Data processing workflow
Map, transform, and validate structured inputs before any write-back.
Keep reading, transformation, and writing in separate steps. Start with a small JSON array as run input. If you later bind a reader, map the selected sheet, document, or app output and verify whether it exposes rows, a file reference, or a summary.
| Task | Configuration |
|---|---|
| Parse raw text | Use csv_parse for CSV text or jsonl_parse for one JSON value per line. Supply the text as a mapped input. |
| Reshape rows | Use a declared Data transform such as pick, rename, or dedupe and choose a stable key for deduplication. |
| Verify the result | Compare input counts and expected rows in run details. Check nulls, missing keys, and records that share a key. |
| Deliver or write back | Use a supported artifact-producing step before project delivery. For a connected write, pass the proposed changes to a Review gate and verify the write’s receipt. |
Start with the deduplication example or see data and flow nodes.