The whole surface is a query in and an answer out, which makes it a drop-in for the arithmetic, unit conversion and factual computation a model should not be doing in its head. Each call stands alone — no session, no history, no stored results — so a follow-up question has to carry its own context. validate_key is the only diagnostic in the set, and it is the right first call when answers come back empty.
A WolframAlpha front end with two answer shapes and a credential check, 3 tools in total.
- ask_llm returns a structured response in multiple formats for a query, which is the shape to use when the working matters and not only the number
- get_simple_answer returns the condensed version, focused on the most relevant information
- validate_key checks the WolframAlpha LLM API key, which is what separates a failing credential from a query that returned nothing
A WolframAlpha LLM API key, which validate_key exists to confirm. The two answer tools take a query and nothing else.
