Current RAG implementations in Microsoft Fabric (via AI Skills, Azure AI Search integration, and OneLake-backed vector stores) rely predominantly on vector similarity search over chunked documents. While this works well for surface-level semantic retrieval, it has significant limitations for enterprise scenarios that require multi-hop reasoning across structured and unstructured data: Loss of relational context: When documents are chunked and embedded, the relationships between entities (customers, products, transactions, regulations, contracts) are lost. Vector search retrieves semantically similar chunks but cannot traverse "Customer A → owns → Account B → linked to → Transaction C → flagged by → Regulation D."   What Would Help A first-class Graph-Based RAG capability natively integrated into Fabric, including: OneLake Graph artifact — a new item type (alongside Lakehouse, Warehouse, Eventhouse) that stores property graphs in an open format (e.g., extending Delta with graph metadata, or adopting an open standard like Apache AGE or GraphAr), with full Purview lineage and OneLake shortcut support. Automatic graph construction pipelines — Data Factory / Fabric Notebook templates that extract entities and relationships from Silver-layer tables and unstructured documents using Azure OpenAI / Phi models, with human-in-the-loop validation.   Why It Matters Accuracy where it counts: Enterprise and regulated customers Unlocks Fabric's strategic differentiator: Fabric already unifies data estates in OneLake. Graph-RAG is the reasoning layer that turns that unified estate into a reasoning estate Reduces architectural sprawl: Customers currently stitch together Azure AI Search + Cosmos DB Gremlin + custom orchestration ... View more

Hi Power BI Team, I am writing this not just as a Data Engineer, but as someone who genuinely loves building stories with data. In my professional life, I work with complex APIs, write Python codes in PyCharm, and manage data structures in VS Code. My MacBook Pro is the heart of my creative and technical world. However, there is one piece missing in this puzzle: Power BI Desktop for Mac. Every time I need to make a quick fix or respond to an urgent data request on a weekend, I find myself struggling with virtual machines or switching back to my corporate Windows laptop just to open a .pbix file. My MacBook M3 has incredible power, and it feels like a waste not to be able to use the world's best BI tool natively on the world's most efficient hardware. I know the future is web-based and Fabric is amazing. But as engineers, we sometimes need that desktop experience: the offline capability, the advanced modeling, and the seamless integration. Please do not make us choose between the hardware we love and the software we rely on. Bringing a native Power BI experience to macOS would not just be a technical update; it would be a huge embrace for the growing community of data professionals who use Mac. With respect and hope, Fatma Nur Kurbanoglu Data Engineer ... View more

Customer wanted to know how they can get a notifications for upcoming Power BI preview features before they are pushed and enabled in the tenant settings. Not using the below options: Preview features are announced externally first (blogs, newsletters, events) Check Power BI updates blog (monthly drop) The most reliable proactive model is: Blog + Newsletter + Enablement sessions + RSS tracking     ... View more

Gen2 V2 CI/CD Dataflows currnelty require Contributor or higher access to the the workspace to consume tables. This will be a violation of permission restrictions to be compliant with CMMC, ITAR, Finance and other governmental infrastructure/data requirements. We need to limit user access to a workspace to Viewer access, but still be able to access Gen2 V2 CI/CD dataflow tables. Gen1 Dataflows allow consumption via Viewer access (which is desired), but they do not support Fast Load. With Fast Load from Gen2 Dataflows, some queries are running over 90% faster with lower overhead on our upstream cloud services/platforms. The alternative solution to share Gen2 V2 CI/CD dataflow tables by extending access via Semantic model is not sufficient as users are unable to use the source in Transforms as they can with Dataflow tables. ... View more

I have the following workspace setup in Microsoft Fabric: Production workspace containing: Lakehouses Semantic models Notebooks Copy Jobs These are used for production workloads. Additional workspaces: Test Development To support environment separation, I created corresponding lakehouses in the Development workspace and configured a Copy Job to copy data from the Production lakehouse to the Development lakehouse. This setup works correctly. Scenario To avoid manually recreating Copy Jobs, I attempted to duplicate an existing Copy Job by: Opening the original Copy Job JSON definition Copying the JSON Updating: The Copy Job name The Copy Job ID The target lakehouse ID Saving/closing the JSON editor After closing the editor, the Copy Job appears correctly configured and displays the expected settings. Problem However, after closing the Copy Job and reopening it, the Copy Job is completely empty — as if it were newly created. This suggests that: The edited JSON definition is not actually being persisted/saved Or the manually modified metadata/IDs invalidate the Copy Job without showing an error Expected Behavior After editing and saving the Copy Job JSON: The updated Copy Job definition should persist Reopening the Copy Job should display the saved configuration If certain properties (such as IDs) are not allowed to be modified, the UI/editor should provide validation errors or warnings Actual Behavior JSON edits appear accepted initially No validation or save errors are shown Reopening the Copy Job results in an empty/default Copy Job configuration Additional Notes This capability would be very useful for promoting or replicating Copy Jobs across Development, Test, and Production workspaces without needing to recreate them manually. ... View more

We would like to request support for a direct private connectivity path between Power BI Service and Azure Databricks, such as Azure Private Link (or an equivalent service-managed private endpoint), that bypasses the need for a VNet Data Gateway seralize/deserialize stage. Currently, when connecting Power BI Service to a Databricks workspace configured with private networking (no public endpoint), the only supported approach is through a VNet Data Gateway or On-premises Data Gateway. While this enables secure connectivity, it introduces additional overhead due to the serialization and deserialization of data through the gateway layer. ... View more

It would be nice if there is a way to send subscriptions directly in Fabric without workarounds based on a condition in the dataset . Example: Only send the email subscription when: MAX(Calendar[Date]) = TODAY() - 1 This would help ensure subscriptions are only sent when the required data is available or conditions are met. Thanks. ... View more

Problem Fabric ontology can be generated from a semantic model with one click — but customers still have to build the semantic model manually. For a lakehouse with 100–200 Gold tables, that's days of click-by-click work to select tables, define relationships, mark date tables, and create measures. Customers with 5,000+ tables and 300–800 ontology candidates often estimate 3–4 weeks of manual effort before they can even start with ontology. This is a major adoption blocker for Fabric IQ. What would help Any form of automation or AI-assisted bulk creation of semantic models directly from lakehouse metadata — using schema, profiling, lineage, or Copilot — would dramatically shorten the path to ontology. Even partial automation (e.g., relationship detection on user-selected tables) would be a meaningful first step. Anyway to automate the creation ? not fully but a base atleast ? Why it matters Snowflake (Semantic Views Autopilot), Databricks (Genie), dbt, and AtScale already automate this. Fabric is behind on this specific capability, despite having stronger downstream features (ontology graph, Activator, agents). Closing this gap would accelerate Fabric IQ adoption for the enterprise segment with the largest data estates. ... View more

While documentation for Dataflows Gen1 exists, clearer guidance on roadmap, support lifecycle, and migration expectations is needed, as Gen1 is now a legacy artifact. Although migration guidance to Gen2 is available, more detailed scenarios—such as migrating Dataflows Gen1 with linked entities (Linked dataflow gen1's) to Gen2—along with improved clarity, additional context and consolidation across documentation, would help plan and transition more effectively. ... View more

Introduce shared / reusable schedule objects that can be referenced by multiple Fabric artifacts (multiple pipelines, notebooks, and other artifacts), similar to SQL Server Agent job schedules. This would allow teams to define execution timing once (for example, a nightly or hourly schedule) and apply it consistently across many workloads, rather than duplicating identical schedules per artifact. Shared schedules would reduce configuration drift, simplify maintenance, and make schedule changes safer and faster by allowing updates in one place. This capability would significantly improve enterprise operations, governance, and migration scenarios for organizations moving from SQL Server Agent–based architectures to Fabric, where coordinated execution timing across many processes is a common requirement. ... View more

The ability to reorder the left hand objects here (see pic) for example, i want to open a notebook monitorting object and i want it to be right next to the actual notebook object for quick transfer.      ... View more

when navigating workspaces, each workspace has icons next to it. the diamond icon that symbolizes a fabric or premium capacity simply shows "Fabric contenct" which is not very usuful. it would be great if on hover you get back the capacity name.  ... View more

The title says it all. Due to the lack of a proper variables library implementation for connections and warehouses, our current development workflow includes switching connections/warehouses from one workspace to another one. Whenever we just switch connections/warehouses the hard coded schema and table names in activities, that can reference warehouses, will be cleared which is a huge paint point which loses us time over and over. We suggest to stop clearing those input fields whenever a connection or warehouse is changed in those activities. ... View more

Would be very helpful to have the drag and drop feature enabled in the Fabric workspaces UI.   ... View more

Due to a common pattern in GCP in which customers use one project for billing and other projects for data, the Fabric Factory Copy Activity Destination tab does not expose enough properties to permit successful writes. The UI currently only exposes dataset-id and table name entered manually.  For proper billing, the Fabric Connection is setup to use the GCP billing project.  Due to the fact that the UI does not permit setting of the data project to define the fully-qualified name of the target write table, at runtime, the target write table becomes {billing project}.{dataset}.{table}.  The billing project intentionally contains no tables at all and is strictly for purposes of allocating egress costs correctly.  Extending the Fabric Factory Copy Activity Destination tab by adding the project context for the write would allow Fabric Copy Activity to align with the common pattern of usage in GCP.  An alternative that might be considered would be to make the Billing Project Id, and Project Id in the Fabric Connector for BigQuery actually permit runtime override of the project default.  Were this change made then the Copy Activity UI change would be unnecessary. ... View more

I'm looking for a way to increase the width and size of the visualization's scroball. Please consider and provide this feature. ... View more

Currently, there is a significant productivity gap in Power BI Desktop's Table View. Unlike Excel, which allows users to instantly access column filters using Alt + Down Arrow, Power BI requires a mouse click to open the filter/search dropdown on a column header. ​While Shift + F10 opens a context menu, it only provides sorting and copying options,it does not allow the user to access the filter checkboxes or the search bar within the column. ... View more

Please update the CICD interface where you can add objects to a 'cart' in order to deploy.   So as you scroll through the list of objects you can click and them them to your deployment cart.   The reason for this request is because if you have many objects and want to search and add them, everytime you search you lose whatever was previously added. ... View more