SQL

The engine is DuckDB, compiled into the app, reading the object where it lives.

SELECT region, count(*), sum(amount)
FROM 's3://sales-eu/orders/2026-08-24.parquet'
GROUP BY region
ORDER BY 2 DESC

A quoted path straight after FROM or JOIN is what names the data. It can be:

• an object in a bucket — s3://, gs://, az://, or a path in one of your roots;
• a folder that is a Delta, Iceberg or Hudi table;
• a Hive-partitioned export, or a YYYY/MM/DD prefix;
• a local file.

Nothing is staged

DuckDB reads the bucket directly, through the same signed requests the explorer uses, so a WHERE over a multi-gigabyte Parquet file touches the row groups it needs and no more. There is no temporary copy in a scratch folder and no upload anywhere.

What is queryable

parquet csv tsv json jsonl xlsx avro, plus SPSS .sav and SAS .sas7bdat .xpt.

ORC, Arrow and HDF5 open as grids and charts too; only SQL over them is missing, and the cell says so.

Running

Shift+Enter, or the run button in the cell's toolbar. The cell's label says what came back — the row count — and the grid under it is virtualized, so a million rows scroll rather than load.

Next: charts.