Generates a per-column data dictionary for a tabular dataset — type, nullability, range, unique-count, semantic class (ID / categorical / continuous / ordinal / text / datetime / boolean), sample values, and inferred role (feature / target / leakage-risk / PII). Use whenever receiving a fresh client / partner / vendor dataset; before fitting any model on a dataset whose columns are not already documented; whenever a column's purpose is ambiguous; or whenever onboarding a teammate to a dataset they have not worked with before.
