Redis ADBC driver
Query Redis with SQL and exchange data as Apache Arrow.
The Redis Arrow Database Connectivity (ADBC) driver lets you query Redis with SQL and exchange data as Apache Arrow record batches. ADBC is an Arrow standard for database drivers, so any tool that can load an ADBC driver can connect to Redis with it. Examples include pandas, Polars, DuckDB, SQLAlchemy, Apache Airflow, and Dagster.
The Redis ADBC driver project is open source under the Apache 2.0 license.
When to use the driver
To build an application, use the Redis client library for your programming language. The ADBC driver is an alternative for work that fits SQL and Arrow, such as analytics, notebooks, dashboards, and data pipelines.
The driver supports:
- SQL reads, inserts, updates, and deletes.
- Analytical queries with filters, grouping, and aggregates.
- Bulk loading of Arrow data.
- Typed Arrow results, with supported filters and aggregations run by the Redis Query Engine.
These pages show examples in Python and Go. The driver also works from other languages and from tools that load ADBC drivers. The project's integration guides cover them.
Tools that can't load an ADBC driver, such as JDBC applications, can connect through the Redis Arrow Flight server instead. The server embeds its own, older copy of the driver, and the two don't share a database.
How it works
You create SQL tables through the driver. Each table has a Query Engine index, and the driver stores each row as a Redis hash under keys that it manages. The key names are internal, so use SQL or the driver's tools to read and change table data rather than reading or writing those keys directly.
Hashes that your applications already write don't appear as SQL tables
automatically. The driver's redis-adbc command-line tool can copy them into a
table or adopt them in place. See
Existing HASH collections
in the driver reference.
The driver doesn't upgrade tables that earlier storage formats created, so a driver upgrade can require you to re-create or reload tables. Check the driver reference before you upgrade a database that holds data you need to keep.
Requirements
The driver needs a Redis database with the Query Engine (Search) enabled. It checks for Search when it connects and refuses to connect without it. The project tests against:
- Redis Open Source 8.6.2, standalone and clustered
- Redis Software, with Redis 8.8 databases
- Redis Cloud, with Redis 8.6.2 databases
For Redis Software and Redis Cloud databases that support it, set the query performance factor (QPF) to 4× and provision enough CPU for it. QPF is a performance recommendation, not a connection requirement.
The driver doesn't support:
- Redis Flex databases.
- Valkey, or Redis 7.2 derivatives such as Amazon ElastiCache and Google Cloud Memorystore.
For tested versions and results, see the project's compatibility details.
Limitations
- The driver is autocommit only. Every statement commits on its own.
BEGINandCOMMITare accepted for compatibility with SQL written for PostgreSQL, but they have no effect.ROLLBACKundoes nothing: a row that you insert betweenBEGINandROLLBACKstays in the table.SAVEPOINTis not supported. - Wheels cover limited platforms. The Python package includes the native driver for Linux (x86-64 and ARM64, glibc 2.28 or later) and macOS on Apple silicon (macOS 15 or later). There are no wheels for Windows, Intel Macs, or musl-based Linux.
- Unconstrained
NUMERICexpressions can break some readers. A calculation such asSUM(quantity * unit_price)returns a 256-bit Arrow decimal whose precision depends on the data. Some readers, including recent Polars versions, can't import it. Cast the result to a fixed precision, such asCAST(SUM(quantity * unit_price) AS NUMERIC(12,2)).
For the complete list of supported SQL and limits, see the driver reference.
Data tools
The project documents how to connect these tools. Each guide lists its own version requirements and limitations, and some tools need custom code rather than a ready-made integration.
| Tool | Guide |
|---|---|
| pandas | pandas guide |
| Polars | Polars guide |
| SQLAlchemy | SQLAlchemy guide |
| Apache Superset | Superset guide |
| DuckDB | DuckDB guide |
| Apache Airflow | Airflow guide |
| Dagster | Dagster guide |
| Prefect | Prefect guide |
| dlt | dlt guide |
The project's ecosystem comparison covers about 50 tools, including which connection paths are tested and which need custom work.