Unity Gateway now supports Amazon Bedrock Mantle

Unity Gateway now supports integration with Amazon Bedrock Mantle as an external model provider service, centralizing credentials, permissions, and costs in Unity Catalog.

Use cases

Imagine you need to integrate Amazon Bedrock Mantle AI models into your Databricks applications without exposing API keys in notebooks. With this feature, you can:

• Centralize encrypted credentials and connection details in a single managed object.
• Control user and service access using standard Unity Catalog permissions.
• Apply rate limits, budgets, and inference tables to requests.

How to use

You create and manage the provider service using SQL in Databricks to register Amazon Bedrock Mantle.

CREATE MODEL PROVIDER SERVICE main.default.bedrock_mantle_prod
TYPE 'bedrock-mantle'
CREDENTIALS (
  aws_access_key_id = 'my_access_key',
  aws_secret_access_key = 'my_secret_key'
);

What happens in this code:

1. Databricks creates a new securable object in the specified catalog and schema.
2. The provider type is set to 'bedrock-mantle'.
3. AWS credentials are securely stored and encrypted.

Limitations and requirements

• Amazon Bedrock Mantle (bedrock-mantle) uses different endpoints compared to the traditional bedrock-runtime provider.
• Users require appropriate Unity Catalog permissions, along with USE CATALOG and USE SCHEMA on the parent catalog and schema.