Data foundation → analytics → optional AI
Data Science Workbench
Profile CSV, TSV, flat JSON arrays or NDJSON, review quality and associations, and explore proportional scenarios. Analysis runs in this browser with no API or AI call.
Works without network access after this page has loaded. Data stays in this tab and is cleared when you leave or reload. Sign-in and optional AI require network access.
Limits: 64 KiB UTF-8, 2000 rows, 24 columns. CSV/TSV require a header. Quote whole fields with double quotes; double embedded quotes. Quoted delimiters/newlines and UTF-8 BOM are supported. Cells are trimmed; empty cells are missing. Text, nonfinite numbers and magnitudes above 1e100 are excluded from numeric statistics. Spreadsheet formulas stay text. JSON requires flat objects with scalar/null values and unique keys; missing keys/nulls are missing cells. Keys are sorted, case must be consistent, numeric strings use the same numeric rules, and nested records are rejected. No schema validation, conversion or whole-file processing is implied.
File compatibility assessment
Choose the format and intended use. This offline assessment reviews declared requirements without opening or uploading a file. Moving original bytes, reading records and converting formats each need separate acceptance checks. Formats and sectors can be extended through reviewed adapters.
Optional AI interpretation
The workbench above works without AI. Explanation sends numeric aggregates, including extrema, with generic column labels to the configured model. It excludes raw records and original column names. AI chat access and network connectivity are required.