Dataset FAQ & fixes
Common dataset questions, and fixes for the errors you hit most
In this page
The questions people ask most about datasets, and the errors they run into, with a direct fix for each, in the table below. If your answer isn’t here, reach out via support.
Common errors and fixes
| Symptom | Cause | Fix |
|---|---|---|
| Upload rejected before it starts | The file isn’t .csv, .xls, .xlsx, .json, or .jsonl, or it’s over 25 MB | Convert or split the file to fit; see Limits & Data Types for every cap |
| ”A dataset with this name already exists in your organization” when creating | Names must be unique per organization, so the clash can be with a dataset you don’t have access to see | Pick a different name |
| Upload finished but the dataset shows no rows yet | The file is still processing in the background | Wait it out; the rows land once it finishes |
| Image, audio, or document cells fill in slowly after upload | Media cells upload in batches with retries, not all at once | No action needed, it catches up on its own |
| A dataset built from Observe traces fills in gradually | Spans convert into rows in chunks, not all at once | Wait it out; the row count climbs while the conversion runs |
| A freshly added eval column sits idle for a few seconds | Evals are picked up by a poller on a short cycle rather than dispatched the instant you add the column | Give it a few seconds; it starts on its own |
| A cell is stuck showing running | The producer behind it, a run prompt, an eval, or another dynamic column method, hasn’t been picked up yet or is still executing | Wait a few seconds; if the column status shows Failed or Error, rerun the column |
| A cell shows error | That row failed whatever produces the column | Rerun the column |
| Synthetic generation fails | It runs as a background job and can fail partway through | Reopen it with Configure Synthetic Data; the drawer returns with your saved configuration, so fix whatever caused the failure and generate again |
| A column or row you expected is gone | Most likely it was deleted; the grid can also hide columns and filter rows. Deleting a column also removes the producer behind it (a run prompt, an eval) and anything derived from it | There’s no copy to restore; recreate the column or add the rows back |
| Add Row or Add Column is greyed out | Both are disabled while the dataset is processing or synthetic generation hasn’t finished; Add Column alone also stays disabled until the dataset has at least one row | Wait for processing or synthetic generation to finish, or add a row first if you’re adding a column to an empty dataset |
| Duplicate or Delete is greyed out | Permission gates both: duplicating a row needs update access, deleting one needs delete access | Check with an org admin about your dataset permission |
Keep exploring
Create a dataset
Every way to get a dataset that exists and has data in it
Manage datasets
Rename, duplicate, export, and delete once the data is in
Static & Dynamic Columns
Where a column’s values come from, and what deleting one takes with it
Limits & Data Types
Every row, column, and file cap in one table
Was this page helpful?
Questions & Discussion