Vision 👀 - Give Bots Sight

[mrWh1te]

Some apps uglify their CSS classes with unique names, on each run, making it impractical to use class names or IDs in the selector to reference specific UI elements

perhaps a better solution is an image processing one. For example, generate a screenshot of the page (from top to bottom), then find the element with image processing that recognizes labels or icons, then draw a rectangle around it to return those coordinates to guide the bot

food for thought, some kind of machine learning that guides the "mouse", with small movements on the screen, to the "element" it wants to "click" instead of trying to manually cause a click event on a specific DOM element