Duplicate feature request finder

Each request is cleaned up first: lowercased, filler words like “please” and “add” removed, and common synonyms folded together, so “dark theme”, “night mode” and “dark mode” read as one idea.

Then every pair of requests gets a similarity score. Words that only a few requests use count for more than words that appear everywhere, a short request that sits inside a longer one counts as a match, and letter patterns catch typos and different word forms.

Pairs above the sensitivity you choose are joined into groups. The request that is most similar to the rest of its group becomes the suggested title. That’s usually the clearest wording to keep when you merge.

Merge the duplicates in your own tool and keep each one as a vote, so the count reflects everyone who asked. The guide to duplicate feature requests covers how to merge without losing voters.