Tool calling: how an LLM uses your tools
When a model "calls a tool", what does it write, and who runs it?
A model never runs your tool. It writes a short request, as text. Your code reads the request, runs the function and sends the result back. Then the model writes its answer.
