koog-agents-optimization/ai.koog.agents.optimization.optimizers.ace

Package-level declarations

Types

class ACEOptimizer<Input, Output, InputLabel>(playbookStoragePath: ResilientPath, onExistingPlaybook: OnExistingPlaybookAction, reflectorModel: LLModel, curatorModel: LLModel, labelExtractor: (datasetItem: TrainSetItem<Input, InputLabel>) -> String) : AgentOptimizer<Input, Output, InputLabel>

Agentic Context Engineering optimizer: improves an agent by curating an ACEPlaybook of bullet insights instead of editing its prompt directly.

Mutable collection of curated PlaybookSections persisted to storagePath.

ACE-specific feature that appends playbook content to the agent's system prompt at strategy start.

Configuration for ACEPlaybookFeature.

Prompt builders for the ACE reflector and curator stages, and for injecting the playbook at run time.

@Serializable

sealed class DeltaUpdate

A single mutation to apply to an ACEPlaybook, as produced by the curator and consumed by ACEPlaybook.applyDelta.

What an ACEOptimizer should do when a playbook already exists at the target storage path.

@Serializable

data class PlaybookBulletItem(var content: String, val id: String, var helpfulCount: Int, var harmfulCount: Int)

A single curated insight in an ACEPlaybook, living inside a PlaybookSection.

@Serializable

data class PlaybookSection(val name: String, val shortName: String, val bullets: MutableList<PlaybookBulletItem>)

A named group of PlaybookBulletItems within an ACEPlaybook.

fun GraphAIAgent.FeatureContext.installACEPlaybook(configure: ACEPlaybookFeatureConfig.() -> Unit)

Installs the ACEPlaybookFeature with the given configuration.