In a Fabric capacity, you can now create Data Agents with your various data artifacts. This solution allows multiple data sources in Fabric to be combined into a RAG-type area for natural language searches.

Within this interface, developers can select the data source and then narrow down the data to tables or queries. The Chat can help start building queries to answer questions. If the question is not answered correctly, the developer can have a conversation to help formulate the query to assist in the results.

In the case in the previous image, a SQL query was generated from lakehouse delta tables to answer a Sales Reason question for Internet Sales without any instructions to the agent. This can be done with Semantic Models (DAX query) or Real-time Intelligence database(KQL).
This allows multiple data sources in one agent to answer a variety of questions from the agent.