Fabric information brokers work with present OneLake implementations, giving them a base set of information to make use of as context to your queries. Along along with your information, they are often fine-tuned utilizing examples or be given particular directions to assist construct queries.
There are some stipulations earlier than you may construct an information agent. The key requirement is an F64 or greater shopper, together with an acceptable information supply. This could be a lake home, an information warehouse, a set of Power BI semantic fashions, or a KQL database. Limiting the sources is sensible, because it reduces the chance of shedding the context related to a question and retains the AI grounded. This helps make sure the agent makes use of a restricted set of recognized question sorts, permitting it to show your questions into the suitable question.
Building AI-powered queries
The agent makes use of person credentials when making queries, so it solely works with information the person can view. Role-based entry controls are the default, preserving your information as safe as attainable. Agents’ operations must keep away from leaking confidential data, particularly in the event that they’re to be embedded inside extra complicated Azure AI Foundry agentic workflows and processes.