This is a very relevant question. Browser-based agentic workloads have some unique requirements around OS-level dependencies, persistent browser processes, dynamic ports, and CDP connectivity that don’t fit neatly into typical Databricks compute patterns. It would be great to know if Databricks has a supported architecture for running Playwright/Chrome workloads, perhaps through an external service integrated with Databricks, rather than requiring everything to run directly on Databricks compute. Any recommended reference architecture or guidance from the Databricks team would be very helpful.