Suppose I want to create 2 clients that extends a base trait

trait Clients {
   def emitter: Emitter
   def push(data: str): Unit
}

case class RestClient(config: RestConfig) extends Clients {
   def emitter: RestEmitter = ??? // where RestEmitter extends Emitter
   def push(data: str): Unit = ???
}

case class KafkaClient(config: KafkaConfig) extends Clients {
   def emitter: KafkaEmitter = ??? // where KafkaEmitter extends Emitter
   def push(data: str): Unit = ???
}

next I have a singleton that returns the caller an instance of the client. Reason I have this singleton is I dont want callers to create multiple instances of the client

object ClientFactory {
   private var client: Option[Emitter] = None // I hate using this mutable state
   
   def getClient: Emitter = {
       if (client.isDefined) client.get
       else throw RuntimeException("client has not init yet")
   }

   def init(config: Config) = {
       config match {
          case rCfg: RestConfig => 
              client = RestClient(rCfg)
          case kCfg: KafkaConfig =>
              client = KafkaClient(kCfg)
       }
   }
}

A caller would use it as

ClientFactory.init(some_config)
ClientFactory.getClient.push(some_data)

Is there any better way of defining this singleton ClientFactory , so that I dont have to use this mutable variable