koog-agents-optimization/ai.koog.agents.optimization.optimizers.gepa/GEPAMergeConfig

GEPAMergeConfig

data class GEPAMergeConfig(val enabled: Boolean = false, val maxMergeInvocations: Int = 5, val maxMergeAttempts: Int = 10, val maxFindSharedAncestorAttempts: Int = 10, val valSetSubsampleSize: Int = 5)

Runtime controls for GEPA's optional merge step.

Merging is experimental: tests reach its proposal step and no further, it has never run end to end, and it may diverge from GEPA's original implementation more significantly.

A merge combines the module one lineage changed with the module the other changed, so it needs an agent with at least two optimizable modules. Enabling it for a single-module agent does nothing.

Constructors

GEPAMergeConfig

constructor(enabled: Boolean = false, maxMergeInvocations: Int = 5, maxMergeAttempts: Int = 10, maxFindSharedAncestorAttempts: Int = 10, valSetSubsampleSize: Int = 5)

Properties

enabled

val enabled: Boolean = false

Whether the optimizer should attempt candidate merges.

maxFindSharedAncestorAttempts

val maxFindSharedAncestorAttempts: Int = 10

Maximum number of candidate-pair samples used while searching for parents with a shared ancestor.

maxMergeAttempts

val maxMergeAttempts: Int = 10

Maximum number of merge proposals to try for one due merge.

maxMergeInvocations

val maxMergeInvocations: Int = 5

Maximum number of accepted merges during one optimization run.

valSetSubsampleSize

val valSetSubsampleSize: Int = 5

Number of validation examples used to decide whether to accept a merge.