Manage AI-Assisted features - Precisely Data Integrity Suite

1. Go to Configuration > AI.
2. Open the Features page.
3. Use the AI Features toggle to turn AI capabilities on or off for your workspace.
   When the toggle is on, supported AI features are available throughout Data Integrity Suite.
4. In the Large language model (LLM) set preference area, select the models that AI Manager can use.
   AI features use the best model for the feature from the selected model set. Prioritization applies only when the same model is present on different connections and in other edge case scenarios.
   If a model from the configured choices is removed ^Preview, AI Manager uses another valid model from those choices. Deprecated models can still be used. If no valid model remains, the feature status changes to Error.
5. To add another LLM connection to the model set, select +.
6. Expand a product area to view the available AI features.
   Product areas can include Catalog, DQ Pipelines, DQ Rules, DQPlus, and Foundation.

1. Go to Configuration > AI.
2. Open the LLM Connections page.
3. Select Add LLM Connection.
   1. Provide a name and a suitable description to the connection.
   2. Select the LLM provider.
   3. Select the associated provider region.
   4. Under Authorization, choose the authentication method.
   5. For access key authentication, enter the AWS access key ID and AWS secret access key.
   6. For AWS Identity and Access Management (IAM) role authentication, enter the Amazon Resource Name (ARN) for the customer IAM role.
   7. For Azure AI Foundry ^Preview, enter the Tenant ID, Client ID, Client Secret, and Project URL.
   8. After you enter all details, select Test to validate the connection details. After successful validation, a confirmation message appears.
   9. Select Add. The configured LLM connection appears in the list.
4. Open the context menu for the associated connection, then select Edit, Test, or Delete.

Set up a customer-owned AWS Identity and Access Management (IAM) role so AI Manager can use temporary credentials to access Amazon Bedrock.

1. Create or choose the customer IAM role.

   The role ARN uses this format: arn:aws:iam::<CUSTOMER_ACCOUNT_ID>:role/<CUSTOMER_ROLE_NAME>.

2. Attach the permissions required for the Bedrock models that the LLM connection uses.

   The following policy includes the Bedrock actions required for model discovery and invocation. Limit the permissions to the required AWS Region and models where possible.

   {
     "Version": "2012-10-17",
     "Statement": [
       {
         "Sid": "Statement1",
         "Effect": "Allow",
         "Action": [
           "bedrock:ListInferenceProfiles",
           "bedrock:ListFoundationModels",
           "bedrock:InvokeModel",
           "bedrock:InvokeModelWithResponseStream"
         ],
         "Resource": "*"
       }
     ]
   }

3. Add a trust policy that allows the Precisely global role to assume the customer role.

   Replace PRECISELY_GLOBAL_ROLE_ARN with the Precisely global role ARN for your environment and workspace region. Replace YOUR_WORKSPACE_ID with the workspace ID provided by Precisely.

   {
     "Version": "2012-10-17",
     "Statement": [
       {
         "Sid": "AllowPreciselyGlobalRole",
         "Effect": "Allow",
         "Principal": {
           "AWS": "<PRECISELY_GLOBAL_ROLE_ARN>"
         },
         "Action": "sts:AssumeRole",
         "Condition": {
           "StringEquals": {
             "sts:ExternalId": "<YOUR_WORKSPACE_ID>"
           }
         }
       }
     ]
   }

4. Copy the customer role ARN.
   The ARN is ready to enter when you create the AWS Bedrock LLM connection in AI Manager.