@krizaka/a11y-audit

The accessibility guard of every Krizaka product's page audit. It runs axe-core (WCAG 2.0/2.1 A and AA, plus axe's best practices: contrast, names, labels, landmarks, headings) on the pages a Playwright crawl already visits, in dark and light, and holds the result with a ratchet: violations are counted per rule in a baseline file; a new one fails the build, the count only goes down.

npm install --save-dev @krizaka/a11y-audit playwright-core

ESM, Node ≥ 22, test tool only (axe-core is MPL-2.0: never import it from an app — import this package).

import { A11yCollector, auditPage, finish } from "@krizaka/a11y-audit";

const a11y = new A11yCollector("my-app");
// in the crawl, once the page is settled (load, fonts, theme), at the phone and the desktop width:
a11y.add({ route: path, theme }, await auditPage(page));
// at the end:
const { problems, notes } = finish(a11y, { baseline: "a11y-baseline.json" }); // --a11y-update lowers, --a11y-init records

| Export | What | | :-- | :-- | | auditPage(page, { exclude?, tags? }) | axe in the page now; one A11yFinding per failing node (rule, impact, target, help) | | A11yCollector | collects a crawl; a node counts once per rule, route, theme and selector (not per width) | | finish(collector, { baseline, update?, init? }) | compares with a11y-baseline.json, returns { problems, notes }; --a11y-update lowers the baseline (never raises it), --a11y-init records today's counts | | compare, lowered, readBaseline | the pure ratchet |

The baseline is { "app", "rules": { "<rule>": { "impact", "count" } } }, committed at the repository root.