Zellio.io

CSV & SQL Data Playground

Open CSV files too big for Excel and query them with SQL — no upload.

Runs locallyWorks offlineShare link carries settings, never your data
Send output toImage to Text (OCR)Batch WorkspaceBuild a workflow from thisOutput stays on this page until you send it.

Frequently asked questions

How large a file can it actually handle?

Comfortably into the hundreds of megabytes, because the file is streamed off disk rather than loaded into memory. The ceiling is your machine's available RAM and the browser's 4 GB per-tab WebAssembly limit, not an upload cap — and a query that only touches two columns never reads the rest.

Is my data uploaded anywhere?

No. The file is opened through the browser's file API and queried locally, so it never leaves your device. You can verify it the same way you would for any tool here: open the Network tab and watch, or simply disconnect from the internet once the page has loaded and run a query.

Which formats and SQL does it support?

CSV, TSV, JSON and Parquet, with types sniffed automatically. The SQL is DuckDB's PostgreSQL-compatible dialect, so joins, window functions, CTEs and aggregates all work — including handy extras like SUMMARIZE for instant column statistics.

Why is the first load slower than the other tools?

Because the database engine itself is around 35 MB of WebAssembly, downloaded the first time you open a file. It is cached afterwards, so subsequent visits start immediately. That download is the trade for doing real analytical work without a server.

Do I need to know SQL?

Not to start. Dropping a file previews it immediately, and the preset buttons cover the common questions — row counts, column types and summary statistics — while showing you the SQL they ran, so you can adapt it rather than write it from scratch.

Pro tips

  • Query the file directly by name, e.g. SELECT * FROM 'data.csv' — no import step is needed.
  • SUMMARIZE SELECT * FROM 'data.csv' gives min, max, null counts and distinct values for every column at once.
  • Press Ctrl/Cmd + Enter to run the query without reaching for the mouse.
  • Parquet is dramatically faster than CSV on large files — convert once if you will query repeatedly.

About CSV & SQL Data Playground

Drop a CSV, TSV, JSON or Parquet file and query it with real SQL, powered by DuckDB compiled to WebAssembly. Because the file is read directly off your disk in byte ranges rather than uploaded or loaded into memory, it handles files in the hundreds of megabytes — well past Excel's 1,048,576-row limit — and nothing ever leaves your machine.

Spreadsheet software hits a wall long before your data does. Excel stops at 1,048,576 rows, and Google Sheets far sooner — so the moment an export grows past a few hundred thousand records, the usual workflow is to upload it somewhere, wait, and hope the service is trustworthy. This tool removes that step entirely: the file is read straight off your disk and queried in place.

It runs DuckDB, a column-oriented analytical database, compiled to WebAssembly. Crucially the file is never loaded into memory in one piece — the engine reads byte ranges on demand as a query needs them, which is what makes files in the hundreds of megabytes practical inside a browser tab. Your data never touches a network, so there is no upload wait, no size cap imposed by someone else, and nothing to delete afterwards.

SQL earns its place here by recording the method rather than the result. A spreadsheet captures what you got; a query captures how you got it, which means you can read it back in a month, hand it to a colleague as a precise description of the analysis, or change one clause and re-run. When a number is later questioned, the query is the answer to how it was produced — a chain of manual sorts and filters performed weeks ago is not.

Converting a large CSV to Parquet once usually pays for itself immediately. Parquet stores data column by column rather than row by row, so a query touching one column of forty reads only that column instead of the whole file, and its per-column type-aware compression typically shrinks the file five to tenfold. It also stores column statistics per chunk, letting a filtered query skip entire sections without decompressing them.

Common use cases

  • Opening a data export that is too large for Excel or Google Sheets
  • Joining two exports on a shared key without loading them into a database
  • Profiling an unfamiliar dataset before deciding what to do with it
  • Filtering a large log or transaction file down to the rows that matter
How it comparesCloud tools like Google Sheets, Airtable or an online CSV viewer all require uploading the file first, which means a wait proportional to your connection speed and your data sitting on someone else's infrastructure. A local database such as PostgreSQL or DuckDB CLI avoids that but needs installing and a schema. This sits between the two: no upload, no install, and no row limit.