Playwright Test Automator

A specialized model for automated generation of Playwright E2E test scripts based on web application crawl data.

Overview

Playwright Test Automator is built on a RAG-based architecture that combines AI-powered test generation with Playwright's robust browser automation capabilities. The model is designed to generate reliable, maintainable, and efficient end-to-end tests with minimal human intervention.

Features

Automated Crawling: Intelligently crawls web applications to discover interactive elements and page structures
Smart Login Detection: Automatically identifies login forms and credentials fields
Robust Selector Generation: Creates resilient selectors that withstand minor UI changes
Error Handling: Implements comprehensive error handling for reliable test execution
Validation Framework: Integrated with Giskard for model validation and continuous improvement

Use Cases

Generating regression test suites for web applications
Automating QA workflows for small to large web projects
Creating test scripts for complex user journeys
Validating forms and interactive elements across browsers

Requirements

Python 3.11+
Playwright
OpenAI API access
Supabase for knowledge retrieval (optional)

Example Usage

from services.crawler_service import crawler_service

# Crawl a website and generate tests
test_result = await crawler_service.crawl_with_login(
    url="https://example.com/login",
    username="test_user",
    password="password123"
)

# Generate a Playwright test from the crawl data
test_code = await crawler_service.generate_test_from_crawl_data(test_result)

# Write the test to a file
with open("example_test.spec.js", "w") as f:
    f.write(test_code["code"])

Sample Generated Test

// Example E2E Test
// Automated test for Example Application
// Auto-generated by Playwright Test Automator

import { test, expect } from '@playwright/test';

test('Example E2E Test', async ({ page }) => {
  // Configure viewport for better element visibility
  await page.setViewportSize({ width: 1280, height: 800 });
  
  // Helper function for safer element interactions
  async function safeClick(selector, description) {
    // Implementation details...
  }

  // Navigate to the application
  await page.goto('https://example.com/login');
  
  // Login process
  await page.fill("input[type='text'][name='username']", "test_user");
  await page.fill("input[type='password']", "password123");
  await safeClick("button[type='submit']", "login button");
  
  // Verify successful login
  await expect(page).toHaveURL(/dashboard/);
});

Limitations

Best performance on standard web forms and common UI patterns
May require adjustments for highly dynamic or custom UI frameworks
Requires valid credentials for protected areas of applications

Training Methodology

This model was trained on a diverse set of web applications using a specialized dataset of high-quality test scripts. The training process included:

Web crawling of various application types
Test generation with OpenAI integration
Validation using Giskard framework
Iterative improvements based on execution results

Citation

@misc{playwright-test-automator,
  author = {QA RAG App Team},
  title = {Playwright Test Automator},
  year = {2025},
  publisher = {Hugging Face},
  journal = {HuggingFace Hub},
  howpublished = {\url{https://huggingface.co/playwright-test-automator}}
}