Sign inGet started


May 16, 2024

Iterative AI editing & faster AI responses

Untitled (63).png

Iterative AI editing

Asking Deepnote AI to edit the content of a Python or SQL block is a powerful feature that can significantly enhance your ability to modify and debug your code. Often, achieving the optimal outcome involves iterative changes: you request a modification, run the code, and then realize further adjustments are needed. This back-and-forth interaction with the AI is typical when working with LLMs. However, until now, it required a workaround in Deepnote. Previously, you could only request a single edit at a time, meaning multi-step changes necessitated duplicating the given block and starting anew with a fresh prompt.

We’re excited to introduce iterative AI editing to streamline this workflow. Now, when you request an AI edit in a code/SQL block and receive a suggestion, you can immediately type in a new prompt as a follow-up. The AI will process this new prompt and provide another edit suggestion. As always, you can instantly run the code for these suggestions to observe how they impact your outputs.

Each AI follow-up suggestion is added as a new item to your edits history list. If you wish to revert to a previous stage in your editing flow, you can simply click on the desired step from the list, returning the block to that specific state. This feature allows you to effortlessly compare results across different prompts and continue with the most promising options.

This iterative editing flow essentially functions as an isolated mini-chat with the AI, ensuring that all the context of your prompts and AI outputs is preserved during your editing. The significant advantage is that you don’t need to repeatedly rewrite your original prompts; you can simply use short instructions as you would in a chat. For instance, if something doesn’t quite work, you can just say "try another option," and the AI will understand how to proceed.

The new editing flow can also help you keep your notebook more organized! Working with generative AI is inherently experimental, and maintaining the in-between stages in your notebook can create a mess of duplicated blocks. With iterative editing, you gain a safe drafting space, allowing you the freedom to commit only the final results to the notebook, ensuring it remains clean and focused for others to see.

See the new AI editing features in action in the video below!

Faster AI responses

You might have noticed that the speed of AI code generation, editing and fixing has significantly increased! This improvement is due to our switch to the latest and most capable OpenAI model. We hope you enjoy the 2x speed boost, which pairs especially well with the new iterative editing flow!

Reactive notebooks & AI in chart blocks

Untitled (62).png

Reactive notebook execution

Traditional data notebooks are executed block-by-block, allowing for a block to be executed out-of-order or even repeatedly. While this is very powerful, it also has the potential to create hidden state that can complicate data reproducibility and clarity.

One common problem is the potential 'staleness' in your notebook. For instance, if you update data at the beginning of your notebook but don’t rerun the whole thing, later sections might still be using old, outdated data and outputs (e.g. a chart that doesn’t reflect your latest filtering logic). Typically you would circumvent these problems by manually re-running your whole notebook (or larger sections of it), which can be an inefficient (and still error-prone) solution.

To resolve these core issues, we've introduced reactive execution mode in our notebooks. Simply put, reactive mode makes notebooks similar to Excel spreadsheets — you just input the data and code, and the whole document updates itself accordingly. No need to click to manually run blocks. This makes reactive notebooks easier to work with and reason about.

To activate this mode, simply click on the option dropdown next to the Run notebook button and select Block and dependent blocks.

reactivity (1).png

Once enabled, reactive execution becomes the default mode. This means that when you execute a specific block, all downstream blocks with dependent variables will automatically be re-executed. For instance, if you update an input block, any other blocks that depend on this input will automatically update. Similarly, if you alter the filtering logic on a dataframe, any charts visualizing this dataframe will automatically refresh to reflect the changes.

Reactive execution mode is a major enhancement to the core notebook experience, boosting your efficiency and helping to ensure that your data and results always remain up-to-date.

Check out this video to see it an action! You can also discover more details in our 👉 docs.

AI in chart blocks

With our latest update to Deepnote AI, you can now generate or edit visualizations in your chart blocks by using natural language. Just type in your question or the specifics of the visualization you need, and hit Submit. In seconds, you'll see a suggested chart. You can approve this suggestion by clicking Done or reject it with Undo.

2024-04-25 13.36.54 (1).gif

The suggested chart comes fully configured, serving as an ideal starting point. From there, you can fine-tune the details to better suit your needs. If you need further modifications, just submit a new prompt, and Deepnote AI will adjust your existing chart accordingly.

2024-04-25 13.45.18 (1).gif

Please note that Chart AI is currently in Beta. While we are excited to offer this feature, some functionality may not be perfect yet. We welcome your feedback to improve the service—use the built-in 👎 button or share your thoughts on our product portal.

For more details on chart blocks, visit our 👉 docs.

IAM roles for S3 and Redshift, notebook locking & custom chart tooltips

Untitled (59).png

Role-based authentication for S3 and Redshift

We've introduced a new, secure way to connect to your S3 buckets and Redshift data warehouse - now available through IAM role authentication! This feature leverages AWS's native role-based access control, providing a robust security enhancement over traditional static keys. By using IAM roles, you can benefit from:

  • Automatic credential rotation: temporary credentials that reduce the risk of long-term exposure.
  • Simplified permissions management: enables more precise control over S3 resources, adhering to the principle of least privilege.
  • Eliminated need for static credentials: decreases the potential for security breaches.
IAM (1).png

To set up your connection, choose the “IAM role” option in the "Authentication" section, input the Amazon Resource Name (ARN) of the role you wish to use, and update the trust policy as outlined. This process applies similarly for both S3 buckets and Redshift clusters.

For step-by-step instructions and more details, visit our S3 and Redshift docs.

Notebook locking

We’ve all been there - you go to run an app or a notebook, and it doesn’t work because someone accidentally changed something they shouldn’t have. To prevent these frustrating mishaps and ensure your work remains stable, we’re introducing notebook locking.

Untitled (57).png

Simply enable the lock from the notebook’s context menu. While locked, the notebook can still be run, shared, or commented on—just no edits allowed. When updates are necessary, users can unlock it, make their changes, and rest assured that it will automatically lock again afterward.

Learn more in our docs.

Custom tooltips on charts

Ever needed to show just a bit more information on a chart without cluttering the entire visualization? Our new custom tooltips feature is here to help. Now you can enrich your charts by selecting additional data points to display from any column in your plotted dataframe—not just those configured on the chart.

It’s perfect for highlighting more information for an unusual data point on a scatterplot or adding context with an extra metric on your bar charts, all without overwhelming your audience.

Simply go to the Display tab and choose the columns you wish to display in the Data selector in the Tooltips section.

custom tooltips (1).png

Scheduled projects view

Stay on top of your automated workflows with our new scheduled project view! This handy new section, located in the workspace navigation sidebar, provides a clear overview of all projects in your workspace that have automation set up, showing exactly when each project is scheduled to run.

schedule (2).png

SQL AI Beta release

SQL AI.png

Writing SQL queries is often one of the first steps in data analysis. With our latest upgrade to Deepnote AI, it can now assist you in getting this essential task done much more easily. Simply click on the AI button in a SQL block and ask a question in plain language - Deepnote AI will interpret it and provide you with a suggested SQL query you can run immediately.

Watch this 2-minute demo to see SQL AI in action for yourself!


You can also use the AI button to help with editing an existing query or fix issues with errors you may encounter.


As with all AI features, the right context is key for accurate suggestions. Deepnote AI utilizes the content of your notebook along with the schema of your connected data warehouse. To ensure the AI is incorporating the most updated schema information, you may want to refresh the given integration’s schema. It might take a minute or two for the updated schema to propagate to Deepnote AI, especially if you are connecting to an integration from your Deepnote workspace for the first time.


Once that’s done, Deepnote AI will be able to take advantage of your schema structure to compose queries with the relevant table and column names. (It may take a few minutes after refreshing the schema to make this information available).

If you prefer to use SQL for data manipulation of Python variables, Deepnote AI can also assist you in that. The AI button is available in Dataframe SQL blocks as well, so you can ask for the AI’s help in writing SQL for any downstream filtering and aggregations.


We hope that SQL AI will make query writing a more efficient, smoother process for you, but as of now, this is a Beta release with some limitations to note.

  • Schemas can get very large, sometimes including thousands of tables. Deepnote AI searches for the relevant parts, but it may not always come up with the right table/columns names for you. If you notice this pattern, try providing guidance in the notebook or in the prompt (e.g. mention in a text block to ‘always use table order_details for order attributes’). This kind of information can greatly help steer the AI to get the right answers for you. Check out our docs for more prompting tricks and tips.
  • SQL AI is only available to assist you in one SQL block at a time. When you use Auto AI to generate multiple blocks for you, SQL blocks won’t be added.
  • Supported integrations include BigQuery, Snowflake, PostgreSQL, Redshift, Athena and Clickhouse, MySQL, MariaDB and Microsoft SQL Server.

We are actively improving SQL AI to make the suggestions more accurate and relevant for you. Your feedback is crucial for us: please let us know if you encounter issues by clicking on the thumbs down button and leaving a note!


Be sure to check out our docs to learn more.

Pie charts, histograms & new blocks for more interactive apps

Charts & blocks.png

Pie charts

Visualization experts will tell you all about why you shouldn't use pie charts to present your data. But let’s be honest, the pie is a staple of dashboards and your stakeholders often expect them. We have good news: you can now easily create those pie charts in our native chart blocks! Simply select Pie chart in the Type selector and configure your slices with two clicks!

The new chart type comes with its own customization options under the Format tab. You can switch between absolute and relative views (switch Show as to ‘Percentage’); or change where you position the value labels in the Position dropdown. What’s more, you can easily turn your pie into a doughnut by adjusting the Center gap in your chart. Happy baking!


We also added another chart type to the roster: histograms! Yes, you could already create histograms in our bar charts, but it was a bit too finicky a process and many of you requested a simpler way to create these visualizations. We listened and made it as easy as possible: just pick the new Histogram chartfrom the type selector and select a single column - that’s it! If you need to customize the chart, you can adjust the binning parameters by clicking on the BIN selector. Here’s to quicker basic data exploration!


New blocks types

Blocks are kind of a big deal in Deepnote and we just added not one, but 4 brand new block types so you can build more appealing and interactive Apps!

Text area input: a more spacious input block that can nicely accommodate multi-line texts.

File upload: this powerful input block enables your consumers to upload and process their own data in your Apps with ease.

Button: a customizable button that makes submitting input changes a lot clearer and easier for your App consumers.

Separator: a handy new text block type for better visual separation of content. Tip: you can also type three dashes ‘—-’ to insert a separator.

We hope these new building blocks will turn your Apps even more powerful and can’t wait to see what you will build with them!

new blocks.gif

Simplified hardware sidebar

We also made a small, but important to change to the environment configuration experience in your notebooks. Gone are the days when you had to hunt for your machine configuration and other environment-related options in a tucked away settings submenu: these often used controls are now always visible in the left side panel.

hardware sidebar.png

SQL auto formatting, always-on AI UI and prompt suggestions

SQL Auto-format, Prompt suggestions.png

Code formatting for SQL blocks

Instantly tidy up your SQL code using the new auto-formatting feature for SQL blocks.

Available through the command palette, block actions menu, or as a keyboard shortcut (option+shift+F for Mac orAlt+Shift+F Windows / Linux), the SQL formatter makes querying your data in Deepnote feel like second nature.

SQL formatting.gif

AI front and center, always

Deepnote AI is now immediately ready to assist at every step of your analyses in Deepnote. With the AI prompt box now always visible at the bottom of your notebook, you no longer have to hunt for the AI button to unlock the power of Deepnote AI.

This update to the AI UI is part of a series of changes we’ve added in recent weeks to make Deepnote an AI-first notebook, which brings us to one more AI update…

AI prompt suggestions

The power of an AI coding assistant is only as good as the prompt you give it. We’re no stranger to stumbling upon writers’ (or coders’) block during the prompting process, so we’ve introduced prompt suggestions to make the Deepnote AI experience even more seamless.

Whether you need fresh ideas during your data exploration process, a way of saving time and keystrokes, or just more assistance from Deepnote AI, prompt suggestions will have your back from beginning to end of your Deepnote projects.

Prompt suggestions.gif

Check out our updated docs to read more about the latest and greatest AI features!

Stronger GPU hardware and visual revamp of charts

GPU & Chart colors.png

More powerful machines now available

We’re welcoming a new, more powerful machine to our cloud-based hardware options. When navigating to the machines menu in the Environment pane, you’ll notice a new GPU hardware, which includes the heavy-duty graphics card Nvidia V100 offering 64 GB of RAM and 8 vCPUs.

This stronger machine is currently only available as a pay-as-you-go hardware, which enables uninterrupted execution of code in your most complex analyses, on your largest datasets.


A brand new aesthetic for chart blocks

Your Deepnote notebooks and apps now have a much more polished look and feel. Our latest visual update includes new color schemes for chart blocks, including new categorical and continuous color palettes.


In addition to new color palettes, you might also notice updated layouts, spacing and fonts applied to your charts, making for a sleeker visualization experience in Deepnote. With these visual changes, we’re setting the stage for even bigger changes to come to chart blocks soon—including new chart types and AI-powered charts, so stay tuned!

Faster apps and new AI UI


Lightning-fast apps

We’ve added some more improvements to the app experience that you won’t want to miss. Apps now stay active for longer, meaning you can step away, do your thing and pick back up where you left off without the executor shutting off the app. Now, the app executor keeps running until the user closes the app's browser window, or if the app has been idle for more than 10 minutes.

You’ll also start to notice Deepnote apps now load significantly faster when you turn on the “Allow reactivity” toggle in your app settings. With this new setting, your app will no longer execute all the contents from top to bottom with every input adjustment like it used to. Instead, only the contents that depend on your changed inputs will be recalculated, making for lightning-fast app interactions. Watch it in action in this video.

We’re launching this same reactivity in notebooks soon, so send us a note if you want to be part of the private beta!

AI front and center

Collaborating with Deepnote AI on your analyses is now a seamless experience thanks to a host of UI updates and new powerful interactions.

Convenient new shortcuts make accessing the new UI a breeze. Use cmd + shift + i to ask the AI for an explanation of any block of code, and cmd + shift + e to edit the code.


Intuitive prompting means your AI collaborator can guess your next move. The UI defaults to edit mode when you’re accessing it in a block with existing code, but generate mode in an empty block.

More space for your AI prompts. The new UI is now front and center so you can start writing your prompt right away, but you can still easily see your notebook contents behind the prompt window with the new Peek Behind UI toggle.


New powerful Auto AI interactions. When leveraging Auto AI to generate and run multiple blocks in your notebook, you now have new commands that enable you to iteratively fine-tune the AI’s output with just a push of a button:

  • Run code allows you to check the results of the AI’s suggestions before you accept or reject them
  • Regenerate gives you the power to restart AI generation with the same prompt if you aren’t happy with the AI’s first suggestion
  • Delete allows you to reject the AI’s results and edit your prompt until the code is generated to your liking.

We hope you enjoy your faster apps and the enhanced AI experience, and stay tuned for more exciting updates to come!

New app building experience and Slack integration


Streamlined in-notebook app building

Calling all app builders: we have an update that will turbocharge the app creation process from within notebooks.

You can now control the visibility of blocks in your app directly from the notebook interface.

Hide/show in app: Clicking the disappearing computer icon will hide or show a block or its contents in your app, but will still be visible in your notebook. Blocks that are currently hidden in your app will appear as faded out in the notebook.

Select multiple blocks: Efficiently manage the visibility of multiple app blocks at the same time by using the multi-select feature available in notebooks. Simply hold SHIFT and click on neighboring blocks to hide multiple blocks’ output and/or code from your app.

Dynamic apps: Once you’ve configured your app exactly to your liking, the visibility settings you’ve set in your notebook will automatically come to life in the app view. Gone are the days of constant back-and-forth between notebook and app interfaces to share your data insights!


Slack integration

Conveniently view the status of your scheduled notebook runs from the comfort of your Slack workspaces with our new Slack integration.

This integration allows you to deliver notifications about your Deepnote analyses at desired checkpoints within your data workflows.


To get started with sending your scheduled notebook runs to Slack, check out the step-by-step instructions in our docs or try it out for yourself by duplicating this example notebook.

BigQuery schema browser improvements & more flexible AI copilot

Both (1).png

Multi-project schema explorer for BigQuery

If you’re a power user of our BigQuery integration, you’ve probably taken advantage of the schema explorer to query data from multiple projects in your warehouse using Deepnote.

That process just became even easier with our new multi-project schema explorer for BigQuery, which gives you a high-level exploration of multiple projects at a time and the columns within them, all from one convenient side panel view within your notebook.

With a more unified view of all your BigQuery data, you’re now ready to tackle multiple projects with less set up time and context switching.


AI Copilot toggle

Maintain total control over the working environment within your notebook with the new AI Copilot toggle. If you’d prefer to kick off your code and queries without any AI assistance, easily disable Copilot with a flip of a switch.

This new feature is a persistent user setting, meaning once you turn it off, you disable it for yourself for all projects. Need to turn Copilot back on later on? Simply toggle it back on and it stays on for you.


Faster HTML cell outputs

Spend less time waiting for your charts to spin up and more time on the parts of your workflow that actually matter.

It’s no secret Deepnote supports your exploratory programming needs with the power of interactive visualizations, whether that’s through code or no-code chart blocks. If you prefer using code to build your visualizations through HTML-based libraries like Plotly, your plots will now load with lightning speeds.

We hope these updates bring you faster data exploration and workflows more tailored to your data needs in the new year. Rest assured, we have plenty more updates in store for you in 2024, so stay tuned!

Next page



  • Integrations
  • Pricing
  • Documentation
  • Changelog
  • Security




  • Privacy
  • Terms

© Deepnote