadth of capabilities (3D + 2D + DFM + AI analysis + reporting) is genuinely differentiated
The UX Problems: Why Good Capabilities Feel Scattered¶
Problem 1: Navigation Is Icon-Heavy, Not Task-Driven¶
Current state: Multiple vertical icon rails (left and right) plus hamburger menu; "what is where" is unclear
Why it matters: For complex tools, engineers don't tolerate "hunt-and-peck navigation." Users need clear mental models of where to find reviews, findings, reports, and DraftLint.
What it looks like: A user wants to find their previous review → must click through multiple panels and remember which icon corresponds to reviews
Problem 2: Too Many Primary CTAs Compete at Once¶
Current state: "Create Drawing," "Compare Models," and "Collaborate" all visible in strong orange simultaneously
Why it matters: Users need one "next best action" at a time. Competing CTAs create decision paralysis and make the tool feel cluttered.
What it looks like: A designer opens RapidDraft and sees three equally important actions, unsure which to take first
Problem 3: Implementation Details Leak Into the UI¶
Current state: Labels like "Open CASCADE STEP translator 7.8" appear as file identities
Why it matters: Backend details harm trust and comprehension. Users should see "Part name / file name / revision," not pipeline internals.
Problem 4: Panel Visual Language Is Inconsistent¶
Current state: Mix of dark navy (part profile), white (analysis panels), light beige (comments/reviews). Feels like multiple UI systems stitched together.
Why it matters: Inconsistency erodes perceived quality and professionalism. Enterprise engineering software depends on visual consistency for trust.
Problem 5: Collaboration Features Feel Underpowered¶
Current state: Reviews/Comments panes show "No items yet"; disconnected from viewer and from each other
Why it matters: Review workflows are the core value prop. Without visible engagement signals ("unseen by you," "unresolved counts," "activity," "participant list"), the collaboration layer feels incomplete.
Recommended Fixes (Priority Matrix)¶
P0: Navigation and Page Model¶
| Issue | Fix | Effort | Impact | Acceptance Criteria |
|---|---|---|---|---|
| Icon-heavy navigation; hierarchy unclear | Implement single primary left sidebar with text labels + icons: Home, Workspaces, Files, Reviews, Findings, Reports, DraftLint, Settings | High | High | New users answer "Where am I?" and "Where are Reviews?" in <5 sec; tree test success ≥80% |
| No context breadcrumbs | Add breadcrumbs + workspace indicator | Medium | Medium | Breadcrumb always shows: Project → Review → File |
P0: Single Primary Action Per Context¶
| Issue | Fix | Effort | Impact | Acceptance Criteria |
|---|---|---|---|---|
| Multiple orange CTAs compete | Define one primary CTA per screen. Example: "Generate Draft" (primary) + secondary actions as outlined/ghost or overflow menu | Medium | High | Visual scan shows only one primary-colored CTA per screen |
| Internal tech terms leak into UI | Replace "Open CASCADE STEP translator 7.8" with "Part name / file name / revision"; put technical details in collapsible "Import details" | Low | Medium–High | No backend component names in primary headers |
P0: Review Container as Mission Control¶
| Issue | Fix | Effort | Impact | Acceptance Criteria |
|---|---|---|---|---|
| Reviews/Comments panes are disconnected | Create Review page with tabs: Files, Findings, Comments, Activity, Export | High | High | Review has: participants, status, unresolved count, activity log; comments deep-link to view states |
| No unified findings model | Introduce Findings schema: title, severity, confidence, evidence, linked geometry; state (New → Triaged → Resolved); actions (Accept, Dismiss, Convert to Issue, Assign) | High | High | Every analysis run emits findings into a table; clicking a finding highlights the geometry |
P1: Panel Ergonomics & Consistency¶
| Issue | Fix | Effort | Impact | Acceptance Criteria |
|---|---|---|---|---|
| Inconsistent panel colors (navy, white, beige) | Standardize panel visual language: left sidebar (consistent styling), right task panels (consistent headers with title, help, close buttons) | Medium | Medium–High | All panels follow same header pattern and color scheme |
| Analysis panels feel "expert-only" | Add "guided mode" + "advanced mode"; inline tooltips; "Recommended defaults" badge | Medium | Medium–High | First-time users can run DFM/Vision/Fusion without opening advanced controls |
P1: Output Clarity & Actionability¶
| Issue | Fix | Effort | Impact | Acceptance Criteria |
|---|---|---|---|---|
| Results unclear; where do they go? | Standardize all analysis outputs into Findings: severity, confidence, evidence, actions (Accept, Convert to Issue, Assign) | High | High | Every finding lands in Findings tab; sortable/filterable; each finding can be resolved or converted to an issue |
| Report builder is functional but unclear | Add live preview ("What your report will contain") + template gallery + duplication | Medium | Medium | User can preview before saving; template dependencies are visible |
P2: Trust and Accountability Signals¶
| Issue | Fix | Effort | Impact | Acceptance Criteria |
|---|---|---|---|---|
| Who ran what, when, and how? | Add Provenance: run history, model versions, rule set version, timestamps, audit info in exports | Medium | High | Every export shows "Generated on… using…"; audit trail is available |
| No personal task queue | Add My Work: assigned reviews/findings/issues; filters for "Needs my response" and "Unseen by me" | Medium | High | Users can clear queue; each item links to exact viewer state |
Detailed PRD Requirements¶
Epic A: Navigation and Information Architecture (P0)¶
Goal: Users immediately understand where they are and how to navigate to Files/Reviews/Outputs.
Requirements: - Replace mixed rails + hamburger with one primary left sidebar - Sidebar includes: workspace context (switcher or dropdown), labeled items (icon + text), notifications, profile/settings - Persistent breadcrumbs showing: Project → Review → File - Consistent focus/selection styling across navigation
CoLab precedent: Updated sidebar + workspace switcher + context-adaptive breadcrumbs
Acceptance criteria: - 5-task tree test (Find: Reviews, Findings, Export report, DraftLint, Settings) success ≥80% - Breadcrumb always shows current container hierarchy - New users can locate primary tasks within 2 clicks
Epic B: Review Container as Primary UX Unit (P0/P1)¶
Goal: All generation tools (DFM/Vision/Fusion/DraftLint) feed into consistent review artifact with findings, issues, comments, and exports.
Requirements: - Review object has: - Name, owner, participants, status, due date (optional) - Files list (multi-file supported) - Findings list (AI-generated) - Comments (human) - Activity feed - Export configuration - From any file, create a review in ≤2 clicks - Review exports include all findings/comments/decisions into single package - Unresolved count visible at file and review level
CoLab precedent: Request reviews + Track issues and reviews; multi-file reviews; exportable review record
Acceptance criteria: - Review can be created from file detail page with one button - Export includes all findings/comments with links to geometry
Epic C: Viewer + Panel Ergonomics (P0/P1)¶
Goal: Keep viewer large while making tool panels discoverable, consistent, and non-overwhelming.
Requirements: - Dockable panels: Left (file/part context + properties), Right (task panel: DFM/Vision/Fusion/Report/DraftLint) - Panels are resizable and collapsible; auto-collapse into drawers on narrow screens - Consistent panel header pattern: title, one-line purpose, help icon, close/dock controls - Viewer never shrinks below minimum usable width on 1366px screens
Acceptance criteria: - Panels collapse and expand smoothly - Minimum viewer width maintained on standard displays - Help icon opens contextual documentation
Epic D: Output Modeling - Findings and Issues (P1)¶
Goal: AI output becomes actionable work items, not just text.
Requirements: - Standard Findings schema: - title, severity, confidence, evidence snippet, linked geometry/view - state: New → Triaged → Resolved (or Equivalent) - actions: Accept, Dismiss, Convert to Issue, Assign, Comment - Findings table with sorting (severity, type, status), filtering, saved views - Clicking a finding highlights geometry or opens relevant 2D region - Findings can be bulk-actioned (accept multiple at once)
CoLab precedent: Track issues/reviews with search/filter/sort; file-level unresolved counts; viewer engagement indicators
Acceptance criteria: - Every analysis run emits findings into a table - Findings are sortable/filterable by status, severity, type - Clicking finding jumps to relevant geometry
Proposed Information Architecture¶
Home
├── Workspaces / Projects
│ ├── Files
│ │ ├── File Detail
│ │ │ ├── Create / Join Review
│ │ │ └── Run Analysis (DFM / Vision / Fusion / DraftLint)
│ │ └── Findings (generated by analysis)
│ ├── Reviews
│ │ ├── Review Container
│ │ │ ├── Viewer (3D/2D)
│ │ │ ├── Findings (AI)
│ │ │ ├── Comments (Human)
│ │ │ ├── Activity / Provenance
│ │ │ └── Export / Reports
│ │ └── Issues / Tasks
│ ├── My Work (personal queue)
│ ├── Reports / Templates
│ ├── DraftLint (2D Review mode)
│ └── Settings / Standards
└── Profile / Account
Accessibility Requirements (WCAG 2.2 AA)¶
Focus and Keyboard Support¶
- Visible focus indicators on all interactive controls
- Focus not hidden behind sticky headers/overlays (WCAG 2.2: "Focus Not Obscured")
- All non-native UI widgets (custom menus, drawers, dialogs) follow keyboard interaction guidance (Escape, Tab trapping, predictable focus return)
Target Sizes¶
- Toolbar icons and buttons meet minimum target size expectations (WCAG 2.2: "Target Size Minimum")
Contrast¶
- Dark panels maintain sufficient contrast for text and UI component boundaries
- Non-text contrast for icons/borders used as affordances
Validation¶
- Keyboard-only walkthrough of 5 critical flows: open file, run DFM, open findings, export report, scan in DraftLint
- Automated checks (axe, lighthouse) plus manual checks for viewer toolbars and complex grids
Performance Targets (Core Web Vitals)¶
Target field data at p75: - LCP ≤ 2.5s (largest contentful paint) - INP ≤ 200ms (interaction to next paint; replaces FID) - CLS ≤ 0.1 (cumulative layout shift)
RapidDraft-Specific Tactics¶
- Use skeleton loaders for panels that load results
- Virtualize tables for findings/issues/comments (large datasets likely)
- Add staged progress for long jobs (upload → convert → analyze → render results)
- Defer non-critical sidebar content until user interaction
Phased Implementation Roadmap¶
Phase 1: Navigation + Panel Consistency (Weeks 1–4)¶
Output: Users can navigate clearly; panels are consistent
Deliverables: - Left sidebar with labeled navigation (Home, Workspaces, Files, Reviews, Findings, DraftLint, Settings) - Breadcrumb navigation - Consistent panel headers across all task panels - Color palette / design system update
Effort: High (UI architecture refactor) Impact: High (immediate UX improvement)
Phase 2: Review Container + Findings Model (Weeks 5–8)¶
Output: Review is the primary unit; findings are actionable
Deliverables: - Review page with tabs (Files, Findings, Comments, Activity, Export) - Findings schema and table with sorting/filtering - Integration of DFM/Vision/Fusion/DraftLint outputs into unified Findings - "Convert finding to issue" workflow
Effort: High (data model + UI integration) Impact: High (core workflow redesign)
Phase 3: Output Clarity + Provenance (Weeks 9–12)¶
Output: Every analysis output is transparent and traceable
Deliverables: - "My Work" personal queue - Activity feed and provenance tracking - Enhanced report builder with live preview - Audit trail in exports
Effort: Medium Impact: Medium–High (trust + accountability)
Phase 4: Advanced Ergonomics (Weeks 13–16, Optional)¶
Output: Advanced users get power-user features
Deliverables: - Custom saved views (favorite filtering/sorting) - Bulk actions on findings - Advanced DraftLint comparison (overlay, side-by-side) - Template customization and sharing
Effort: Medium Impact: Medium (polish + power-user stickiness)
Success Metrics¶
Track these after each phase:
- Navigation: Task completion time for "find reviews," "navigate to findings," "export report" (target: <30 sec per task)
- Panel consistency: User perception of "professional/cohesive" UI (qualitative feedback)
- Findings adoption: % of findings that get triaged/resolved (target: >70% within 2 weeks of generation)
- Review engagement: # of participants per review, comment/issue creation rate
- Performance: LCP, INP, CLS at p75 (target: meet Core Web Vitals thresholds)
- User satisfaction: NPS or satisfaction survey post-implementation (target: >50 NPS)