The Apply refiner step reformats extracted data according to your specifications, using a linked refiner module.
Refiner is a post-processing tool that takes raw model or agent extraction output and transforms it into the format required by downstream systems and validation rules. It operates on a spreadsheet-like interface where each row is a processed record and each column is a field.
Refiner provides a rich function library organized by category:
String functions — trimming, splitting, replacing, formatting text values
Numerical functions — arithmetic, rounding, unit conversions
Table functions — manipulating extracted tabular data
Logical functions — conditional logic, branching, null handling
Parsing functions — date parsing, currency parsing, structured text parsing
Validation functions — type checks, range checks, format validation
UDFs can also be called from within refiner for custom logic that exceeds the built-in function library. Learn more in Creating refiner programs.
Input file type — Specifies the source file format for the data you’re refining. The default, IBOCR, is a JSON file that represents the output for most flow steps. In typical flows, IBOCR is the appropriate choice. If you’re refining data that you’ve exported to CSV files, select CSV.
Refiner program — Select a refiner program. The refiner must be listed in the Modules panel to display as an option.
