liang

zenian

A self-initiated Excel cleanup sample using fictional order data. From 28 source rows, 19 are validated, 7 are held for clarification and 2 exact duplicates are archived. The workbook keeps the raw export and a review log, with totals limited to validated Paid rows. The attached PDF shows the decisions and results. The Excel file is available on request.

An offline Python prototype using nine synthetic messages. It extracts labelled invoice fields from JSON, preserves source IDs and line references, identifies duplicates and conflicting versions, and keeps currency totals separate with Decimal arithmetic. The attached PDF shows the input and actual demo output. Source code, fixtures and six passing tests are available. This is a self-initiated prototype; it has no live email integration.