usearch
Smaller & Faster Single-File Vector Search Engine from Unum
Decision gist · record as of 2026-08-14
Yes. USearch is production-stable (Development Status 5), actively maintained, has no known vulnerabilities, and offers a lightweight alternative to heavier vector search libraries. The permissive Apache-2.0 license and broad platform support (Linux, macOS, Windows) make it suitable for most projects. Install friction is moderate due to compiled wheels, but pre-built binaries are available for current Python versions (3.10+). Verify that the numkong dependency is intentional before deploying.AI-flagged interpretation of the facts on this page — verify before relying
Before you install
- Requires Python 3.10 or later; compiled wheels are platform-specific (Linux, macOS, Windows with various architectures).
- Medium install friction due to compiled wheels across multiple platforms and Python versions (3.10+).
- Active maintenance with recent releases and 4264 repository stars suggest solid ongoing support.
License · maintenance · safety
Apache-2.0 (permissive) — Apache-2.0 permissive license allows commercial and private use with minimal restrictions, making it suitable for most production deployments.
last release 2026-07-10 (35 days) · last repo commit 2026-07-10 · 4,264 stars
0 known vulnerabilities (OSV.dev, 2026-08-14) · 486,107 downloads/mo, #6,392 on PyPI
Alternatives
Verify before relying
pip install usearch
import numpy as np
from usearch.index import Index
index = Index(ndim=3)
vector = np.array([0.2, 0.6, 0.4])
index.add(42, vector)
matches = index.search(vector, 10)- Whether numkong dependency is a typo or an actual required package; fact sheet lists it but it is not mentioned in the description.
- Performance claims (e.g., '10x faster than FAISS') are stated in the description but not independently verified in the fact sheet.
What it is and what it does
USearch is a vector search engine that builds approximate nearest-neighbor indexes using the HNSW algorithm. It is designed as a lightweight, single-file C++11 library with Python bindings and support for custom distance metrics. The package lets you index vectors, search for similar vectors efficiently, and optionally serve large indexes from disk without loading them entirely into memory.
Typical usage involves creating an Index with a specified dimensionality and metric, adding vectors with keys, and then searching for nearest neighbors. It supports various data types (f32, f16, bf16, i8, and others) for memory efficiency, works across Linux, macOS, and Windows, and integrates with numpy. The runtime dependencies are numpy, tqdm, and numkong.
Use it for
- Building semantic search systems that find similar embeddings from language models or vision models in large document or image collections.
- Implementing recommendation engines that match user embeddings to candidate item embeddings in real time.
- Clustering millions of vectors for data analysis, with support for on-disk indexes to reduce memory costs.
- Genomics and chemistry applications using binary Tanimoto and Sorensen coefficients for molecular similarity.
- Hybrid search combining vector similarity with custom filtering predicates or external data structures.
Worth the install?
AI-flagged interpretation of the facts on this page. Verify before relying on it.
Yes.
USearch is production-stable (Development Status 5), actively maintained, has no known vulnerabilities, and offers a lightweight alternative to heavier vector search libraries. The permissive Apache-2.0 license and broad platform support (Linux, macOS, Windows) make it suitable for most projects. Install friction is moderate due to compiled wheels, but pre-built binaries are available for current Python versions (3.10+). Verify that the numkong dependency is intentional before deploying.
Install
usearch on PyPI
Before you install
Medium install friction due to compiled wheels across multiple platforms and Python versions (3.10+). Active maintenance with recent releases and 4264 repository stars suggest solid ongoing support.
Requires Python 3.10 or later; compiled wheels are platform-specific (Linux, macOS, Windows with various architectures).
License in practice
Apache-2.0 permissive license allows commercial and private use with minimal restrictions, making it suitable for most production deployments.
Quickstart
pip install usearch
import numpy as np
from usearch.index import Index
index = Index(ndim=3)
vector = np.array([0.2, 0.6, 0.4])
index.add(42, vector)
matches = index.search(vector, 10)Verify before relying
- Whether numkong dependency is a typo or an actual required package; fact sheet lists it but it is not mentioned in the description.
- Performance claims (e.g., '10x faster than FAISS') are stated in the description but not independently verified in the fact sheet.
Package facts
| License | Apache-2.0 permissive |
| Python support | Supports the current Python release >=3.10 |
| Install friction | Medium. Platform-specific wheel |
| Runtime dependencies | 3 packagesnumpytqdmnumkong |
| Maintenance | Actively maintained 35 days since the last release |
| Last repo commit | |
| First released | |
| Downloads | 486,107 / month, #6,392 on PyPI 30-day window, as of 2026-08-14 |
| Known vulnerabilities | None known OSV.dev, checked 2026-08-14 |
| Classifiers | Development Status :: 5 - Production/StableIntended Audience :: DevelopersIntended Audience :: Information TechnologyNatural Language :: EnglishOperating System :: MacOSOperating System :: Microsoft :: WindowsOperating System :: UnixProgramming Language :: C++Programming Language :: JavaProgramming Language :: JavaScriptProgramming Language :: Objective CProgramming Language :: OtherProgramming Language :: Python :: 3 :: OnlyProgramming Language :: Python :: Implementation :: CPythonProgramming Language :: RustTopic :: Database :: Database Engines/ServersTopic :: Scientific/Engineering :: Artificial IntelligenceTopic :: System :: Clustering |
Evidence: usearch-2.26.0-cp310-cp310-macosx_10_9_universal2.whl; usearch-2.26.0-cp310-cp310-macosx_10_9_x86_64.whl; usearch-2.26.0-cp310-cp310-macosx_11_0_arm64.whl; usearch-2.26.0-cp310-cp310-manylinux_2_26_aarch64.manylinux_2_28_aarch64.whl; usearch-2.26.0-cp310-cp310-manylinux_2_26_x86_64.manylinux_2_28_x86_64.whl; usearch-2.26.0-cp310-cp310-musllinux_1_2_aarch64.whl; usearch-2.26.0-cp310-cp310-musllinux_1_2_x86_64.whl; usearch-2.26.0-cp310-cp310-win_amd64.whl; usearch-2.26.0-cp310-cp310-win_arm64.whl; usearch-2.26.0-cp311-cp311-macosx_10_9_universal2.whl; usearch-2.26.0-cp311-cp311-macosx_10_9_x86_64.whl; usearch-2.26.0-cp311-cp311-macosx_11_0_arm64.whl; usearch-2.26.0-cp311-cp311-manylinux_2_26_aarch64.manylinux_2_28_aarch64.whl; usearch-2.26.0-cp311-cp311-manylinux_2_26_x86_64.manylinux_2_28_x86_64.whl; usearch-2.26.0-cp311-cp311-musllinux_1_2_aarch64.whl; usearch-2.26.0-cp311-cp311-musllinux_1_2_x86_64.whl; usearch-2.26.0-cp311-cp311-win_amd64.whl; usearch-2.26.0-cp311-cp311-win_arm64.whl; usearch-2.26.0-cp312-cp312-macosx_10_13_universal2.whl; usearch-2.26.0-cp312-cp312-macosx_10_13_x86_64.whl
Tags
Let your AI agent find packages like this
Example. Real query, live index.
You found this page by searching. An agent finds it by wishing: SkillFed indexes 14,416 PyPI packages by what they can do, searchable in plain language.
wish › “HNSW indexing”
- usearchUSearch provides approximate nearest-neighbor vector search using…
- hnswlibFast approximate nearest neighbor search using the HNSW algorithm…
- voyagerVoyager performs fast approximate nearest-neighbor searches on…
Give your agent the search over MCP, or paste the wish link into any chat.
More Artificial Intelligence packages
LiteLLM provides a unified Python interface to call 100+ LLM providers (OpenAI, Anthropic, Gemini, Bedrock, Azure, and others) using OpenAI-compatible API format, available as both a Python SDK and a self-hosted AI Gateway proxy server.
Install it if you need to work with multiple LLM providers or want to centralize LLM routing in your organization.
Client library and CLI tool for downloading, uploading, and managing models, datasets, and repositories on the Hugging Face Hub platform.
Install it if you work with Hugging Face Hub models or datasets.
LangChain provides a framework for building agents and LLM-powered applications by composing language models, tools, and memory through a unified API that abstracts over multiple model providers.
hf-xet provides chunk-based deduplication and efficient file transfer for the Hugging Face Hub, enabling faster uploads and downloads of large files with local disk caching.
Tokenizers converts raw text into token sequences for NLP models, with support for training custom vocabularies and using pre-built tokenizers (BPE, WordPiece) optimized for speed via Rust.
Transformers provides a unified framework for loading, fine-tuning, and running state-of-the-art pretrained models across text, vision, audio, video, and multimodal tasks using PyTorch, JAX, or TensorFlow.
Install it if you need to run or train any transformer-based model for NLP, vision, audio, or multimodal tasks.
See also pyspark-hnsw · voyager · cuvs-cu12 · annoy · nmslib · faiss-gpu · hnswlib · simsimd · upstash-vector · nano-vectordb