A Swift client for the OpenAI API.

Mattt

Last update: Dec 26, 2022

Related tags

Overview

OpenAI

A Swift client for the OpenAI API.

Requirements

Swift 5.3+
An OpenAI API Key

Example Usage

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

"]) { result in guard case .success(let response) = result else { return } response.answers.first // "puppy A." } ">

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."
}

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: "0.1.0"
    ),
  ]
)

Then run the swift build command to build your project.

License

MIT

Contact

Mattt (@mattt)

Comments

Added support for more engines.

Added support for newer engines in the Instruct series and Codex, so that these models can be quickly referenced via the enum. Updated enum case descriptions to reflect this.

opened by zcutlerf 10
Adding user field in completions

OpenAI requested that I include a user field in a recent app pre-launch review. Wanted to add it in here in case other developers are also being requested to include the field. Here's the note provided by the production review team:

Please pass a uniqueID for every user w/ each API call (both for Completion & the Content Filter) e.g. user= $uniqueID. This 'user' param can be passed in the request body along with other params such as prompt, max_tokens etc.

opened by benhynes 5

Alamofire Error: invalid_request_error - Engine not found

Alamofire throws the error invalid_request_error - Engine not found.

Setup:

class Model: ObservableObject {
	@Published var input = ""
	@Published var output = ""
	
	private let client: Client
	
	init() {
		let apiKey = "<redacted>"
		client = Client(apiKey: apiKey)
	}
	
	func go() {
		client.completions(
			engine: "davinci-codex",
			prompt: input,
			sampling: .temperature(0.0),
			numberOfTokens: ...256,
			numberOfCompletions: 1,
			echo: false,
			stop: ["//"],
			presencePenalty: 0.0,
			frequencyPenalty: 0.0,
			bestOf: 1
		) { [self] result in
			guard case .success(let completions) = result else {
				fatalError("\(result)")
			}
			
			let choices = completions.flatMap(\.choices)
			
			for choice in choices {
				print("\(choice.text)")
			}
			
			output = choices.first?.text ?? ""
		}
	}
}

opened by melgu 3

Document best practices for using available engines
Related to #8 and #9

As discussed in https://github.com/mattt/OpenAI/pull/9#issuecomment-1117913832, there's no way for this library to provide enumeration cases for each of the OpenAI engines available.

So as an alternative, this PR documents the pattern of extending Engine.ID to provide convenience APIs for any engines used in a codebase.

extension Engine.ID { static var babbageSearchQuery: Self = "babbage-search-query" }

This PR also updates the existing testEngines() to run against a mocked response, and updates the mocked response of GET /v1/engines/ada with new fields.
opened by mattt 2
Adding dedicated contentFilter function
It was required by OpenAI for a project I've been working on, so I just wanted to simplify usage a little bit. Now there's a contentFilter function that can be called as:

client.contentFilter(prompt: prompt, completion: { rating in switch rating { 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 result") } } })

It effectively wraps a completion request instead of implementing a different request, so I think the error handling should be sufficient with the completion of the the completions function with Result<[Completion], Swift.Error>. This is handled with the guard case .success(let completions) = result else { fatalError("\(result)") } within contentFilter, but I'm interested in input as to whether this is an acceptable way of dealing with this.

Just a note that I rebased my commits to modify a commit name for clarity, in case there were any questions regarding why there were 4 commits made at the same time.
opened by benhynes 2

failure(Alamofire.AFError.sessionDeinitialized)

When I try the recommended code

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)")
    }
}

I get the following error:

TestWithOpenAI/ViewController.swift:31: Fatal error: failure(Alamofire.AFError.sessionDeinitialized)
2022-05-08 20:08:33.976354-0700 TestWithOpenAI[15129:4948609] TestWithOpenAI/ViewController.swift:31: Fatal error: failure(Alamofire.AFError.sessionDeinitialized)

opened by scott-lydon 1

Some parameters in the client.completion(... etc not passed on to the session call
Hi - I want to set the presencePenalty and frequencyPenalty parameters to my own custom values, but they don't seem to be carried forward into the parameters flatmap before the session call.

The parameters exposed in the high level client method completions that should ideally be carried through in to the URL session call.

e.g.:

Parameters in client.completion():

presencePenalty: Double? = nil, frequencyPenalty: Double? = nil, bestOf: Int? = nil,

Not used in:

var parameters: [String: Any?] = [ "prompt": prompt, "max_tokens": numberOfTokens?.upperBound, "user": user, "n": numberOfCompletions, "echo": echo, "stop": stop /// missing penalty and bestOf parameters ideally added here... ]

Net result is that you think you are changing certain values in the call to client.completions(), but in reality they aren't getting changed.
opened by paulostergaard 1

Alamofire throws fatal error in SwiftUI

Occurs when executing this code block:

let apiKey: String = "<<API KEY>>"
let client = Client(apiKey: apiKey)

client.completions(engine: .davinci,
                           prompt: prompt,
                           numberOfTokens: ...300,
                           numberOfCompletions: 1,
                           echo: false,
                           stop: ["\n\n"],
                           presencePenalty: 0.1,
                           frequencyPenalty: 0.1,
                           bestOf: 1) { result in
            guard case .success(let completions) = result else { fatalError("\(result)") }
            
            for choice in completions.flatMap(\.choices) {
                    print("\(choice.text)")
            }
}

Error: Fatal error: failure(Alamofire.AFError.sessionDeinitialized)

This is with iOS 14.0 as a deployment target and Xcode 13.0 as a SwiftUI project. The problem seems to be caused when the client is declared within the local namespace of the function. When the declaration is moved out to global namespace, outside of the view struct, the issue is resolved.

It seems SwiftUI might excessively garbage collect in this situation. It may make sense to add a footnote for SwiftUI to the docs (I'm happy to help contribute).

opened by benhynes 1

Undefined symbol: nominal type descriptor for OpenAI.Completion
Download OpenAI package.

On the Project level, add OpenAI package into my project through Package Dependencies. OpenAI-main package shows up in the folder Packages; Alamofire 5.4.4 and AnyCodable 0.4.1 show up in Package Dependencies.

On the Target level, add OpenAI product into Build Phases->Dependencies

in my project, code "import OpenAI" in ContentView.swift without errors.

run the build, the errors show:

Undefined symbols for architecture arm64: "nominal type descriptor for OpenAI.Completion.Choice", referenced from: _symbolic Say_____G 6OpenAI10CompletionV6ChoiceV in ContentView.o "nominal type descriptor for OpenAI.Completion", referenced from: _symbolic Say_____G 6OpenAI10CompletionV in ContentView.o "type metadata accessor for OpenAI.Client", referenced from: one-time initialization function for client in ContentView.o ForTest.ContentView.followUpPrompt(prompt: Swift.String) -> () in ContentView.o "OpenAI.Client.completions(engine: OpenAI.Engine.ID, prompt: Swift.String?, sampling: OpenAI.Sampling?, numberOfTokens: Swift.PartialRangeThrough<Swift.Int>?, numberOfCompletions: Swift.Int?, echo: Swift.Bool?, stop: [Swift.String]?, user: Swift.String?, presencePenalty: Swift.Double?, frequencyPenalty: Swift.Double?, bestOf: Swift.Int?, completion: (Swift.Result<[OpenAI.Completion], Swift.Error>) -> ()) -> ()", referenced from: ForTest.ContentView.followUpPrompt(prompt: Swift.String) -> () in ContentView.o "OpenAI.Client.__allocating_init(apiKey: Swift.String, organization: Swift.String?) -> OpenAI.Client", referenced from: one-time initialization function for client in ContentView.o ForTest.ContentView.followUpPrompt(prompt: Swift.String) -> () in ContentView.o ld: symbol(s) not found for architecture arm64 clang: error: linker command failed with exit code 1 (use -v to see invocation)

ld: symbol(s) not found for architecture arm64 clang: error: linker command failed with exit code 1 (use -v to see invocation)

Undefined symbol: nominal type descriptor for OpenAI.Completion.Choice

Undefined symbol: nominal type descriptor for OpenAI.Completion

Undefined symbol: type metadata accessor for OpenAI.Client

Undefined symbol: OpenAI.Client.completions(engine: OpenAI.Engine.ID, prompt: Swift.String?, sampling: OpenAI.Sampling?, numberOfTokens: Swift.PartialRangeThrough<Swift.Int>?, numberOfCompletions: Swift.Int?, echo: Swift.Bool?, stop: [Swift.String]?, user: Swift.String?, presencePenalty: Swift.Double?, frequencyPenalty: Swift.Double?, bestOf: Swift.Int?, completion: (Swift.Result<[OpenAI.Completion], Swift.Error>) -> ()) -> ()

Undefined symbol: OpenAI.Client.__allocating_init(apiKey: Swift.String, organization: Swift.String?) -> OpenAI.Client
opened by truman0505 0
Lacking new endpoints

OpenAI API changes

As OpenAI added new endpoints to their API and plans to deprecate some of the already existing ones, there is a need to take action.

Namely, this package should include new Edits and Embeddings endpoints.

Please also note that endpoints Searches, Classifications, Answers and Engines 'are deprecated and will be removed on December 3rd, 2022`. Reference - OpenAI API Reference

opened by syrekable 0
Deprecate engine cases

Made a few changes here. Please let me know if any of these could be reworked. -Deprecated the enum cases, which produces some warnings in the package. I haven't figured out how to "deprecate in the future." -After re-reading OpenAI's documentation, it seems like Completion should be used with the newer Instruct Series, and the other methods should be used with the older Base Series. I've streamlined the readme to reflect this. -I also moved the extension suggestion higher in the readme so it's more noticeable.

opened by zcutlerf 0

Owner

Mattt

GitHub

Fetch Multiple Rest API using Swift 5.5 Async Await with Task, TaskGroup, Continuation API

Source code for Tutorial on experimenting with Swift Async Await to fetch multiple REST API endpoints and eliminate Pyramid of Doom callback hell to improve code readability and maintanability

14 Dec 7, 2022

Swift client for Kubernetes

Table of contents Overview Compatibility Matrix Examples Usage Creating a client Configuring a client Client authentication Client DSL Advanced usage

94 Dec 14, 2022

A (really) native and powerful macOS Telegram client built using SwiftUI, optimized for moderating large communities and personal use.

Moc A (really) native and powerful macOS Telegram client, optimized for moderating large communities and personal use. This client is currently in dev