Microsoft Fabric Updates Blog

Expanded Data Agent Support for Large Data Sources

We are continuously enhancing data agents in Fabric to deliver more powerful and flexible data experiences. In February of this year, we introduced a host of new improvements coming to the AI Skill—including support for additional Data Sources such as Eventhouse KQL and Semantic Models. Initially, integration to data sources was limited to sources with fewer than 1,000 tables or under 100 columns plus measures, which restricted many users from fully leveraging LLMs for data analysis and reasoning.

We are excited to announce that these schema size restrictions have now been lifted. You can now seamlessly integrate large-scale data sources–including Kusto, Semantic Models, Lakehouse, and Warehouse datasets–with over 1,000 tables and more than 100 columns and measures into Fabric data agent. This update significantly broadens the scope of what users can achieve with Fabric data agent, enabling deeper insights, richer semantic modeling, and more robust AI-powered data experiences.

The example demonstrates how to add a data source with a large schema to data agent:

The table is named patient_medical_records. This table contains a total of 103 columns, which exceeds the limit of what Fabric data agent used to support for data sources.
Despite the larger schema size, we are now able to add the patient_medical_records table to Fabric data agent. When we select the table as an input to the data agent, we receive a warning stating that the accuracy of my results may vary with larger schema sizes. Regardless, we are also able to select this table as an input to the data agent.

While this expansion unlocks exciting new possibilities, we want to be transparent about performance expectations. With larger schema sizes, the reliability of results may vary. We are actively working to enhance reliability across Fabric data agent, and targeted improvements for handling larger schemas are well under way. Stay tuned for more updates and new feature releases!

Next Steps

To learn more about Fabric data agents, please explore our Data Agent Documentation.

For guidance on improving your data agent’s reliability, please refer to the Best Practices for Configuring your Data Agent guide.

Give us your feedback on Fabric Ideas.

Entradas de blog relacionadas

Expanded Data Agent Support for Large Data Sources

enero 8, 2026 por Adi Eldar

What if generating embeddings in Eventhouse didn’t require an external endpoint, callout policies, throttling management, or per‑request costs? That’s exactly what slm_embeddings_fl() delivers: a new user-defined function (UDF) that generates text embeddings using local Small Language Models (SLMs) from within the Kusto Python sandbox, returning vectors that you can immediately use for semantic search, similarity … Continue reading “Create Embeddings in Fabric Eventhouse with built-in Small Language Models (SLMs)”

diciembre 16, 2025 por Alex Powers

As 2025 ends, we’re taking a moment to reflect on Microsoft Fabric’s second year in the market and the collective progress made alongside our community, customers, and partners. What began as a unified vision for data and AI has grown into a platform adopted by more than 28,000 organizations worldwide, anchored by OneLake and shaped … Continue reading “Microsoft Fabric 2025 holiday recap: Unified Data and AI Innovation”