\b is *the worst*. From MDN:
Note: JavaScript's regular expression engine defines a specific set of
characters to be "word" characters. Any character not in that set is considered
a word break. This set of characters is fairly limited: it consists solely of
the Roman alphabet in both upper- and lower-case, decimal digits, and the
underscore character. Accented characters, such as "é" or "ü" are,
unfortunately, treated as word breaks.
We fix this by matching on whitespace instead, but then need to tweak the
replace() code since that bluntly replaces the entire match (which now includes
whitespace). It all works now and I can happily tab-complete non-ascii names.