What problem does it solve?
Visual Ralph turns an intended frontend UI reference into a measurable, fail/pass-driven implementation loop instead of subjective visual tweaking, so teams can reliably reach pixel-level match.
Core Features & Use Cases
- Visual evidence-driven iteration: Uses Visual Ralph verdict output and pixel diff as measurable guidance for each editing cycle.
- Reference-first workflow: Builds a visual truth source from an approved mockup, static reference, or a URL-derived baseline (without routing new URL cloning through the deprecated $web-clone).
- Design-system extraction: Ensures the outcome produces reusable tokens/components (colors, spacing, typography, radii, elevation, variants) rather than one-off screenshots.
- Guardrailed completion criteria: Enforces screenshot commands, viewport/state recording, verdict thresholds (score >= 90), and repo build/lint/test checks.
Quick Start
Ask the assistant to run the visual-ralph workflow to implement your approved UI reference and iteratively verify it with Visual Ralph verdict and pixel-diff until it matches (score >= 90), then extract reusable design tokens into the repository.