VELORYZ STUDIO / KNOWLEDGE

Testing & accessibility

Test matrix, release gates and accessibility checks.

08 — QA plan, accessibility requirements and release gates

Automated commands

From `VELORYZ-Studio/`:

```sh npm run test:creator npm run test:project-ai npm run test:all npm run build ```

These checks require a compatible Node version and installed project dependencies; do not interpret a source syntax check as a passing production build.

Functional regression matrix

| ID | Scenario | Required result | | --- | --- | --- | | NAV-01 | Open /create/ directly | Welcome screen only, no original guided project panel below | | NAV-02 | Click Start a project | Discover question opens in same tab | | NAV-03 | On each question click top Continue | One screen advances, never a hidden bottom button | | NAV-04 | Click top Back | Previous question restored without data loss | | NAV-05 | Browser Back/Forward | URL and screen remain in sync | | NAV-06 | Return from Review/Edit | Changed answer persists and Review shortcut works | | NAV-07 | Enter a malformed email | Inline validation, no request submitted | | NAV-08 | Submit with valid consent | Actual POST succeeds, server reference shown | | NAV-09 | API fails/timeout | Visible error; draft preserved and safe retry possible | | NAV-10 | Completed request | Draft cleared, reference remains visible on confirmation | | LIB-01 | Open studio template library | 12 concepts and category/search controls present | | LIB-02 | Open external reference | Original credited URL opens without granting a license | | COST-01 | Business versus app budget | Distinct category baselines, clear nonbinding language | | COST-02 | Published promotion | Accurate eligibility/expiry, opt-in only, no auto deduction | | AI-01 | Widget on Welcome | Button visible at bottom-left and opens | | AI-02 | Wizard route transition | Panel remounts after navigation; chat history retained in page memory | | AI-03 | Working model | Verified live answer from the intended provider, not another persona | | AI-04 | No model/503 | Honest offline status and labeled prewritten guidance | | AI-05 | Contact-data boundary | Email input not automatically included in AI context | | PORTAL-01 | Confirmation copy | Reference distinguished from separate invitation | | PORTAL-02 | 12-hour copy | No guaranteed 12-hour delivery or immediate token | | PORTAL-03 | Visit /client/ | Portal expects a real team-issued invite, not project reference | | ISO-01 | Create route | Legacy expert launcher, old planner, footer and offer widgets hidden | | ISO-02 | Exit Create | Other routes still show their original navigation/widgets as intended |

Browser/device matrix

Desktop: recent Chrome, Edge, Firefox and Safari where supported. Mobile: iOS Safari and Android Chrome; test portrait and landscape, virtual keyboard open, 320/375/430 px widths, touch hit areas, 200% and 400% zoom. Test keyboard-only navigation and a screen reader on at least one desktop/mobile setup.

Accessibility review

Target WCAG 2.2 AA where applicable. Examine:

  • SC 2.4.11: focused elements not entirely obscured by sticky top bar — https://www.w3.org/WAI/WCAG22/Understanding/focus-not-obscured-minimum
  • Focus visibility; contrast; logical tab order; name/role/value; form errors and assistance; landmark structure; reduced motion; accessible dialog/panel behavior.
  • AI messages have a live log and their provenance labeled; do not move keyboard focus unexpectedly on rerender.
  • No icon-only button without a meaningful accessible name.

Release gate

Ready to deploy only when code review and automated tests pass, asset version hashes are refreshed, release notes list changes and rollback target is recorded.

Ready to declare live only after checking the deployed `/create/` HTML, stylesheet and scripts, visiting in a real browser, completing a test enquiry against real backend, checking Project AI provider availability, and confirming no legacy content below the wizard.

Ready to promise email delivery only after a real test invitation shows accepted provider status and recipient inbox receipt for a verified sending domain. Even then, do not guarantee a specific time without operational capacity and monitoring.

What was and was not verified at code authoring

Source syntax, markup assertions and simulated journeys may pass without proving hosting has deployed. A test stub returning `{reference:'VZ-EXAMPLE-123'}` is not an actual production enquiry. Model-binding stubs are not live AI evidence. Record real browser screenshots/results in a release checklist before changing these statuses.

Let’s talk
Make your project