Skip to main content

Chapter 66: Agentic Function Calling

Teach models to call tools reliably. This chapter builds an agentic-function-calling skill to create datasets, tune, and evaluate function/tool use for your Task API agents.

Goals​

  • Define tool schemas and structured output expectations
  • Generate/tune datasets for function calls and arguments
  • Evaluate tool-use accuracy and robustness
  • Orchestrate multi-tool flows safely
  • Capture reusable prompts and tests in a skill

Lesson Progression​

  • Build the function-calling skill
  • Tool schema design and dataset creation
  • Fine-tuning/evaluation for function use
  • Multi-tool orchestration patterns
  • Capstone: tool-capable Task API model; finalize the skill

Outcome & Method​

You finish with a function-calling–capable model plus a skill that standardizes tool schemas, data, and evaluations.

Prerequisites​

  • Chapters 63-65 (data, SFT, persona)
  • Part 6 agent/tool patterns for context