Fabric

Use Fabric User Data Functions with Pandas DataFrames and Series in Notebooks

Use Fabric User Data Functions with Pandas DataFrames and Series in Notebooks A major upgrade to Notebook integration with Fabric User Data Functions (UDFs) is now available: Pandas DataFrames and Series can now be used as input and output types—thanks to native integration with Apache Arrow! This update brings higher performance, improved efficiency, and greater scalability to your Fabric Notebooks—enabling seamless function reuse for large-scale data processing across Python, PySpark, Scala, and R. With this release, Pandas DataFrames and Series are now supported as first-class input and output types for UDFs, enabled by deep integration with Apache Arrow, a highly efficient columnar memory format optimized for analytics workloads.

Microsoft Fabric APIs Specification I’m excited to share that we’ve successfully published the Microsoft Fabric APIs Specification in the microsoft/fabric-rest-api-specs GitHub repository!

The Functions portal includes a Generate invocation code feature that allows for automatic generation of an Open API specification for Fabric User Data Functions.

OpenAPI spec generation in Fabric User Data Functions (Preview) The Functions portal includes a Generate invocation code feature that allows for automatic generation of an Open API specification for Fabric User Data Functions.

Expanded Data Agent Support for Large Data Sources (Preview) Data Agent is officially lifting restrictions on adding Data Sources with larger schema sizes. Users can now add Kusto, Semantic Model, Lakehouse, and Warehouse Data sources that contain over 100 Columns + Measures and more than 1000 Tables to the Data Agent. This change allows users to bring larger-scale databases and semantic models into Fabric’s Data Agent, unlocking deeper insights and enhanced capabilities.

Bulk delete query: This feature enables users to delete multiple saved queries at once, eliminating the need to remove them individually. It was introduced in response to user feedback highlighting the difficulty of managing large lists of saved queries without a multi-select delete option. In the query editor’s ‘Queries’ folders, users can now hold Shift, select multiple queries, and right-click to delete them all in a single action. This streamlines the cleanup process, making it easier to maintain an organized and clutter-free workspace with minimal effort.

Open your database in SQL Server Management Studio (SSMS): This feature integrates the Fabric SQL web-based editor with SSMS, allowing for a smooth transition to the desktop environment. With a single click from the query editor, SSMS launches and automatically fills in the connection details for your Fabric SQL database — no manual copy-paste or setup required. This streamlines the workflow for users who prefer or need the advanced capabilities and richer UI of SSMS, making it faster and easier to switch between tools while working with Fabric SQL.

Use Python Notebooks to Read/Write to Fabric SQL Databases (Preview) You can now read from and write to SQL databases in Microsoft Fabric using Python Notebooks, thanks to the new integration with the T-SQL magic command. This highly requested feature enables users to run powerful T-SQL queries directly within notebooks—combining scripting, visualizations, and explanatory text in one collaborative workspace. It supports rich, interactive charts, automated workflows, scheduled jobs, and secure sharing, making it easier than ever to analyze and operationalize SQL data seamlessly across the Fabric platform.

Microsoft Fabric APIs Specification

Notebook snapshot for running Notebooks

OpenAPI spec generation in Fabric User Data Functions (Preview)

Expanded Data Agent Support for Large Data Sources (Preview)

Use Python Notebooks to Read/Write to Fabric SQL Databases (Preview)

Optimizing Query Management: New Controls in the Editor

Create CI/CD-Enabled Dataflow Gen2 from Existing Dataflow Gen2 (Generally Available) Integrated Run History and Validation Feedback in Dataflow Gen2 Editor Improvements to SharePoint as a destination in Dataflow Gen2

New Category Filters Added to Template Gallery

Accelerate app development: Cosmos DB (NoSQL) in Fabric now in preview

This removes the barriers between users and insights, enabling everyone—from business analysts to data scientists—

engage in natural language conversations with their data across multiple reports and semantic models

This chat with your data experience will allow users to ask broader questions and intelligently retrieve the most relevant data

Coming soon in public preview, Fabric data agents can be added to any custom agent built in Microsoft Copilot Studio

Once connected, the custom agent uses the Fabric data agent to retrieve insights from OneLake, respecting data access permissions

Developers can also define actions (e.g., send an email or trigger workflows) to automate processes

For teams tasked with building new AI and analytics solutions, finding and accessing the necessary data across a sea of disconnected data services

Shortcut transformations

These updates reinforce our commitment to Fabric’s four core pillars: A complete, AI-powered data platform. An open, AI-ready data lake. Empowering AI-enabled business users. A mission-critical foundation.

The general availability of the Native Execution Engine enables Spark queries to run natively on your lakehouse—up to 6x faster with no code changes or vendor lock-in. This release brings built-in optimizations and resource profiles for faster, more cost-effective data engineering at scale.

real-time endpoints for ML models

The preview of Warehouse Snapshots enables users to access a consistent view of data from a specific point in time, even during ETL processes

Copilot for SQL Analytics Endpoint in the context of your business Picture this: you need to tie customer orders from your mirrored CRM to fulfillment data in the Warehouse – something that normally requires digging through schemas, writing complex joins, and double-checking table relationships. With Copilot, you skip the heavy lifting. A simple prompt – ‘Get customer orders from CRM and join with fulfillment data from the warehouse‘ – returns a ready-to-run query that pulls exactly what you need, no manual schema deep dives required. Or maybe you’re looking to blend product metadata in a Lakehouse with sales numbers stored in a Warehouse. Instead of bouncing between storage formats and trying to reconcile differences yourself, you ask Copilot: ‘Show me top-selling products by category using product metadata from the Lakehouse and sales from the warehouse.‘ Copilot does the heavy lifting, stitching sources together behind the scenes so you can focus on the insight, not the integration. And when you’re under pressure to deliver quick insights – like pulling revenue by region for a quarterly review – Copilot has you covered there, too. Even if you’ve never touched the Finance Lakehouse, you can ask: ‘Show me total revenue by region for the last quarter using the Finance Lakehouse.‘ Copilot identifies the right tables, applies the correct filters, and generates the SQL you need without the typical back-and-forth. Copilot transforms how you work with data: no more searching, stitching, or second-guessing. Complex environments become accessible. Insights flow faster. Data silos disappear. With Copilot in SQL Analytics Endpoint, you don’t just save time, you unlock the full power of Fabric and move from question to impact without missing a beat.