Process pool with custom availability logic

I have a GenServer (I will call it A) that I connect to a external source (Actually it is a WebSockex process since I connect using websocket) and provide an API to communicate with that source.

A normal API call for A would be something like subscribe_to_channel(channel, callback) which means that it would subscribe that connection to a specific channel and pass all messages from that channel to the callback.

This works great, but the external source limits the number of channel subscriptions to a certain number per connection, let’s say 10. So, after 10 calls to subscribe_to_channel with specific channels, I would need to create a new A process with a new connection and use it to subscribe to new channels.

I was thinking in how to automatize this, what comes to mind is something like poolboy, where I can tell it to initialize like 5 A processes in a pool (and, as result, 5 websocket connections to the external source) and let it handle the subscribe_to_channel calls.

The idea is that A would have another method called available_subscriptions which returns the number of available subscriptions this process can handle. So, for example, if A have 1 channel subscribed, then available_subscriptions will return 9, for 2 channels, it will return 8, and so on.

From that, poolboy could check that function and decide to send the subscribe call to the process in the pool with the least subscriptions currently.

Now, as far as I know, I don’t think I can configure a behavior like that in poolboy, so my question is, is there any library that implements something like that? Or do I need to implement my custom pool with the above described logic?

Thanks!