Enum ChunkQualityProfile

Namespace

LMKit.Retrieval.ChunkQuality

Assembly

LM-Kit.NET.dll

Calibrated sensitivity presets for the SemanticChunkQualityGate. Each profile selects a frozen set of decision thresholds validated against a labeled corpus of real-world extraction noise and multilingual clean content.

public enum ChunkQualityProfile

Fields

Lenient = 0

Drops only unequivocal garbage (binary noise, hard mojibake, exact duplicates). Borderline chunks are indexed. Choose this when recall matters more than index purity.

Balanced = 1

The recommended default: rejects noise with high confidence while protecting short, numeric, tabular, and low-resource-language content. Calibrated so that clean chunks are essentially never dropped.

Aggressive = 2

Additionally flags weak boilerplate and navigation fragments that Balanced lets through. Choose this for high-volume ingestion pipelines where recurring low-value chunks measurably degrade search precision.