Seems like a great fit for LLM - extensive documentation, plus a live system it can probe against freely to match the behavior.