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

There should be a new "Last Updated" Timestamp column in Synapse Data Science when viewing your created notebooks. This would allow users to see when others in their team have last made changes to their notebooks to see what the most up to date working notebook is. ... View more

When training a Data Agent I think it would be good if you included a checkbox where a user could indicate the SQL or DAX query generated correctly answers the prompt. ... 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

All you products must have a dark mode. Light blue on white is impossible to see. ... 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 T-SQL notebook feature in Microsoft Fabric lets you write and run T-SQL code within a notebook. It allows direct execution of T-SQL on connected warehouse or SQL analytic endpoint. Real time data in Fabric is often hosted in Evenhouses, which support T-SQL as well. Being able to use the T-SQL familiar experience to unlock raw and aggregated data is Eventhouses, tables, views and materialized views will complement the story. ... View more

Would love to have the option of a sound chime and/or browser notifications when using Notebooks in Fabric for the following: Successfully starting a new session Completing a run (successful or failure) of a code block ... 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

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

We now have high concurrency for notebooks in pipeline in public preview. This is great news for us using custom pools (and all using private endpoint and therefor using custom pools). However for high concurrency to take place we off course need a session to be active. The problem is we cannot choose how long a session is active and they time out very fast. My suggestion is to let user control the time to keep a session alive after no one using it. An ideal place to make this would be in the environment.  ... View more

Today code completion for notebooks does not benefit of the advanced code recommendations that Github Copilot gives, integration with Copilot will speed up the use of the product for both existing and new users. ... View more

Idea: Fabric notebooks should support checkpointing and autosave checkpoints. Also, there should be a maximum of one connection with cell editing permission to a notebook from a given user at a time. Short version: Most implementations of Jupyter Notebooks autosave multiple (five?) checkpointed versions of your file as you work to help with recovery, as well as have a manual "save check point" menu option to save a dedicated quick reversion point for your file. This behaviour/feature is in Azure Machine Learning notebooks, but isn't in Fabric. It would be nice to have it for consistency of experience. The existing Fabric autosave is an unrecoverable hammer and has caused me more issues than it's solved. Sure, you can revert back to the last thing you pushed to a Git repo --- but unless you're comitting every time you write a line of code, there's a chance that you might accidentally wipe some cells that you wrote between commits and have the default autosave make it not revertable. As for the second connection thing: the same user logged into a notebook set to autosave on two different machines interacts with the existing autosave in a destructive and nonfunctional way. Any edits made by the second connection may be immediately undone after they close the notebook if the first connection is still active. The first connection hasn't necessarily updated the state of the notebook from the changes made by the second, and it autosaves its (old) version of the content over all changes that were made, which is then unrecoverable as far as I know. The long version: I just lost a day's work, so I'm grumpy. It's mostly my fault due to bad practices (I should have just turned off autosave, I shouldn't leave notebooks open etc.), but if Fabric notebooks were checkpointed like AML notebooks literally already are, and/or if Fabric logged out existing connections from editing notebooks upon a second connnection from the same user it could have been avoided. The situation: sometimes I work from home, sometimes I work from the office, each on a different machine. Unfortunately for me, sometimes I forget to close my browser window editing a Fabric notebook on one machine at the end of a workday, and continue working on the same notebook the next day on the other machine. I wrote a bunch of edits/new cells to the notebook at the office PC (Notebook Connection #2) March 4th, unaware that I had left the same notebook open on my home PC (Notebook Connection #1). Today, March 5th, I logged into my home PC, saw an open browser window with notebook connection 1 on my home PC, and found all my March 4th edits overwritten and unrecoverable. To add insult to injury, I even thought I had commited all of my March 4th changes to a git repo at the end of the day from my office computer. However, in the 5 minute span between me closing Notebook Connection 2 on my office computer and then using the Fabric Workspace GUI to commit changes, it appears that Notebook Connection 1 autosaved and overwrote all of the days edits in the notebook, and the commit simply registered what was effectively the March 3rd version of the file. I didn't notice the discrepancy until digging into the repo today as the GUI commit doesn't show any details while it's being applied other than which files are being synced. I don't know what scenario there could be where having multiple connections simultaneously connected autosaving over each other is a benefit, as all it does is immediately erase any edits. I know I shouldn't have put myself in a position where this could have happened, but it seems silly that it's even a possibility. It's a bit of a rare scenario, but with the number of hybrid workers these days I'm sure I'm not the only one who's ever run into it. ... 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

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

Hi, In the Microsoft Fabric when I am in the Synapse Data Science Notebook on the left pane I am only seeing an option to add Lakehouse to the Lakehouse explorer. But as a Data Scientist if I want to explore DataWarehouse as well how do I do that? Can we also add an option to add warehouse also the explorer ? ... View more

Its very common to need to integrate Fabric PySpark into office to download excel files. Currently this is difficult to do and if we do a pip install at Library level, the startup times is too slow. Can we get the following PIP installs into the standard Spark "Started" image ? pip install msoffcrypto-tool pip install Office365-REST-Python-Client This would add support for reading and writing files from SharePoint using Notebooks and also to support password encrypted excel files ... 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

The base environment for SparkR doesn't seem to have "less" installed on to it - which causes several off-the-shelf R libraries to break. Could you please include "less" in the base OS image? ... View more