SwiftyChat

SwiftyChat is an observable chat helper with an async send(_:) method. It is useful when you want to drive chat from your own task instead of using the AIChat.send() fire-and-forget helper.

Public state	Meaning
`messages`	Visible chat history owned by the helper
`isStreaming`	True while `send(_:)` is consuming the stream
`error`	Last stream or model error captured by the helper

State Model

@State private var chat = SwiftyChat(
    model: "openai/gpt-4o-mini",
    systemPrompt: "You are a helpful support assistant.",
    maxMessages: 30
)

SwiftyChat can also be initialized with a direct AIStreamModel.

let provider = OpenAICompatibleProvider(
    baseURL: "https://api.openai.com/v1",
    apiKey: openAIKey,
    model: "gpt-4o-mini"
)
 
let chat = SwiftyChat(model: provider)

Sending

Task {
    try await chat.send("How do I reset my API key?")
}

send(_:) appends the user message, builds the model history, appends an empty assistant message, and fills that assistant message as stream chunks arrive. If another send is already streaming, it returns without starting a second stream.

Clearing

chat.stop()
chat.clear()

clear() stops the current stream, clears messages, and clears error. maxMessages is constructor configuration; it prunes the history sent to the model while keeping the visible messages array as app state.

Related docs

Use AIChat when you want an input-bound @MainActor helper with send() and reset(). Use message management for chat history and ContextWindow.