Installing the Browsers

Two separate installs

Playwright is unusual compared to most Python packages, because installing it doesn't fully set it up. uv sync installs the playwright Python package, the code that knows how to talk to a browser, but it does not install any actual browser. That's a second, separate step, run through Playwright's own command line tool:

uv run playwright install chromium

Think of it like the difference between installing psycopg (the Postgres client library, from the pgvector course) and installing Postgres itself. One is a library your code imports, the other is the real program doing the work. Here, playwright (the Python package) is the library, and Chromium (the browser) is the real program.

Why a separate download

Playwright doesn't use whatever browser happens to already be on your machine, your regular Chrome or Firefox. Instead, it downloads its own pinned copies of Chromium, Firefox, and WebKit, matched exactly to the version of the playwright package you have installed. This matters because browser internals change constantly, and Playwright's automation code is written against a specific, known version. Using your everyday browser instead would mean things silently break every time that browser auto-updates.

This course only uses Chromium, the open-source engine behind Google Chrome, which is why every install command in this course says playwright install chromium and not the bare playwright install (which would download all three engines).

Where it goes

On Linux and macOS, downloaded browsers live in ~/.cache/ms-playwright/, in a folder named after the specific browser and build, e.g. chromium-1234. You never need to touch this folder directly, Playwright manages it, but it's worth knowing it exists so "where did that 150 MB go" has an answer.

The code, piece by piece

CACHE_DIR = Path.home() / ".cache" / "ms-playwright"

Path.home() gives your home directory as a Path object, and / is overloaded here to join path segments, this is pathlib's way of building a path without manually gluing strings together with slashes.

def browser_appears_installed() -> bool:
    if not CACHE_DIR.exists():
        return False
    return any(CACHE_DIR.glob("chromium-*"))

.glob("chromium-*") finds every folder in the cache directory whose name starts with chromium-. any(...) returns True the moment it finds one match, without needing to check all of them, a common Python idiom for "does at least one exist."

sys.stdout.flush()

Python buffers what print() writes, holding it briefly before it actually appears, while subprocess.run() writes straight to the terminal, bypassing that buffer entirely. Without flushing first, a print() call just before a subprocess.run() call can end up appearing after the subprocess's own output, out of the order it was actually written in. sys.stdout.flush() forces anything queued up to appear immediately, keeping the output in the order you'd expect.

result = subprocess.run(
    [sys.executable, "-m", "playwright", "install", "chromium"],
    check=False,
)

subprocess.run() runs another program from inside Python and waits for it to finish, exactly as if you'd typed the equivalent command yourself. sys.executable is the path to the exact Python interpreter running this script, using it here (rather than assuming a python command exists on the PATH) makes sure the install runs inside the same uv-managed environment as everything else in this project. check=False means "don't raise an exception automatically if this fails", we check result.returncode ourselves instead, so we can give a clearer error message pointing at this project's root Troubleshooting section.

subprocess.run(
    [sys.executable, "-m", "playwright", "install", "chromium", "--dry-run"],
    check=False,
)

--dry-run asks Playwright to report what it would install (or confirm what's already installed) without downloading anything. It's a safe way to verify the install worked, and it prints useful details like the exact Chrome for Testing version and the install path.