Squirreling is a SQL engine built natively for the browser. It ships at about 9 KB minified and gzipped with zero external dependencies, and it is designed for instant startup and constant memory usage on streaming queries. The goal is interactive, AI-friendly querying over data that already lives in the browser, without standing up a backend server.
The project exists because existing browser SQL options compile a full analytical engine to WebAssembly. The announcement post points at DuckDB-Wasm as an example: a footprint above 4 MB, a synchronous execution model that limits true streaming, and data copying between WebAssembly linear memory and the JavaScript heap. Squirreling takes the opposite approach, staying small and dependency free so queries can start instantly and stream results. It was built for Hyperparam, a browser based tool for transforming and analyzing massive LLM datasets, where users query large text heavy columns such as chat logs alongside structured labels and scores.
Squirreling is open source under the hyparam GitHub organization. Its author, Hyperparam's Kenny Daniel, records that he made the first commit on November 15, open sourced it on November 22 and had it live inside Hyperparam by November 26, crediting AI assistance for the speed. It is a young, single purpose library, so evaluate its SQL coverage against your query shapes before adopting it.
