fast-csv-import

Use @fast-csv/parse to stream-parse the uploaded file in an API route, validate each row with a Zod schema, then upsert products into the database. Return a result object with imported count, skipped count, and per-row error details.

Define a Zod schema matching your CSV columns using z.coerce.number() for numeric strings, then call safeParse on each row in the parser's data event. Failed rows are collected into an errors array with row number, SKU, and issue messages instead of being inserted.

Yes, @fast-csv/parse supports stream-based parsing via Node.js Readable streams, processing rows incrementally rather than loading the entire file into memory. Pipe the buffer through Readable.from() into the parse stream for constant memory usage.

Use a database upsert keyed on a composite unique constraint of tenantId and SKU. Re-importing the same file updates existing product rows with new prices and stock instead of creating duplicates.

CSV parsers return all values as strings, so numeric columns fail strict number validation. Use z.coerce.number() in the Zod schema to convert string values, and store prices as integer paise to avoid floating-point rounding errors.