Product / Files
Your files stay files
Most desktop analytics tools swallow a spreadsheet on import and forget where it came from. CLAIR keeps the link, so repeated work on the same file stays cheap and honest.
1,204 → 1,318 rows (+114)
What it does
A multi-sheet Excel workbook, .xlsx or the legacy .xls format, imports through a sheet picker: one card per sheet with measured row and column counts and a preview, so you choose what to bring in. A sheet you skip is remembered, not discarded, and can be imported later without re-picking the file off disk. Import through the native file picker and CLAIR remembers where the file actually lives on your disk; a file you dragged in or pasted has no path to remember, and CLAIR labels that honestly rather than treating it as an error.
How it works
Change detection is a plain file-system check, modified time and size, which costs nothing: a content hash is only computed when those two actually differ, so re-saving a file without editing it never reports a false change. Refreshing re-reads the file using the same settings it was imported under and tells you what moved, for example 1,204 to 1,318 rows, and whether any columns changed. Every refresh writes a change-log entry with a restorable snapshot, so "the numbers changed because the file changed" stays answerable, and undoable, later. If columns were added, removed or retyped, refresh stops and asks instead of proceeding silently.
Honest limits
- Single-file upload size scales with your machine's memory rather than a fixed number: roughly 300 MB on an 8 GB machine, up to 1.5 GB or more on 32 GB. Column count is capped at 2,000 regardless of RAM, checked before import starts.
- Only manual refresh, "ask me first", is wired today. Automatic refresh is stored as a preference but nothing acts on it yet.
- Cancelling a long import takes effect once the file-parsing step itself finishes, not instantly, because the reader has no point where it can stop mid-parse.
Related
Chat over your data with cited, verified answers.
Resumable AutoML that runs while your machine is idle.
A model for one outcome column, with a range, not just a number.
Background jobs that make chat and analysis sharper after upload.
HTML, PowerPoint, PDF, and publication-ready figures.
Boards of charts and pinned findings, exported as one PDF.