Open-source AI Security Operations Center: alert fusion, LLM-agent triage, MITRE ATT&CK investigation, and a replayable decision ledger for every agent step. Self-hostable, runs with no API keys, MIT licensed. Ships an MCP server for Claude, Cursor and Continue.
Python agentic-aiai-securityalert-triageblue-teamcybersecuritydetection-engineeringincident-responselanggraphllm-agentsmcpmitre-attackmodel-context-protocolpurple-teamsecurity-operationsself-hostedsiemsoarsocthreat-detectionthreat-intelligence
/**
* Visual regression suite for AiSOC's Storybook design system.
*
* Phase 4.7 — Storybook 10 already enumerates every design-system story
* we have under `apps/web/stories/**` and `apps/web/src/**`. Rather
* than maintain a hand-rolled list of routes (which would require
* mocking the entire backend on every release), we build Storybook
* once, serve it as a static site, and screenshot each story id at
* desktop + mobile viewports. The result is a deterministic, fully
* offline visual snapshot of every UI primitive the console exposes.
*
* Baselines live next to this spec in
* ``visual.spec.ts-snapshots/<story-id>-<viewport>.png``. They are
* captured on Linux (the CI runner) so locally-developed snapshots
* may differ slightly in font hinting; the CI image is the source of
* truth and PRs only update the Linux baselines.
*
* To refresh baselines intentionally (e.g. after a deliberate visual
* change), run:
*
* pnpm --filter @aisoc/web exec playwright test \
* --project=visual --update-snapshots
*
* and commit the updated PNGs.
*/
import { promises as fs } from "node:fs";
import path from "node:path";
import { expect, test } from "@playwright/test";
// Mirrors the layout produced by `pnpm build-storybook`. The
// `index.json` (formerly `stories.json`) sidecar emits one entry per
// story with the routable id we need to feed into `iframe.html`.
const STORYBOOK_INDEX = path.resolve(
__dirname,
"..",
"..",
"storybook-static",
"index.json",
);
interface StorybookIndexEntry {
id: string;
name: string;
title: string;
type?: "story" | "docs";
tags?: string[];
}
interface StorybookIndex {
v: number;
entries: Record<string, StorybookIndexEntry>;
}
// Some stories are deliberately animated / inherently non-deterministic
// (loading shimmer, theme transition, toast portals). We skip them so
// the suite reports real regressions instead of frame-timing noise.
//
// Keep the list short and reviewable; entries here should always come
// with a comment explaining *why* the story is excluded so future
// maintainers can revisit the decision.
const SKIP_STORY_IDS = new Set<string>([
// Loading shimmer animates continuously; capture is non-deterministic.
"primitives-skeleton--shimmering",
// Theme story flips colors mid-transition; the static snapshot races
// the transition's first frame and produces flaky diffs.
"foundations-theme--switch",
]);
const VIEWPORTS = [
{ name: "desktop", width: 1440, height: 900 },
{ name: "mobile", width: 390, height: 844 },
] as const;
async function loadIndex(): Promise<StorybookIndex> {
try {
const raw = await fs.readFile(STORYBOOK_INDEX, "utf8");
return JSON.parse(raw) as StorybookIndex;
} catch (err) {
throw new Error(
`Storybook index not found at ${STORYBOOK_INDEX}. ` +
`Run \`pnpm --filter @aisoc/web build-storybook\` before the visual suite. ` +
`(${(err as Error).message})`,
);
}
}
// Stylesheets injected to neutralise non-deterministic effects: caret
// blink, intro animations, CSS transitions, smooth scrolling, etc.
// We apply them after navigation so they take precedence over the
// per-story styles. The values here intentionally over-reach: it is
// fine if a story's "blink" demo no longer blinks in the snapshot, the
// point is that the diff is pixel-stable across runs.
const FREEZE_CSS = `
*, *::before, *::after {
animation-duration: 0s !important;
animation-delay: 0s !important;
animation-iteration-count: 1 !important;
transition-duration: 0s !important;
transition-delay: 0s !important;
caret-color: transparent !important;
}
html { scroll-behavior: auto !important; }
`;
const indexPromise = loadIndex();
test.describe("visual regression — Storybook stories", () => {
// Resolve story IDs once and emit a `test()` per (story, viewport)
// combo. We do this inside `describe` so Playwright shows a clean
// hierarchy in the HTML reporter.
test.beforeAll(async () => {
await indexPromise;
});
for (const viewport of VIEWPORTS) {
test.describe(`viewport=${viewport.name}`, () => {
test(`enumerates and snapshots stories`, async ({ page }) => {
const index = await indexPromise;
const storyIds = Object.values(index.entries)
.filter(
(entry) =>
entry.type !== "docs" && !SKIP_STORY_IDS.has(entry.id),
)
.map((entry) => entry.id)
.sort();
expect(
storyIds.length,
"Storybook index returned zero stories — did the build succeed?",
).toBeGreaterThan(0);
await page.setViewportSize({
width: viewport.width,
height: viewport.height,
});
for (const storyId of storyIds) {
const url = `/iframe.html?id=${encodeURIComponent(
storyId,
)}&viewMode=story`;
await page.goto(url, { waitUntil: "networkidle" });
// Wait for web-fonts to settle so glyph rasterisation matches
// across runs. `document.fonts.ready` is the canonical signal.
await page.evaluate(async () => {
await (document as Document & {
fonts: { ready: Promise<unknown> };
}).fonts.ready;
});
await page.addStyleTag({ content: FREEZE_CSS });
// Brief settle to let the layout reflow after the freeze CSS.
await page.waitForTimeout(50);
await expect(page).toHaveScreenshot(
`${storyId}--${viewport.name}.png`,
{
fullPage: true,
animations: "disabled",
caret: "hide",
// 0.5% pixel-diff tolerance per story is enough to ride
// over GPU-driven anti-aliasing rounding without masking
// real regressions (which are typically 10-40% diffs).
maxDiffPixelRatio: 0.005,
threshold: 0.2,
},
);
}
});
});
}
});