After you have deployed your solution, you can add Azure Data Agent to get data analysis features from Fabric built-in AI capabilities. You can follow Microsoft's Fabric Data Agent Guide to create and set up your Fabric Data Agent, and use the materials provided here to set up your agent quickly. For example, you can create a new item, search for 'data agent', you will be presented the clickable resource to create a Fabric Data Agent.
- Create a Fabric Data Agent with a name, for example,
RTI_Data_Agent, - Add the KQL Database created in the Fabric workspace as your data source,
- Use the Agent configuration files provided below to set up your Fabric Data Agent.
Note
The Fabric Data Agent SDK is currently in Preview. After deployment, table selections may not persist — you may need to manually select the required tables in the Fabric portal under the Data tab and save. Once saved through the UI, the selections persist and the agent works as expected.
This folder contains essential configuration files for setting up your Fabric Data Agent to deliver optimal intelligence based on your data.
Agent Instructions - Master Prompt
Contains the primary instructions and behavior guidelines for your Fabric Data Agent. This file establishes the agent's role, scope, and response patterns.
Data Source Description
Provides a description of the data source.
Data Source Instructions
Provides detailed information about your data structure, table relationships, and query patterns. This helps the agent understand your data and deliver more accurate, efficient responses.
Example Question and KQL Query Set
You can provide example questions and corresponding KQL query code as well. In the docs/fabric_data_agent folder, you will find the below files as starters:
- example_question_1.md and example_kql_code_1.md
- example_question_2.md and example_kql_code_2.md
- example_question_3.md and example_kql_code_3.md
Once the Data Agent is set up with the above agent configuration files, you will be able to chat with the agent right away. You can also publish your data agent for others to use. Below is the user interface you can expect (link to more sample test questions):
- Replace sample data references with your organization's actual data structure
- Add domain-specific terminology and business context to the master instructions
- Include your most common business questions in the training examples
- Update query patterns to match your specific analytics needs

