## Connect

Connect to Live without query parameters. Send session.start with model and session configuration, then wait for session.started. Follow the [Live prompting guide](https://developers.openai.com/api/docs/guides/live-prompting) before writing instructions.

**WebSocket** `wss://api.openai.com/v1/live/sessions`

### Example

```ruby
require "openai"
require "async"

# The session model is sent in the first event.
client = OpenAI::Client.new
Sync do |task|
  # Stop waiting if the peer never sends the expected events.
  task.with_timeout(60) do
    # The block always releases this session's transport.
    client.live.connect() do |connection|
      connection.send_event(type: "session.start", session: { model: "gpt-live-1" })
      complete = false
      close_requested = false
      connection.each do |event|
        case event
        when OpenAI::Live::ErrorEvent
          raise "Live reported a session error."
        when OpenAI::Live::SessionStartedEvent
          # Wait for startup before closing the session this example owns.
          puts "Session ready: #{event.session.id} #{event.session.model}"
          connection.send_event(type: "session.close")
          close_requested = true
        when OpenAI::Live::SessionClosedEvent
          unless close_requested && event.reason == :close_requested
            raise "Session ended unexpectedly: #{event.reason}; usage=#{event.usage}"
          end

          puts "Final usage: #{event.usage}"
          complete = true
          break
        end
      end
      raise "Connection closed before session.closed; final usage unconfirmed" unless complete
    end
  end
end
```