Endpoint: https://playwright.val.run

AI-powered service that fetches any URL's DOM and generates Playwright-native locators. The AI model does all selector generation — there is zero algorithmic/semantic logic in the code itself. The code only fetches HTML, extracts raw DOM context, and passes it to the model.

Every POST request requires the X-Api-Key header matching the X_API_KEY environment variable set on the val.

curl -X POST https://playwright.val.run \
  -H "Content-Type: application/json" \
  -H "X-Api-Key: your_endpoint_key" \
  -d '{"url": "https://example.com"}'

Requests without a valid key receive 401 Unauthorized.

Set these on the val (all required):

Health check.

{ "status": "ok", "service": "playwright-selector-gen" }

Headers:

Body:

{
  "url": "https://example.com",
  "totalElements": 5,
  "elements": [
    {
      "index": 0,
      "tag": "h1",
      "locators": [
        {
          "method": "getByRole",
          "playwrightCode": "page.getByRole('heading', { name: 'Example Domain' })",
          "confidence": "high"
        },
        {
          "method": "getByText",
          "playwrightCode": "page.getByText('Example Domain', { exact: true })",
          "confidence": "medium"
        }
      ]
    },
    {
      "index": 1,
      "tag": "a",
      "locators": [
        {
          "method": "getByRole",
          "playwrightCode": "page.getByRole('link', { name: 'Learn more' })",
          "confidence": "high"
        },
        {
          "method": "getByText",
          "playwrightCode": "page.getByText('Learn more', { exact: true })",
          "confidence": "medium"
        },
        {
          "method": "locator:css",
          "playwrightCode": "page.locator('a[href=\"https://iana.org/domains/example\"]')",
          "confidence": "medium"
        }
      ]
    }
  ]
}

The AI model follows the official Playwright locator recommendation:

Deep nested selectors are never generated. Each element gets up to 3 locators ranked best → fallback.

POST { url } + X-Api-Key
        │
        ▼
┌─ Fetch target URL HTML ─┐
│                          │
├─ Parse DOM (deno_dom)    │
│                          │
├─ Extract raw elements:   │
│   tag, attributes,       │
│   outerHTML snippet,     │
│   text content           │
│                          │
│   (NO selector logic)    │
└──────────┬───────────────┘
           │
           ▼
┌─ AI Model (Z.AI) ───────┐
│                          │
│  Receives raw DOM data.  │
│  Generates ALL locators. │
│  Ranks by Playwright     │
│  priority. Returns JSON. │
│                          │
└──────────┬───────────────┘
           │
           ▼
     JSON response

The code contains no selector generation logic whatsoever. It extracts raw DOM context (tag names, attributes, outer HTML snippets, text content) and sends it to the AI model. The model produces every locator.

Interactive: a, button, input, select, textarea, form, label, details, summary, dialog

Landmarks: nav, header, footer, main, aside, section, article

Content: img, h1–h6, table, th, td, tr, ul, ol, li, span, p, div

By attribute: Any element with role, data-testid, data-test-id, data-cy, aria-label, tabindex, or contenteditable

Capped at 300 elements per request. Elements are batched in groups of 30 for the AI call.