RapidFuzz name-candidate practice — October 28, 2026 Requires existing Python 3.11+ and rapidfuzz==3.14.6. Tested Windows Python 3.13.15. Inspect name-candidates.py, then create a virtual environment and install the pinned dependency: python -m venv .venv .\.venv\Scripts\python.exe -m pip install rapidfuzz==3.14.6 .\.venv\Scripts\python.exe .\name-candidates.py The script contains five fictional records and three fictional queries, compares a subset pair and raw/processed strings, and prints three WRatio shortlists. No file reads/writes, network calls, model calls or merges occur in the script. Installation needs network access; a virtual environment isolates dependencies, not untrusted code. The project's Windows installation guidance describes a Visual C++ runtime prerequisite if DLL loading fails. Recorded results are from our local authored fixture, not model outputs. token_set_ratio is 100 for the short/long-name pair; ordinary ratio is 76.92. For Q01, processed ratio is 100 and raw ratio 32.26. Q03 has three displayed candidates tied at 90; order is not a confidence distinction. automatic_merges remains zero. Python version can differ in your report. Scores use different definitions and are not interchangeable probabilities. Normalization may discard meaningful punctuation or case. limit=3 is an illustrative display choice, not a validated retrieval threshold or complete candidate set. The script has no real-data input interface or production validation. No model extraction, identity verification, accuracy/performance/security evaluation, other OS or real records tested. Source: https://rapidfuzz.github.io/RapidFuzz/Usage/fuzz.html Source: https://rapidfuzz.github.io/RapidFuzz/Usage/process.html Source: https://rapidfuzz.github.io/RapidFuzz/Usage/utils.html Full guide: https://jabsystems.io/ai-anonymous/rapidfuzz-review-the-match