Alibaba Cloud Database · Open Source

Alibaba's MySQL Branchwith DuckDB Analyticsand Native Vector Search

MySQL / InnoDB at the core, with DuckDB columnar analytics and native HNSW indexes.

Get started Explore the architecture

MySQL Client / SQL

AliSQLMySQL Service Layer

InnoDBOLTP

Transactions · Vector search

HNSWNative Flashback

DuckDBOLAP

Columnar analytics

Columnar storageVectorized execution

Latest feature releaseAliSQL 8.0.44-2MySQL 8.0.44 base

GitHub5.9K+ Stars900+ Forks

LicenseGPL-2.0Public source and docs

Technology paths

Three core capabilities

A MySQL-compatible core, plus opt-in DuckDB analytics and native vector search.

Transactional kernel

MySQL 8.0-compatible kernel

Based on MySQL 8.0.44 and InnoDB, preserving familiar protocols, transactions, and row storage.

Explore MySQL / InnoDB

Columnar analytics

DuckDB analytical path

DuckDB 1.4.4 provides columnar storage and vectorized execution for analytical tables and replicas.

Explore DuckDB

Native retrieval

Native HNSW indexes in InnoDB

InnoDB stores VECTOR data and HNSW indexes, queried through SQL with transaction visibility.

Explore Vector Index

Kernel extensions

Binlog in RedoEligible binlog events share InnoDB redo durability.

Native FlashbackRead retained InnoDB history with AS OF TIMESTAMP.

SQL example

One MySQL entry point, three data paths

Use one MySQL connection for InnoDB, DuckDB, and Vector Index.

InnoDB runs directly; enable duckdb_mode or Vector Index for the other examples.

Full quick start

InnoDB transactions

CREATE TABLE orders (
  id BIGINT PRIMARY KEY,
  customer_id BIGINT NOT NULL,
  amount DECIMAL(18,2) NOT NULL,
  status VARCHAR(16) NOT NULL
) ENGINE=InnoDB;

START TRANSACTION;
INSERT INTO orders VALUES (1001, 42, 299.00, 'PAID');
SELECT id, amount, status FROM orders WHERE customer_id = 42;
COMMIT;