Testing

In one line: Royal Glow follows a testing diamond — most tests at the integration boundary — across 13 layers, all on free tiers at $0/month. CI gates get stricter from dev to prod, and Lighthouse requires performance ≥ 95 with accessibility/best-practices/SEO = 100.

Philosophy

Royal Glow uses a testing diamond (not the traditional pyramid) because the integration boundary is where most bugs live in a fullstack app:

        ┌─────┐
       │ E2E  │  ← Critical paths only (golden path verification)
      ┌┴───────┴┐
     │Integration│ ← MOST tests live here (API + DB + services)
      └┬───────┬┘
       │ Unit  │  ← Pure logic only (no mocking infrastructure)
        └─────┘
       ┌───────┐
      │AI-Powered│ ← Autonomous regression detection (Meticulous)
       └───────┘

Core principles:

Tests are production code — same quality, same review rigor, same TypeScript strictness
Test behaviour, not implementation — if you refactor and tests break, the tests were wrong
Deterministic or nothing — flaky tests get quarantined immediately, fixed within 24h, or deleted
Fast feedback loop — unit tests < 10s, integration < 60s, E2E < 5 min (local)

Testing Layers

Layer	Tool	Scope	Hard Gate?
Static Analysis	TypeScript 5.8 strict	Type safety	✅ CI blocks
Linting + Format	Biome + Ultracite	Code style, bugs, a11y	✅ CI blocks
Unit	Vitest	Pure functions, business logic	✅ CI blocks
Integration	Vitest + Neon test branch	API routes, DB queries	✅ CI blocks
Component	Vitest + React Testing Library	UI components	✅ CI blocks
End-to-End	Playwright	Critical user journeys	✅ CI blocks
Visual Regression	Meticulous AI	Pixel-perfect UI diff	⚠️ Warn (review)
Accessibility	axe-core + Lighthouse	WCAG 2.1 AA	✅ CI blocks
Performance	Lighthouse CI	Core Web Vitals	✅ CI blocks
Load & Stress	k6	Concurrent user simulation	✅ CI blocks (pprd+)
Security (SAST)	Trivy + Semgrep	Dependency vulns + code	✅ CI blocks
Security (DAST)	OWASP ZAP	Runtime vulnerability scan	✅ CI blocks (pprd+)
Synthetic Monitoring	BetterStack + Checkly	Production health	🚨 Alert

Tool Decisions

Locked-In Stack

STATIC:      TypeScript strict + Biome + Ultracite
UNIT:        Vitest + @faker-js/faker
COMPONENT:   Vitest + React Testing Library + MSW
INTEGRATION: Vitest + Neon test branch + MSW
E2E:         Playwright (5 browsers) + axe-core
VISUAL:      Meticulous AI (zero-effort, session replay)
PERF:        Lighthouse CI + Unlighthouse + @next/bundle-analyzer
LOAD:        k6 (local execution)
SECURITY:    Trivy + Semgrep + OWASP ZAP + Socket.dev
MONITORING:  BetterStack (uptime) + Checkly (synthetic) + Sentry (errors)
MUTATION:    Stryker (quarterly)
TOTAL COST:  $0/month (all free tiers)

Tools Rejected

Cypress

Slower than Playwright, no WebKit, worse Next.js App Router support.

Jest

Vitest is a drop-in replacement, 10x faster with Bun.

Chromatic

Requires Storybook — RGSS doesn’t use Storybook.

Datadog

31/host/mo — BetterStack + Sentry covers all needs at

Snyk

Trivy + Socket.dev covers the same ground for free.

Unit Testing (Vitest)

What to unit test:

Module	What to Test
`packages/business/invoicing/gst.ts`	GST back-calculation, CGST/SGST split
`packages/business/loyalty/gems.ts`	Earning rules, redemption rules, balance validation
`packages/business/offers/discount.ts`	Percentage, flat, combo discount calculations
`packages/business/utils/currency.ts`	Indian number formatting (₹1,00,000.00)
`packages/business/utils/date.ts`	IST timezone handling, DD/MM/YYYY formatting
`packages/business/booking/booking-number.ts`	Booking number format generation
`packages/business/invoicing/invoice-number.ts`	Financial year calculation, invoice number format

Integration Testing

Integration tests run against the Neon test branch — a real database with seeded fixtures. The test branch is wiped and reseeded before each CI run.

// tests/integration/api/bookings.test.ts
describe('POST /api/bookings', () => {
  it('creates a booking and returns 201', async () => {
    const res = await fetch('/api/bookings', {
      method: 'POST',
      body: JSON.stringify({ serviceIds: ['haircut-basic'], date: '2026-06-15', time: '10:00' }),
    })
    expect(res.status).toBe(201)
    const json = await res.json()
    expect(json.success).toBe(true)
    expect(json.data.bookingNumber).toMatch(/^BK-RS-/)
  })
})

E2E Testing (Playwright)

Critical paths covered:

Homepage loads → Book Now opens dialog → 4-step booking flow → submission
Sign in with Google → onboarding → profile complete
Admin dashboard → approve booking → mark complete → invoice generated
Customer views booking history → cancels upcoming booking
Admin creates membership → customer views membership page

# Run E2E tests
bun run test:e2e

# Run with UI (headed mode)
bunx playwright test --ui

Accessibility Testing

axe-core integrated into Playwright E2E tests — runs on every page
Lighthouse CI — Accessibility score = 100 required to merge
Keyboard navigation — booking flow completable via keyboard only
Screen reader — manual testing with VoiceOver (macOS) before each release

Performance Testing

Lighthouse CI

// lighthouserc.json
{
  "ci": {
    "assert": {
      "assertions": {
        "categories:performance": ["error", { "minScore": 0.95 }],
        "categories:accessibility": ["error", { "minScore": 1 }],
        "categories:best-practices": ["error", { "minScore": 1 }],
        "categories:seo": ["error", { "minScore": 1 }]
      }
    }
  }
}

k6 Load Testing

Target: 50 concurrent users on pprd. Runs on every PR to pprd.

// tests/load/booking-flow.js
export const options = {
  vus: 50,
  duration: '2m',
  thresholds: {
    http_req_duration: ['p(95)<500'],  // 95% of requests under 500ms
    http_req_failed: ['rate<0.01'],    // Less than 1% failure rate
  },
}

Security Testing

Tool	What It Catches	When
Trivy	Known CVEs in dependencies	Every PR
Semgrep	SQL injection, XSS patterns, hardcoded secrets	Every PR
OWASP ZAP	Runtime XSS, CSRF, injection	PR to pprd+
Socket.dev	Malicious npm packages, typosquatting	Every PR

CI Pipeline Integration

Gates are cumulative — each stage adds checks on top of the previous one.

PR to dev
PR to test
PR to pprd
PR to prod

Lint + Format (Biome)
Type check (tsc —noEmit)
Unit tests (Vitest)

All dev checks + Integration tests + Playwright E2E + Lighthouse CI

All test checks + k6 load test + OWASP ZAP + smoke test suite

Coverage Requirements

Module	Statements	Branches	Functions
`packages/business/`	95%	90%	95%
`apps/web/src/lib/`	95%	90%	95%
`apps/web/src/app/api/`	90%	85%	90%

Commands

bun run test              # Unit tests (excludes integration + E2E)
bun run test:unit         # Unit tests only
bun run test:integration  # Integration tests (requires Neon test branch)
bun run test:e2e          # Playwright E2E
bun run test:coverage     # Unit tests with coverage report

Meticulous AI — Visual Regression

Meticulous records real user sessions on pprd, then replays them as visual regression tests on every PR. Zero test code required — it learns from real usage. What Meticulous catches that Playwright doesn’t:

Subtle font rendering changes
1px layout shifts after CSS changes
Colour/spacing regressions in components you didn’t touch
Responsive breakpoint issues on real device viewports

Overview

Product

Features

Engineering

System design

Engineering standards

API reference

Application surface

Platform

Quality & security

Operations

Reference

Testing

Testing

Philosophy

Testing Layers

Tool Decisions

Locked-In Stack

Tools Rejected

Unit Testing (Vitest)

Integration Testing

E2E Testing (Playwright)

Accessibility Testing

Performance Testing

Lighthouse CI

k6 Load Testing

Security Testing

CI Pipeline Integration

Coverage Requirements

Commands

Meticulous AI — Visual Regression

​Testing

​Philosophy

​Testing Layers

​Tool Decisions

​Locked-In Stack

​Tools Rejected

​Unit Testing (Vitest)

​Integration Testing

​E2E Testing (Playwright)

​Accessibility Testing

​Performance Testing

​Lighthouse CI

​k6 Load Testing

​Security Testing

​CI Pipeline Integration

​Coverage Requirements

​Commands

​Meticulous AI — Visual Regression

Testing

Philosophy

Testing Layers

Tool Decisions

Locked-In Stack

Tools Rejected

Unit Testing (Vitest)

Integration Testing

E2E Testing (Playwright)

Accessibility Testing

Performance Testing

Lighthouse CI

k6 Load Testing

Security Testing

CI Pipeline Integration

Coverage Requirements

Commands

Meticulous AI — Visual Regression