agents-core/ai.koog.agents.core.agent.context

Package-level declarations

Types

class AgentContextData(messageHistory: List<Message>, nodePath: String, lastInput: JSONElement? = null, lastOutput: JSONElement? = null, rollbackStrategy: RollbackStrategy, additionalRollbackActions: suspend (AIAgentContext) -> Unit = {})

AIAgentContext

interface AIAgentContext

The AIAgentContext interface represents the context of an AI agent in the lifecycle. It provides access to the environment, configuration, LLM context, state management, storage, and other metadata necessary for the operation of the agent. Additionally, it supports features for custom workflows and extensibility.

AIAgentFunctionalContext

class AIAgentFunctionalContext(environment: AIAgentEnvironment, agentId: String, runId: String, agentInput: Any?, config: AIAgentConfig, llm: AIAgentLLMContext, stateManager: AIAgentStateManager, storage: AIAgentStorage, strategyName: String, pipeline: AIAgentFunctionalPipeline, executionInfo: AgentExecutionInfo, parentContext: AIAgentContext? = null) : AIAgentFunctionalContextBase<AIAgentFunctionalPipeline>

Represents the context for an AI agent of FunctionalAIAgent type, serving as the execution environment and state holder while an agent operates within a predefined pipeline. It extends AIAgentFunctionalContextBase and is designed to allow configuration, state management, and storage for an agent's functional operations.

AIAgentFunctionalContextBase

expect abstract class AIAgentFunctionalContextBase<Pipeline : AIAgentPipeline> : AIAgentFunctionalContextBaseCommon<Pipeline>

Base class for functional/planner context implementations across all targets.

actual abstract class AIAgentFunctionalContextBase<Pipeline : AIAgentPipeline> : AIAgentFunctionalContextBaseCommon<Pipeline>

Base class for functional/planner context implementations across all targets.

actual abstract class AIAgentFunctionalContextBase<Pipeline : AIAgentPipeline> : AIAgentFunctionalContextBaseCommon<Pipeline>

Base class for functional/planner context implementations across all targets.

AIAgentFunctionalContextBaseCommon

open class AIAgentFunctionalContextBaseCommon<Pipeline : AIAgentPipeline> : AIAgentContext

Base AIAgentContext implementation providing functionality common across platforms

AIAgentGraphContext

class AIAgentGraphContext(environment: AIAgentEnvironment, val agentId: String, val agentInputType: TypeToken, val agentInput: Any?, val config: AIAgentConfig, llm: AIAgentLLMContext, stateManager: AIAgentStateManager, storage: AIAgentStorage, val runId: String, val strategyName: String, val pipeline: AIAgentGraphPipeline, executionInfo: AgentExecutionInfo, val parentContext: AIAgentGraphContextBase?) : AIAgentGraphContextBase

Implements the AIAgentGraphContext interface, providing the context required for an AI agent's execution. This class encapsulates configurations, the execution pipeline, agent environment, and tools for handling agent lifecycles and interactions.

AIAgentGraphContextBase

interface AIAgentGraphContextBase : AIAgentContext

The AIAgentGraphContextBase interface extends the AIAgentContextBase interface to provide a foundational context specifically tailored for AI agents operating within a graph structure.

AIAgentLLMContext

expect open class AIAgentLLMContext : AIAgentLLMContextCommon

Constructs a new instance of AIAgentLLMContext with the provided parameters.

actual open class AIAgentLLMContext : AIAgentLLMContextCommon

Constructs a new instance of AIAgentLLMContext with the provided parameters.

actual open class AIAgentLLMContext : AIAgentLLMContextCommon

Constructs a new instance of AIAgentLLMContext with the provided parameters.

AIAgentLLMContextCommon

abstract class AIAgentLLMContextCommon

Common AIAgentLLMContext implementation shared across platforms.

AIAgentPlannerContext

class AIAgentPlannerContext(environment: AIAgentEnvironment, agentId: String, runId: String, agentInput: Any?, config: AIAgentConfig, llm: AIAgentLLMContext, stateManager: AIAgentStateManager, storage: AIAgentStorage, strategyName: String, pipeline: AIAgentPlannerPipeline, executionInfo: AgentExecutionInfo, parentContext: AIAgentContext? = null) : AIAgentFunctionalContextBase<AIAgentPlannerPipeline>

Represents a context for the AI agent of PlannerAIAgent type, responsible for managing execution pipelines, configurations, and other contextual data required for agent operations.

DetachedPromptExecutorAPI

annotation class DetachedPromptExecutorAPI

Annotation for marking APIs as detached prompt executors within the AIAgentLLMContext.

RollbackStrategy

enum RollbackStrategy : Enum<RollbackStrategy>

SubtaskBuilder

class SubtaskBuilder(val context: AIAgentFunctionalContextBase<*>, val taskDescription: String)

A builder class responsible for creating and managing subtasks within the context of an AI agent's functional operations.

SubtaskBuilderWithOutput

class SubtaskBuilderWithOutput<Output : Any>(val context: AIAgentFunctionalContextBase<*>, val taskDescription: String, val output: OutputOption<Output>, var tools: List<ERROR CLASS: Symbol not found for Tool<*, *>>? = null, var llmModel: ERROR CLASS: Symbol not found for LLModel?? = null, var llmParams: ERROR CLASS: Symbol not found for LLMParams?? = null, var responseProcessor: ERROR CLASS: Symbol not found for ResponseProcessor?? = null, var runMode: ToolCalls = ToolCalls.SEQUENTIAL, var assistantResponseRepeatMax: Int? = null, var executorService: ExecutorService? = null)

Builder class to create and configure a subtask with specified input and output types.

Properties

agentContextDataAdditionalKey

val agentContextDataAdditionalKey: AIAgentStorageKey<AgentContextData>

A storage key used for associating and retrieving AgentContextData within the AI agent's storage system.

Functions

agentInput

inline fun <T> AIAgentContext.agentInput(): T

Utility function to get AIAgentContext.agentInput and try to cast it to some expected type.

feature

inline fun <TFeature : Any> AIAgentContext.feature(feature: AIAgentFeature<*, TFeature>): TFeature?

fun <TFeature : Any> AIAgentContext.feature(featureClass: KClass<TFeature>, feature: AIAgentFeature<*, TFeature>): TFeature?

Retrieves a feature from the AIAgentContext.pipeline associated with this context using the specified key.

featureOrThrow

inline fun <TFeature : Any> AIAgentContext.featureOrThrow(feature: AIAgentFeature<*, TFeature>): TFeature

Retrieves a feature from the AIAgentContext.pipeline associated with this context using the specified key or throws an exception if it is not available.

getAgentContextData

fun AIAgentContext.getAgentContextData(): AgentContextData?

Retrieves the agent-specific context data associated with the current instance.

removeAgentContextData

fun AIAgentContext.removeAgentContextData(): Boolean

Removes the agent-specific context data associated with the current context.

rootContext

fun AIAgentContext.rootContext(): AIAgentContext

Provides the root context of the current agent. If the root context is not defined, this function defaults to returning the current instance.

store

fun AIAgentContext.store(data: AgentContextData)

Stores the given agent context data within the current AI agent context.

with

inline fun <T> AIAgentContext.with(executionInfo: AgentExecutionInfo, block: (executionInfo: AgentExecutionInfo, eventId: String) -> T): T

Executes a block of code with a modified execution context.

inline fun <T> AIAgentContext.with(partName: String, block: (executionInfo: AgentExecutionInfo, eventId: String) -> T): T

Executes a block of code with a modified execution context, creating a parent-child relationship between execution contexts for tracing purposes.