[grimmai143]

Async is definitely the right approach for this, but how are you handling the event loop blocking if a local model takes way too long to generate a heavy sequence?