OutspeedSwift

Swift SDK for the Outspeed Live API that enables real-time voice conversations using WebRTC.

Features

• Real-time voice conversations with AI
• Support for both Outspeed and OpenAI providers
• WebRTC-based audio streaming
• Customizable voice and model selection

Installation

1. Open Your Project in Xcode

2. Go to File > Add Packages...

3. Enter Repository URL in the Search Bar: https://github.com/outspeed-ai/OutspeedSwift

4. Import the SDK

   import OutspeedSDK

Important

Ensure NSMicrophoneUsageDescription is added to your Info.plist to explain microphone access.

Requirements

• iOS 15.2 or later
• Swift 6.1+

Usage

Basic Setup

import OutspeedSDK

// Create a session configuration
let config = OutspeedSDK.SessionConfig()

// Create callbacks to handle events
let callbacks = OutspeedSDK.Callbacks()
callbacks.onMessage = { message, role in
    print("Received message from \(role.rawValue): \(message)")
}

callbacks.onError = { message, error in
    print("Error: \(message)")
}

callbacks.onStatusChange = { status in
    print("Status changed to: \(status.rawValue)")
}

// Start the conversation
Task {
    do {
        let conversation = try await OutspeedSDK.Conversation.startSession(
            config: config,
            callbacks: callbacks
            apiKey: "<YOUR_OUTSPEED_API_KEY>"
        )

        // When done with the conversation
        conversation.endSession()
    } catch {
        print("Failed to start conversation: \(error)")
    }
}

ElevenLabs Swift Compatibilty

OutspeedSDK is fully compatible with Elevenlabs Swift SDK specifications, but registering tools is not supported yet.

To switch from ElevenLabsSDK:

1. Replace all occurrences of "ElevenLabsSDK" with "OutspeedSDK". So for example:

   import ElevenLabsSDK
   
   let config = ElevenLabsSDK.SessionConfig(agentId: "testagent")

   becomes

   import OutspeedSDK
   
   let config = OutspeedSDK.SessionConfig(agentId: "testagent") // you can even skip agentId

2. Add your Outspeed API key to startSession:

   let conversation = try await ElevenLabsSDK.Conversation.startSession(
       config: config,
       callbacks: callbacks
   )

   becomes

   let conversation = try await OutspeedSDK.Conversation.startSession(
       config: config,
       callbacks: callbacks
       apiKey: "<YOUR_OUTSPEED_API_KEY>" // required
   )

Configuring System Prompt, First Message, and Voice

Configuration Options

• AgentConfig: Customize the AI's behavior and initial response

  • prompt: System instructions that define the AI's personality and behavior
  • firstMessage: Optional initial message the AI will speak when the conversation starts

• TTSConfig: Configure voice settings

  • voiceId: Select from available voices (e.g., OutspeedSDK.Voice.david.rawValue)

Note

All configuration objects (AgentConfig, TTSConfig, and ConversationConfigOverride) are fully compatible with ElevenLabs SDK specifications.

Configuration Examples

System Prompt:

let agentConfig = OutspeedSDK.AgentConfig(
    prompt: "You are a helpful assistant with a witty personality."
)

let config = OutspeedSDK.SessionConfig(
    overrides: OutspeedSDK.ConversationConfigOverride(agent: agentConfig)
)

First Message:

let agentConfig = OutspeedSDK.AgentConfig(
    firstMessage: "Hey there, how can I help you with Outspeed today?"
)

let config = OutspeedSDK.SessionConfig(
    overrides: OutspeedSDK.ConversationConfigOverride(agent: agentConfig)
)

Voice Selection:

let ttsConfig = OutspeedSDK.TTSConfig(voiceId: OutspeedSDK.Voice.david.rawValue)

let config = OutspeedSDK.SessionConfig(
    overrides: OutspeedSDK.ConversationConfigOverride(tts: ttsConfig)
)

Combined Configuration

Example snippet to customize the AI's behavior, initial message, and voice together using configuration objects:

// Configure the AI agent's behavior and initial message (ElevenLabs compatible)
let agentConfig = OutspeedSDK.AgentConfig(
    prompt: "You are a helpful assistant with a witty personality.",
    firstMessage: "Hey there, how can I help you with Outspeed today?"
)

// Configure voice selection (also ElevenLabs compatible)
let ttsConfig = OutspeedSDK.TTSConfig(voiceId: OutspeedSDK.Voice.david.rawValue)

// Create session configuration with overrides
let config = OutspeedSDK.SessionConfig(
    overrides: OutspeedSDK.ConversationConfigOverride(
        agent: agentConfig,
        tts: ttsConfig
    )
)

// Set up callbacks
var callbacks = OutspeedSDK.Callbacks()
callbacks.onMessage = { message, role in
    print("Received message from \(role.rawValue): \(message)")
}

callbacks.onError = { message, error in
    print("Error: \(message)")
}

callbacks.onStatusChange = { status in
    print("Status changed to: \(status.rawValue)")
}

// Start the conversation with custom configuration
Task {
    do {
        let conversation = try await OutspeedSDK.Conversation.startSession(
            config: config,
            callbacks: callbacks,
            apiKey: "<YOUR_OUTSPEED_API_KEY>",
        )
    } catch {
        print("Failed to start conversation: \(error)")
    }
}

Examples

You can find examples of the SDK here: https://github.com/outspeed-ai/outspeed-swift-examples