Automated testing

Useful when the thing under test behaves differently for traffic it thinks is automated:

• Sites that serve a different page to bots
• Your own bot detection, checked against something that actually evades it
• Geography-dependent behavior, through a proxy with a matching timezone
• Layout at screen sizes real devices have

import httpx
from playwright.sync_api import sync_playwright

PROXY = "http://USERNAME:[email protected]:8000"
KEY = "rb_live_YOUR_KEY"

resp = httpx.get(
    "https://browser.rayobyte.com/connect",
    params={
        "os": "windows",
        "headless": "true",
        "proxy": PROXY,
        "screen_width_min": 1280,
        "screen_height_min": 720,
    },
    headers={"x-api-key": KEY},
    timeout=120,
)
if resp.status_code != 200:
    raise SystemExit(f"{resp.status_code}: {resp.text}")
cdp_url = resp.text.strip()
session_id = resp.headers["x-session-id"]

try:
    with sync_playwright() as p:
        browser = p.chromium.connect_over_cdp(cdp_url)
        context = browser.contexts[0]
        page = context.pages[0] if context.pages else context.new_page()
        page.goto("https://quotes.toscrape.com/login")

        assert page.title() == "Quotes to Scrape"
        assert page.locator("input[type=submit]").is_visible()
finally:
    httpx.post(
        "https://browser.rayobyte.com/api/browser/close",
        json={"browserId": session_id},
        headers={"x-api-key": KEY},
        timeout=30,
    )

Fingerprints differ between runs

Each session draws a different device profile, so values such as the screen size and the font list are not the same twice. A test that asserts on navigator.userAgent or on a pixel-exact screenshot fails on the second run for reasons that have nothing to do with your application.

Pin what you need: screen_width_min, screen_height_min, os and the browser version parameters all constrain the profile. For a fully fixed device, email [email protected].

CI concurrency

5 concurrent browsers is the account limit, not a per-pipeline one. Test suites running in parallel across branches share it, and the sixth job gets 429 Concurrent browser limit reached rather than a queue. Cap your parallelism below 5, or email [email protected].