Edit

Add and configure datasources in Fabric data agent

Data Agent in Microsoft Fabric enables organizations to build conversational experiences over their enterprise data. By connecting Fabric artifacts to a Data Agent, users can translate natural language questions into precise queries, empowering stakeholders, from analysts to executives, to unlock insights without writing a single line of code. This article walks through every data source that Data Agent supports today and the configuration capabilities available for each data source.

Overview

Data Agent supports the following data source categories:

Category Artifacts Query Language Key Scenario
SQL Lakehouse, Data Warehouse, SQL Database, Mirrored Databases T-SQL Structured analytics over relational and Delta Lake data
Eventhouse Eventhouse KQL Database KQL Real-time intelligence and time-series analytics
Semantic Model Power BI Semantic Models DAX Business logic, calculated measures, and curated metrics
Graph Graph Model GQL Relationship-rich data exploration and graph analytics
Ontology Fabric Ontology Ontology-native Domain knowledge and semantic context for data integration
Azure AI Search Azure AI Search Index Natural language + search Unstructured data retrieval (PDFs, text, enriched content)

Tip

A single Data Agent can combine up to five data sources in any combination, letting you blend structured, real-time, semantic, and unstructured data in one conversational experience.

Supported Artifacts

  • Lakehouse—Delta Lake tables surfaced through the SQL Analytics Endpoint.
  • Data Warehouse—Full Fabric warehouse with T-SQL surface area.
  • SQL Database—Fabric-native SQL databases.
  • Mirrored Databases—External databases mirrored into Fabric (for example, Azure SQL, Cosmos DB, Snowflake).

Every SQL source in Fabric comes with a SQL Analytics Endpoint, a high-performance, read-only T-SQL query surface over OneLake Delta data. Data Agent leverages a built-in NL2SQL service that:

  1. Translates the user's natural language question into a T-SQL query using user provided selected schema, instructions, and example queries.
  2. Validates the generated query against the schema selection to ensure it only references approved tables and views.
  3. Executes the query through the SQL Analytics Endpoint and returns human-readable results.

Configure Your Datasource

SQL Sources Supported Configurations

Configuration Supported Details
Schema Selection ✅ Yes Select specific Tables, Views, and Functions to scope the agent.
Agent Instructions ✅ Yes Guide the agent on when and how to route questions to this source.
Datasource Instructions ✅ Yes Provide table descriptions, join logic, key column details, and business terminology to NL2SQL.
Datasource Description ✅ Yes Description that helps the agent determine whether this data source is relevant to the user's question.
Example Queries ✅ Yes Supply natural-language/SQL pairs so the agent can learn complex query patterns. Top examples are automatically retrieved via vector similarity.

Learn more