Data Compare Engine

Problem Being Solved

When two CSV/XLSX files come from different systems, export formats, or manual edits, manually checking column names, row counts, and amount discrepancies can be costly. The problem is sharper when column names are not identical but refer to similar business concepts. Data Compare Engine targets this kind of structured data reconciliation, turning column matching, difference counting, amount-difference calculation, and report generation into one workflow.

How The Skill Works

The skill accepts two uploaded files, reads them with pandas, attempts to detect encoding, and stores them under /tmp/data-compare-engine/. It then uses an LLM via environment variables such as OPENAI_API_KEY or ANTHROPIC_API_KEY to infer likely corresponding columns between the two files. If automatic matching is not precise enough, you can manually specify the Join Key. During comparison, it reports fully matched rows, rows only in the left file, rows only in the right file, difference ratio, and, when an amount column is specified, the total amount discrepancy. The final Markdown report includes a statistical summary, an AI explanation of the differences, grouped details, and an optional CSV export of the discrepancy results.

Boundaries And Notes

This skill is best suited to tabular CSV/XLSX files, not direct database queries, message streams, or unstructured logs. Column matching depends on the LLM's interpretation of field names, so ambiguous columns should be handled with an explicit Join Key. Amount comparison only works if the relevant amount column is identified; otherwise the output remains row-level difference statistics. Different plans may also impose call-count and row-size limits, so large files should be checked against the applicable tier before running the comparison.