Episode 315 – Fabric November 2025 Feature Summary (Part 3)

January 7, 2026

Welcome back to the final installment of our November 2025 Fabric feature summary marathon! Jason and John are wrapping up what turned into quite the trilogy, finishing the rest of Warehouse, Real-Time Intelligence, and diving deep into Data Factory. As John put it right at the start – they’re determined to make it through this before the end of the year, and they did it!

What’s Ahead for BIFocal

Before jumping into features, Jason teased some exciting developments. The hosts have compiled their 2024 predictions from last year and will be grading themselves on how those turned out – should be fun to see what they got right (or wildly wrong). They’re also making their 2026 predictions, discussing upcoming conferences, and teasing some “fun things” they have in the works for next year.

If you’ve been following along, you know Jason just got back from Ireland, and as he joked, he completely lost track of date and time. Fair enough when you’re crossing that many time zones!

Warehouse: Small But Mighty Updates

Jason started off thinking Warehouse had a lot to cover, but it turned out to be more about quality over quantity. Microsoft delivered some really solid how-to articles alongside a few key features.

Identity Columns Hit Preview

For those coming from traditional SQL backgrounds, this one’s a relief. Identity columns are now in preview, giving you that familiar auto-increment functionality we’ve had forever in SQL Server.

As John explained, while warehouses typically handle bulk loads differently than transactional databases, having identity columns makes things feel more familiar for SQL practitioners. Jason had an amusing moment thinking this was about implementing some new authentication system (imagine his confusion thinking they were replacing Entra ID!), but no – this is good old-fashioned surrogate key generation.

The system-managed approach ensures uniqueness across Fabric’s distributed engine, even when multiple ingestion jobs run in parallel. No more worrying about key duplication or integrity issues.

Data Clustering for Performance Wins

The backend optimization updates include data clustering, which enables aggressive file pruning. The blog shows some impressive gains – a query dropping from what would normally take much longer down to about 20 seconds in their example.

As Jason pointed out with his usual candor about animated GIFs in blog posts, the visual shows the performance improvement pretty dramatically. Data clustering organizes rows with similar values together during ingestion, meaning queries only scan files that match their predicates. Smart stuff from the Warehouse team.

Warehouse Snapshots Go GA

Warehouse Snapshots have officially hit general availability. Microsoft’s calling this “true time travel database” functionality – and to their credit, it’s a pretty innovative capability.

The concept isn’t entirely new (we’ve had transaction logs and restore points forever), but Microsoft has wrapped this in a nice UI that makes it genuinely easy to use. You can roll back to previous versions of your data warehouse, which is invaluable for:

• Stable reporting during ETL: Keep dashboards consistent while pipelines run
• Financial close processes: Lock KPIs for month/quarter-end without blocking operations
• Data audits: Track changes and access different versions at any point in time
• Historical analysis: Schedule snapshots hourly, daily, or weekly

As Jason noted, this was one of those features he needed but didn’t have access to early on, which limited how he could use Warehouse. Having this capability available changes the game for real-world implementations.

VARCHAR(MAX) Support Finally Arrives

The previous limitation on large columns has been lifted. VARCHAR(MAX) and VARBINARY(MAX) support is now available, allowing up to 16MB per cell. For anyone who hit that 8KB truncation wall with text, logs, JSON, or spatial data – this one’s for you.

Jason was transparent about this being a blocker for some of his association work, so it’s great to see it resolved. Real-world data doesn’t always fit in neat little boxes, and having this flexibility matters.

Real-Time Intelligence: Where John Comes Alive

As Jason said, “this is your bread and butter, this is right your heart and soul” – and John did not disappoint with his enthusiasm for these updates.

HTTP Connector: The Game-Changer

John was genuinely excited about the new HTTP connector for Event Stream. And with good reason – this thing is actually a big deal.

The connector lets Event Stream pull from any HTTP endpoint, which opens up a world of IoT device integrations. As John explained, he’s been collecting weather data for years using various methods – Power Automate flows, Azure Logic Apps, Azure Functions – all just to hit weather station endpoints and get that data into Event Stream. This new connector should eliminate all of that middleware.

The connector handles:

• No-code configuration for streaming data from REST APIs
• Automatic JSON parsing into clean, structured events
• Predefined public APIs with auto-populated headers and parameters
• Scheduled polling to continuously pull data

Jason made an interesting observation about JSON vs. JSOL (JSON Lines) – he’s been seeing JSOL pop up more and more as the more efficient format. The parsing capabilities here handle both, which is smart future-proofing.

MongoDB CDC Joins the Party

Right alongside HTTP, the MongoDB CDC connector brings change data capture for MongoDB deployments. It’s exactly what you’d expect – pick up changes to a MongoDB database and take action. John called it “fit and finish” – not revolutionary, but necessary and well-executed.

The feature began rolling out in November and should be available in all regions by mid-December (so it’s definitely available now if you’re reading this in early 2026).

Cribble Integration via Kafka

Speaking of sources, there’s also a new Cribble source. Neither Jason nor John had used Cribble, but the interesting technical detail here is that it uses Event Stream’s Kafka endpoints. As John noted, this is the first time they’ve seen one of these explicitly called out to work that way – you set up a custom action in Event Stream and have it behave as Kafka. Good to know that option exists.

Event Stream Activator Destination Hits GA

This one required some unpacking during the episode. The Event Stream Activator destination is now generally available, which means you can have your streams processed by Activator and trigger actions based on the data.

Jason got a bit turned around trying to understand whether this was “the only instantiation of activator” (it’s not – Activator exists in multiple places), but eventually got his head wrapped around it. As John explained, instead of persisting your stream in a database, it gets processed by Activator which can then take various actions, including calling user-defined functions.

The key understanding: this is Activator listening to Event Stream. It’s one way (of several) to take action using Activator, but it makes sense as a logical integration point between Event Stream and Activator items in your workspace.

Capacity Overview Events in Preview

Capacity overview events can now stream into Event Stream, giving administrators real-time insights into capacity health and utilization. You get two event types:

• Capacity Summary: Point-in-time snapshot of capacity usage based on smoothed utilization metrics
• Capacity State: Captures transitions like Paused or Overloaded states

As Jason noted, this is very much an admin function – end users aren’t going to think much about capacity states. But for administrators, being able to trigger Activator when your capacity is overloaded? That’s proactive monitoring at its finest. You could potentially automate capacity scaling or at least get alerts before users start screaming.

John also mentioned a new real-time dashboard for monitoring capacity that ties into these events – giving you visual monitoring of what’s happening with your capacity consumption in near real-time.

KQL Database Entity Diagrams

Entity diagrams for KQL databases are now available, and John loves this addition (as someone who appreciates the diagram/lineage view in Power BI, this makes perfect sense).

The feature shows you the flow of data through your database – from bronze to silver to gold in a medallion architecture, including update policies, functions, and materialized views. John was searching for the term during the recording and eventually landed on “lineage view” – that’s exactly what this is, but for Event House and KQL databases.

You can view ingestion details, spot schema violations, and understand the relationships between tables and update policies. For anyone working with Event House regularly, this is going to make life significantly easier.

Operations Agent Arrives in Preview

The Operations Agent is something Jason called out as feeling like a big deal, and it does. Users can create autonomous agents that monitor data for goals and recommend actions.

As John described it, think of this as an agent version of Activator. Instead of setting up rules in a UI, you just tell the agent what to watch and what to do, and it handles the monitoring and response. The agent lives in the Real-Time Intelligence workload, which makes sense given its relationship to Activator-type functionality.

Requirements to use it:

• Enable Copilot and AI at tenant level
• Enable the Operations Agent
• Workspace backed by Fabric capacity
• Can’t use it in trial capacity

The visualization they showed works through Teams, with notifications keeping “the human in the loop” as Microsoft puts it. Jason wondered if it also works in the native Copilot tab in Fabric, which seems like a reasonable expectation.

Maps Updates: Imagery Files and Data Labeling

Maps continue to get love with two preview features: imagery file support (GeoTIFF, GeoJSON, plus raster files) and data labeling for polygons and lines.

Jason raised a valid point that he’d like to see these visualization capabilities in Power BI rather than siloed over in RTI. “Visuals belong in Power BI. That’s reporting,” as he put it. Maps as a separate item in RTI might not get the love and adoption it deserves, especially when we’ve spent 10 years getting people comfortable with Power BI. As Jason shared the story about someone at their Ireland session saying “I’ve been scared of Power BI and I’m not anymore” after a decade – do we really want to tell people to go check out another completely different interface for mapping?

John agreed it might not get the adoption it should living on the side like this. Fair criticism from both hosts.

Real-Time Dashboards Get Copilot

Copilot-assisted real-time data exploration is now available for real-time dashboards. The capability existed before in KQL and Kusto, but as John noted, it was “clunky at best.” It could interpret your question and spit out KQL, but then you had to run it yourself.

Now it’s much smoother – a proper sidebar where you ask questions about the data, it detects what’s on your dashboard, queries intelligently in the backend, and just gives you the result. All that KQL generation happens behind the scenes. Much better user experience.

Data Factory: The Power Query Empire Expands

The Data Factory section started with enterprise readiness items (Snowflake key pair auth, manual gateway updates, VNet gateway certificates and proxy support) that Jason and John acknowledged as important but not particularly interesting to discuss.

Error Insight Summaries and Expression Generation

Instead of copying errors out to Claude (as Jason admitted doing), you can now ask Copilot for Pipelines directly: “What does this error mean?” The in-context evaluation helps you get over problem humps without leaving the interface.

Similarly, Copilot can now generate expressions in pipelines, and more importantly, it can explain expressions to you. As Jason noted, that’s the feature he really likes – when you’re looking through someone else’s code trying to figure out what a section does, being able to ask for a summary is incredibly helpful. Once you get past about 10,000 lines of code, the human eye needs some assistance.

Monitoring Hub Gets Hierarchical Organization

The monitoring hub can now organize pipeline runs hierarchically, showing multiple runs for a given pipeline with upstream and downstream run tracking. Previously, you just had a big monolithic list to sort through.

Jason’s wish: “I’d honestly like to see a tree view almost in this situation…almost like a lineage view for pipelines, John.” John’s response: “Getting there.” Sounds like they’re thinking along the same lines!

Modern Get Data Hits Excel Desktop

Here’s where things get interesting for the bigger picture. Modern Get Data experience is now available in Excel Desktop.

As John explained, we’ve had Power Query in Excel for ages, but it was its own implementation separate from Power BI Desktop, which was separate from the online experience. Microsoft’s been working toward a “modern get data experience” with consistent UI across all platforms. This finally brings that to Excel Desktop.

Jason made a really insightful observation: “I’m going to go back a couple of years where we talked about the fact that we thought that Power Query was going to get its own sort of top level brand. This highlights that that’s what Data Factory really has become is sort of the hub for Power Query.”

The fact that Excel Desktop improvements are showing up in the Fabric blog under Data Factory? That’s no accident. Power Query is everywhere, and Data Factory is the umbrella bringing it all together.

Importing Data Flow in Modern Get Data Copilot

The heading here is a grammatical train wreck (“new importing data flow supported in modern get data copilot”), but Jason figured out the issue: there’s a space in “data flow” – it’s about the flow of data, not a Data Flow Gen2 artifact.

This is about new capabilities for importing data using Copilot in the modern get data experience. You can point Copilot at sources like OneLake catalog and it knows what to do without having to guess.

AI Functions in Data Flow Gen2

Fabric AI functions integration with Data Flow Gen2 brings those AI capabilities (summarize, sentiment analysis, etc.) that were already available in notebooks into your data flows.

You can add a column and call an AI function that depends on another column – it executes when you run the data flow. John was particularly excited about the translation capabilities. Imagine having a language resource table with English terms that you need translated into multiple languages. Instead of sending it out for translation, you run Power Query against all the terms and get back the language equivalents. Since it’s using modern AI (not old machine translation), it understands context reasonably well.

Power Query Language Service IntelliSense

IntelliSense for Power Query in Data Flow Gen2 brings that typeahead completion we’ve had in Desktop to the service. Jason’s played with it and reports it’s “pretty damn accurate” – often knowing what he’s trying to say before he does (which, as he jokes, “honestly isn’t that hard”).

When you’re writing M code, having that auto-completion makes things significantly faster and reduces syntax errors.

Apache Airflow Job Management APIs

New APIs for <a href=”https://blog.fabric.microsoft.com/en-us/blog/fabric-november-2025-feature-summary/” target=”_blank”>Apache Airflow job and file management</a>. As John admitted, “I know nothing about it.” If you use Airflow, this might matter to you. Moving on!

Mirroring: Expanding the Universe

The mirroring updates continue to add data sources:

• Snowflake Iceberg tables (generally available)
• SAP (preview)
• SQL Server on-premises now supports 2025 (previously only through 2022)
• Cosmos DB (generally available)
• PostgreSQL (generally available)

There’s also user-assigned managed identity (UAM) support for Azure SQL Database mirroring. As Jason noted, this is a behind-the-scenes improvement as Microsoft moves further away from SQL authentication toward managed identities. Good security practice, even if most people won’t notice the change.

Copy Job Gets More Capable

The copy job continues to mature with:

• CDC support for triggering when source data changes
• Additional sources and destinations including SAP, Snowflake, and Google BigQuery
• Query-based incremental strategies for defining how incremental copies work
• Truncate destination before full copy (as John explained, much faster than deleting rows individually)
• Multiple folder copying in a single copy job
• Variable library parameterization

As Jason noted, “Variable library, all the things.” It’s getting tendrils everywhere in Fabric, which makes sense for building reusable, maintainable data pipelines.

Developer Tooling: Going Open Source

The finale: Fabric VS Code extension is now open source. The community can contribute to the GitHub repository, improving the extension for everyone.

As Jason admitted, he’s probably never going to contribute anything unless it’s documentation. John’s “certainly not” coding either. But for those who want to extend or improve the VS Code Fabric experience, the doors are now open.

Wrapping Up the Trilogy

And that, folks, is it. Three episodes to cover one massive feature release. As Jason noted, this was the only one they went long on, but they wanted to make sure they got through everything. Mission accomplished.

Links

Microsoft Fabric Blog Posts Referenced:

• Fabric November 2025 Feature Summary
• Fabric Data Warehouse Goes All-In on Enterprises at Ignite
• Warehouse Snapshots in Microsoft Fabric (Generally Available)
• Introducing the HTTP and MongoDB CDC Connectors for Eventstream

Previous Episodes:

• Episode 314 – Fabric November 2025 Feature Summary (Part 2)
• Episode 313 – Fabric November 2025 Feature Summary (Part 1)

Subscribe:SoundCloud | iTunes | Spotify | TuneIn | Amazon Music