Data Quality Intelligence API
Profile supplied CSV and JSON tables, point at missing cells and duplicate rows, and propose explicit normalization without changing the source.
Infer a column type from the supplied CSV or JSON values. A column with more than one non-numeric type stays mixed. No value is rewritten.
Count present, missing, and distinct values for each column and report the minimum and maximum of the supplied values.
List the observed types in each column. A value that differs from the most common type is referenced by row and column and is not changed. A tie is not blamed on a row.
Find empty or whitespace-only cells and give each one a row and column reference. The row stays in the table.
Group records that match after trimming each compared cell. When key columns are supplied, only those columns are compared. No record is removed.
Flag numeric values outside a Tukey fence of 1.5 times the interquartile range when a column has at least eight numbers. Each flag is a possible outlier, not a confirmed error.
Propose trim, email lowercase, boolean tokens, and unambiguous thousands separators. Each change names its rule and keeps the original cell. Dates that are not already YYYY-MM-DD are not rewritten.
Combine the schema, profile, missing cells, duplicate groups, possible outliers, and proposed normalization for one supplied table. No record is removed and no outlier is confirmed.