Preprocessors transform loaded entries before scanners run. CerbIA Core exports
SpeculativeDecodingPreprocessor and
WhitespaceNormalizationPreprocessor. Both preserve input-entry cardinality
and order, and retain preprocessing lineage metadata. Configure preprocessors
in order: each preprocessor’s output becomes the next one’s input.
-
Speculative decoding searches for plausible encoded or obfuscated text.
-
Whitespace normalization reduces whitespace patterns that can split or hide text.
Preprocessing is conservative. An entry that needs no change is passed through unchanged as the original entry object. A changed entry retains its source, content type, and lineage; it records the original text and transformation metadata. See Architecture for its position in the full pipeline and Configuration for the order of component records.