Playwright export

Upliftr is AI debugging agents that debug your app, read-only by design. When an agent confirms a flow, it saves the check as a plain-English suite. You can compile that suite into a standalone Playwright .spec.ts and drop it into any repo with @playwright/test, no Upliftr runtime, no LLM at runtime.

This neutralizes the most common objection to AI debugging agents, "non-deterministic, not reproducible, locked-in." The answer: let the agents reproduce and root-cause the failure, then run deterministic Playwright at run time when you want it.

Export via the API

curl http://127.0.0.1:8400/api/v1/suites/signup-login.upliftr.yaml/export

The endpoint returns the spec as text/plain. Save it into your Playwright project:

curl http://127.0.0.1:8400/api/v1/suites/signup-login.upliftr.yaml/export \
  -o tests/signup-login.spec.ts

What you get

Each case becomes a test(...); the suite becomes a test.describe(...). Steps become a comment plus a best-effort action; expectations become real Playwright assertions:

// Generated from a Upliftr suite. Portable Playwright spec, no Upliftr
// runtime required. Locators and assertions are best-effort; review before use.
import { test, expect } from '@playwright/test';

test.describe('Signup & Login', () => {
  test('A new user can sign up', async ({ page }) => {
    await page.goto('http://localhost:8500');
    // step: Open the sign-up page
    // step: Enter email and password
    // await page.getByLabel('field').fill('value');
    // step: Click 'Create account'
    await page.getByRole('button', { name: 'Create account' })
      .or(page.getByText('Create account')).first().click();

    // assert: a confirmation 'Account created' is shown
    await expect(page.locator('body')).toContainText('Account created');
  });
});

How checks compile

Deterministic checks map straight to Playwright assertions:

Check	Compiles to
url (path_is / contains)	await expect(page).toHaveURL(...)
visible_text (contains)	await expect(locator).toContainText(...)
visible_text (equals + selector_hint)	await expect(locator).toHaveText(...)
element_visible (selector_hint)	await expect(locator).toBeVisible()
absent_text (any_of)	await expect(body).not.toContainText(...) per string

Assertions with no deterministic check export as a clearly-marked // TODO assert: comment carrying the plain-English text, so a human can finish them. Likewise, steps export their action when the target is obvious (a quoted phrase or a Capitalized run) and a // (no obvious target, fill in the locator) hint otherwise.

When to use it

• You want determinism in CI. Let the agents confirm the check in Upliftr; run the exported spec in your existing Playwright pipeline.
• You're migrating in. Keep your Playwright project; let the Upliftr agents generate first drafts.
• You want zero lock-in. The export is plain @playwright/test, yours to keep, edit, and run with no Upliftr dependency.

Best-effort, review before relying on it

The compiler derives locators and actions from prose, so the output is a strong starting point, not a guaranteed-green spec. Review the generated locators and finish any // TODO assertions before wiring it into a gate.

Related: URL → test generation · Checks reference · CI