Dataset Reconciliation Workbench

Compare two local CSV or flat-object JSON datasets without uploading them or silently collapsing duplicate keys.

Left dataset

FormatData

Right dataset

FormatData

Keys, alignment, and comparison

Left key columns (comma-separated)Right key columns (same count)Field mappings, one left:right per line

Numeric toleranceDate tolerance (days)

Trim keys and compared strings Compare strings case-insensitively Compare numeric-looking values numerically Compare ISO date-looking values by timestamp (timezone-free values use UTC)Merge-candidate precedence

Limits per side: 3 MiB, 50,000 rows, 200 columns. Preview: 50 groups/page. Missing-key, duplicate-key, and conflicting groups remain unresolved and are excluded from merge candidates.

Reconciliation results

Matched key groups and row-level differences
Class	Key	Left rows	Right rows	Field differences