Docs · Step 3
Import a file
CLAIR reads your file on this machine and tells you what it found before you ask anything.
1. Drop in a file
- Drag a
.csv,.xlsxor legacy.xlsfile onto the CLAIR window, or use the file picker. - A file dragged or pasted in has no known location on disk, which is a normal state, not an error. Import through the native file picker instead if you want CLAIR to remember where the file lives, so it can tell you later when the source has changed.
2. Pick sheets, if it's a workbook
A multi-sheet .xlsx or .xls opens a sheet picker: one card per sheet, with a measured row and column count and a preview. Tick the sheets you want; each becomes its own dataset in CLAIR, kept together under the same file. Sheets you skip are remembered, not discarded, and can be imported later without re-picking the file off disk.
3. Read the import notices
CLAIR checks the file for the kinds of problems that quietly produce wrong numbers, and shows what it found rather than silently fixing or ignoring it:
- A completely blank row is removed, and the removal is disclosed with a count.
- A row that looks like a subtotal or grand total is flagged, never auto-deleted, so a sum can't silently double-count it.
- A row with more or fewer values than the header expects is flagged rather than silently mis-read.
- The column delimiter (comma, semicolon, or pipe) is detected rather than assumed, so a European-style export opens as columns, not one glued field.
Accepted with notes
A file with any of the above still imports. "Accepted with notes" means CLAIR kept a value you can see rather than guessing silently, not that something failed.
4. Let background processing finish
After import, CLAIR profiles every column, checks for data-quality issues, and prepares your file for questions, all in the background. You can start using the dataset immediately; a short delay before every feature is ready is normal on a larger file.
Honest limits
- File size scales with your machine's RAM rather than a fixed number: roughly 300 MB on an 8 GB machine, up to 1.5 GB or more on a 32 GB machine.
- Column count is capped at 2,000 regardless of RAM, enforced before import starts with a clear error naming the count and the limit.
- A long import can be cancelled. CLAIR checks the cancel flag twice, including immediately before the final database write, so nothing partial is ever saved. The one exception: the file-parsing step itself can't be interrupted mid-parse, so cancelling during parse takes effect as soon as parsing ends rather than instantly.