Hello Microsoft Team,   Currently, Power Automate offers a trigger for Power BI dataflow refresh completion, but there is no native trigger for Power BI semantic model (dataset) refresh completion. According to community and documentation, this gap forces users to rely either on data‑driven alerts, which cannot detect failure states, or REST API polling, which requires Premium HTTP actions.   Many business workflows depend on knowing whether a dataset refresh succeeded or failed, especially when downstream automation or data quality processes rely on up‑to‑date semantic model data.   I would like to request a new built‑in trigger:  “When a Power BI semantic model (dataset) refresh completes” with outputs such as:   Refresh status (Succeeded / Failed) Error details Start and end time Workspace ID and Dataset ID This would eliminate the need for complex REST API workarounds and significantly improve automation reliability for Power BI + Power Automate users. Thank you. ... View more

Hello Power BI Team, I would like to suggest a new feature for the Date Slicer. Today, when a report is refreshed, the date slicer keeps the last selected value instead of automatically updating to the latest date available in the data model. This creates a problem for dashboards that are updated daily, because the slicer continues to show an outdated date unless the user manually resets it. Requested Enhancement: Add an option in the Date Slicer properties allowing the user to set the slicer to automatically default to the most recent date present in the related table. For example: “Auto-select latest date in data” = On/Off When enabled, after every refresh, the slicer should automatically adjust itself to filter the report by the latest date loaded into the dataset. This behavior would be extremely helpful for operational dashboards that are updated every day and always need to reflect the newest data without user interaction. Why this feature is important: It improves usability and avoids outdated information being shown when a user opens the report. It eliminates the need for complex DAX workarounds or visual-level filtering tricks. It supports real-time and daily operational dashboards where the “latest available date” is always the default expected view. This feature would make the Date Slicer more intuitive and aligned with business scenarios where the newest data must always be displayed automatically. Thank you for considering this improvement! ... View more

Hi team, I would like to propose a feature enhancement around Fabric Data Agents and their integration with Power BI.   Problem Statement Currently, Fabric Data Agents are primarily accessible through the Copilot interface. While this works well for exploratory analysis, it introduces limitations for production-grade and embedded analytics scenarios where UX control is critical. In real-world deployments (especially client-facing dashboards), relying on the Copilot pane: restricts customization of the user experience limits integration with report visuals makes it difficult to embed conversational analytics seamlessly within applications Proposed Enhancement Enable native integration of Fabric Data Agents directly within Power BI reports/dashboards, without requiring the Copilot interface. This could include: A dedicated “AI Visual” or component that can be placed on the report canvas Ability to bind a Data Agent to a semantic model used in the report(Currently, we have in Copilot, but adding a data agent directly would be a plus) Support for interactive querying directly within the dashboard (e.g., chat or prompt-based input) Mechanisms for agent responses to interact with visuals, such as: dynamically filtering visuals generating insights tied to selected data context augmenting dashboards with narrative explanations Why this matters There is a growing demand for conversational + visual analytics in a unified experience. Many enterprise use cases (e.g., customer segmentation, profitability analysis, behavioral insights) require: structured dashboards for KPIs and an AI layer for ad-hoc reasoning Currently, developers are forced to build custom hybrid architectures (Power BI Embedded + external AI layer) to achieve this, which increases complexity and reduces adoption of Fabric-native capabilities.   Suggested Capabilities / API Considerations Programmatic access to Data Agents (REST / SDK) Event hooks between report interactions and agent context Embedding support for agent interaction in Power BI Embedded scenarios Ability to pass filter context from report → agent and vice versa Impact This would: significantly enhance Power BI as an AI-first BI platform reduce the need for external custom implementations unlock new enterprise use cases for embedded analytics and intelligent dashboards I believe this feature would be a major step toward making Fabric a truly unified platform for data, AI, and business intelligence. Would love to hear if anything like this is on the roadmap or being considered. Thank you. ... View more

Currently, Microsoft Fabric offers three certifications: PL-300 (Power BI Data Analyst) DP-600 (Fabric Analytics Engineer) DP-700 (Fabric Data Engineer Fabric is also becoming a powerful platform for Data Science workflows, especially with capabilities around Lakehouses, Notebooks, Spark, ML integration, and advanced AI tools.   I would like to suggest adding a dedicated Fabric Data Science certification. This certification could cover areas such as: Building and managing data science environments in Fabric Using Lakehouse and Spark for advanced analytics Integrating Machine Learning models with Fabric pipelines Applying responsible AI practices within Fabric End-to-end workflows from data preparation to model deployment Such a certification would provide recognition for data scientists using Fabric and help organizations validate expertise in this growing area. Looking forward to the community’s thoughts on this.  Rufyda Rahma | LinkedIn ... View more

We want the ability for a creator to customize the data agent suggested prompts so that users can select from the applicable prompt suggestions. ... View more

Summary: Please add support for running Marimo notebooks in Microsoft Fabric. Marimo is a new reactive Python notebook framework that enables fully reproducible, Git-friendly, and app-like data workflows.   Why it matters: Fabric notebooks are great for Spark workloads but limited for lightweight, interactive data apps. Marimo would let users:   Build reactive notebooks that update dynamically Run small workloads without a Spark cluster Version control notebooks as plain .py files Deploy notebooks as interactive web apps (WASM or server mode)     Requested capability:   Allow installing and running the marimo package in Fabric notebooks or Lakehouses (Future) Add a “Marimo Notebook” item type or integration similar to existing notebooks     Impact: Adds a modern, flexible notebook experience in Fabric — bridging traditional notebooks and lightweight web apps, and encouraging open-source integration.   References:   https://docs.marimo.io ... View more

For deep learning tasks it would be great to have an easy way to use cuda devices in MS Fabric notebooks. ... View more

Currently the amount of data that shows up on Fabric data agent after a prompt is very limited. It currently outputs a maximum of 25-30 rows of data. It would be super helpful to output 1000+ rows of data as summarized tables and data can easily reach 1000 rows sometimes across various dimensions. This will significantly help advance the use case of Fabric data agents.  ... View more

Could we include chart / visualization rendering capability within Fabric data agent. This will be super helpful to visualize the output results while doing Q&A on the data.  ... View more

As a Python developer I need to run unit tests on all my PySpark notebook code. All the functionality already exists to run these tests automatically but the naming convention of Notebook files completely ignore python module and package naming rules. There are two issue that would be fantastic to address: The Notebook folder created from git integration should not have periods in the name, eg "my_notebook.Notebook". The notebook python script created from git integration should not have dashes in the name, eg "notebook-content.py" My idea is to simply replace these symbols with underscores. I hope it's a small enough change and it would have a massive impact on enabling automated testing for all Notebook scripts ... View more

The new Rich dataframes in Fabric notebooks are nice but the Chart View still leaves much to be desired for compared other notebook products such as Databricks, HEX and notable. Here are a few proposals and ideas: Similar to Cell and Markdown, create a "Chart" cell type which lets the user pick a dataframe and display it. Users do not need to wrap display(..) every single time. Visuals available in display are very limited and do not allow any formatting or configurations. e.g. you cant change the axis colors, change axis scale. While histogram is available, density plots, scatterplots etc for any serious and meaningful data exploration once created the visuals created w/ display do not persist. e.g. if the notebook author creates a visual, saves the notebook and shares it with someone else, the other users don't see the original chart Fabric already has Quickvisualize and powerbiclient which allow use of Power BI in notebooks but a deeper, native integration to leverage Power BI inside Fabric notebooks would be great especially for those who are new to notebooks as well data scientists who can use UI option in the notebooks for visualization. ... View more

The great thing about Fabric is that we can share data between Lakehouse and Warehouse in one workspace. However, the utility is massively unlocked when we can share workload/processing/pipeline across the workspace too. This means the ability to run a bit of Spark notebook, then a stored proc, then another notebook, then stored proc again etc. Usually a mixed pipeline would need a secondary tool like Airflow or Prefect. However, if we can execute T-SQL stored procs from the Notebook, then we can orchestrate a pipeline across lakehouse/warehouse from within Fabric itself. This will be massive. Please add the ability to easily execute Warehouse T-SQL from Spark notebook. Presumably we can try to do this by setting up the Apache Spark connector but it would be quite complicated. ... View more

Integrate a chat UI for AI skills into organizational apps and make it available to users. Developers can provide a query interface to users without needing to set up Copilot Studio or AI Foundry ... View more

It would be extremely helpful to have a built-in option to export or download notebooks in Microsoft Fabric directly as PDF files. This feature would support better sharing of work, documentation, and reporting, especially for stakeholders who prefer non-editable formats. A native "Export as PDF" option (similar to JupyterLab or Databricks) would significantly improve usability and collaboration for data teams working within Fabric ... View more

Currently we can run Open AI services for in Azure Synapse Analytics in lowest possible spark compute, which is most cost effective for us to have a dev playground. Spinning up an F64 just so that people can test it is not the best solution available. ... View more

It would be helpful to the significant group of R Users if Fabric had support for "R Environments" in the same way that Python is supported. With Python environments, I can specify, download, and use libraries (including the specific versions) that I would like to use in my sessions. This is facilitated via the "Environment" object in the Fabric workspace. This functionality does not exist for R. While it is possible to download and use libraries (including the specific versions) in the CRAN repository, I need to re-download the libraries whenever I start a new session and this can take some time, slowing down my analysis, especially if my session is prone to crashing. Essentially, it would be great if there was a functionality to create R environments without needing to do session-level installs. Is this on the radar? If so, how soon until this is implemented? Thanks! ... View more

In our investigation of the new AI data agents, a common question that our users want to ask is something along the lines of "What was the total sales for such and such product this year?" However, the agent would consistently use the year 2023 as the current year. Out of curiousity, we asked it "What is the current year?", and again it answered "The current year is 2023." If we ask it about the current time, it doesn't even know how to answer. It's odd that the current date and time are not included by default. ... View more

The management of Spark jobs is done via the GUI Create an Apache Spark job definition - Microsoft Fabric | Microsoft Learn This is fine for ad-hoc development, but not suitable for serious operationalising/automation of Spark pipelines. Please support a REST API and/or python SDK for developers to create/execute/monitor Spark jobs programmatically ... View more