Best practices for data source connector configuration in Amazon Q Business - Amazon Q Business

Best practices for data source connector configuration in Amazon Q Business

The following list describes best practices for setting up and configuring your Amazon Q Business data source connector:

  • Each document in an index must be unique. Check that there are no duplicate documents in a data source, or across any data sources, that you plan to connect to an Amazon Q Business retriever.

  • If you change your authentication type and credentials, you must update your IAM role to access the correct AWS Secrets Manager secret ID.

  • We recommend that you regularly refresh or rotate your credentials and secret. Provide only the necessary access level for your own security. We don't recommend the re-use of credentials and secrets across data sources.

  • IAM roles used for retrievers can't be used for data sources. If you're unsure if an existing role is used for a retriever or data source, create a new IAM role to avoid errors.

  • If you use AWS KMS keys for the application, ensure that the IAM for your application is given the permission to describe, encrypt, and decrypt data using this key.

  • For on-premises or server data source connectors, Amazon Q Business checks if the endpoint information included in Secrets Manager is the same as the endpoint information specified in your data source configuration details. This helps protect against the confused deputy problem, which is a security issue. The problem occurs when a user doesn’t have permission to perform an action. But, by using Amazon Q Business as a proxy, the user can access the configured secret and perform the action.

    If you change your endpoint information later, you must create a new secret to sync this information.

  • Most data sources use regular expression patterns, which are inclusion or exclusion patterns referred to as filters.

    If you specify an inclusion filter, only content that matches the inclusion filter is indexed. If you specify an inclusion and exclusion filter, documents that match the exclusion filter aren't indexed, even if they match the inclusion filter.