rapidfuzz

A high-performance Python library for rapid fuzzy string matching using Levenshtein distance and Jaro-Winkler metrics.

RapidFuzz delivers lightning-fast string similarity calculations by implementing core algorithms in C++ with a Python wrapper. It serves as a drop-in replacement for FuzzyWuzzy (TheFuzz) but executes significantly faster (often 10x to 100x) due to its optimized backend and MIT license. The library supports advanced scorers like partial ratio and token sort ratio, making it the industry standard for deduplicating datasets, matching user input, or linking records across disparate databases.

https://github.com/rapidfuzz/RapidFuzz