Import and manage test cases

Use an import when your cases already live in a file. Download ticket-routing.json for a five-case starting point, or follow the format reference to prepare your own.

Import a file

Open the prompt's dataset import control, select a JSON or TSV file, and inspect the preview. Check that inputs and references landed in the correct fields and that JSON fields retain their format.

Choose Append to keep existing cases and add the new rows. Choose Replace only when the file should become the dataset; replacement removes the previous rows from future evaluations. Review row counts, rejected entries, duplicates, and remaining capacity before committing the import. Exact duplicate cases are skipped. If every row is invalid, correct the file before retrying.

Screenshot placeholder — Dataset import preview Show the tutorial file, append/replace choice, mapped input/reference fields, and accepted, duplicate, and invalid row counts.

Edit and organize

Edit a cell to correct its content or format. Reorder cases to put a useful smoke-test set in the first 10 positions used by Quick evaluations. Delete selected cases only after reviewing the selection. Existing completed runs keep their captured inputs and results.

Changes use the current dataset revision. If someone edits the dataset while you are importing or rearranging it, reload the dataset and review the change rather than blindly retrying against the new revision.

Import through automation

The dashboard accepts a JSON array of cases. The management API and CLI wrap that same array in a body containing rows, mode, and the current revision. Read the dataset first and use the revision from that response. See the CLI walkthrough for a complete example.

Files may be up to 5 MB and an import request may contain up to 1,000 rows, but your dataset's plan capacity still applies. A large file does not raise the number of cases the prompt can store.