Your Playwright tests, checked by Delta-QA. One line per page to check.

Three steps in your tests

1. Create an API key in your Delta-QA workspace and store it in a CI secret. The SDK reads it from DELTAQA_API_KEY; the workspace is the key's.
2. Add a check where your test must verify the page: vision.check("Payment"). Its name matches it with its baseline.
3. Push. On the first run, every test is "New": its first screenshot becomes the baseline. On the second, with the site unchanged, no difference is reported.

import { test } from "@playwright/test";
import { Vision } from "@delta-qa/vision";

const vision = new Vision(); // reads DELTAQA_API_KEY

test("Checkout", async ({ page }) => {
  await vision.open(page, { id: "checkout", name: "Checkout" });
  await page.goto("https://shop.example.com/checkout");
  await vision.check("Payment");
  await vision.close(); // fails the test on an unreviewed difference
});

The SDK exposes only settings Delta-QA applies: none is accepted without effect. The tag and the reference tag are optional; without them, the SDK infers them.

One baseline per branch, no automatic merge

• The tag is inferred from your git branch or the CI, and the reference tag from the PR's target, otherwise main. A new branch is compared with main from its first run, and the batch shows "compared with main".
• "Intended" applies to the tags you tick, "this tag only" by default: other branches see nothing change.
• A test's history filters by environment and by tag; any scan can be set as the baseline, for the tags you tick.
• The tests of one CI run form a batch, shared by machines running in parallel; the batch list reads like your CI.

Create your scenarios with an AI agent: the MCP server