Swift Vector Database (SVDB)

A new fast local on-device vector database for Swift Apps.

Built for those building the next-generation of user experiences only possible with on-device intelligence.

Local on-device vector databases are just the beginning.

Installation

To install it using the Swift Package Manager, either directly add it to your project using Xcode 11, or specify it as dependency in the Package.swift file:

// ...
dependencies: [
    .package(url: "https://github.com/Dripfarm/SVDB.git", from: "1.0.0"),
],
//...

Usage

1. Create Embeddings

let document = "cat"

ChatGPT:

I find This Swift OpenAI package to be the best

import OpenAI

func embed(text: String) async -> [Double]? {
	let query = EmbeddingsQuery(model: .textEmbeddingAda, input: text)

	let result = try! await openAI.embeddings(query: query)

	return result.data.first?.embedding
}

let wordEmbedding = embed(text: document)

NLEmbeddings

import NaturalLanguage

let embedding: NLEmbedding? = NLEmbedding.wordEmbedding(for: .english)

let wordEmedding = embedding?.vector(for: document) //returns double array

2. Add Documents

let animalCollection = SVDB.shared.collection("animals")

SVDB.shared.addDocument(text: document, embedding: wordEmbedding)

3. Search

let dogEmedding = embedding?.vector(for: "dog")

let results = animalCollection.search(query: dogEmedding)

Demo

Check out the demo Demo

Todo

Not sure. I want to make it easier to add documents and take care of the embeddings for you. Any suggestions?