Improved AI code editing & fixing

Diff viewing for AI editing and fixing

We’re shaking things up with Deepnote’s AI code editing and error fixing workflow.

Auto AI is not just for generating code, but also for editing and fixing that generated code. Now with AI 2.0, changes made by the autonomous agent will be displayed in a diff viewer so that you can see what was added, deleted and updated, side-by-side with the original code.

The AI-generated content in the “new” side of the diff viewer is automatically written into your code block so that you can immediately see the impact of the AI’s change on your analysis. Change your mind? Just undo it with a click of a button using the new Done/Undo button.

Notebook variables as AI context

Auto AI has always had a deep understanding of your project context, including your content in your blocks, outputs and filenames. But now we’ve extended its awareness to also cover all the variables in your notebook. This will provide even more accurate suggestions, particularly in cases when the given variable has not already appeared in outputs. What does this mean for improvements to your coding workflows? No more “name ‘variable’ is not defined” errors.

With the AI’s new awareness of available DataFrame variables, Auto AI is now equipped with the metadata it needs to work with you in lockstep as a true collaborator.

Visual updates coming soon to a screen near you

You’ll be noticing some changes in Deepnote’s UI soon. We’ve been working meticulously to create consistent, beautiful UI throughout Deepnote to add a more polished and sleek feel to your experience. The most highly requested update? Dark mode. Stay tuned for a more official announcement soon!

Introducing: Fully autonomous AI

The ultimate AI-powered notebook has arrived.

We’ve been keeping this one under wraps for several months now, but we’re excited it’s finally here: a completely autonomous, self-sufficient virtual collaborator. How powerful is it?

Give your AI teammate a task and it will continue working until it completes it. It generates text and Python blocks, adapts to prior outputs, and fixes its own errors all in the same run. After the code is generated, it automatically executes the code for you, even while you’re working in other notebooks.

It's aware of your context from other blocks and variables in your notebook and files.

While it can operate completely independently, you still maintain control. Switch it off using the Auto AI toggle, or even stop it mid-run if you’d like. Of course, it’s built with collaboration in mind. Any other collaborators working in your notebook can see its progress the same way you can.

Read more in docs →

Enhanced apps: Comments and multi-select inputs

Comments in apps

App users now have the ability to add comments directly in the app interface.

Comments added to your apps will be reflected in your notebook and vice versa, giving you the power to seamlessly and collaboratively share your thoughts from beginning to end of your analysis.

This highly requested feature brings the collaborative capabilities that that you know and love from Deepnote notebooks now to the app level too. This means no more screenshots of apps that often lack context and are not very interactive.

Select multiple options for your input blocks

We’ve added a small but mighty addition to input blocks used in both notebooks and published apps.

You can now select multiple options in input blocks, which will make your stunning apps and polished dashboards even more interactive from the ability to filter app outputs to reflect your input selections.

Multiple-option selection joins the four other types of inputs that are available: text, select, slider and date input. All five of these input types allow you to use your inputs as variables in your code or SQL blocks, thus further empowering your colleagues and stakeholders to be able to interact with and understand insights from your notebooks and apps, regardless of their coding abilities.

Read more in docs →

Create full notebooks with AI & GPT-4 support

Generate multiple blocks with AI

We’ve taken AI code generation in Deepnote to a whole new level with multi-block generation, which will give you the ability to generate entire notebooks with just one prompt and a click of a button.

Whether you’re tackling routine data tasks or more complex projects, this feature automatically generates multiple blocks of code that build on each other, while still allowing users to make iterative edits and maintain control over the flow of your notebook.

With this multi-block generation capability, simply provide your prompts, questions, or instructions, and watch as an entire data notebook—including code, text and plots—materializes before your eyes.

Check out our quick demo of Deepnote AI's multi-block generation feature.

GPT-4 to power AI code suggestions

Deepnote AI just got radically smarter with GPT-4.

The power of GPT-4 is now fully embedded into the entire process of creating a notebook from end-to-end, including the code generation, editing, explaining and fixing steps, as well as the ability to learn from the cumulative context of the contents within the project you are building together with your trusty AI sidekick.

With this enhancement to our growing suite of AI tools, you can now expect more accurate and succinct recommendations that are sure to make Deepnote AI an indispensable part of all your data projects.

Read more in docs →

Bulk actions & Deepnote AI improvements

Multi-select block actions

Bulk actions have arrived in your notebook! Effortlessly select multiple blocks and perform a variety of tasks with them at once.

When you’re in a block, simply hold SHIFT and click on neighbouring blocks to include them in your selection. Alternatively, tap Esc within a block, followed by SHIFT+↓/↑ to expand your selection range.

Once the blocks are selected, you can access a range of actions in the block actions menu. You can run, duplicate, delete and move the selected blocks up/down. In addition, you can also copy-paste the selected blocks by pressing Cmd+C and Cmd+V. Say goodbye to the days of handling blocks individually and enjoy this little productivity boost!

Richer context for Deepnote AI

Deepnote AI now has a more profound understanding of the context of your project, leading to even smarter and more relevant suggestions.

When you use Deepnote AI to generate or edit code, it now incorporates a broader view of your project. This includes the entire content of your notebook, whether it's code, SQL, or text blocks. This means Deepnote AI is fine-tuned to your project, providing you with a more customized assistance.

But that's not all! We're introducing a new option that allows you to decide if you'd like to share the outputs of block executions with Deepnote AI. By simply toggling the 'Provide access to block outputs' switch in Settings & Members → Project settings, Deepnote AI can read the outputs in your notebook. This greatly improves the relevance of its suggestions and enables a more conversational interaction. For instance, you can simply ask the AI to do the next step by referring to previous results, streamlining your workflow and sparing you from detailing every prompt.

Read more in docs →

AI Copilot in SQL blocks

Deepnote's AI Copilot just got a major boost! With added support for SQL and an improved understanding of your project's variables, it's ready to take on more of your coding tasks.

SQL support

AI Copilot now makes writing queries against your database or data warehouse even easier with new SQL completions. By pulling context from your notebook's content and relevant SQL blocks from across your workspace, it aligns suggestions with your SQL flavour and schema. And if you're working with DataFrame SQL blocks, it helps there too, smoothing the switch between Python and SQL. It's all about making your coding a bit more fluid across both languages.

Runtime variables as context

We've brought variable explorer content into the context shared with the AI Copilot. Now it recognizes variable and column names, which means more accurate and relevant suggestions when you're working with existing variables. No more guessing or double-checking variable names; this change helps you work faster and with more confidence in your ongoing projects.

Introducing Deepnote apps

Turn your notebooks into apps

Some analyses are best presented with the journey to get there, while others call for a more polished interface. You may wish to hide some code blocks so they don’t distract from the story you are trying to tell, or parametrize the notebook with a couple of inputs so your audience knows which knobs to turn. With Deepnote, you can now create stunning apps that will impress your CEO.

Published apps just got a lot more powerful

Apps now execute within your projects providing support for all environment options you are used to. You can use stronger machines for more compute or RAM, install additional packages or use custom Docker environments.

Improved layout options with columns

Arrange blocks in your notebooks into multiple columns with drag-and-drop. This allows you to create a beautiful and organized notebook layouts that will be immediately reflected in your app. You can also make the apps full-width to showcase charts on larger screens. The possibilities are endless.

Read more in docs →

Next steps

Your existing dashboards will continue to function for the next 6 months without any changes. When you open the projects, Deepnote will guide you through the transition process, typically by creating a copy of the published notebook. In some cases, automatic conversion of dashboards may be possible, but others may require your review.

Stay tuned for more improvements to apps! The next big one we are working on are comments in apps.

Streamlined project view & autonomous AI sneak-peek

Welcome the new layout

We've been hard at work lately, putting your feedback into action, and we're excited to show a reimagined Deepnote navigation and workspace access. This isn't just a tune-up, we've rethought the whole experience to better serve you.

Now, your project view is crisp and clean, with a single sidebar on the left showcasing all your notebooks, files, and integrations. You can easily access what the rest of your team is working on and browse other projects by clicking the top left menu icon. But, rather than describing in-length all the details of what changed, we're more excited to hear your thoughts after you've given the new look a try. To share your feedback, all you need to do is click the in-app question mark in the bottom right.

And there’s more!

• Drag to resize the sidebars, or drag them all the way to the side to hide them. You can also press Ctrl + .(⌘ + . on Apple devices) to hide the UI.
• Quickly search and jump between projects by hitting Ctrl + K (⌘ + K on Apple devices). You can even easily search any kind of text and we will look it up for you in project’s code or text.

From earlier: Autonomous Deepnote AI in development

We are working on an exciting AI update - an autonomous AI assistant.

You can give Deepnote AI a goal, and it will work in our notebooks, writing code, running it, looking at the results, and continuing based on what are the results. This turns out to be extremely powerful, similar to ChatGPT code interpreter. As an added bonus, all the code and explanations persist in the notebook for you to review, tweak, and share with colleagues.

We’ll be shipping this feature soon, with priority access given to our Team and Enterprise customers.

Check out the demo video

Faster notebook startup & images in notebooks

Faster and simpler notebook initialization

We are excited to announce a significant improvement to Initialization (’Init’) notebooks - the special configuration notebooks that automatically execute each time you start your project, allowing you to set up your preferred environment effortlessly.

In our latest update, we have made the 'Init' notebooks more transparent and user-friendly. Previously operating behind the scenes, we have now given them more prominence in our UI, making it easier for you to configure and manage them.

One of the key changes is that we no longer automatically create an initialization notebook for new projects by default. This means that projects not relying on custom environments will now start up much quicker than before!

Creating an ‘Init’ notebook is a breeze. Simply navigate to your project's Environment section in the project sidebar, click on the settings icon, and then press the Create an initialization notebook button. Alternatively, you can also create a new notebook by clicking the plus button in your Notebook list and naming it 'Init'.

Once created, the 'Init' notebook will be conveniently located at the top of your Notebook list, ensuring easy access and quick configuration for your projects.

Read more in docs ->

Image blocks

Deepnote now natively supports images in the notebook editor, supercharging your storytelling capabilities. Simply drag and drop an image to the notebook, or upload it to files and hit Add as an Image block in the file right click menu. This is supported for .png, .jpg, .jpeg, .gif, .svg and .webp files.

You can also adjust the width of the image or the alignment by clicking the block actions menu on the right.

Search your workspace & Deepnote annual plans

Full-text workspace search

We have overhauled our search experience and now every piece of Deepnote information you need is always at your fingertips. Save countless hours and build on your organizational knowledge instead of starting from scratch every time.

Whether it's metadata, code, titles, or comments, you can now effortlessly find the exact information you need by hitting Ctrl + K (or ⌘ + K on Mac) and typing in plain language what keywords you are looking for.

Annual plans

In addition to monthly payments, you can now upgrade to our annual plans and enjoy the full power of Deepnote at a 20% cheaper price than the monthly option.

Learn more at our pricing page ->