liveBook · Manning

Call

As explained in The Art of AI Product Development

Function calling is a model interaction capability in which a language model produces a structured request to invoke a specified software function or tool.

Function calling lets a language model select and instantiate an appropriate software function, turning a user request into a structured action such as querying a database, running a calculation, or accessing another model. section 9.1

Website analytics report

A content application can issue a function call to gather website analytics, then use the returned data to create a report tailored to the client. The same system can expose functions for user-specific rewriting and analytical tasks. section 5.1.3

Reliable function calling requires attention to the execution boundary:

  • Connect the model to downstream functions or plug-ins that impose defined interfaces and constraints. section 5.2.2
  • Guard execution against faulty or harmful outputs, such as a generated DELETE query replacing a read-only SELECT query. section 5.2.2
  • Improve structural reliability through function-calling models, constrained inference, schema validation, or provider APIs. section 5.2.3
  • Evaluate models with function-calling benchmarks and application-specific tests rather than relying only on general language benchmarks. section 5.4.2
Figure 5.9 LMs can generate code that is automatically executed by downstream components.
A blue and black text

AI-generated content may be incorrect.

A process diagram showing how a language model generates code or another structured request for automatic execution by downstream components. section 5.2.2

How The Art of AI Product Development covers it

How deep it goes
Explains selection, structured execution, reliability options, and risks, but stops short of detailed implementation code.
Examples
Uses website analytics retrieval, database queries, calculations, machine-learning access, and function-calling benchmarks as examples.
What it assumes
Assumes familiarity with language models and software functions while explaining how functions become agent tools.
What you can do after
Operate/configure model-tool integrations with structural constraints, validation, security measures, and application-specific evaluation.
How it connects
Links language models with external tools, AI agents, Text2SQL, structured outputs, schema validation, and guardrails.

Also explained in

Compare how 7 books explain call →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of The Art of AI Product Development. The section links point to the parts of the book it draws on.