Unity Catalog: Open, Multi-modal Catalog for Data & AI

Unity Catalog is the most open and interoperable catalog for data and AI

• Multimodal interface supports any format, engine, and asset
  • Multi-format table support - Delta Lake, Apache Iceberg as UniForm, Apache Parquet, CSV, etc.
  • Beyond tables - Unstructured data (Volumes) and AI assets (ML models, Gen AI tools)
  • Plugin support - extensible to Iceberg REST Catalog and HMS interface for client compatibility, plus additional plugins (e.g., integration with a new AI framework)
  • Interoperates with the Delta Sharing open protocol for sharing tabular and non-tabular assets cross-domain
• Fully Open - OpenAPI spec and OSS implementation (Apache 2.0 license)
• Unified governance for data and AI - Asset-level access control is enforced through temporary credential vending via REST APIs

Quickstart - Hello UC!

Let's take Unity Catalog for spin. In this guide, we are going to do the following:

• In one terminal, run the UC server.
• In another terminal, we will explore the contents of the UC server using a CLI. An example project is provided to demonstrate how to use the UC SDK for various assets as well as provide a convenient way to explore the content of any UC server implementation.

Prerequisites

You have to ensure that you local environment has the following:

• Clone this repository.
• Ensure the JAVA_HOME environment variable your terminal is configured to point to JDK11+.
• Compile the project using build/sbt package

Run the UC Server

In a terminal, in the cloned repository root directory, start the UC server.

bin/start-uc-server

For the rest of the steps, continue in a different terminal.

Operate on Delta tables with the CLI

Let's list the tables.

bin/uc table list --catalog unity --schema default

You should see a few tables. Some details are truncated because of the nested nature of the data. To see all the content, you can add --format jsonPretty to any command.

Next, let's get the metadata of one those tables.

bin/uc table get --full_name unity.default.numbers

You can see that it is a Delta table. Now, specifically for Delta tables, this CLI can print snippet of the contents of a Delta table (powered by the Delta Kernel Java project). Let's try that.

bin/uc table read --full_name unity.default.numbers

Operate on Delta tables with DuckDB

For trying with DuckDB, you will have to install it (at least version 1.0). Let's start DuckDB and install a couple of extensions. To start DuckDB, run the command duckdb in the terminal. Then, in the DuckDB shell, run the following commands:

install uc_catalog from core_nightly;
load uc_catalog;
install delta;
load delta;

If you have installed these extensions before, you may have to run update extensions and restart DuckDB for the following steps to work.

Now that we have DuckDB all set up, let's trying connecting to UC by specifying a secret.

CREATE SECRET (
      TYPE UC,
      TOKEN 'not-used',
      ENDPOINT 'http://127.0.0.1:8080',
      AWS_REGION 'us-east-2'
 );

You should see it print a short table saying Success = true. Then we attach the unity catalog to DuckDB.

ATTACH 'unity' AS unity (TYPE UC_CATALOG);

Now we ready to query. Try the following

SHOW ALL TABLES;
SELECT * from unity.default.numbers;

You should see the tables listed and the contents of the numbers table printed. TO quit DuckDB, run the command Ctrl+D or type .exit in the DuckDB shell.

Full Tutorial

You can read Delta Uniform tables from Spark via Iceberg REST Catalog APIs, operate on volumes and functions from the CLI, and much more. See the full tutorial for more details.

APIs and Compatibility

• Open API specification: The Unity Catalog Rest API is documented here.
• Compatibility and stability: The APIs are currently evolving and should not be assumed to be stable.

Compiling and testing

• Install JDK 11 by whatever mechanism that is appropriate for your system, and set that version to be the default Java version (e.g., by setting env variable JAVA_HOME)
• To compile all the code without running tests, run the following:

  build/sbt clean compile
  

• To compile and execute tests, run the following:

  build/sbt clean test
  

• To update the API specification, just update the api/all.yaml and then run the following:

  build/sbt generate
  

  This will regenerate the OpenAPI data models in the UC server and data models + APIs in the client SDK.