Swift Faiss

Use Faiss in Swift.

Based on Faiss Mobile and OpenMP Mobile.

Usage

Command line demo

$ swift run swift-faiss <subcommand> <options>

Available subcommands:

• flat: create a FlatIndex, add vectors to it and search for the most similar sentences.
• ivfflat: create an IVFFlatIndex, train and add vectors to it and search for the most similar sentences.
• pq: create an PQIndex, train and add vectors to it and search for the most similar sentences.
• clustering: k-means clustering example.

Command line help

$ swift run swift-faiss --help

In your own code

import SwiftFaiss

let embeddings: [[Float]] = [
    [0.1, 0.2, 0.3],
    [0.4, 0.5, 0.6],
    [0.7, 0.8, 0.9],
    [1.0, 1.1, 1.2],
    [1.3, 1.4, 1.5],
    [1.6, 1.7, 1.8]
]
let d = embeddings[0].count
let index = try FlatIndex(d: d, metricType: .l2)
try index.add(embeddings)

let result = try index.search([[0.1, 0.5, 0.9]], k: 2)
// do something with result

https://github.com/jkrukowski/UseSwiftFaiss contains an iOS example.

Installation

Swift Package Manager

You can use Swift Package Manager and specify dependency in Package.swift by adding:

.package(url: "https://github.com/jkrukowski/SwiftFaiss.git", from: "0.0.7")

Format code

$ swift package plugin --allow-writing-to-package-directory swiftformat

Tests

$ swift test

More info

• Faiss: The Missing Manual
• Faiss C API