OpenAIStreamingCompletions

Streaming text generartion using the OpenAI APIs.

Supports streaming results via ObservableObject or AsyncStream, and non-streaming results via async/await.

Supports message-based models (e.g. ChatGPT) and text-based models (e.g. davinci).

Installation via Swift Package Manager

You can either:

• Add this line to your Package.swift dependencies array: .package(url: "https://github.com/nate-parrott/openai-streaming-completions-swift", from: "1.0.1")
• Use Xcode’s File -> Add Packages and paste the URL to this repository

Calling ChatGPT

Provide API key and prompt

let messages: [OpenAIAPI.Message] = [
    .init(role: .system, content: "You are a helpful assistant. Answer in one sentence if possible."),
    .init(role: .user, content: prompt)
]
let api = OpenAIAPI(apiKey: key)

Option A: Generate text (streaming, asyncStream)

Task {
    let stream =  try api.completeChatStreaming(.init(messages:  promptMessages))
    for await message in stream {
        print("\(message.content)") // each message contains a small part of the response
    }
}

Option B: Generate text (non-streaming, async/await)

Task {
    do {
        self.completedText = try await api.completeChat(.init(messages: messages))
    } catch {
        print("Error: \(error)")
    }
}

Calling other text APIs

See OpenAIAPI+TextCompletion.swift