Docs/jAIn/Upload

jAIn

Upload your data

One file starts a session. These are the rules it has to satisfy, and what to do when it doesn't.

What we accept

Two formats, decided by the file's extension: .csv (comma separated) and .tsv (tab separated). Matching is case-insensitive and only the last extension counts, so sales.2026.CSV is read as a CSV.

The limit is 100 MB per file.

Spreadsheets

Spreadsheet files — .xlsx, .xls, .xlsm, .xlsb, .ods and .numbers — are not read directly. Uploading one is not a mysterious parse failure; you get told to export the sheet as CSV, which is a menu item in every one of those programs (typically File → Export or Save As, then pick CSV).

Export one sheet at a time. A workbook with several tabs has no single answer to "which table is the data", so the choice is yours rather than ours.

What the file should contain

  • A header row naming the columns. The header is checked as soon as the file arrives.
  • One row per observation, and one column you want predicted — two outcomes reads as binary classification, a handful as multiclass, a number as regression. Those three are the whole of what jAIn searches for; what it does not do is on the product page.
  • Blanks are fine — missing values are part of real data and are handled.

If something is rejected

Rejections happen before any compute starts, so a bad file costs you nothing but the upload. The message says which rule was missed — the wrong extension, a file over the size limit, or a file with no extension at all, which leaves nothing to dispatch on.

One run at a time

Each account runs one session at a time. Starting a second while the first is going gets you a clear "you already have a run going" rather than a queue that silently grows. Wait for the first to finish, or stop it.

Your file is deleted automatically after 30 days. What the run made from it — your report, your trained model and its evidence receipt — is kept. See what we keep.