🚀 Auto Playwright Python

Automating Playwright steps using ChatGPT(OpenAI) or DeepSeek in Python
This library is inspired by Luc Gagan's auto-playwright node.js library.

Setup

Install the library using pip:

pip install autoplaywright-python

Supported Browsers

Autoplaywright supports all browsers supported by Playwright, including:

• Chromium
• Firefox
• WebKit

Usage

Below are examples demonstrating how to use the library:

▶️ Async Example

import asyncio
from playwright.async_api import async_playwright
from pydantic import BaseModel
from autoplaywright.async_api.openai_client import AutoPlaywrightOpenAI

# Create an instance of the AI client using the GPT-4o-mini model
# Remember to replace 'OPENAI_API_KEY_HERE' with you your actual openai api key
ai = AutoPlaywrightOpenAI(model="gpt-4o-mini", api_key="OPENAI_API_KEY_HERE")


# Define a Pydantic model to store the price of the plan
class Plan(BaseModel):
    price: float


# The main function that performs the sequence of operations in the browser
async def main():
    # Launch Playwright in an asynchronous context
    async with async_playwright() as playwright:
        chromium = playwright.chromium

        # Launch the Chromium browser in non-headless mode (i.e., the browser UI is visible)
        browser = await chromium.launch(headless=False)
        page = await browser.new_page()

        # Open the test e-commerce website
        await page.goto("https://webscraper.io/test-sites/e-commerce/allinone")

        # Click the "Pricing" link in the header of the page using the AI command
        # The AI model performs the task "Click on pricing" on the page
        await ai.auto("Click on pricing", page)

        # Select the most expensive plan and return its price as a Pydantic model (Plan)
        # The AI model analyzes the page and extracts the relevant price
        plan = await ai.auto("Return price of most expensive plan", page, Plan)

        # Print the price of the most expensive plan to the console
        print(plan.price)

        # Close the page
        await page.close()


# Run the main function asynchronously
asyncio.run(main())

▶️ Sync Example

from typing import List

from playwright.sync_api import sync_playwright
from pydantic import BaseModel
from autoplaywright.sync_api.openai_client import AutoPlaywrightOpenAI


# Create an instance of the AI client using the GPT-4o-mini model
# Remember to replace 'OPENAI_API_KEY_HERE' with you your actual openai api key
ai = AutoPlaywrightOpenAI(model="gpt-4o-mini", api_key="OPENAI_API_KEY_HERE")


# Define a Pydantic model for a person structure
class Person(BaseModel):
    first_name: str
    last_name: str
    username: str


# Define a Pydantic model to store the people from the table
class Table(BaseModel):
    people: List[Person]


# The main function that performs the sequence of operations in the browser
def main():
    # Launch Playwright in an asynchronous context
    with sync_playwright() as playwright:
        chromium = playwright.chromium

        # Launch the Chromium browser in non-headless mode (i.e., the browser UI is visible)
        browser = chromium.launch(headless=False)
        page = browser.new_page()

        # Open the test e-commerce website
        page.goto("https://webscraper.io/test-sites/tables")

        # Select people from the table and return data as a Pydantic model (Table)
        # The AI model analyzes the page and extracts the relevant price
        table = ai.auto("Return people from the table", page, Table)

        # People in the table
        print(table.people)

        # Close the page
        page.close()


# Run the main function synchronously
main()

Supported actions

Action

ai.auto("Click on pricing button", page)

Query / Assert

# Define a Pydantic model for a person structure
class Person(BaseModel):
    first_name: str
    last_name: str
    username: str


# Define a Pydantic model to store the people from the table
class Table(BaseModel):
    people: List[Person]

# Select people from the table and return data as a Pydantic model (Table)
# The AI model analyzes the page and extracts the relevant price
table = ai.auto("Return people from the table", page, Table)

# People in the table
print(table.people)

Supported Playwright Actions

Here is a list of supported actions:

• add_init_script
• bring_to_front
• click
• close
• dblclick
• dispatch_event
• evaluate
• fill
• focus
• go_back
• go_forward
• goto
• hover
• press
• reload
• screenshot
• set_content
• set_viewport_size
• tap
• type
• uncheck
• wait_for_event
• wait_for_load_state
• wait_for_selector
• wait_for_timeout
• wait_for_url

Why use Auto Playwright Python?

Implementation

The library optimizes AI query costs by shortening the HTML code before sending the query, reducing the token count and lowering costs associated with OpenAI and DeepSeek models.

Links

• Documentation (soon)
• GitHub

Made with ❤️ by Nohet.