Creating and selecting a retriever for an Amazon Q Business application
After creating your Amazon Q Business application, you create and select the retriever and provision the index that will power your generative AI web experience. The retriever pulls data from the index in real time during a conversation.
Amazon Q provides retrievers for Amazon Kendra indexes and also for a native index. You can choose between selecting an Amazon Q retriever and a Amazon Q native index or using an already configured Amazon Kendra index as a retriever.
To select a retriever, you use the AWS Management Console or the CreateRetriever API operation. If you use the console and choose to use a Amazon Q retriever, Amazon Q creates an index for you as part of the application configuration process. You can then configure provisioning for the created index. For easy tracking, you can tag both the retriever and index. If you use the API to create a Amazon Q retriever, you must first use the CreateIndex API operation to create and provision an Amazon Q index, and then use CreateRetriever to create your Amazon Q retriever.
If you choose to use an Amazon Q native index, you must configure provisioning for it. Depending on your use case, you can choose to provision one of the following types of indexes:
- Enterprise index
-
This index type is suitable for production workloads. It is deployed across three Availability Zones for fault tolerance. Scales upto 1 million documents.
- Starter index
-
This index type is suitable for proof-of-concept or developer workloads. It is deployed in a single Availability Zone. Scales upto 100,000 documents.
Important
You can't change the retriever or index type for your application after your application has been created. To change your retriever or index type, you must create a new application.
Note
The data sources and indexes available to connect to your application change depending on your retriever choice.
For instructions on how to select a retriever and an index, choose a topic based on your retriever preference for Amazon Q.
