Stoolap

Stoolap is an embedded SQL database implemented in Rust, designed to be shipped directly with applications as a single binary. It targets developers who need a high-performance, embeddable database engine with advanced concurrency, analytical, and AI/ML features. Stoolap supports multi-version concurrency control (MVCC) with snapshot isolation, allowing readers and writers to operate without blocking each other and enabling time-travel queries through the AS OF clause.

The database includes a cost-based query optimizer inspired by PostgreSQL, capable of adaptive execution that learns from actual cardinalities and re-optimizes at runtime. Performance enhancements include bloom filter propagation, zone map pruning, and semantic query caching to minimize redundant work. Stoolap automatically parallelizes query execution using a work-stealing scheduler, distributing operations like filters, joins, sorts, and distinct computations across available CPU cores. Chunked processing with configurable thresholds is used to maintain efficiency for both small and large queries.

For AI and machine learning use cases, Stoolap offers a native VECTOR type and HNSW indexes, supporting sub-linear approximate nearest neighbor search with L2, Cosine, and Inner Product distance metrics. An optional built-in sentence-transformer model allows direct text-to-vector embedding within the database, removing the need for external APIs or Python dependencies. Additional index types include B-tree, Hash, and Bitmap indexes. Advanced SQL features are supported, such as window functions (ROW_NUMBER, RANK, DENSE_RANK, LAG, LEAD, NTILE), recursive common table expressions, and complex constructs like JOINs, subqueries, ROLLUP, CUBE, GROUPING SETS, UNION, INTERSECT, and EXCEPT. Durability is achieved through write-ahead logging with snapshot-based crash recovery.

Stoolap can be embedded as a library in applications or accessed via a standalone command-line interface. It also compiles to WebAssembly, enabling use directly in the browser with no server or installation required. First-party drivers are available for a wide range of programming languages, including Python, Go, Node.js, Java, C, PHP, Ruby, Swift, C, and WebAssembly, making it accessible from most major development environments.