Privacy

Status: the data tools are not released yet. This page describes the architecture they are being built on and will be updated, and checked against the shipped tools, as each one launches.

How the tools work

• Files you open in a tool are read by your browser and processed by DuckDB-Wasm running in your browser. The tools are designed so that file contents are not uploaded to our servers.
• The site is a frontend application. We do not run a database for user data, and there are no accounts.
• The processing engine, including its WebAssembly code and the Parquet and JSON components, is served from this site itself, not from a third-party CDN.
• The site tells your browser to refuse requests from the page and the engine to any other website (a Content-Security-Policy). In automated tests that start the engine, run queries and open test files, the only requests the browser made were to this site.

What we do not do

• We do not store the contents of your files.
• We do not log dataset contents or column values.
• We do not send file contents or column values to analytics or any third party.

Analytics

No analytics are installed at the moment. If we add privacy-conscious analytics later, they will record only coarse events such as which tool was opened, never file contents, column values or other dataset information, and this page will say so.

Limits of these guarantees

• Like any website, the hosting provider receives standard request information such as your IP address and user agent when you load a page. That is about loading the site, not about your files.
• Local processing depends on your browser and device. Very large files may be slow or fail because browsers limit memory. Tools will warn you about large files instead of promising unlimited size.
• The engine may write error messages, including part of a query and values from your data, to your browser's developer console. This stays in your browser, but keep it in mind when sharing screenshots or logs.
• These checks cover our own code and tests. They cannot account for browser extensions or software on your device.
• If a future feature ever needs server-side processing, it will be clearly separated from the local tools and labelled before you use it.