agents-features-event-handler/ai.koog.agents.local.features.eventHandler.feature/EventHandlerConfig

EventHandlerConfig

class EventHandlerConfig : FeatureConfig(source)

Configuration class for the EventHandler feature.

This class provides a way to configure handlers for various events that occur during the execution of an agent. These events include agent lifecycle events, strategy events, node events, LLM call events, and tool call events.

Each handler is a property that can be assigned a lambda function to be executed when the corresponding event occurs.

Example usage:

handleEvents {
    onToolCall = { stage, tool, toolArgs ->
        println("Tool called: ${tool.name} with args $toolArgs")
    }

    onAgentFinished = { strategyName, result ->
        println("Agent finished with result: $result")
    }
}

Constructors

EventHandlerConfig

constructor()

Properties

messageProcessor

val messageProcessor: List<FeatureMessageProcessor>

onAfterLLMCall

var onAfterLLMCall: suspend (response: String) -> Unit

Handler called after a response is received from the language model.

onAfterLLMWithToolsCall

var onAfterLLMWithToolsCall: suspend (response: List<Message.Response>, tools: List<ToolDescriptor>) -> Unit

Handler called after a response with tool calls is received from the language model.

onAfterNode

var onAfterNode: suspend (node: AIAgentNodeBase<*, *>, context: AIAgentContextBase, input: Any?, output: Any?) -> Unit

Handler called after a node in the agent's execution graph has been processed.

onAgentFinished

var onAgentFinished: suspend (strategyName: String, result: String?) -> Unit

Handler called when an agent finishes execution.

onAgentRunError

var onAgentRunError: suspend (strategyName: String, throwable: Throwable) -> Unit

Handler called when an error occurs during agent execution.

onBeforeAgentStarted

var onBeforeAgentStarted: suspend (strategy: AIAgentStrategy, agent: AIAgent) -> Unit

Handler called when an agent is created.

onBeforeLLMCall

var onBeforeLLMCall: suspend (prompt: Prompt) -> Unit

Handler called before a call is made to the language model.

onBeforeLLMWithToolsCall

var onBeforeLLMWithToolsCall: suspend (prompt: Prompt, tools: List<ToolDescriptor>) -> Unit

Handler called before a call is made to the language model with tools.

onBeforeNode

var onBeforeNode: suspend (node: AIAgentNodeBase<*, *>, context: AIAgentContextBase, input: Any?) -> Unit

Handler called before a node in the agent's execution graph is processed.

onStrategyFinished

var onStrategyFinished: suspend (strategyName: String, result: String) -> Unit

Handler called when a strategy finishes execution.

onStrategyStarted

var onStrategyStarted: suspend (strategy: AIAgentStrategy) -> Unit

Handler called when a strategy starts execution.

onToolCall

var onToolCall: suspend (tool: Tool<*, *>, toolArgs: Tool.Args) -> Unit

Handler called when a tool is about to be called.

onToolCallFailure

var onToolCallFailure: suspend (tool: Tool<*, *>, toolArgs: Tool.Args, throwable: Throwable) -> Unit

Handler called when a tool call fails with an exception.

onToolCallResult

var onToolCallResult: suspend (tool: Tool<*, *>, toolArgs: Tool.Args, result: ToolResult?) -> Unit

Handler called when a tool call completes successfully.

onToolValidationError

var onToolValidationError: suspend (tool: Tool<*, *>, toolArgs: Tool.Args, value: String) -> Unit

Handler called when a validation error occurs during a tool call.

Functions

addMessageProcessor

fun addMessageProcessor(processor: FeatureMessageProcessor)