voy

App Listed Dormant

πŸ•ΈοΈπŸ¦€ A WASM vector similarity search written in Rust

Checked Β· Sources

Downloads
Not measured
Dependents
Not applicable
Stars
1,076
Forks
39
Contributors
2
Commits
88

About

From the project README Β·

A WASM vector similarity search engine written in Rust

voy: a vector similarity search engine in WebAssembly

  • Tiny: 75KB gzipped, 69KB brotli.
  • Fast: Create the best search experience for the users. Voy uses k-d tree to index and provide fast search
  • Tree Shakable: Optimize bundle size and enable asynchronous capabilities for modern Web API, such as Web Workers.
  • Resumable: Generate portable embeddings index anywhere, anytime.
  • Worldwide: Designed to deploy and run on CDN edge servers.

🚜 Work in Progress

Voy is under active development. As a result, the API is not stable. Please be aware that there might be breaking changes before the upcoming 1.0 release.

A sneak peek of what we are working on:

  • [ ] Built-in text transformation in WebAssembly: As of now, voy relies on JavaScript libraries like transformers.js to generate text embeddings. See Usage for more detail.
  • [x] Index update: Currently it's required to re-build the index when a resource update occurs.
  • [x] TypeScript support: Due to the limitation of WASM tooling, complex data types are not auto-generated.

Installation

# with npm
npm i voy-search

# with Yarn
yarn add voy-search

# with pnpm
pnpm add voy-search
Read the full README

APIs

class Voy

The Voy class encapsulates an index and exposes all the public methods Voy has to offer.

class Voy {
  /**
   * By instantiating with a resource, Voy will construct the index. If the resource is
   * absent, it will construct an empty index. Calling Voy.index() later on will override
   * the empty index.
   * @param {Resource | undefined} resource
   */
  constructor(resource?: Resource);
  /**
   * Index given resource. Voy.index() is designed for the use case where a Voy instance
   * is instantiated without a resource. It will override the existing index. If you'd like
   * to keep the existing index, you can use Voy.add() to add your resource to the index.
   * @param {Resource} resource
   */
  index(resource: Resource): void;
  /**
   * Search top k results with given query embedding.
   * @param {Float32Array} query: Query Embedding
   * @param {number} k: Number of items in the search result
   * @returns {SearchResult}
   */
  search(query: Float32Array, k: number): SearchResult;
  /**
   * Add given resource to the index.
   * @param {Resource} resource
   */
  add(resource: Resource): void;
  /**
   * Remove given resource from the index.
   * @param {Resource} resource
   */
  remove(resource: Resource): void;
  /**
   * Remove all resources from the index.
   */
  clear(): void;
  /**
   * @returns {number}
   */
  size(): number;
  /**
   * Serialize a Voy instance.
   * @returns {string}
   */
  serialize(): string;
  /**
   * Deserialize a serialized index into a Voy instance.
   * @param {string} serialized_index
   * @returns {Voy}
   */
  static deserialize(serialized_index: string): Voy;
}

interface Resource {
  embeddings: Array<{
    id: string; // id of the resource
    title: string; // title of the resource
    url: string; // url to the resource
    embeddings: number[]; // embeddings of the resource
  }>;
}

interface SearchResult {
  neighbors: Array<{
    id: string; // id of the resource
    title: string; // title of the resource
    url: string; // url to the resource
  }>;
}
Individual Functions

Besides the Voy class, Voy also exports all the instance methods as individual functions.

index(resource: Resource): SerializedIndex

It indexes the given resource and returns a serialized index.

Parameters

interface Resource {
  embeddings: Array<{
    id: string; // id of the resource
    title: string; // title of the resource
    url: string; // url to the resource
    embeddings: number[]; // embeddings of the resource
  }>;
}

Return

type SerializedIndex = string;
search(index: SerializedIndex, query: Query, k: NumberOfResult): SearchResult

It deserializes the given index and search for the k nearest neighbors of the query.

Parameter

type SerializedIndex = string;

type Query = Float32Array; // embeddings of the search query

type NumberOfResult = number; // K top results to return

Return

interface SearchResult {
  neighbors: Array<{
    id: string; // id of the resource
    title: string; // title of the resource
    url: string; // url to the resource
  }>;
}
add(index: SerializedIndex, resource: Resource): SerializedIndex

It adds resources to the index and returns an updated serialized index.

Parameter

type SerializedIndex = string;

interface Resource {
  embeddings: Array<{
    id: string; // id of the resource
    title: string; // title of the resource
    url: string; // url to the resource
    embeddings: number[]; // embeddings of the resource
  }>;
}

Return

type SerializedIndex = string;
remove(index: SerializedIndex, resource: Resource): SerializedIndex

It removes resources from the index and returns an updated serialized index.

Parameter

type SerializedIndex = string;

interface Resource {
  embeddings: Array<{
    id: string; // id of the resource
    title: string; // title of the resource
    url: string; // url to the resource
    embeddings: number[]; // embeddings of the resource
  }>;
}

Return

type SerializedIndex = string;
clear(index: SerializedIndex): SerializedIndex

It removes all items from the index and returns an empty serialized index.

Parameter

type SerializedIndex = string;

Return

type SerializedIndex = string;
size(index: SerializedIndex): number;

It returns the size of the index.

Parameter

type SerializedIndex = string;

Usage

With Transformers

As of now, voy relies on libraries like transformers.js and web-ai to generate embeddings for text:

import { TextModel } from "@visheratin/web-ai";

const { Voy } = await import("voy-search");

const phrases = [
  "That is a very happy Person",
  "That is a Happy Dog",
  "Today is a sunny day",
];
const query = "That is a happy person";

// Create text embeddings
const model = await (await TextModel.create("gtr-t5-quant")).model;
const processed = await Promise.all(phrases.map((q) => model.process(q)));

// Index embeddings with voy
const data = processed.map(({ result }, i) => ({
  id: String(i),
  title: phrases[i],
  url: `/path/${i}`,
  embeddings: result,
}));
const resource = { embeddings: data };
const index = new Voy(resource);

// Perform similarity search for a query embeddings
const q = await model.process(query);
const result = index.search(q.result, 1);

// Display search result
result.neighbors.forEach((result) =>
  console.log(`✨ voy similarity search result: "${result.title}"`)
);
Multiple Indexes
import { TextModel } from "@visheratin/web-ai";

const { Voy } = await import("voy-search");
const phrases = [
  "That is a very happy Person",
  "That is a Happy Dog",
  "Today is a sunny day",
  "Sun flowers are blooming",
];
const model = await (await TextModel.create("gtr-t5-quant")).model;
const processed = await Promise.all(phrases.map((q) => model.process(q)));

const data = processed.map(({ result }, i) => ({
  id: String(i),
  title: phrases[i],
  url: `/path/${i}`,
  embeddings: result,
}));
const resourceA = { embeddings: data.slice(0, 2) };
const resourceB = { embeddings: data.slice(2) };

const indexA = new Voy(resourceA);
const indexB = new Voy(resourceB);

License

Licensed under either of

  • Apache License, Version 2.0, (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)
  • MIT license (LICENSE-MIT or http://opensource.org/licenses/MIT)

at your option.

Sponsor

Contribution

Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.

License
Apache-2.0
Crate facts
Not measured
  • Runs on, Native code and Python bindings read crates on crates.io. This entry lists no crate.

Compare with similar projects

Order: strongest tier first, then most downloads, then name, with unmeasured values last. A library counts downloads over 30 days. An app counts binary downloads per release. An app with none counts its crate downloads over 30 days. ML Rust runs this catalogue, so its closest project always gets a row.

voy beside the projects that share a topic with it
ProjectDownloadsDependentsStarsLatest releaseRuns on (GPU)Native codeLicense
voy This page App Listed Dormant

πŸ•ΈοΈπŸ¦€ A WASM vector similarity search written in Rust

Not measured Not applicable 1,076 Not measured Not measured Not measured Apache-2.0
usearch Featured

Smaller & Faster Single-File Vector Search Engine from Unum

242,089 downloads over 30 days 70 4,373 2.26.6 GPU support unknown Binds native code: cxx, link-cplusplus, numkong (feature numkong). Build uses a native toolchain: cxx-build Apache-2.0
qdrant App Featured

Qdrant - High-performance, massive-scale Vector Database and Vector Search Engine for the next generation of AI. Also available in the cloud https://cloud.qdrant.io/

34,017 binary downloads per release Not applicable 35,478 Not measured Not measured Apache-2.0
ruvector Listed

High-performance Rust vector database core with HNSW indexing

39,411 downloads over 30 days 51 4,675 2.5.0 CUDA, WebGPU, Metal Binds native code: libsqlite3-sys, napi-sys, ort-sys, pgrx-pg-sys, pqcrypto-internals and 5 more MIT
faiss-rs Listed

High-level bindings for Faiss, the vector similarity search engine

19,449 downloads over 30 days 5 250 0.13.0 GPU support unknown Binds native code: faiss-sys, faiss_c. Build uses a native toolchain: cmake (feature static) Apache-2.0 OR MIT
splintr Listed

Tokenizer for Rust and Python: pure Rust by default, with 18 bundled vocabularies and four backends covering BPE, SentencePiece and WordPiece. Its output matches Hugging Face tokenizers id for id on 14 model families, including GPT-2, Qwen, Whisper, T5 and BERT.

Python bindings

8,204 downloads over 30 days 7 63 0.22.2 GPU support unknown Binds native code: pcre2-sys (feature pcre2) MIT
velesdb Listed

High-performance vector database engine written in Rust

2,396 downloads over 30 days 0 102 8.1.0 WebGPU Binds native code: tauri, tauri-plugin-velesdb, atk-sys, gdk-pixbuf-sys, gdk-sys and 9 more Not measured
Why ListedNo archive, New or Featured rule places it, so it is Listed.

What it lacks for Featured

What else the check found

  • The last activity is on 20 Sep 2023, 1146 days ago. That reaches the 365 days the rules expect, so the project is dormant. Dormancy rule
  • The oldest outside issue waits 1340 days for a reply, past the limit of 270. No maintainer replied in 365 days. Rot signals rule

Not measured: Latest release, Unpatched vulnerabilities past the grace window, First release.

Exact rules: Listed, Featured app. The check on applied thresholds version 9.

Every rule outcome, as the check records it
  • expected-interval: app: E = 365 days
  • dormancy: last activity 2023-09-20, 1146 days ago, E = 365: dormant
  • rot: oldest unanswered issue 1340 days (limit 270), no maintainer reply in 365 days
  • featured: not featured: dormant and not stable
  • listed: listed: no archive, new or featured rule applies

Measurements and sources

Each signal row shows the raw value, its source and the fetch date. The Sources list gathers every distinct source on this page. Movement needs more history.

Sources4 sources
AdoptionNot measured
Adoption measurements
SignalValueSourceFetchedCounts
Active dependentsNot measuredYes
New dependents (365 days)Not measuredYes
Peak active dependentsNot measuredYes
Active dependents as a share of the peakNot measuredYes
Binary downloads per releaseNot measuredYes
MaintenanceLatest human commit: 3 years ago8 not measured
Maintenance measurements
SignalValueSourceFetchedCounts
Latest human commit2023-09-20GitHub repositoryYes
Latest releaseNot measuredYes
First releaseNot measuredYes
Latest versionNot measuredYes
Days the oldest outside issue waits for a reply1340GitHub repositoryYes
Maintainer replies over 365 days0GitHub repositoryYes
Repository archivedfalseGitHub repositoryYes
Repository gonefalseGitHub repositoryYes
Release commit or manifest in the repoNot measuredYes
Published docsNot measuredYes
Latest release fails cargo checkNot measuredYes
Build check dateNot measuredYes
Build check toolchainNot measuredYes
SecurityNot measured
Security measurements
SignalValueSourceFetchedCounts
Unpatched vulnerabilities past the grace windowNot measuredYes
RustSec unmaintained advisoryNot measuredYes
Deprecation noticeNot measuredYes
Deprecated requirementNot measuredYes
Distribution0 binary downloads over the last 3 releases10 not measured
Distribution measurements
SignalValueSourceFetchedCounts
ReleasesNot measuredYes
Downloads over 30 daysNot measuredYes
Binary downloads over the last 3 releases0GitHub releasesYes
Homebrew installs on request over 365 daysNot measuredYes
Repology repository familiesNot measuredYes
Docker pull growth over 90 daysNot measuredYes
Docker pulls at the growth baselineNot measuredYes
Every version a prereleaseNot measuredYes
Every version 0.0.xNot measuredYes
Every version yankedNot measuredYes
Crate deleted from crates.ioNot measuredYes
Repository1,076 stars, 39 forks, 2 contributors, 88 commits. Stars alone never qualify a project.1 not measured
Repository measurements
SignalValueSourceFetchedCounts
GitHub stars1076GitHub repositoryYes β€” one part of the repository route
GitHub forks39GitHub repositoryYes β€” one part of the repository route
Contributors with a GitHub account2GitHub contributorsYes β€” one part of the repository route
Commits on the default branch88GitHub repositoryNo
Repository's main crate left out of the entryNot measuredYes
Facts that never count
Facts that never count measurements
SignalValueSourceFetchedCounts
LicenseApache-2.0GitHub repositoryNo
Runs on, from listed crates and their direct dependenciesNot measuredNo
What shows each backendNot measuredNo
Native code, from listed crates and their dependenciesNot measuredNo
Python bindings, from direct dependencies of listed cratesNot measuredNo
Crates that depend on a listed crate directlyNot measuredNo

Corrections and review history

Report a correction

No correction applies to this entry.

Judgments

A judgment enters only through an approved issue and a merged pull request.

No judgment applies to this entry.