AI Agent Architecture — ReAct, Memory, Planning and Multi-Agent Systems · Tool-Augmented LLM — Tools, Functions and the Environment
JSON Schema of a Tool — the Contract Between LLM and Environment
Tool-Augmented LLM — Tools, Functions and the Environment
Introduction
A tool is not just a Python function — it is a contract describing what it does, what arguments it accepts, and what value constraints apply. JSON schema is the standard for this contract: it defines parameter types, required fields, enums, and semantic descriptions. These descriptions are what the model reads at inference time when deciding whether and how to call a tool. This lesson analyses the anatomy of a tool JSON schema (OpenAI function calling format), the impact of description quality on LLM routing accuracy, and common pitfalls in schema design.