Congrats on the launch, and thanks for open-sourcing it! We recently gave it a go to browse third-party API docs and were surprised by the accuracy of the summaries! I'm curious, why do you think the actor model works better for LLMs?