Defining a Tool JSON Schema

Cost and Latency of Multi-Turn Tool Use

Security Risks of Tool Execution

Error Handling and Retries in Agent Loops

Agent Frameworks Overview

Stopping Conditions and Max Iterations

Agent Memory and State Across Steps

Multi-Step Task Decomposition by Agents

Common Tool Types for LLM Agents

Tools vs Functions vs Agents

Agent Reasoning Loops: The ReAct Pattern

What an LLM Agent Is

Structured Output via Function Calling

Handling Invalid or Hallucinated Tool Arguments

Controlling Tool Choice

Parallel and Multiple Tool Calls

The Full Function Calling Loop

Returning Tool Results Back to the Model

Executing the Function in Your Code

Parsing the Model's Tool Call Response

Token Cost of Tools and Loops

Handling Hallucinated Calls

What an LLM Agent Is

The ReAct Loop

Agents vs Function Calling

Common Agent Tools

Agent Memory and State

Stopping Conditions

Returning Tool Results

The Tool Message Role

Agent Frameworks Overview

Error Handling and Retries

Security Risks of Tool Calls

What Function Calling Is

Why LLMs Need Tools

Defining a Tool Schema

JSON Schema for Parameters

Writing Function Descriptions

Passing Tool Definitions

How the Model Decides to Call

Reading the Tool Call

Parsing Call Arguments

Multi-Step Function Loop

Parallel Tool Calls

Forcing Tool Choice

Structured Output vs Function Calling

Executing the Function

How the Model Decides to Call a Tool

Passing Tool Definitions in an API Request

Tool Definitions and JSON Schema

What Function Calling Is and Why It Exists

Forcing a specific tool_choice

Validating arguments against allowed values

Building the tool-call loop

Writing function descriptions

Handling malformed arguments

Matching tool_call_id to result

Handling parallel tool calls

Parsing tool call arguments

Distinguishing final vs tool response

Parameter types and required fields

Returning tool errors to the model

Dispatching tool names to handlers

Preventing infinite loops

Sending tool results back

Passing the tools array

Appending the tool result message

Executing the local function

Detecting a tool call

Preserving the assistant tool_calls message

Stopping Conditions and Limits

Parsing the Tool Call Response

Agent Frameworks Overview

Executing and Returning Results

Parallel Function Calls

Handling Call Errors

Structured Output via Function Calling

What an LLM Agent Is

Planning and Task Decomposition

Security Risks of Tool Execution

Why LLMs Need Tools

Passing Tools in a Request

The Full Function Calling Loop

Tool Choice Control

Writing Effective Tool Descriptions

The ReAct Pattern

Agent Memory and State

The Agent Loop

Multi-Tool Orchestration

Function Calling vs RAG vs Fine-Tuning

Cost and Latency of Agent Loops

What Function Calling Is

Defining a Tool Schema

Describing Parameters and Types

How the Model Picks a Tool