A converged multimodal database engine integrating Relational SQL, AI-Native Vector RAG, NoSQL JSON Documents, and PL/SQL Procedural Scripting into a single storage engine with zero native C/C++ dependencies.
| Click Run Query to execute this query live inside your browser via the pure Dart engine WebAssembly runtime. |
No separate database servers. No external vector search daemons. Everything runs through a unified execution planner and storage engine.
First-class VECTOR data type with Hierarchical Navigable Small World (HNSW) graphs. Execute cosine, Euclidean, and inner product similarity queries across 768-dim embeddings in ~6 ms with >99% Recall@10.
Store schema-free JSON payloads and query nested attributes directly using intuitive dotted syntax and PostgreSQL arrow operators.
Full procedural block execution with DECLARE, BEGIN, LOOP, cursor iterations, and atomic transaction savepoints.
Standard SQL-92/2016 dialect with cost-based optimization, MVCC lock-free read isolation, slotted page organization, CRC32 checksums, and crash-safe WAL recovery.
Tested on local NVMe disk with WAL durability on a 100,000-row structured dataset.
Connect natively in Flutter, deploy as a backend microservice, or query from Python, Node.js, Go, Rust, or Docker.
# Add to your Dart or Flutter application
dart pub add ultsql
# Usage in code
import 'package:ultsql/ultsql.dart';
final db = Database('./app.db');
await db.init();
final interpreter = Interpreter(db);
final result = await interpreter.executeScript('SELECT * FROM users;');