Glad you see the potential! We do have an existing function "call graph" view that uses traditional static analysis methods with LSP calls. Augmenting this view with LLMs seems very promising, and we are very much headed in that direction! Are there use cases you can see this being helpful for?