Prerequisites for Knowledge Base

Before you can add a knowledge base, you need to create an Oracle Cloud Infrastructure Object Storage bucket in which to upload files, create an OCI Search with OpenSearch Cluster to associate with the knowledge base, create dynamic groups and policies to satisfy Resource Principal Session Token (RPST) prerequisites, and create a private endpoint in Oracle Integration.

Prerequisites for OCI Object Storage

Before you can add a knowledge base, you need to create an Oracle Cloud Infrastructure Object Storage bucket in which to upload files and create dynamic groups and policies to satisfy Resource Principal Session Token (RPST) prerequisites.

1. Create an Oracle Cloud Infrastructure Bucket to Upload Documents

You need storage to upload documents and store temporarily to be ingested into the knowledge base. Information you'll need for your knowledge base:
  • Compartment
  • Bucket name

You can use an existing Oracle Cloud Infrastructure Object Storage bucket or create a new one.

To create a new Oracle Cloud Infrastructure Object Storage bucket, follow the instructions in Creating an Object Storage Bucket.

Note:

Oracle Cloud Infrastructure Object Storage bucket must reside in the same tenancy as the Oracle Integration instance.

2. Create a Dynamic Group and Assign a Policy to satisfy Resource Principal Session Token (RPST) Prerequisites

To invoke Oracle Cloud Infrastructure Object Storage from Oracle Integration, you must satisfy Resource Principal Session Token (RPST) prerequisites in the Oracle Cloud Console. Create the required dynamic group and assign a policy to that group to allow your Oracle Integration instance to access Oracle Cloud Infrastructure Object Storage. The policy defines permissions for the dynamic group and determines which operations the dynamic group can perform on Oracle Cloud Infrastructure Object Storage. For instructions, see Prerequisites to Invoke Oracle Cloud Infrastructure Object Storage from an Integration

Prerequisites for OCI Search with OpenSearch

A knowledge base is associated with a database in which documents are ingested. Before you can add a knowledge base, you need to create an OCI Search with OpenSearch Cluster and a private endpoint in Oracle Integration.

1. Create a Private Endpoint for Your OCI Search with OpenSearch Cluster

You must create a private endpoint to make sure your OpenSearch cluster is not publicly accessible. A private endpoint is required.

See Configure a Private Endpoint for Your Instance.

2. Create an OCI Search with OpenSearch Cluster to Associate with the Knowledge Base

Once you upload documents, you ingest them into the vector database that is associated with your knowledge base.

Supported software version:
  • OCI Search with OpenSearch version: 2.19 or higher

Note:

When you create the OCI Search with OpenSearch Cluster, make note of the user and password you use to create it. You will need this for configuring the connection to OpenSearch in Oracle Integration.
  1. Create an OCI Search with OpenSearch Cluster for your knowledge base. For instructions, see Create an OCI Search with OpenSearch Cluster in the Oracle Cloud Console.

    Configure Nodes: Minimum Required Configuration

    When creating the OCI Search with OpenSearch cluster, these are the minimum configurations for knowledge base. Use a Development profile if you are trying out the system and use Application Search profile for production environments. Whichever profile you use for your OpenSearch cluster, make sure you change the default values to meet the minimum as indicated:

    Profile Leader Node Configuration Data Node Configuration
    Development
    • Dedicated leader nodes: 3 (minimum)

    Note:

    You always want an odd number. If one goes down, you still have two.
    • Dedicated data nodes: 3 (minimum)
    • Data node memory (GB): 40 GB (minimum)

      Note:

      Do not use the default data node memory value because you will not be able to load the default models. The selected model for the knowledge base is loaded into memory the first time you ingest. Models can get large. If you try different models to see which one is best for you, you have to have enough memory to use the models.
    • Data node OCPU count:
      • 4 (minimum)
      • 10 (recommended)

      Note:

      The fewer CPUs you have, the longer ingestion takes.
    Application Search You can use the default values for Leader Node Configuration. You can use default values for Data Node Configuration except for the following:
    • Data node storage (GB) - 500 GB (minimum)

      Note:

      Set this depending on your usage. The default will not be enough. The size of the document is not what is being stored. What takes more space is the vectors that are computed.
    Log Analytics You can use the default values for Leader Node Configuration. You can use default values for Data Node Configuration except for the following:
    • Data node storage (GB) - 500 GB (minimum)

      Note:

      Set this depending on your usage. The default will not be enough. The size of the document is not what is being stored. What takes more space is the vectors that are computed.
    Configure Networking
      • The VCN must be in the same region as your Oracle Integration instance.
      • The VCN and subnet can be in any compartment within the tenancy.
      • The subnet can be public or private. Use a private subnet for maximum isolation and security.

        Note:

        A private endpoint is required.

Next step: Create a Knowledge Base.