Current documentation mentions "A major advantage of Google Colab is free GPU in the cloud." which could be expanded, with LLM integration discussed.
Also curious why wrapping a local fiji is prioritized. I wonder if it's necessary for some reason in Colab? If possible I would lean towards starting with dynamic imagej/fiji initialization and moving to a wrapped fiji for custom old plugins.