USearch

Smaller & Faster Single-File
Similarity Search & Clustering Engine for Vectors & 🔜 Texts

Spatial • Binary • Probabilistic • User-Defined Metrics
C++11 • Python 3 • JavaScript • Java • Rust • C99 • Objective-C • Swift • C# • Go • Wolfram
Linux • macOS • Windows • iOS • Android • WebAssembly • SQLite

ISCC Foundation Fork -- This is a maintained fork of USearch by the ISCC Foundation, published on PyPI as usearch-iscc. The Python import name remains usearch for compatibility. Install with: pip install usearch-iscc

Fork divergence from upstream:

128-bit key support (Python): Index(ndim=..., key_kind="uuid") for packed 16-byte keys

Multi-index UUID support (Python): Indexes works with both u64 and uuid-keyed shards

NPHD metric (all bindings): Normalized Prefix Hamming Distance for length-prefixed binary vectors

Build: published as usearch-iscc on PyPI with independent release cycle

✅ 10x faster HNSW implementation than FAISS.
✅ Simple and extensible single C++11 header library.
✅ Trusted by giants like Google and DBs like ClickHouse & DuckDB.
✅ SIMD-optimized and user-defined metrics with JIT compilation.
✅ Hardware-agnostic f16 & i8 - half-precision & quarter-precision support.
✅ View large indexes from disk without loading into RAM.
✅ Heterogeneous lookups, renaming/relabeling, and on-the-fly deletions.
✅ Binary Tanimoto and Sorensen coefficients for Genomics and Chemistry applications.
✅ NPHD metric for variable-length binary fingerprint comparison.
✅ Space-efficient point-clouds with uint40_t, accommodating 4B+ size.
✅ Compatible with OpenMP and custom "executors" for fine-grained parallelism.
✅ Semantic Search and Joins.
🔄 Near-real-time clustering and sub-clustering for Tens or Millions of clusters.

Technical Insights and related articles:

Comparison with FAISS

FAISS is a widely recognized standard for high-performance vector search engines. USearch and FAISS both employ the same HNSW algorithm, but they differ significantly in their design principles. USearch is compact and broadly compatible without sacrificing performance, primarily focusing on user-defined metrics and fewer dependencies.

	FAISS	USearch	Improvement
Indexing time ⁰
100 Million 96d `f32`, `f16`, `i8` vectors	2.6 · 2.6 · 2.6 h	0.3 · 0.2 · 0.2 h	9.6 · 10.4 · 10.7 x
100 Million 1536d `f32`, `f16`, `i8` vectors	5.0 · 4.1 · 3.8 h	2.1 · 1.1 · 0.8 h	2.3 · 3.6 · 4.4 x

Codebase length ¹	84 K SLOC	3 K SLOC	maintainable
Supported metrics ²	9 fixed metrics	any metric	extendible
Supported languages ³	C++, Python	10 languages	portable
Supported ID types ⁴	32-bit, 64-bit	32-bit, 40-bit, 64-bit	efficient
Filtering ⁵	ban-lists	any predicates	composable
Required dependencies ⁶	BLAS, OpenMP	-	light-weight
Bindings ⁷	SWIG	Native	low-latency
Python binding size ⁸	~ 10 MB	< 1 MB	deployable

⁰ Tested on Intel Sapphire Rapids, with the simplest inner-product distance, equivalent recall, and memory consumption while also providing far superior search speed. ¹ A shorter codebase of usearch/ over faiss/ makes the project easier to maintain and audit. ² User-defined metrics allow you to customize your search for various applications, from GIS to creating custom metrics for composite embeddings from multiple AI models or hybrid full-text and semantic search. ³ With USearch, you can reuse the same preconstructed index in various programming languages. ⁴ The 40-bit integer allows you to store 4B+ vectors without allocating 8 bytes for every neighbor reference in the proximity graph. ⁵ With USearch the index can be combined with arbitrary external containers, like Bloom filters or third-party databases, to filter out irrelevant keys during index traversal. ⁶ Lack of obligatory dependencies makes USearch much more portable. ⁷ Native bindings introduce lower call latencies than more straightforward approaches. ⁸ Lighter bindings make downloads and deployments faster.

Base functionality is identical to FAISS, and the interface must be familiar if you have ever investigated Approximate Nearest Neighbors search:

# pip install usearch

import numpy as np
from usearch.index import Index

index = Index(ndim=3)               # Default settings for 3D vectors
vector = np.array([0.2, 0.6, 0.4])  # Can be a matrix for batch operations
index.add(42, vector)               # Add one or many vectors in parallel
matches = index.search(vector, 10)  # Find 10 nearest neighbors

assert matches[0].key == 42
assert matches[0].distance <= 0.001
assert np.allclose(index[42], vector, atol=0.1) # Ensure high tolerance in mixed-precision comparisons

More settings are always available, and the API is designed to be as flexible as possible. The default storage/quantization level is hardware-dependant for efficiency, but bf16 is recommended for most modern CPUs.

index = Index(
    ndim=3, # Define the number of dimensions in input vectors
    metric='cos', # Choose 'l2sq', 'ip', 'haversine' or other metric, default = 'cos'
    dtype='bf16', # Store as 'f64', 'f32', 'f16', 'i8', 'b1'..., default = None
    connectivity=16, # Optional: Limit number of neighbors per graph node
    expansion_add=128, # Optional: Control the recall of indexing
    expansion_search=64, # Optional: Control the quality of the search
    multi=False, # Optional: Allow multiple vectors per key, default = False
)

128-bit Keys (UUID Mode)

By default, USearch uses 64-bit unsigned integer keys. This fork adds support for 128-bit keys via key_kind="uuid", allowing you to pack structured identifiers (e.g. content hashes, chunk pointers) directly into the key.

import numpy as np
from usearch.index import Index

# Create an index with 128-bit keys
index = Index(ndim=128, metric='cos', key_kind='uuid')

# Keys are 16-byte values: single keys as bytes, batches as numpy V16 arrays
batch_size = 1000
keys = np.empty(batch_size, dtype='V16')
vectors = np.random.randn(batch_size, 128).astype(np.float32)

for i in range(batch_size):
    body   = i.to_bytes(8, 'big')          # 8 bytes: content identity
    offset = (i * 16).to_bytes(4, 'big')   # 4 bytes: chunk offset
    size   = (1024 + i).to_bytes(4, 'big')  # 4 bytes: chunk size
    keys[i] = body + offset + size          # 16 bytes total

index.add(keys, vectors)
matches = index.search(vectors[0], count=5)

for match in matches:
    print(match.key, match.distance)  # match.key is bytes(16)

# Single-key operations use bytes(16)
single_key = keys[0].tobytes()
index.contains(single_key)  # bool
index.get(single_key)       # np.ndarray or None
index.remove(single_key)

# Save/load preserves key kind; mismatched load raises ValueError
index.save('index.usearch')
restored = Index.restore('index.usearch')  # auto-detects uuid mode

Note: Auto-generated keys are not supported in uuid mode — you must always pass explicit keys to add().

NPHD Metric (Normalized Prefix Hamming Distance)

NPHD is a built-in distance metric for comparing length-prefixed binary vectors. Each vector's first byte stores the data length in bytes. The metric computes the Hamming distance over the common prefix of two vectors and normalizes by the shorter vector's bit count, returning a value in [0.0, 1.0].

This is useful for content identification systems like ISCC where binary fingerprints may have variable-length prefixes. Previously this required a custom Numba @cfunc metric (~500MB of dependencies) and change_metric() hacks after every load()/view(). The native metric eliminates both.

import numpy as np
from usearch.index import Index, MetricKind, ScalarKind

# Vector layout: [length_byte, data_byte_0, data_byte_1, ..., padding...]
# ndim is total size in bits, including the length byte.
ndim = 264  # 33 bytes = 1 length byte + up to 32 data bytes

index = Index(ndim=ndim, metric=MetricKind.NPHD, dtype=ScalarKind.B1)

def make_vector(length, data_bytes):
    """Build a length-prefixed binary vector."""
    vec = np.zeros(ndim // 8, dtype=np.uint8)
    vec[0] = length
    vec[1:1 + len(data_bytes)] = data_bytes
    return vec

a = make_vector(4, [0xAA, 0xBB, 0xCC, 0xDD])
b = make_vector(4, [0xAA, 0xBB, 0xCC, 0x00])

index.add(0, a)
index.add(1, b)

matches = index.search(a, 2)
print(matches[0].key, matches[0].distance)  # 0, 0.0
print(matches[1].key, matches[1].distance)  # 1, ~0.15625

# Save/load preserves the metric — no change_metric() needed
index.save("nphd_index.usearch")
restored = Index.restore("nphd_index.usearch")
assert str(restored.metric_kind) == "MetricKind.NPHD"

Key details:

Only valid with dtype=ScalarKind.B1 (binary vectors).
The length byte encodes the number of data bytes (not bits), excluding itself.
When vectors have different lengths, only the common prefix is compared.
A length byte of 0 yields distance 0.0 (no data to compare).

Serialization & Serving `Index` from Disk

USearch supports multiple forms of serialization:

Into a file defined with a path.
Into a stream defined with a callback, serializing or reconstructing incrementally.
Into a buffer of fixed length or a memory-mapped file that supports random access.

The latter allows you to serve indexes from external memory, enabling you to optimize your server choices for indexing speed and serving costs. This can result in 20x cost reduction on AWS and other public clouds.

index.save("index.usearch")

index.load("index.usearch")
view = Index.restore("index.usearch", view=True, ...)

other_view = Index(ndim=..., metric=...)
other_view.view("index.usearch")

Exact vs. Approximate Search

Approximate search methods, such as HNSW, are predominantly used when an exact brute-force search becomes too resource-intensive. This typically occurs when you have millions of entries in a collection. For smaller collections, we offer a more direct approach with the search method.

from usearch.index import search, MetricKind, Matches, BatchMatches
import numpy as np

# Generate 10'000 random vectors with 1024 dimensions
vectors = np.random.rand(10_000, 1024).astype(np.float32)
vector = np.random.rand(1024).astype(np.float32)

one_in_many: Matches = search(vectors, vector, 50, MetricKind.L2sq, exact=True)
many_in_many: BatchMatches = search(vectors, vectors, 50, MetricKind.L2sq, exact=True)

If you pass the exact=True argument, the system bypasses indexing altogether and performs a brute-force search through the entire dataset using SIMD-optimized similarity metrics from SimSIMD. When compared to FAISS's IndexFlatL2 in Google Colab, USearch may offer up to a 20x performance improvement:

faiss.IndexFlatL2: 55.3 ms.
usearch.index.search: 2.54 ms.

User-Defined Metrics

While most vector search packages concentrate on just two metrics, "Inner Product distance" and "Euclidean distance", USearch allows arbitrary user-defined metrics. This flexibility allows you to customize your search for various applications, from computing geospatial coordinates with the rare Haversine distance to creating custom metrics for composite embeddings from multiple AI models, like joint image-text embeddings. You can use Numba, Cppyy, or PeachPy to define your custom metric even in Python:

from numba import cfunc, types, carray
from usearch.index import Index, MetricKind, MetricSignature, CompiledMetric

ndim = 256

@cfunc(types.float32(types.CPointer(types.float32), types.CPointer(types.float32)))
def python_inner_product(a, b):
    a_array = carray(a, ndim)
    b_array = carray(b, ndim)
    c = 0.0
    for i in range(ndim):
        c += a_array[i] * b_array[i]
    return 1 - c

metric = CompiledMetric(pointer=python_inner_product.address, kind=MetricKind.IP, signature=MetricSignature.ArrayArray)
index = Index(ndim=ndim, metric=metric, dtype=np.float32)

Similar effect is even easier to achieve in C, C++, and Rust interfaces. Moreover, unlike older approaches indexing high-dimensional spaces, like KD-Trees and Locality Sensitive Hashing, HNSW doesn't require vectors to be identical in length. They only have to be comparable. So you can apply it in obscure applications, like searching for similar sets or fuzzy text matching, using GZip compression-ratio as a distance function.

Filtering and Predicate Functions

Sometimes you may want to cross-reference search-results against some external database or filter them based on some criteria. In most engines, you'd have to manually perform paging requests, successively filtering the results. In USearch you can simply pass a predicate function to the search method, which will be applied directly during graph traversal. In Rust that would look like this:

let is_odd = |key: Key| key % 2 == 1;
let query = vec![0.2, 0.1, 0.2, 0.1, 0.3];
let results = index.filtered_search(&query, 10, is_odd).unwrap();
assert!(
    results.keys.iter().all(|&key| key % 2 == 1),
    "All keys must be odd"
);

Memory Efficiency, Downcasting, and Quantization

Training a quantization model and dimension-reduction is a common approach to accelerate vector search. Those, however, are only sometimes reliable, can significantly affect the statistical properties of your data, and require regular adjustments if your distribution shifts. Instead, we have focused on high-precision arithmetic over low-precision downcasted vectors. The same index, and add and search operations will automatically down-cast or up-cast between f64_t, f32_t, f16_t, i8_t, and single-bit b1x8_t representations. You can use the following command to check, if hardware acceleration is enabled:

$ python -c 'from usearch.index import Index; print(Index(ndim=768, metric="cos", dtype="f16").hardware_acceleration)'
> sapphire
$ python -c 'from usearch.index import Index; print(Index(ndim=166, metric="tanimoto").hardware_acceleration)'
> ice

In most cases, it's recommended to use half-precision floating-point numbers on modern hardware. When quantization is enabled, the "get"-like functions won't be able to recover the original data, so you may want to replicate the original vectors elsewhere. When quantizing to i8_t integers, note that it's only valid for cosine-like metrics. As part of the quantization process, the vectors are normalized to unit length and later scaled to [-127, 127] range to occupy the full 8-bit range. When quantizing to b1x8_t single-bit representations, note that it's only valid for binary metrics like Jaccard, Hamming, etc. As part of the quantization process, the scalar components greater than zero are set to true, and the rest to false.

USearch uint40_t support

Using smaller numeric types will save you RAM needed to store the vectors, but you can also compress the neighbors lists forming our proximity graphs. By default, 32-bit uint32_t is used to enumerate those, which is not enough if you need to address over 4 Billion entries. For such cases we provide a custom uint40_t type, that will still be 37.5% more space-efficient than the commonly used 8-byte integers, and will scale up to 1 Trillion entries.

`Indexes` for Multi-Index Lookups

For larger workloads targeting billions or even trillions of vectors, parallel multi-index lookups become invaluable. Instead of constructing one extensive index, you can build multiple smaller ones and view them together.

from usearch.index import Indexes

multi_index = Indexes(
    indexes=[index_a, index_b],  # Merge in-memory shards
    paths=["shard_a.usearch", "shard_b.usearch"],  # Or load from disk
    view=False,
    threads=0,
)
multi_index.search(query_vectors, 10)

Indexes supports both u64 and uuid key kinds. The key kind is auto-detected from the first merged shard or path, or can be set explicitly:

# Auto-detect from shards
indexes = Indexes([uuid_index_a, uuid_index_b])

# Auto-detect from paths
indexes = Indexes(paths=["uuid_shard.usearch"])

# Explicit key kind
indexes = Indexes(key_kind="uuid")
indexes.merge(uuid_index)

# Incremental loading
indexes = Indexes()
indexes.merge_path("shard.usearch")

Clustering

Once the index is constructed, USearch can perform K-Nearest Neighbors Clustering much faster than standalone clustering libraries, like SciPy, UMap, and tSNE. Same for dimensionality reduction with PCA. Essentially, the Index itself can be seen as a clustering, allowing iterative deepening.

clustering = index.cluster(
    min_count=10, # Optional
    max_count=15, # Optional
    threads=..., # Optional
)

# Get the clusters and their sizes
centroid_keys, sizes = clustering.centroids_popularity

# Use Matplotlib to draw a histogram
clustering.plot_centroids_popularity()

# Export a NetworkX graph of the clusters
g = clustering.network

# Get members of a specific cluster
first_members = clustering.members_of(centroid_keys[0])

# Deepen into that cluster, splitting it into more parts, all the same arguments supported
sub_clustering = clustering.subcluster(min_count=..., max_count=...)

The resulting clustering isn't identical to K-Means or other conventional approaches but serves the same purpose. Alternatively, using Scikit-Learn on a 1 Million point dataset, one may expect queries to take anywhere from minutes to hours, depending on the number of clusters you want to highlight. For 50'000 clusters, the performance difference between USearch and conventional clustering methods may easily reach 100x.

Joins, One-to-One, One-to-Many, and Many-to-Many Mappings

One of the big questions these days is how AI will change the world of databases and data management. Most databases are still struggling to implement high-quality fuzzy search, and the only kind of joins they know are deterministic. A join differs from searching for every entry, requiring a one-to-one mapping banning collisions among separate search results.

Exact Search	Fuzzy Search	Semantic Search ?
Exact Join	Fuzzy Join ?	Semantic Join ??

Using USearch, one can implement sub-quadratic complexity approximate, fuzzy, and semantic joins. This can be useful in any fuzzy-matching tasks common to Database Management Software.

men = Index(...)
women = Index(...)
pairs: dict = men.join(women, max_proposals=0, exact=False)

Read more in the post: Combinatorial Stable Marriages for Semantic Search 💍

Functionality

By now, the core functionality is supported across all bindings. Broader functionality is ported per request. In some cases, like Batch operations, feature parity is meaningless, as the host language has full multi-threading capabilities and the USearch index structure is concurrent by design, so the users can implement batching/scheduling/load-balancing in the most optimal way for their applications.

	C++ 11	Python 3	C 99	Java	JavaScript	Rust	Go	Swift
Add, search, remove	✅	✅	✅	✅	✅	✅	✅	✅
Save, load, view	✅	✅	✅	✅	✅	✅	✅	✅
User-defined metrics	✅	✅	✅	❌	❌	✅	❌	❌
Batch operations	❌	✅	❌	✅	✅	❌	❌	❌
Filter predicates	✅	❌	✅	❌	❌	✅	❌	✅
Joins	✅	✅	❌	❌	❌	❌	❌	❌
Variable-length vectors	✅	❌	❌	❌	❌	❌	❌	❌
4B+ capacities	✅	❌	❌	❌	❌	❌	❌	❌

Application Examples

USearch + UForm + UCall = Multimodal Semantic Search

AI has a growing number of applications, but one of the coolest classic ideas is to use it for Semantic Search. One can take an encoder model, like the multi-modal UForm, and a web-programming framework, like UCall, and build a text-to-image search platform in just 20 lines of Python.

from ucall import Server
from uform import get_model, Modality
from usearch.index import Index

import numpy as np
import PIL as pil

processors, models = get_model('unum-cloud/uform3-image-text-english-small')
model_text = models[Modality.TEXT_ENCODER]
model_image = models[Modality.IMAGE_ENCODER]
processor_text = processors[Modality.TEXT_ENCODER]
processor_image = processors[Modality.IMAGE_ENCODER]

server = Server()
index = Index(ndim=256)

@server
def add(key: int, photo: pil.Image.Image):
    image = processor_image(photo)
    vector = model_image(image)
    index.add(key, vector.flatten(), copy=True)

@server
def search(query: str) -> np.ndarray:
    tokens = processor_text(query)
    vector = model_text(tokens)
    matches = index.search(vector.flatten(), 3)
    return matches.keys

server.run()

Similar experiences can also be implemented in other languages and on the client side, removing the network latency. For Swift and iOS, check out the ashvardanian/SwiftSemanticSearch repository.

A more complete demo with Streamlit is available on GitHub. We have pre-processed some commonly used datasets, cleaned the images, produced the vectors, and pre-built the index.

Dataset	Modalities	Images	Download
Unsplash	Images & Descriptions	25 K	HuggingFace / Unum
Conceptual Captions	Images & Descriptions	3 M	HuggingFace / Unum
Arxiv	Titles & Abstracts	2 M	HuggingFace / Unum

USearch + RDKit = Molecular Search

Comparing molecule graphs and searching for similar structures is expensive and slow. It can be seen as a special case of the NP-Complete Subgraph Isomorphism problem. Luckily, domain-specific approximate methods exist. The one commonly used in Chemistry is to generate structures from SMILES and later hash them into binary fingerprints. The latter are searchable with binary similarity metrics, like the Tanimoto coefficient. Below is an example using the RDKit package.

from usearch.index import Index, MetricKind
from rdkit import Chem
from rdkit.Chem import AllChem

import numpy as np

molecules = [Chem.MolFromSmiles('CCOC'), Chem.MolFromSmiles('CCO')]
encoder = AllChem.GetRDKitFPGenerator()

fingerprints = np.vstack([encoder.GetFingerprint(x) for x in molecules])
fingerprints = np.packbits(fingerprints, axis=1)

index = Index(ndim=2048, metric=MetricKind.Tanimoto)
keys = np.arange(len(molecules))

index.add(keys, fingerprints)
matches = index.search(fingerprints, 10)

That method was used to build the "USearch Molecules", one of the largest Chem-Informatics datasets, containing 7 billion small molecules and 28 billion fingerprints.

USearch + POI Coordinates = GIS Applications

Similar to Vector and Molecule search, USearch can be used for Geospatial Information Systems. The Haversine distance is available out of the box, but you can also define more complex relationships, like the Vincenty formula, that accounts for the Earth's oblateness.

from numba import cfunc, types, carray
import math

# Define the dimension as 2 for latitude and longitude
ndim = 2

# Signature for the custom metric
signature = types.float32(
    types.CPointer(types.float32),
    types.CPointer(types.float32))

# WGS-84 ellipsoid parameters
a = 6378137.0  # major axis in meters
f = 1 / 298.257223563  # flattening
b = (1 - f) * a  # minor axis

def vincenty_distance(a_ptr, b_ptr):
    a_array = carray(a_ptr, ndim)
    b_array = carray(b_ptr, ndim)
    lat1, lon1, lat2, lon2 = a_array[0], a_array[1], b_array[0], b_array[1]
    L, U1, U2 = lon2 - lon1, math.atan((1 - f) * math.tan(lat1)), math.atan((1 - f) * math.tan(lat2))
    sinU1, cosU1, sinU2, cosU2 = math.sin(U1), math.cos(U1), math.sin(U2), math.cos(U2)
    lambda_, iterLimit = L, 100
    while iterLimit > 0:
        iterLimit -= 1
        sinLambda, cosLambda = math.sin(lambda_), math.cos(lambda_)
        sinSigma = math.sqrt((cosU2 * sinLambda) ** 2 + (cosU1 * sinU2 - sinU1 * cosU2 * cosLambda) ** 2)
        if sinSigma == 0: return 0.0  # Co-incident points
        cosSigma, sigma = sinU1 * sinU2 + cosU1 * cosU2 * cosLambda, math.atan2(sinSigma, cosSigma)
        sinAlpha, cos2Alpha = cosU1 * cosU2 * sinLambda / sinSigma, 1 - (cosU1 * cosU2 * sinLambda / sinSigma) ** 2
        cos2SigmaM = cosSigma - 2 * sinU1 * sinU2 / cos2Alpha if not math.isnan(cosSigma - 2 * sinU1 * sinU2 / cos2Alpha) else 0  # Equatorial line
        C = f / 16 * cos2Alpha * (4 + f * (4 - 3 * cos2Alpha))
        lambda_, lambdaP = L + (1 - C) * f * (sinAlpha * (sigma + C * sinSigma * (cos2SigmaM + C * cosSigma * (-1 + 2 * cos2SigmaM ** 2)))), lambda_
        if abs(lambda_ - lambdaP) <= 1e-12: break
    if iterLimit == 0: return float('nan')  # formula failed to converge
    u2 = cos2Alpha * (a ** 2 - b ** 2) / (b ** 2)
    A = 1 + u2 / 16384 * (4096 + u2 * (-768 + u2 * (320 - 175 * u2)))
    B = u2 / 1024 * (256 + u2 * (-128 + u2 * (74 - 47 * u2)))
    deltaSigma = B * sinSigma * (cos2SigmaM + B / 4 * (cosSigma * (-1 + 2 * cos2SigmaM ** 2) - B / 6 * cos2SigmaM * (-3 + 4 * sinSigma ** 2) * (-3 + 4 * cos2SigmaM ** 2)))
    s = b * A * (sigma - deltaSigma)
    return s / 1000.0  # Distance in kilometers

# Example usage:
index = Index(ndim=ndim, metric=CompiledMetric(
    pointer=vincenty_distance.address,
    kind=MetricKind.Haversine,
    signature=MetricSignature.ArrayArray,
))

Integrations & Users

ClickHouse: C++, docs.
DuckDB: post.
ScyllaDB: Rust, presentation.
TiDB & TiFlash: C++, announcement.
YugaByte: C++.
Google: UniSim, RetSim paper.
MemGraph: C++, announcement.
LanternDB: C++, Rust, docs.
LangChain: Python and JavaScript.
Microsoft Semantic Kernel: Python and C#.
GPTCache: Python.
Sentence-Transformers: Python docs.
Pathway: Rust.
Vald: GoLang.

Citations

@software{Vardanian_USearch_2023,
doi = {10.5281/zenodo.7949416},
author = {Vardanian, Ash},
title = {{USearch by Unum Cloud}},
url = {https://github.com/unum-cloud/usearch},
version = {2.24.0},
year = {2023},
month = oct,
}

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distributions

No source distribution files available for this release.See tutorial on generating distribution archives.

Built Distributions

If you're not sure about the file name format, learn more about wheel file names.

The dropdown lists show the available interpreters, ABIs, and platforms. Enable javascript to be able to filter the list of wheel files.

usearch_iscc-2.24.6-cp314-cp314-win_arm64.whl (430.4 kB view details)

Uploaded Aug 3, 2026 CPython 3.14Windows ARM64

usearch_iscc-2.24.6-cp314-cp314-win_amd64.whl (424.7 kB view details)

Uploaded Aug 3, 2026 CPython 3.14Windows x86-64

usearch_iscc-2.24.6-cp314-cp314-musllinux_1_2_x86_64.whl (2.9 MB view details)

Uploaded Aug 3, 2026 CPython 3.14musllinux: musl 1.2+ x86-64

usearch_iscc-2.24.6-cp314-cp314-musllinux_1_2_aarch64.whl (2.8 MB view details)

Uploaded Aug 3, 2026 CPython 3.14musllinux: musl 1.2+ ARM64

usearch_iscc-2.24.6-cp314-cp314-manylinux_2_26_x86_64.manylinux_2_28_x86_64.whl (2.8 MB view details)

Uploaded Aug 3, 2026 CPython 3.14manylinux: glibc 2.26+ x86-64manylinux: glibc 2.28+ x86-64

usearch_iscc-2.24.6-cp314-cp314-manylinux_2_26_aarch64.manylinux_2_28_aarch64.whl (2.7 MB view details)

Uploaded Aug 3, 2026 CPython 3.14manylinux: glibc 2.26+ ARM64manylinux: glibc 2.28+ ARM64

usearch_iscc-2.24.6-cp314-cp314-macosx_11_0_arm64.whl (577.2 kB view details)

Uploaded Aug 3, 2026 CPython 3.14macOS 11.0+ ARM64

usearch_iscc-2.24.6-cp314-cp314-macosx_10_15_x86_64.whl (609.3 kB view details)

Uploaded Aug 3, 2026 CPython 3.14macOS 10.15+ x86-64

usearch_iscc-2.24.6-cp314-cp314-macosx_10_15_universal2.whl (1.1 MB view details)

Uploaded Aug 3, 2026 CPython 3.14macOS 10.15+ universal2 (ARM64, x86-64)

usearch_iscc-2.24.6-cp313-cp313-win_arm64.whl (419.0 kB view details)

Uploaded Aug 3, 2026 CPython 3.13Windows ARM64

usearch_iscc-2.24.6-cp313-cp313-win_amd64.whl (412.0 kB view details)

Uploaded Aug 3, 2026 CPython 3.13Windows x86-64

usearch_iscc-2.24.6-cp313-cp313-musllinux_1_2_x86_64.whl (2.9 MB view details)

Uploaded Aug 3, 2026 CPython 3.13musllinux: musl 1.2+ x86-64

usearch_iscc-2.24.6-cp313-cp313-musllinux_1_2_aarch64.whl (2.8 MB view details)

Uploaded Aug 3, 2026 CPython 3.13musllinux: musl 1.2+ ARM64

usearch_iscc-2.24.6-cp313-cp313-manylinux_2_26_x86_64.manylinux_2_28_x86_64.whl (2.8 MB view details)

Uploaded Aug 3, 2026 CPython 3.13manylinux: glibc 2.26+ x86-64manylinux: glibc 2.28+ x86-64

usearch_iscc-2.24.6-cp313-cp313-manylinux_2_26_aarch64.manylinux_2_28_aarch64.whl (2.7 MB view details)

Uploaded Aug 3, 2026 CPython 3.13manylinux: glibc 2.26+ ARM64manylinux: glibc 2.28+ ARM64

usearch_iscc-2.24.6-cp313-cp313-macosx_11_0_arm64.whl (580.0 kB view details)

Uploaded Aug 3, 2026 CPython 3.13macOS 11.0+ ARM64

usearch_iscc-2.24.6-cp313-cp313-macosx_10_13_x86_64.whl (612.1 kB view details)

Uploaded Aug 3, 2026 CPython 3.13macOS 10.13+ x86-64

usearch_iscc-2.24.6-cp313-cp313-macosx_10_13_universal2.whl (1.1 MB view details)

Uploaded Aug 3, 2026 CPython 3.13macOS 10.13+ universal2 (ARM64, x86-64)

usearch_iscc-2.24.6-cp312-cp312-win_arm64.whl (418.9 kB view details)

Uploaded Aug 3, 2026 CPython 3.12Windows ARM64

usearch_iscc-2.24.6-cp312-cp312-win_amd64.whl (412.1 kB view details)

Uploaded Aug 3, 2026 CPython 3.12Windows x86-64

usearch_iscc-2.24.6-cp312-cp312-musllinux_1_2_x86_64.whl (2.9 MB view details)

Uploaded Aug 3, 2026 CPython 3.12musllinux: musl 1.2+ x86-64

usearch_iscc-2.24.6-cp312-cp312-musllinux_1_2_aarch64.whl (2.8 MB view details)

Uploaded Aug 3, 2026 CPython 3.12musllinux: musl 1.2+ ARM64

usearch_iscc-2.24.6-cp312-cp312-manylinux_2_26_x86_64.manylinux_2_28_x86_64.whl (2.8 MB view details)

Uploaded Aug 3, 2026 CPython 3.12manylinux: glibc 2.26+ x86-64manylinux: glibc 2.28+ x86-64

usearch_iscc-2.24.6-cp312-cp312-manylinux_2_26_aarch64.manylinux_2_28_aarch64.whl (2.7 MB view details)

Uploaded Aug 3, 2026 CPython 3.12manylinux: glibc 2.26+ ARM64manylinux: glibc 2.28+ ARM64

usearch_iscc-2.24.6-cp312-cp312-macosx_11_0_arm64.whl (579.9 kB view details)

Uploaded Aug 3, 2026 CPython 3.12macOS 11.0+ ARM64

usearch_iscc-2.24.6-cp312-cp312-macosx_10_13_x86_64.whl (612.1 kB view details)

Uploaded Aug 3, 2026 CPython 3.12macOS 10.13+ x86-64

usearch_iscc-2.24.6-cp312-cp312-macosx_10_13_universal2.whl (1.1 MB view details)

Uploaded Aug 3, 2026 CPython 3.12macOS 10.13+ universal2 (ARM64, x86-64)

usearch_iscc-2.24.6-cp311-cp311-win_arm64.whl (416.8 kB view details)

Uploaded Aug 3, 2026 CPython 3.11Windows ARM64

usearch_iscc-2.24.6-cp311-cp311-win_amd64.whl (405.3 kB view details)

Uploaded Aug 3, 2026 CPython 3.11Windows x86-64

usearch_iscc-2.24.6-cp311-cp311-musllinux_1_2_x86_64.whl (2.9 MB view details)

Uploaded Aug 3, 2026 CPython 3.11musllinux: musl 1.2+ x86-64

usearch_iscc-2.24.6-cp311-cp311-musllinux_1_2_aarch64.whl (2.7 MB view details)

Uploaded Aug 3, 2026 CPython 3.11musllinux: musl 1.2+ ARM64

usearch_iscc-2.24.6-cp311-cp311-manylinux_2_26_x86_64.manylinux_2_28_x86_64.whl (2.8 MB view details)

Uploaded Aug 3, 2026 CPython 3.11manylinux: glibc 2.26+ x86-64manylinux: glibc 2.28+ x86-64

usearch_iscc-2.24.6-cp311-cp311-manylinux_2_26_aarch64.manylinux_2_28_aarch64.whl (2.7 MB view details)

Uploaded Aug 3, 2026 CPython 3.11manylinux: glibc 2.26+ ARM64manylinux: glibc 2.28+ ARM64

usearch_iscc-2.24.6-cp311-cp311-macosx_11_0_arm64.whl (572.5 kB view details)

Uploaded Aug 3, 2026 CPython 3.11macOS 11.0+ ARM64

usearch_iscc-2.24.6-cp311-cp311-macosx_10_9_x86_64.whl (598.4 kB view details)

Uploaded Aug 3, 2026 CPython 3.11macOS 10.9+ x86-64

usearch_iscc-2.24.6-cp311-cp311-macosx_10_9_universal2.whl (1.1 MB view details)

Uploaded Aug 3, 2026 CPython 3.11macOS 10.9+ universal2 (ARM64, x86-64)

usearch_iscc-2.24.6-cp310-cp310-win_arm64.whl (415.6 kB view details)

Uploaded Aug 3, 2026 CPython 3.10Windows ARM64

usearch_iscc-2.24.6-cp310-cp310-win_amd64.whl (404.2 kB view details)

Uploaded Aug 3, 2026 CPython 3.10Windows x86-64

usearch_iscc-2.24.6-cp310-cp310-musllinux_1_2_x86_64.whl (2.9 MB view details)

Uploaded Aug 3, 2026 CPython 3.10musllinux: musl 1.2+ x86-64

usearch_iscc-2.24.6-cp310-cp310-musllinux_1_2_aarch64.whl (2.7 MB view details)

Uploaded Aug 3, 2026 CPython 3.10musllinux: musl 1.2+ ARM64

usearch_iscc-2.24.6-cp310-cp310-manylinux_2_26_x86_64.manylinux_2_28_x86_64.whl (2.8 MB view details)

Uploaded Aug 3, 2026 CPython 3.10manylinux: glibc 2.26+ x86-64manylinux: glibc 2.28+ x86-64

usearch_iscc-2.24.6-cp310-cp310-manylinux_2_26_aarch64.manylinux_2_28_aarch64.whl (2.7 MB view details)

Uploaded Aug 3, 2026 CPython 3.10manylinux: glibc 2.26+ ARM64manylinux: glibc 2.28+ ARM64

usearch_iscc-2.24.6-cp310-cp310-macosx_11_0_arm64.whl (571.3 kB view details)

Uploaded Aug 3, 2026 CPython 3.10macOS 11.0+ ARM64

usearch_iscc-2.24.6-cp310-cp310-macosx_10_9_x86_64.whl (597.0 kB view details)

Uploaded Aug 3, 2026 CPython 3.10macOS 10.9+ x86-64

usearch_iscc-2.24.6-cp310-cp310-macosx_10_9_universal2.whl (1.1 MB view details)

Uploaded Aug 3, 2026 CPython 3.10macOS 10.9+ universal2 (ARM64, x86-64)

File details

Details for the file usearch_iscc-2.24.6-cp314-cp314-win_arm64.whl.

File metadata

Download URL: usearch_iscc-2.24.6-cp314-cp314-win_arm64.whl
Upload date: Aug 3, 2026
Size: 430.4 kB
Tags: CPython 3.14, Windows ARM64
Uploaded using Trusted Publishing? No
Uploaded via: uv/0.12.1 {"installer":{"name":"uv","version":"0.12.1","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}

File hashes

Hashes for usearch_iscc-2.24.6-cp314-cp314-win_arm64.whl
Algorithm	Hash digest
SHA256	`d01bb955d17dce1dfff2850faa9ce6095908b23689589f831bc9ccf770296f91`
MD5	`0ab8273a3518d4231c4c20b0f47b3313`
BLAKE2b-256	`b5aecf848507af0304536875e509528f97b097f35cb0c067169947cd575c421a`

Algorithm	Hash digest
SHA256	`3920c9031fd9ab7477ac3922ff0726ed1b32a7d156b3ce38cc4e7dea730d450e`
MD5	`b5dd7d0393990ac7a1c50f67bc9f09c0`
BLAKE2b-256	`248e7553575cee987d84cd3f7e5f1287d744dadb80cf543b5c45b3b2638508b7`

Algorithm	Hash digest
SHA256	`1ca778e7b6b1518c25947d25e11231a9f3546bbad0f3ebd55e917817048ce6b1`
MD5	`29c2407537911e4f981cd2035eb580cd`
BLAKE2b-256	`7ef4704c814228eda8c16b2b3efad587cc27d29407fb800a3cf93e900f444475`

Algorithm	Hash digest
SHA256	`d8379c41c4d24cad7efc9510b99f05708c778d7f1e7bfb7a4926a3c98198ffb8`
MD5	`e436f9f051806e6521ea6efecb10ca5d`
BLAKE2b-256	`9f3388442dec8f5c24df5d08cf2055a440c2fe79ce89086cb6566011b2a96cf7`

Algorithm	Hash digest
SHA256	`5927103b083ac2aefe27478d9f67960c8c40c100d7d4c299bf28145ebd186567`
MD5	`63fa8eb14644a5fa50805f32e4e89517`
BLAKE2b-256	`4ea0581a718ba6e12faf36d32cf8d2b32ed3193ed65490ae63c84fcc3da1e9cc`

Algorithm	Hash digest
SHA256	`603556d461fc203bfdc2cceb7794767ba35e9a6a8010b92e1e3dd6dd1766dc13`
MD5	`0525bbe6e8f18ffa643805dc6ac89d3d`
BLAKE2b-256	`441e5e8957c87458dbe66a6b15cbd2b457fdafc8ec7f8ed5fbb28aed14bb9bbb`

Algorithm	Hash digest
SHA256	`8d3dc7159d50009daa71f72e1a74b2b607337f5843d3bc6219ce15625e792964`
MD5	`83a1682536e0949bc9980ed63471c84d`
BLAKE2b-256	`a01c0a179739de5422ae4602dfb14a24b1636fe13ef0c437ce84b348af368906`

Algorithm	Hash digest
SHA256	`ee5132dc1fb14c9b7d105eed6f523f24dec8e1f68b2817eb92a6d956a1571ce2`
MD5	`6e706ed4ec0ce626c8b792c0ec78af7a`
BLAKE2b-256	`7ff3f72cf23d7a970160b3faed88f548f2947b78a01e279732846f2e5a81f084`

Algorithm	Hash digest
SHA256	`8162482934f7e320a770172f31c0baa4b18c5f566834e3c9f5dcf4688c7e651c`
MD5	`6058673b5acc78f87824c20599ae4102`
BLAKE2b-256	`7c6b588ea27364faf880f57ec6660a00eb5f575ae164752f70348f6351cc228d`

Algorithm	Hash digest
SHA256	`681882535f3a5b5677bee54eb9f7e4328113efc6f7d8bc9909c221a1cdf97662`
MD5	`4f9ae694b8c8ac1a93bd673480d5181d`
BLAKE2b-256	`53fcab9d0df4fd9cb38388a49b33108257d60effc67cec0ae812d3c09c9e5259`

Algorithm	Hash digest
SHA256	`7d754e13b1cccaf6510dd79b85630591821fa1bff253101296ac253c7489f36d`
MD5	`763619cddd8c21936e0c4f656260c42b`
BLAKE2b-256	`359a31642e13383da005b16e28a3aeb40b9d5f45490376603cfb334da2309b5a`

Algorithm	Hash digest
SHA256	`548cbf1a9bcd809700fbd85d208611ff12d23695d600bda7d781eb9531ef511b`
MD5	`b7b4e6045bf624d58b781eef3dab53ce`
BLAKE2b-256	`010edc9e7eeadee650d271980dc7abba5811d8d9bccc419fc6df3ce1604e726d`

Algorithm	Hash digest
SHA256	`1343be5846ca0819b22a6e5bdcf4522893ed543bc2edb4cf11751c99cf3544a2`
MD5	`8e142e5d67d301db8a35c07031667c53`
BLAKE2b-256	`67cb6caaeeb17149187498dde64c78527c60df134c2bc240ddbc9ed2cad9c9d2`

Algorithm	Hash digest
SHA256	`e9b467c192fd7c91483f681116c15748af826b4d140b9820cd3bc11e8caaf3c6`
MD5	`d0fbfd5ea79de4e21cbf5df786b18f11`
BLAKE2b-256	`4cb8ec9e490010596290e1e10c1700cfcc5e1759bd4b805bed5250fb84cc497d`

Algorithm	Hash digest
SHA256	`c01727e9bd70a3f314c78490d0e0e1bca3f6666c1d7bb233cf23e7b63f27e316`
MD5	`4b81e3c64b01816077a15729286b8614`
BLAKE2b-256	`b308902f7a70eb9b3a5a2605e3fc9f28df20d9bbd79def73cc5ef520de2aaae2`

Algorithm	Hash digest
SHA256	`db1f64acf72c366b8c48ab0882fe7ace904ee7516c5bf697d1c0df4753bc2010`
MD5	`6eb713c0a18a02e38a7c5e4999cfadf2`
BLAKE2b-256	`d380af01df8a1b6c30c0276e5076e4cfb7887811ab28b525d9379596cbb1e55b`

Algorithm	Hash digest
SHA256	`d8b613446c7f4160ff969f3124fbe6c2165000a902eb0f3b9e56d9f1435afdc4`
MD5	`9be96a19e5aea3e6f88ffc787869510b`
BLAKE2b-256	`dd31cdde9f09c48b649bc38d98927d4fbb5d46b57fb9efa6ce73af280c1f9b74`

Algorithm	Hash digest
SHA256	`7e7bec5cdc178793d26bcc81781d1a8dd8ca83c72449f419e8b5424222da5b42`
MD5	`c43b0d3d73356e0d1a064c90450f427d`
BLAKE2b-256	`36a71bff8dc0e7f860e699c1c358e95eca31ef33d37fe8f6842f70b3b25b07e5`

Algorithm	Hash digest
SHA256	`12fe2592a8a47ec482c28e3222a00ca04e0fec814bf9e74bb73e4f74348caffb`
MD5	`99a2a2bd0c8582d7eb7d57308bd25fcb`
BLAKE2b-256	`c3bdb2041cc55b0c5f7d1b14ce9df09720ccd3287e9d971911cb6e3da22c17fd`

Algorithm	Hash digest
SHA256	`384dc4787f3d3f57eaf99c6ace774ce4972076bc681c7bff40e516ddd2d6b66b`
MD5	`284487d2af1e66cb20d76dabe4f59d1d`
BLAKE2b-256	`413430a36615b81b755e8b63bc7be3858e5d1fef6d24d967e391aec702cd7921`

Algorithm	Hash digest
SHA256	`7e9a753f48afc33eadb9aa67bd515899a33355c9c23510033f78eba22c6cf86b`
MD5	`a1f0aff52923c0bc860b7df2bf8d79c9`
BLAKE2b-256	`2d5ed122ded86e79434f3a20ba95e6a0170255a26d150dbec746f8ff229407b6`

usearch-iscc 2.24.6

USearch

Smaller & Faster Single-File Similarity Search & Clustering Engine for Vectors & 🔜 Texts

Comparison with FAISS

128-bit Keys (UUID Mode)

NPHD Metric (Normalized Prefix Hamming Distance)

Serialization & Serving Index from Disk

Exact vs. Approximate Search

User-Defined Metrics

Filtering and Predicate Functions

Memory Efficiency, Downcasting, and Quantization

Indexes for Multi-Index Lookups

Clustering

Joins, One-to-One, One-to-Many, and Many-to-Many Mappings

Functionality

Application Examples

USearch + UForm + UCall = Multimodal Semantic Search

USearch + RDKit = Molecular Search

USearch + POI Coordinates = GIS Applications

Integrations & Users

Citations

Project links

Key dates

1 maintainer

Credits

License

Classifiers

Download files

Source Distributions

Built Distributions

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

Smaller & Faster Single-File
Similarity Search & Clustering Engine for Vectors & 🔜 Texts

Serialization & Serving `Index` from Disk

`Indexes` for Multi-Index Lookups

Algorithm	Hash digest
SHA256	`79e69f32c54206432c4b4f61542a6ae84f6e7003b8e4a2a895e91a4dfe6717ce`
MD5	`7016fc67d3d52029758b15917e669820`
BLAKE2b-256	`022217dfe388ea87376803588e3c7cd2ff35d8faac76e89b1f72e3d5c7710825`

Algorithm	Hash digest
SHA256	`d2c754c46d86098646c7975d32f0ec0b8e6dc630ba4801631b20f3ad25eaff4a`
MD5	`2ee559de2306e220a636422a4eb9a29a`
BLAKE2b-256	`af4f6daf52db4ef1bde3bdaae807167f7e34c9ed3409305e81e6dbb193396265`

Algorithm	Hash digest
SHA256	`a5892731ab03cd3f9d32153761c26709e3d9fa1f1dca33e0b9b9f580b42fd13d`
MD5	`0f0a2e6f483cf1029f55b89dc07cd53a`
BLAKE2b-256	`c5e35e51d3a1d10901eafb5e3170f955491213549baa9cf64c3d46f64a4d1f71`

Algorithm	Hash digest
SHA256	`5224c6d139a00b40fcbfe4f44591e7336eab6fe2f408a201620ef31bade3aab6`
MD5	`66a1892b8f3246e702e502725aee8b18`
BLAKE2b-256	`892b9c2f021192eb12e2d72954beeef45627997011c8aeac2eabbf402d55fdc0`

Algorithm	Hash digest
SHA256	`1d9d4a1d40f45bc2a31e257e4aec2c9d2b63ad20c35fbdbe57d297f9a6652fae`
MD5	`7796d62960bb5a6b42d36698c799e50b`
BLAKE2b-256	`457bdc7198031c6c62d6b160ee9692d0109023915c8f587ad942093caf0f45c2`

Algorithm	Hash digest
SHA256	`f3973365071e04f6b09feabf4ada0579779bfbbcd13c99192971e173a7d28165`
MD5	`3f40dde61bcf3a188ed10b2b2bc9316f`
BLAKE2b-256	`804e4ad3cdcd326cf6f2935d6845286b40e3cf99d6ef7c5bcb573a6619693c3c`

Algorithm	Hash digest
SHA256	`ae9c30eb8113a38335b1eba663c5c904540b50e08aa10399d62a6ff2b28bc85a`
MD5	`8ad18aad2ff6ed6cedc7af9abc6ba8fc`
BLAKE2b-256	`320e0ef726d3910e99262f051283f25eb1c9471b2bf5c869a974d070ee88cebb`

Algorithm	Hash digest
SHA256	`162c953eb6416fd7d3f8905b42fc4f888d65c9fdebd3497a43a992d951af343f`
MD5	`c692b3ef9d1745ce0eaa356f189f2dc3`
BLAKE2b-256	`3439b41961bb1713fb45bcfbcae08fb475f4158eeb299db6fac0e2f06a1e5452`

Algorithm	Hash digest
SHA256	`2229274b6699ed9b6f4b155d1d515f7aa8c376f9661e4ce8eb3f0ae2d3472da3`
MD5	`983dfa0bc562bad051b763ca3866757a`
BLAKE2b-256	`ab962af1cb95b94d5ffc2b960ce9e9460354f34a7f395ae0402a6c51a1dc222a`

Algorithm	Hash digest
SHA256	`7744527e6e8304f0f98e1fdcf5827626a836521c23f65bb88bbdc2626ba8c0b9`
MD5	`0197a0f1be4c84e7b668f0fd8ec341a2`
BLAKE2b-256	`ab396998222ee010549f9f6d69a714370bb8ed645d786944e6b5b9676277a1c8`