iOS & macOS Quickstart

This guide shows how to add the Swift AI SDK to an iOS or macOS app, call models, and stream results into your UI.

1) Install via Swift Package Manager

In Xcode: File → Add Package Dependencies…

• Package URL: https://github.com/teunlao/swift-ai-sdk
• Add products: SwiftAISDK, OpenAIProvider (and other providers as needed)

2) Configure API Keys

Set environment variables in your scheme or via a secure configuration:

OPENAI_API_KEY=sk-...

The default provider shortcuts (e.g., openai) read standard env vars automatically.

3) Minimal call (blocking)

import SwiftAISDK
import OpenAIProvider

let result = try await generateText(
  model: openai("gpt-4o"),
  prompt: "Hello from Swift!"
)
print(result.text)

4) Streaming in SwiftUI

import SwiftUI
import SwiftAISDK
import OpenAIProvider

@MainActor
final class ChatViewModel: ObservableObject {
  @Published var output: String = ""
  @Published var isLoading = false

  func send(_ prompt: String) {
    isLoading = true
    output = ""
    Task {
      do {
        let stream = try streamText(
          model: openai("gpt-4o"),
          prompt: prompt
        )
        for try await delta in stream.textStream {
          output += delta
        }
      } catch {
        output = "Error: \(error)"
      }
      isLoading = false
    }
  }
}

struct ChatView: View {
  @StateObject var vm = ChatViewModel()
  @State var input = "Write a haiku about vectors"

  var body: some View {
    VStack(alignment: .leading) {
      TextEditor(text: $input).frame(height: 120)
      Button(vm.isLoading ? "Streaming…" : "Send") { vm.send(input) }
      ScrollView { Text(vm.output).monospaced() }
    }
    .padding()
  }
}

5) Tools (optional)

struct EchoPayload: Codable, Sendable {
  let message: String
}

let echo = tool(
  description: "Echo back the input data",
  inputSchema: EchoPayload.self
) { payload, _ in
  EchoPayload(message: payload.message.uppercased())
}

let result = try await generateText(
  model: openai("gpt-4o"),
  tools: ["echo": echo.eraseToTool()],
  prompt: "Call echo with {\"message\":\"Hello!\"}"
)

Keep long‑running work off the main actor; UI updates should be marshalled on the main thread.