DICE API Documentation/com.embabel.dice.storage.autoconfigure/CollectorProperties

CollectorProperties

data class CollectorProperties(val enabled: Boolean = true, val matchThreshold: Double = 0.6, val signals: Map<String, CollectorProperties.SignalProperties> = emptyMap(), val sweep: CollectorProperties.SweepProperties = SweepProperties(), val trace: CollectorProperties.TraceProperties = TraceProperties())

Configuration for the multi-signal duplicate collector.

embabel.dice.collector.enabled is the master switch (on by default). Each signal — the built-in pair sources and scorers CollectorAutoConfiguration wires up — gets its own entry under signals, keyed by its short name (vector, lexical, entity-overlap, grounding-overlap, provenance-overlap, polarity-veto), so ops can disable one or tune its weight without touching code.

Constructors

CollectorProperties

constructor(enabled: Boolean = true, matchThreshold: Double = 0.6, signals: Map<String, CollectorProperties.SignalProperties> = emptyMap(), sweep: CollectorProperties.SweepProperties = SweepProperties(), trace: CollectorProperties.TraceProperties = TraceProperties())

Types

SignalProperties

data class SignalProperties(val enabled: Boolean = true, val weight: Double? = null, val similarityThreshold: Double? = null, val topK: Int? = null)

One signal's override. weight, similarityThreshold and topK are nullable so "not set" is distinguishable from "set to the scorer's own default" — a null weight falls back to the scorer's own neutral default instead of forcing a specific number.

SweepProperties