To be more consistent with the rest of the gem, would you mind changing this to use a JS heredoc? This is helpful because many editors support code highlighting in the language specified:

Original:

VS:

Subl:

Code:

I haven't tried the code out yet, but I'm looking forward to this feature!

it makes sense, will change, thanks!

I was feeling like this arguments list was a little crowded, and also noticed this function could be incorrectly called without raising an exception: one of css or xpath is required, but not handled. What do you think about something like this:

      def wait_for_selector(css: nil, xpath: nil, timeout: 3000, interval: 100)
        raise ArgumentError.new("Provide :css or :xpath, but not both") if css && xpath
        raise ArgumentError.new("Provide :css or :xpath") unless css || xpath

        evaluate_func(<<~JS, css || xpath, !!xpath, timeout, interval, awaitPromise: true)
          function(selector, isXpath, timeout, interval) {
            var attempts = 0;

The guard clauses protect the arguments from being called incorrectly, and also simplify the resolution to isXpath, since it can now only be one or the other.

well, it looks like too much for one method to be responsible for selector/XPath and also for validations.
However, I see the confusing point here as well. In this case, I think we should use simple methods for more clear API, like those:

      def wait_for_selector(css, **options)
        wait_for_element(css: css, **options)
      end
      def wait_for_xpath(xpath, **options)
        wait_for_element(xpath: xpath, **options)
      end

The point to merge two ways of element-find logic was a follow a DRY principle for a common JS function, especially this part:

              var element = isXpath
                ? document.evaluate(selector, document, null, XPathResult.FIRST_ORDERED_NODE_TYPE, null).singleNodeValue
                : document.querySelector(selector);

therefore, I think, we can find a way to refactor it with a private function by passing a js function, gimme see closer, will do an update soon.

Nice one, yes, I think this is a better direction. Consider wait_for_css rather than wait_for_selector as a selector could generically mean css or xpath. It makes the expected arg more clear.