Upload a spreadsheet directly to Klairr. CSV (.csv), Excel (.xlsx), and OpenXML files (.xls) are all accepted.
When to use this
- You have an export from another system (NetSuite, ERP, custom tool) and want answers without piping it into a warehouse first.
- The data is small enough to be answered as a snapshot — under ~100k rows.
For larger or live-updating data, point Klairr at the source database instead.
Setup
- Open Settings → Connectors → Add connector → Excel & CSV
- Pick or drag-drop the file
- (Excel only) Choose the sheet name if your workbook has multiple tabs
- Confirm whether the first row is a header
Klairr parses the file at upload, infers column types from the first 100 data rows, and stores the parsed rows encrypted on Klairr’s side. The original file is not retained.
Limits
- 100 MB per file
- 100,000 rows per file (rows beyond the cap are truncated)
- Mixed-type columns are flagged in the schema (e.g. “87% number, 13% text”)
Defence-in-depth
Cells starting with =, +, -, or @ are escaped with a leading apostrophe at parse time so a malicious export cannot trigger an Excel formula on a downstream download. Klairr never evaluates spreadsheet formulas.
Querying
The LLM emits a JSON DSL the connector executes in-memory: filters (eq, gt, contains, …), groupBy, aggregate (count, sum, avg, min, max), sortBy, limit. No SQL runs against uploaded files in the design-partner phase; DuckDB-backed SQL is on the roadmap for files that outgrow the in-memory path.
Refreshing
To update the data, re-upload the file. The connector keeps its name and schema cache; only the rows are replaced.