greek-accentuation.accentuation has a method persistent used for placing accents that are persistent, that is, they stay in place through different inflections as much as is allowed by basic accentuation rules.

The persistent function takes both the unaccented word to be accented and a lemma or base form that IS accented.

The first step is seeing on which syllable the accent is on this base form and what type of accent it is. Note that the position of the accent is determined by the syllable position counting from the left, not the right. The code syllabifies both the word-to-be-accented and the base form. It then works out which three (or fewer) syllable placements are allowed on the word-to-be-accented based on the basic accentuation rules. This is provided by another function possible_accentuations.

Now the first thing that’s tried is whether the exact syllable position and accent type of the base is in the possible accentuations for the word-to-be-accented. If so, we’re done. If not, however, we try changing the accent type from acute to circumflex while keeping it in the same position. If that’s still not allowed, we iterate back, trying to place an acute on each successively later syllable until it’s an accentuation allowed by the basic rules.

However, this algorithm hit a problem with accenting Ἰουδαιων using the base Ἰουδαῖος.

The first thing is tries is Ἰουδαῖων which of course is not permitted so it immediately jumps to an acute on the next position: Ἰουδαιών. However this is incorrect. The bug was that only a change from acute to circumflex was attempted before trying later positions. In this case, the correct thing to do was try an acute in the same position as the original circumflex.

This was an easy addition and results in the correct answer: Ἰουδαίων

You can pip install greek-accentuation==1.0.2. The repo is at https://github.com/jtauber/greek-accentuation.