linaldb

SQL-style querying and native tensor math, in one engine. No separate library, no round-tripping through pandas.

Why linaldb

linaldb is a database engine (persistence, SQL-style querying, joins) with real linear algebra and vector search built natively into the same engine, not stitched together from separate tools. It’s built for scientific and engineering research broadly, not just ML training: anywhere a workflow today means a relational database for the structured data, numpy/scipy for the math, and a separate vector index for similarity search, glued together by hand.

Getting Started

Install the CLI, embed via Python/R, or run the HTTP server, and write your first .lnl query.

DSL Reference

The full query language: joins, window functions, vector/matrix ops, decompositions, and more.

Use Cases

Real workloads against real data: gravitational-wave detection, single-cell PCA, recommender systems, manufacturing QC, and more.

Playground

Run real DSL queries against a live engine, right in your browser.