DuckDB

Query your CSV, TSV, and Parquet files with plain SQL. Nothing to install and no server to start — the DuckDB engine runs inside the app.

What you can do

• Write SQL in a real editor with completion that knows your schema, and run it with ⌘/Ctrl+Enter
• Run only the part you selected, cancel a long query, and see how long it took
• Read results in a fast, sortable grid that handles large result sets, shows column types, and lets you copy a cell or a whole row
• Double-click a .parquet file, or click a file in the sidebar's Sources section, and get SELECT * … LIMIT 1000 straight away
• Open .sql files directly into the workbench
• Browse your tables and columns, keep saved queries, and go back through your query history
• Export a result as a CSV or JSON download, or write the full query result back into your workspace

Notes and limits

• The database is in memory. The files you query are remembered and re-registered the next time you open the workspace
• Parquet and dataset files need the desktop app. CSV and TSV in your workspace work everywhere
• Files are copied into memory, so anything larger than 512 MB is refused
• Integers beyond 2⁵³ come back as strings so no precision is lost
• The engine downloads once, around 34 MB, on your first query. The extension itself starts instantly

Getting started

Double-click a .csv or .parquet file in the explorer, or run DuckDB: New SQL Query from the command palette.