Pyversity is a Python library for fast diversification of search and retrieval results, providing five algorithms (MMR, MSD, DPP, COVER, SSD) to re-rank items balancing relevance and diversity. Its core API is minimal: `diversify(embeddings, scores, k, strategy, diversity)` returns a `DiversificationResult` with selected indices and scores. The library is dependency-light (only NumPy required) and designed for easy integration into retrieval pipelines. The repository also contains a comprehensive benchmark suite evaluating strategies on real-world recommendation datasets (MovieLens-32M, Last.fm, Amazon Video Games, Goodreads), with metrics for both relevance (nDCG, MRR) and diversity (ILAD, ILMD). Benchmark results show DPP offers the best diversity gains with minimal relevance loss. While primarily a software library, the inclusion of reproducible benchmarks, research citations, and systematic evaluation makes it a hybrid application/research repository.