DOCSEARCHbyAlgolia
Submit Docs

EventSource

repository·main·Indexed 41 days ago

https://github.com/recouse/eventsource

A concurrency-safe Swift package for Server-Sent Events (SSE) that utilizes async/await and AsyncSequence to handle real-time updates from persistent HTTP connections. Supports standard SSE and a data-only mode optimized for streaming APIs like OpenAI.

Tokens
1.2K
Snippets
5
Records
6
Agent score
1%

What's inside EventSource

  2. Use EventSource to receive real-time updates

    main

    To connect to an SSE endpoint, create an instance of EventSource, initiate a dataTask(for:) with a URLRequest, and iterate over the events using an AsyncSequence via dataTask.events().

    Note: To explicitly close the connection, call task.cancel() on the parent Task. Note that if you cancel the task, the .closed event will not be emitted.

    import EventSource

let task = Task {
    let eventSource = EventSource()
    let dataTask = eventSource.dataTask(for: urlRequest)
    
    for await event in dataTask.events() {
        switch event {
        case .open:
            print("Connection was opened.")
        case .error(let error):
            print("Received an error:", error.localizedDescription)
        case .event(let event):
            print("Received an event", event.data ?? "")
        case .closed:
            print("Connection was closed.")
        }
    }
}

  4. Install EventSource via Swift Package Manager

    main

    Add the package to your Package.swift dependencies and include it in your target's dependencies:

    // In dependencies
.package(url: "https://github.com/Recouse/EventSource.git"),

// In your target
.target(name: "<target>", dependencies: [
    .product(name: "EventSource", package: "EventSource"),
]),
    .package(url: "https://github.com/Recouse/EventSource.git"),

.target(name: "<target>", dependencies: [
    .product(name: "EventSource", package: "EventSource"),
]),

  5. Use EventSource in Data-only mode

    main

    For APIs that primarily stream data (like OpenAI's completions API), initialize EventSource with mode: .dataOnly. This mode is optimized for handling streams where you only care about the data payload within the event.

    Task {
    var urlRequest = URLRequest(url: URL(string: "https://api.openai.com/v1/chat/completions")!)
    urlRequest.allHTTPHeaderFields = [
        "Content-Type": "application/json",
        "Authorization": "Bearer \(accessToken)"
    ]
    urlRequest.httpMethod = "POST"
    urlRequest.httpBody = """
    {
        "model": "gpt-4o-mini",
        "messages": [
            {"role": "user", "content": "Why is the sky blue?"}
        ],
        "stream": true
    }
    """.data(using: .utf8)!
    
    let eventSource = EventSource(mode: .dataOnly)
    let dataTask = eventSource.dataTask(for: urlRequest)
    
    var response: String = ""
    
    for await event in dataTask.events() {
        switch event {
        case .event(let event):
            if let data = event.data?.data(using: .utf8) {
                let chunk = try? JSONDecoder().decode(ChatCompletionChunk.self, from: data)
                let string = chunk?.choices.first?.delta.content ?? ""
                response += string
            }
        default:
            break
        }
    }
    
    print(response)
}