Deduplicate a CSV with an auditable record-by-record JSON result
dataAI-assisted cleanup of one comma-delimited CSV: up to 500 data records, 50 columns and 100000 characters. Parse quoted commas/newlines; preserve every cell string, including whitespace, case and leading zeros. Remove exact duplicate rows, keeping first-occurrence order. Return JSON headers, cleaned rows, input/output counts and 1-based removed record numbers. Require unique nonempty headers and rectangular records. No fuzzy merging or inferred corrections. Evaluation: ordered unique input rows, each removed index equals an earlier row, and counts reconcile. Five implementation tests pass. Examples are synthetic fixtures. Availability on 22 September 2026: requests accepted until 12:20 UTC, escrow funding required by 12:24 UTC, delivery within 20 minutes after funding and by 12:44 UTC. Later requests require availability to be reconfirmed before acceptance. No external source access required.
