How a Real SQLite Engine Runs Inside a Browser Tab
SQLite is normally a native, compiled database engine - not something you'd expect to run inside a web page. WebAssembly (WASM) changes that: SQLite's actual C source code is compiled to WASM once (via the open-source sql.js project) and loaded from a CDN, giving the browser a genuine, full SQL query engine running in-process. Your .sqlite file's bytes are loaded directly into that WASM-compiled engine's memory - there's no server-side database involved at any point, and no data ever has to leave your machine for a real SQL query to execute against it.
This is meaningfully different from a JavaScript reimplementation of SQL parsing - it's the actual SQLite codebase, so the SQL dialect and behavior match a normal SQLite installation exactly.
What Makes a File "Table-Like" to This Tool
A .sqlite file already has real, explicit table structure built in. The other supported formats don't, so this tool has to infer a reasonable table shape from each one: a JSON array of objects becomes one table (each object a row, each shared key a column); a JSON object whose values are themselves arrays becomes multiple tables, one per key; CSV and NDJSON each naturally map one record per row; and an Excel workbook's sheets each become their own separate table, matching how a workbook is already organized.
Whichever format you drop in, the result is the same browsable grid experience - only SQLite additionally supports running arbitrary SQL against it, since it's the only format with a real query engine attached.
Frequently Asked Questions
SQL querying requires an actual database engine to execute against - that's what sql.js provides for a .sqlite file. CSV, JSON, NDJSON, and Excel have no such engine; this tool parses them directly into a browsable table structure, which supports viewing but not arbitrary querying.
No - it's a fixed, open-source PROGRAM (not your data) fetched once from a CDN, the same way this page's icon font or any other shared library asset loads. Your actual database file's contents are never part of that request and are never uploaded anywhere.
Because that's how the workbook itself is already organized - each sheet is naturally its own dataset with its own columns, so mapping one sheet to one browsable table (rather than merging everything into one undifferentiated table) preserves the structure you already set up in the spreadsheet.
If the top level isn't an array of objects, and isn't an object whose values are arrays, there's no reasonable way to infer a table structure from it - the tool reports that rather than guessing at a table shape that wouldn't actually represent your data correctly.