Supplier CSV Price & Stock Delta Guard — pilot

Compare two supplier CSV exports before passing changed price/stock rows into an import workflow. Runs entirely on supplied text; no storefront access, crawling, merchant passwords or data mutation. This pilot is not currently a paid public listing.

Supply previousCsv and currentCsv, using the same headers and currency/unit conventions. Defaults are sku,price,stock; columns accepts exact alternate header names. SKUs remain case sensitive and retain leading zeroes. Numbers must use unsigned plain decimal notation, period decimal separator, no currency symbols/grouping, and stock must be integral. Blank, negative and ambiguous values block the report.

The report blocks all import output when either feed has duplicates, invalid numbers/rows, empty data, more than 10% missing SKUs, price changes over 30%, or more than 20% of formerly positive-stock matched items becoming zero. Fractions are configurable in [0,1]. New and missing products are reported but never emitted as import instructions. Missing SKUs are never treated as zero stock. These are anomaly thresholds, not proof the remaining data is correct.

Outputs: OUTPUT JSON contains status, counts, issues, source line references, SHA256 hashes and exact change records, including signed priceDelta and stockDelta for changed existing SKUs. UPDATES.csv is produced only for ready reports and contains changed existing SKUs. Formula-leading SKUs are apostrophe escaped for spreadsheet safety; this changes the identifier in the CSV. Use the JSON exact SKU if your downstream importer requires the original identifier. Do not import escaped identifiers blindly. No platform-specific import compatibility is claimed.

Limits: comma-separated UTF-8 CSV text only, 2 MB and 20,000 rows per feed; no URL fetching, Excel, PDFs, currency conversion, variant creation, automatic baseline selection or store writes. Your automation supplies both snapshots each time. A passing report does not guarantee profitability, correct currency, tax treatment or accurate supplier data. Only upload data you are authorized to process; inputs/results are stored in your Apify run storage under Apify retention settings. No feed content is printed in logs.

Local use: python3 main.py example-input.json. Tests: python3 -m unittest -v test_guard.py. In Apify, inspect OUTPUT and require status == "ready" before considering UPDATES.csv. Never build downstream import logic around run success alone: blocked and invalid-input reports can finish normally.

Commercial hypothesis: pay per completed comparison, proposed $0.25, unvalidated. No payment event is implemented in this private prototype. No customers, subscriptions, income, or payout readiness are claimed. Category demand comes from historical paid supplier-import testimony; standalone willingness to pay remains unproven.