tommy_bordas
2025 · VizProof

VizProof - Visual Regression Testing SaaS

FastifyPrismaPostgreSQLReactVitePlaywrightpixelmatchOpenAI

Key Results

3 engines
Browser coverage
Chromium, Firefox, WebKit
pixelmatch
Diff engine
Pixel-by-pixel screenshot comparison
Web + API + CLI + WP
Ecosystem
SaaS platform connected to a WordPress plugin

Challenges

  • Capturing stable web renderings across browsers
  • Comparing screenshots with a reliable diff engine
  • Operating VRT runs at scale across API
  • UI
  • and CLI
  • Balancing product UX
  • quotas
  • and billing constraints

Solutions

  • pnpm monorepo split into server/web/cli/shared packages
  • Playwright capture pipeline (Chromium
  • Firefox
  • WebKit)
  • pixelmatch + pngjs diffing with AI classification
  • Fastify API with Prisma/PostgreSQL and JWT/refresh auth

VizProof - Visual Regression Testing SaaS

Context

VizProof is a Visual Regression Testing (VRT) SaaS. It captures web screenshots, compares visual renderings between runs, and detects UI regressions before release. Production URL: https://vizproof.com.

Product scope

The goal was not only to generate screenshots, but to build a complete platform:

  • web interface to configure and review runs
  • reliable backend orchestration for capture and diff
  • CI/CD integration through a dedicated CLI
  • WordPress integration via plugin

Architecture

The product runs in a pnpm monorepo structured as:

  • packages/server: Fastify API + Prisma + PostgreSQL
  • packages/web: React + Vite + TypeScript
  • packages/cli: @vizproof/cli for pipeline integration
  • packages/shared: shared types and helpers
  • wordpress-plugin/vizproof-timeline: WordPress integration plugin

This setup keeps business logic consistent across API/frontend/CLI while speeding up delivery.

VRT pipeline

VizProof core workflow:

  1. run trigger (manual, scheduled, webhook)
  2. Playwright captures across engines (Chromium, Firefox, WebKit)
  3. rendering stabilization before screenshots
  4. baseline vs current comparison with pixelmatch/pngjs
  5. AI-assisted diff classification and summaries

The output is directly usable by product, QA, and engineering teams to validate UI changes quickly.

Security and operations

  • JWT auth with in-memory access token + httpOnly refresh cookie
  • plan limits and usage quota management
  • webhook integrations (GitHub, Lemon Squeezy, etc.)
  • integrated notifications and alerts

The backend is designed for frequent runs while keeping incident/debug visibility high.

Conclusion

VizProof is a full VRT platform, not a screenshot script. It combines robust backend architecture, multi-browser Playwright execution, reliable visual diffing, and a product experience tailored for real CI/CD workflows.