Data Explorer

It accepts CSV with a header row, JSON arrays of objects, and single JSON objects. The parser auto-detects JSON from .json filenames or content that starts with { or [, and you can still force CSV or JSON manually with the format toggle.

No. Parsing and profiling happen locally in your browser. The pasted or uploaded content is not sent to a server, not stored in the URL, and disappears when you reload the page.

The page rejects pasted or uploaded text over roughly 10 MB, and profiling is capped at the first 50,000 rows to keep the browser responsive. When truncation happens, the page clearly says so and scopes the stats to the profiled rows.

Each column is scanned for non-empty values and labeled as string, number, boolean, date-like, null/empty, or mixed. Null, empty-string, and missing cells are counted separately so nullable numeric or date columns can still infer correctly.

It is intentionally lightweight. The tool flags columns that are almost entirely one value except for a few outliers, and numeric columns with values that sit far from the median using a simple median absolute deviation heuristic. It is a quick triage signal, not a statistically rigorous quality audit.

Each profiled row is converted into a stable string representation of all column values. If the same full row appears again, it counts as a duplicate row.

Yes. The page links out to CSV Viewer, CSV to SQL, JSON to Schema, and the Mock Data Generator as next steps, but those links intentionally open blank so your local data stays private.