Function Calling, Tools & Agents

95 lessons in AI Engineering / Working with LLMs

• Defining a Tool JSON SchemaCode Quiz
• Cost and Latency of Multi-Turn Tool UseSlides / Video
• Security Risks of Tool ExecutionSlides / Video
• Error Handling and Retries in Agent LoopsSlides / Video
• Agent Frameworks OverviewSlides / Video
• Stopping Conditions and Max IterationsSlides / Video
• Agent Memory and State Across StepsSlides / Video
• Multi-Step Task Decomposition by AgentsSlides / Video
• Common Tool Types for LLM AgentsSlides / Video
• Tools vs Functions vs AgentsSlides / Video
• Agent Reasoning Loops: The ReAct PatternSlides / Video
• What an LLM Agent IsSlides / Video
• Structured Output via Function CallingSlides / Video
• Handling Invalid or Hallucinated Tool ArgumentsSlides / Video
• Controlling Tool ChoiceSlides / Video
• Parallel and Multiple Tool CallsSlides / Video
• The Full Function Calling LoopSlides / Video
• Returning Tool Results Back to the ModelSlides / Video
• Executing the Function in Your CodeSlides / Video
• Parsing the Model's Tool Call ResponseSlides / Video
• Token Cost of Tools and LoopsQuiz
• Handling Hallucinated CallsQuiz
• What an LLM Agent IsQuiz
• The ReAct LoopQuiz
• Agents vs Function CallingQuiz
• Common Agent ToolsQuiz
• Agent Memory and StateQuiz
• Stopping ConditionsQuiz
• Returning Tool ResultsQuiz
• The Tool Message RoleQuiz
• Agent Frameworks OverviewQuiz
• Error Handling and RetriesQuiz
• Security Risks of Tool CallsQuiz
• What Function Calling IsQuiz
• Why LLMs Need ToolsQuiz
• Defining a Tool SchemaQuiz
• JSON Schema for ParametersQuiz
• Writing Function DescriptionsQuiz
• Passing Tool DefinitionsQuiz
• How the Model Decides to CallQuiz
• Reading the Tool CallQuiz
• Parsing Call ArgumentsQuiz
• Multi-Step Function LoopQuiz
• Parallel Tool CallsQuiz
• Forcing Tool ChoiceQuiz
• Structured Output vs Function CallingQuiz
• Executing the FunctionQuiz
• How the Model Decides to Call a ToolSlides / Video
• Passing Tool Definitions in an API RequestSlides / Video
• Tool Definitions and JSON SchemaSlides / Video
• What Function Calling Is and Why It ExistsSlides / Video
• Forcing a specific tool_choiceCode Quiz
• Validating arguments against allowed valuesCode Quiz
• Building the tool-call loopCode Quiz
• Writing function descriptionsCode Quiz
• Handling malformed argumentsCode Quiz
• Matching tool_call_id to resultCode Quiz
• Handling parallel tool callsCode Quiz
• Parsing tool call argumentsCode Quiz
• Distinguishing final vs tool responseCode Quiz
• Parameter types and required fieldsCode Quiz
• Returning tool errors to the modelCode Quiz
• Dispatching tool names to handlersCode Quiz
• Preventing infinite loopsCode Quiz
• Sending tool results backCode Quiz
• Passing the tools arrayCode Quiz
• Appending the tool result messageCode Quiz
• Executing the local functionCode Quiz
• Detecting a tool callCode Quiz
• Preserving the assistant tool_calls messageCode Quiz
• Stopping Conditions and LimitsFlashcard
• Parsing the Tool Call ResponseFlashcard
• Agent Frameworks OverviewFlashcard
• Executing and Returning ResultsFlashcard
• Parallel Function CallsFlashcard
• Handling Call ErrorsFlashcard
• Structured Output via Function CallingFlashcard
• What an LLM Agent IsFlashcard
• Planning and Task DecompositionFlashcard
• Security Risks of Tool ExecutionFlashcard
• Why LLMs Need ToolsFlashcard
• Passing Tools in a RequestFlashcard
• The Full Function Calling LoopFlashcard
• Tool Choice ControlFlashcard
• Writing Effective Tool DescriptionsFlashcard
• The ReAct PatternFlashcard
• Agent Memory and StateFlashcard
• The Agent LoopFlashcard
• Multi-Tool OrchestrationFlashcard
• Function Calling vs RAG vs Fine-TuningFlashcard
• Cost and Latency of Agent LoopsFlashcard
• What Function Calling IsFlashcard
• Defining a Tool SchemaFlashcard
• Describing Parameters and TypesFlashcard
• How the Model Picks a ToolFlashcard