Automation: AI-Powered Game Scraper & PDF Reporter

Stack: Python 3.10, Playwright, Pillow (pHash), FPDF2, Stealth-mode.
The Challenge
Most automation systems fail against online casinos because:
• Lack of DOM: Games are rendered on HTML5 Canvas. There are no buttons in the code — it's just a set of pixels.
• Protection: Aggregators (for example, slotslaunch.com) use Cloudflare Turnstile, which blocks regular bots.
• Chaos of interfaces: Each provider (Pragmatic, Spinomenal, Playtech) has its own menu logic and rules icons.
The Solution
I developed a hybrid system that combines browser automation and intelligent image processing:
1. Bypassing blocks: Using playwright-stealth and dynamic proxies to simulate a real user.
2. Preset system (Rule-based): Instead of unreliable AI searching for coordinates, I implemented a JSON profile architecture. This allows the script to know exactly where to look for the "i" or "Menu" button for each specific provider.
3. Computer Vision (pHash): The script compares hashes of screenshots in real-time. This allows:
• Accurately determining game loading (bypassing Splash Screens).
• Automatically scrolling through the rules pages, stopping when the content starts to duplicate.
4. Report generator: Automatically compiling screenshots into branded PDF files while preserving the hierarchy of providers.
The Result
• Automation of 90% of games from the top 50 global providers.
• Scalability: Adding a new provider takes