USearch

Overview of USearch

What it is

USearch is a fast open-source search and clustering engine for vectors and arbitrary objects. It is implemented as a single C++11 header library and provides bindings for C++, C, Python, JavaScript, Rust, Java, Objective-C, Swift, C#, Go, and Wolfram. It uses the HNSW (Hierarchical Navigable Small World) algorithm for approximate nearest neighbor search.

Key features

• 10x faster HNSW implementation than FAISS (as stated in the README).
• Simple and extensible single C++11 header library.
• Trusted by Google and databases like ClickHouse and DuckDB.
• SIMD-optimized and user-defined metrics with JIT compilation.
• Hardware-agnostic support for bf16, e5m2, and i8 half-precision and quarter-precision types.
• Ability to view large indexes from disk without loading into RAM.
• Heterogeneous lookups, renaming/relabeling, and on-the-fly deletions.

Use cases

• Semantic search and joins.
• Near-real-time clustering and sub-clustering for tens or millions of clusters.
• Genomics and chemistry applications using binary Tanimoto and Sorensen coefficients.
• GIS and custom metrics for composite embeddings from multiple AI models.
• Hybrid full-text and semantic search.

Getting started

Install the Python package:pip install usearchExample usage in Python:pythonimport numpy as npfrom usearch.index import Indexindex = Index(ndim=3)vector = np.array([0.2, 0.6, 0.4])index.add(42, vector)matches = index.search(vector, 10)assert matches[0].key == 42assert matches[0].distance < 0.001For C++:cpp#include using namespace unum::usearch;auto index = index_32_t::make(3);index.add(42, {0.2, 0.6, 0.4});auto results = index.search({0.2, 0.6, 0.4}, 10);For other languages, see the repository README.

License and pricing

Licensed under Apache-2.0. Open source (no pricing stated).

Based on the project's public README and repository metadata.