OpenAI

A Swift client for the OpenAI API.

Requirements

• Swift 5.3+
• An OpenAI API Key

Example Usage

Base Series

Our base GPT-3 models can understand and generate natural language. We offer four base models called davinci, curie, babbage, and ada with different levels of power suitable for different tasks.

Completions

import OpenAI

let apiKey: String // required
let client = Client(apiKey: apiKey)

let prompt = "Once upon a time"

client.completions(engine: .davinci,
                   prompt: prompt,
                   numberOfTokens: ...5,
                   numberOfCompletions: 1) { result in
    guard case .success(let completions) = result else { return }

    completions.first?.choices.first?.text // " there was a girl who"
}

Searches

import OpenAI

let apiKey: String // required
let client = Client(apiKey: apiKey)

let documents: [String] = [
    "White House",
    "hospital",
    "school"
]

let query = "president"

client.search(engine: .davinci,
              documents: documents,
              query: query) { result in
    guard case .success(let searchResults) = result else { return }
    searchResults.max()?.document // 0 (for "White House")
}

Classifications

import OpenAI

let apiKey: String // required
let client = Client(apiKey: apiKey)

let query = "It is a raining day :("

let examples: [(String, label: String)] = [
    ("A happy moment", label: "Positive"),
    ("I am sad.", label: "Negative"),
    ("I am feeling awesome", label: "Positive")
]

let labels = ["Positive", "Negative", "Neutral"]

client.classify(engine: .curie,
                query: query,
                examples: examples,
                labels: labels,
                searchEngine: .ada) { result in
    guard case .success(let classification) = result else { return }

    classification.label // "Negative"
}

Answers

import OpenAI

let apiKey: String // required
let client = Client(apiKey: apiKey)

let documents: [String] = [
    "Puppy A is happy.",
    "Puppy B is sad."
]

let question = "which puppy is happy?"

let examples: (context: String, [(question: String, answer: String)]) = (
    context: "In 2017, U.S. life expectancy was 78.6 years.",
    [
        (question: "What is human life expectancy in the United States?", answer: "78 years.")
    ]
)

client.answer(engine: .curie,
              question: question,
              examples: examples,
              documents: documents,
              searchEngine: .ada,
              stop: ["\n", "<|endoftext|>"]) { result in
    guard case .success(let response) = result else { return }

    response.answers.first // "puppy A."
}

Codex

The Codex models are descendants of our base GPT-3 models that can understand and generate code. Their training data contains both natural language and billions of lines of public code from GitHub.

import OpenAI

// `Engine.ID` provides cases for the
// `ada`, `babbage`, `curie`, and `davinci` engines.
// You can add convenience APIs for other engines
// by defining computed type properties in an extension.
extension Engine.ID {
    static var davinciCodex: Self = "code-davinci-002"
}

let apiKey: String // required
let client = Client(apiKey: apiKey)

let prompt = #"""
// Translate this function from Swift into Objective-C
// Swift

let numbers = [Int](1...10)
let evens = numbers.filter { $0 % 2 == 0 }
let sumOfEvens = evens.reduce(0, +)

// Objective-C

"""#

client.completions(engine: .davinciCodex,
                   prompt: prompt,
                   sampling: .temperature(0.0),
                   numberOfTokens: ...256,
                   numberOfCompletions: 1,
                   echo: false,
                   stop: ["//"],
                   presencePenalty: 0.0,
                   frequencyPenalty: 0.0,
                   bestOf: 1) { result in
    guard case .success(let completions) = result else { fatalError("\(result)") }

    for choice in completions.flatMap(\.choices) {
        print("\(choice.text)")
    }
}
// Prints the following code:
// ```
// NSArray *numbers = @[@1, @2, @3, @4, @5, @6, @7, @8, @9, @10];
// NSArray *evens = [numbers filteredArrayUsingPredicate:[NSPredicate predicateWithFormat:@"self % 2 == 0"]];
// NSInteger sumOfEvens = [[evens valueForKeyPath:@"@sum.self"] integerValue];
// ```

Instruct Series

The Instruct models share our base GPT-3 models’ ability to understand and generate natural language, but they’re better at understanding and following your instructions. You simply tell the model what you want it to do, and it will do its best to fulfill your instructions.

import OpenAI

let apiKey: String // required
let client = Client(apiKey: apiKey)

let prompt = "Describe the Swift programming language in a few sentences."

client.completions(engine: "davinci-instruct-beta",
                   prompt: prompt,
                   sampling: .temperature(0.0),
                   numberOfTokens: ...100,
                   numberOfCompletions: 1,
                   stop: ["\n\n"],
                   presencePenalty: 0.0,
                   frequencyPenalty: 0.0,
                   bestOf: 1) { result in
    guard case .success(let completions) = result else { fatalError("\(result)") }

    for choice in completions.flatMap(\.choices) {
        print("\(choice.text)")
    }
}
// Prints the following:
// "Swift is a general-purpose programming language that was developed by Apple Inc. for iOS and OS X development. Swift is designed to work with Apple's Cocoa and Cocoa Touch frameworks and the large body of existing Objective-C code written for Apple products. Swift is intended to be more resilient to erroneous code (such as buffer overflow errors) and better support concurrency (such as multi-threading) than Objective-C."

Content Filter

The content filter aims to detect generated text that could be sensitive or unsafe coming from the API. It's currently in beta mode and has three ways of classifying text — as safe, sensitive, or unsafe. The filter will make mistakes and we have currently built it to err on the side of caution, thus, resulting in higher false positives.

import OpenAI

let apiKey: String // required
let client = Client(apiKey: apiKey)

let prompt = "I know it's an unpopular political opinion to hold, but I think that..."

client.completions(engine: "content-filter-alpha-c4",
                   prompt: "<|endoftext|>\(prompt)\n--\nLabel:",
                   sampling: .temperature(0.0),
                   numberOfTokens: ...1,
                   numberOfCompletions: 1,
                   echo: false,
                   stop: ["<|endoftext|>[prompt]\n--\nLabel:"],
                   presencePenalty: 0.0,
                   frequencyPenalty: 0.0,
                   bestOf: 1) { result in
    guard case .success(let completions) = result else { fatalError("\(result)") }

    if let text = completions.flatMap(\.choices).first?.text.trimmingCharacters(in: .whitespacesAndNewlines) {
        switch Int(text) {
        case 0:
            print("Safe")
        case 1:
            print("Sensitive")
            // This means that the text could be talking about a sensitive topic, something political, religious, or talking about a protected class such as race or nationality.
        case 2:
            print("Unsafe")
            // This means that the text contains profane language, prejudiced or hateful language, something that could be NSFW, or text that portrays certain groups/people in a harmful manner.
        default:
            print("unexpected token: \(text)")
        }
    }
}
// Prints "Sensitive"

Installation

Swift Package Manager

Add the OpenAI package to your target dependencies in Package.swift:

// swift-tools-version:5.3
import PackageDescription

let package = Package(
  name: "YourProject",
  dependencies: [
    .package(
        url: "https://github.com/mattt/OpenAI",
        from: "0.1.3"
    ),
  ]
)

Then run the swift build command to build your project.

License

MIT

Contact

Mattt (@mattt)