The reasoning part is still dependent on an LLM, how the user input is interpreted and how the response from tool is interpreted.

→ View original post on X — @akshay_pachaar