SuperAPI AI Directory

line-normalize

Data & analytics

View repository

Updated 11-03 10:02

About this skill

Line Normalize

Purpose

Make lines machine-comparable (CS-24 -> CS24, "5 hundred" -> 500) while preserving the customer's intent verbatim in a raw field.

When to use

After order-ingest in every run.

Inputs

order.json.

Steps

  1. Normalise tokens; keep raw_value alongside every normalised field.
  2. Never resolve an alias to a product or "fix" a price - normalisation is formatting, not judgment.

Output

order.json with normalised lines.

Grounding requirements

raw_value preserved per field.

Constraints

  • Formatting only; ambiguity survives normalisation for the validator to catch.

Escalation / uncertainty

Token it cannot normalise deterministically -> left raw + flagged.