playsmart 0.1.0

Simple and efficient compagnion tool for Playwright. Write reusable and maintainable E2E tests like a human.

Playwright! Playsmart!

End the never ending game of having to manually record, inspect, and update your E2E tests with Playwright.

This chunk of code[...]

page.locator("#dkDj87djDA-reo").fill("[e-mail]")
page.locator("#dflfkZkfAA-reo").fill("$ecr!t")
page.get_by_role("button", name="Log In").click()

will become

from playsmart import Playsmart

smart_hub = Playsmart(
    browser_tab=page,
)

with smart_hub.context("login page"):
    smart_hub.want("fill email input with [e-mail]")
    smart_hub.want("fill password input with $ecr!t")
    smart_hub.want("click on login")

nicer, isn't it?

Get started!

Install Playsmart via PyPI

pip install playsmart

requires Python 3.10+

Before you get started, either:

• export OPENAI_API_KEY
• or set openai_key=... parameter within Playsmart class constructor.

Here's the minimum runnable example:

from playwright.sync_api import sync_playwright
from playsmart import Playsmart

driver = sync_playwright().start()
chrome = driver.chromium.launch(headless=False)
page = chrome.new_page()

page.goto("https://huggingface.co/docs")

smart_hub = Playsmart(
    browser_tab=page
)

with smart_hub.context("docs page"):
    smart_hub.want("click on PEFT doc section")

Caching

We know how painful consuming needlessly tokens can be. That's why Playsmart have a tiny caching layer that helps with keeping LLM hints.

You may at any moment disable the cache for a specific instruction as:

smart_hub.want("click on PEFT doc section", use_cache=False)

A discrete file, named .playsmart.cache will be created. You are encouraged to share this file across your teams! Commit it!

You may choose a filename at your own convenience via the cache_path=... parameter within the Playsmart class constructor.

The 'want' method in a nutshell

Basically, everything revolve around Playsmart.want(...) as you would have already guessed.

There's two types of action you can execute:

A) Immediate action: e.g. I want to click on something B) Deferred action: e.g. How many orders are marked as 'pending'?

For the case A) you should never expect the method to return anything (aside from empty list).

Finally, for the case B) Playsmart will always translate your query to a (or many) usable playwright.Locator.

Here is a solid example for B):

with smart_hub.context("dashboard"):
    locators = smart_hub.want("how many orders are labelled as 'pending'?")

    print(f"we have {locators[0].count()} order(s) pending")

Yet, another one:

with smart_hub.context("dashboard"):
    locators = smart_hub.want("list every fields in the form")

    for locator in locators:
        ... # your logic for each 'input<text/select/...>'

Debug runtime

If you are asking yourself "How did we arrive at that result?", use the handy function context_debug.

from playsmart import context_debug, Playsmart

smart_hub = Playsmart(
    browser_tab=...
)

with context_debug():
    smart_hub.want("click on PEFT doc section")

It will stream a list of detailed events to help you debug your test.

Disclaimer

This (heuristic) software is still at an early stage and has not been battle tested (yet). Although we envision a great future for it, it would be unwise to replace your entire E2E suite with it.

We encourage its incremental adoption and positive feedbacks to help us improve this.

Finally, note that the library is not thread safe.