Skip to content

ElyraSQL

A robust, MySQL-compatible SQL server written in Rust.

ElyraSQL stores an entire database in a single ACID file, speaks the MySQL wire protocol so existing clients and drivers work unchanged, and adds first-class vector search and parallel analytical (OLAP) aggregation — all under one brand.

- :material-database: **Single file, ACID** The whole database is one crash-safe file with multi-version concurrency. - :material-connection: **MySQL compatible** Connect with `mysql`, DBeaver, or any MySQL driver in any language. - :material-vector-line: **Vector native** `VECTOR(n)` columns with exact and HNSW approximate nearest-neighbour search. - :material-lightning-bolt: **Fast** Sub-millisecond point lookups, index nested-loop joins, and cached ANN.

Highlights

  • Full SQL surface — DDL, CRUD, JOIN (INNER/LEFT/RIGHT/FULL/CROSS), WHERE, ORDER BY, LIMIT/OFFSET, aggregation with GROUP BY.
  • Rich types — BIGINT, DOUBLE, TEXT, BLOB, BOOL, DATE, DATETIME, TIME, DECIMAL, JSON, and VECTOR.
  • Indexing — clustered primary keys (single or composite), secondary B-tree indexes, range scans, and HNSW vector indexes.
  • Transactions — BEGIN / COMMIT / ROLLBACK with snapshot isolation.
  • Security — mysql_native_password auth, TLS, and read/write/admin roles.
  • Operations — single static binary, systemd unit, and a ~15 MB Docker image.

Quick taste

CREATE TABLE docs (id BIGINT PRIMARY KEY, title TEXT, embedding VECTOR(3));
INSERT INTO docs VALUES (1, 'cat', '[1,0,0]'), (2, 'car', '[0,1,0]');

-- nearest neighbours to a query vector
SELECT id, title, VEC_DISTANCE(embedding, '[0.9,0.1,0]') AS dist
FROM docs ORDER BY dist LIMIT 5;

Continue to Getting Started.

Project status

ElyraSQL is young and moving fast. The feature set below is implemented and tested, but see Limitations & Roadmap for known gaps before relying on it in production.