About the Data Sources Used in an AI Agent

In Oracle Analytics, you can associate any data source with an AI agent.

You can add one data source to an AI agent, but you can add multiple data sources to an AIDP Deep Insights agent. You can use a data source based on a file, a table, or multi-table from a connection, a local subject area, or subject area.

It's important to note that when you associate a dataset with an AI agent, you must ensure the dataset is indexed for use with the Oracle Analytics AI Assistant. For information on making data available for the Oracle Analytics AI Assistant, see Make Data Available for Oracle Analytics AI Assistant



In the AI agent editor, you can preview the dataset's metadata, which includes the full list of available columns and the Index Type and Synonyms configured for each column. Use this editor to select a subset of columns and define agent specific filters.



Column Subsetting

You can activate or deactivate any indexed column to control which fields the agent can reference when answering questions. If a user asks about a column that isn’t activated, the Oracle Analytics AI Assistant won’t respond using that data. Columns that aren’t indexed in the dataset can’t be enabled for the agent and won’t be available for selection.

Data Agent Filters

You can define data filters specific to a data agent. These filters are automatically applied to all user questions processed by the data agent and can’t be overridden by consumers. This ensures that all interactions remain within the intended data scope.

Configured agent filters aren’t supported for AIDP Deep Insights agents. However, AIDP Deep Insights agents can interpret filter criteria included in a user’s question.