It could get pricey to use the same model for all of the LLM tasks, so if someone knows how to get rid of the bug, be my guest 🙂