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

Package-level declarations

Types

open class AIAgent<Input, Output>(val inputType: KType, val outputType: KType, val promptExecutor: PromptExecutor, strategy: AIAgentStrategy<Input, Output>, val agentConfig: AIAgentConfigBase, val id: String = Uuid.random().toString(), val toolRegistry: ToolRegistry = ToolRegistry.EMPTY, val clock: Clock = Clock.System, installFeatures: AIAgent.FeatureContext.() -> Unit = {}) : AIAgentBase<Input, Output> , AIAgentEnvironment, Closeable

Represents an implementation of an AI agent that provides functionalities to execute prompts, manage tools, handle agent pipelines, and interact with various configurable strategies and features.

AIAgentBase

interface AIAgentBase<Input, Output>

Represents a basic interface for AI agent.

AIAgentException

open class AIAgentException(problem: String, throwable: Throwable? = null) : Exception

Represents a custom exception class for use in AI Agent-related processes.

AIAgentTool

class AIAgentTool<Input, Output>(agent: AIAgentBase<Input, Output>, agentName: String, agentDescription: String, inputDescriptor: ToolParameterDescriptor, inputSerializer: KSerializer<Input>, outputSerializer: KSerializer<Output>, json: Json = Json.Default) : Tool<AIAgentTool.AgentToolArgs, AIAgentTool.AgentToolResult>

AIAgentTool is a specialized tool that integrates an AI agent for processing tasks by leveraging input arguments and producing corresponding results.

ToolCalls

enum ToolCalls : Enum<ToolCalls>

Enum representing the modes in which a single-run strategy for an AI agent can be executed.

Functions

AIAgent

inline fun <Input, Output> AIAgent(promptExecutor: PromptExecutor, strategy: AIAgentStrategy<Input, Output>, agentConfig: AIAgentConfigBase, id: String = Uuid.random().toString(), toolRegistry: ToolRegistry = ToolRegistry.EMPTY, clock: Clock = Clock.System, noinline installFeatures: AIAgent.FeatureContext.() -> Unit = {}): AIAgent<Input, Output>

Convenience builder that creates an instance of AIAgent, automatically deducing AIAgent.inputType and AIAgent.outputType from Input and Output

fun AIAgent(executor: PromptExecutor, llmModel: LLModel, id: String = Uuid.random().toString(), strategy: AIAgentStrategy<String, String> = singleRunStrategy(), systemPrompt: String = "", temperature: Double = 1.0, numberOfChoices: Int = 1, toolRegistry: ToolRegistry = ToolRegistry.EMPTY, maxIterations: Int = 50, installFeatures: AIAgent.FeatureContext.() -> Unit = {}): AIAgent<String, String>

Convenience builder that creates an instance of an AIAgent with string input and output and the specified parameters.

asTool

inline fun <Input, Output> AIAgentBase<Input, Output>.asTool(agentName: String, agentDescription: String, inputDescriptor: ToolParameterDescriptor, inputSerializer: KSerializer<Input> = serializer(), outputSerializer: KSerializer<Output> = serializer(), json: Json = Json.Default): Tool<AIAgentTool.AgentToolArgs, AIAgentTool.AgentToolResult>

Converts the current AI agent into a tool to allow using it in other agents as a tool.

singleRunStrategy

fun singleRunStrategy(runMode: ToolCalls = ToolCalls.SINGLE_RUN_SEQUENTIAL): AIAgentStrategy<String, String>

Creates a single-run strategy for an AI agent. This strategy defines a simple execution flow where the agent processes input, calls tools, and sends results back to the agent. The flow consists of the following steps: