AI chat & AI code completion in files & input block improvements

AI chat: your data thinking partner

While our AI has always been excellent at executing tasks, we recognize that sometimes you need a thinking partner rather than a doer. That's why we're introducing a dedicated AI chat interface for more exploratory conversations about your data challenges. Check out our quick video below to see it in action!

Located in the bottom right corner of your projects, this new chat window creates a space where you can freely brainstorm ideas, discuss analytical approaches, and explore potential solutions before committing to code. It's perfect for those moments when you're still formulating your thoughts or need guidance on the best way forward.

The chat interface is fully integrated with your project environment - it understands your data, notebooks, and overall context. Ask it how to tackle a specific analysis problem, request suggestions for improving your notebooks, or explore different methodological approaches, all through natural conversation.

We've also enhanced it with comprehensive knowledge about Deepnote itself. Need help finding a feature or understanding how something works? The AI can quickly answer your Deepnote-related questions, saving you a trip to the documentation and keeping you focused on your work.

This new chat interface is just the first step in our effort to create a unified and more powerful AI agent experience in Deepnote. We don't want to spoil surprises, but here's a tiny sneak peek: what if the AI could turn your messy exploration into a structured, executable data story with a single prompt? That's all we'll say for now - watch this space!

AI code completion: now in the file editor

We've brought our popular AI code completion to the file editor! Now you can enjoy intelligent code suggestions while editing Python scripts, including Streamlit apps.

This enhancement gives you the same great experience you're used to in notebooks. The completions are lightning-fast as you type and contextually aware - understanding your project files and integrations to provide relevant suggestions. It's particularly handy when building Streamlit apps that connect to data - code for accessing your files and setting up data connections will be automatically suggested.

Working on scripts outside of notebooks just got a whole lot smoother. Whether you're building a Streamlit application or maintaining utility files, our AI assistant is ready to help speed up your coding.

Input block improvements

We've shipped several upgrades to our interactive inputs based on your requests! These improvements give you more flexibility when building data apps.

First, we're introducing a new date range input block. Now you can select both start and end dates from a single widget instead of managing them separately. We've also added relative date options like "last 7 days" or "last month," making it easier for your stakeholders to navigate dashboards.

You can now define specific default values for your input blocks too. Simply go to the block settings and set a default value. When someone opens your app in 'run app from scratch' mode, they'll see these defaults right away, giving you better control over their initial experience.

We've also added support for empty states in select inputs. Previously, you always needed to have a value selected in dropdown menus. Now you can enable the 'allow empty values' setting, letting users remove all selections when needed. This is particularly useful for workflows that start from a blank state or for removing filters. Remember: if you enable empty selection, make sure your code or SQL can handle None values.

Query chaining & checkbox inputs

Chained SQL

Build complex SQL queries the notebook way - one block at a time! Our new chained SQL feature lets you break down massive queries into simple, manageable steps without worrying about memory limitations. See how it works in this quick demo:

If you've ever built a complex SQL query using Common Table Expressions (CTEs), you know the challenges. Large subqueries can crash with out-of-memory errors, forcing you to upgrade hardware or implement tedious workarounds. Debugging means commenting out sections or running separate queries, then stitching everything back together - it's a headache.

Chained SQL solves this elegantly. When running SQL queries, you'll now see two output options:

• The familiar DataFrame mode that returns full results
• Our new query preview mode that returns a lightweight query object (just 100 rows)

The magic happens when you reference these query objects in subsequent SQL blocks.

Here's a simple example: first, we query our USERS table with query preview mode. Then, in a separate block, we query the USERS_AB table the same way.

Finally, in a third block, reference both query objects to join them together and see which user was in which variant - all with clean, simple SQL in each block.

We automatically build CTEs in the background, linking your blocks together while keeping memory usage low. Your blocks stay clean and focused, showing only the SQL you need for each step. Need to see the full query? Just click "Show compiled SQL" in block actions.

When you've perfected your query chain and want the complete results, simply switch the final block to DataFrame mode. This gives you complete control over when to pull data into memory.

Chained SQL brings the notebook philosophy to complex queries - build incrementally, debug easily, and manage memory efficiently. Visit our documentation for all the details.

Checkbox inputs

We've added a useful new feature to make your notebooks and apps more interactive: checkbox inputs! These new input blocks let you easily toggle Boolean values on and off with a simple click. Creating one is straightforward - just click on the Input button in the notebook footer and select Checkbox input. You can customize it with an optional display name that shows on your apps and configure the label text that appears next to the checkbox. They're perfect for filtering data on Boolean columns, or adding toggle options to your interactive dashboards.

Snowflake key-pair authentication

We've added support for Snowflake's key-pair authentication, giving you a more secure way to connect to your Snowflake data. This authentication method eliminates the need for password-based access by using cryptographic keys instead. Simply generate your private and public key pair, register the public key with your Snowflake user account, and use your private key when connecting from Deepnote. It's more secure and just as convenient once set up. Check out our documentation for step-by-step instructions on generating keys and configuring your connection.

Redesigned native charts & AI with modules

Revamped chart builder

We've completely rebuilt our charting experience from the ground up! Charts are now simpler, faster to create, and more powerful than ever before. See all the improvements in action in this quick video:

The first thing you'll notice is our cleaner layout. Instead of jumping between three separate tabs to configure your charts, everything is now in one place. Data, style, and global settings are all neatly organized on a single screen, making chart creation faster and more intuitive.

Creating specific chart types is now much more straightforward. We've added 14 purpose-built chart templates, each with just the options you need for that specific visualization. Want a pie chart or histogram? Just select it and you'll see only the relevant settings - nothing more, nothing less.

Composite charts have been totally reimagined. Say goodbye to managing complex layers - now you can simply add additional series to your chart for powerful combinations. We've even added a dedicated chart type specifically for building these multi-series visualizations.

You'll discover plenty of quality-of-life improvements too. Changing colors and renaming series is now just a click away. Our AI support has been enhanced to handle more complex requests, including building combo charts for you. There are many more small touches waiting for you to discover.

Go ahead and try out the new charting experience - we think you'll find it makes data visualization in Deepnote faster and more enjoyable than ever. Check out our documentation for all the details.

Deepnote AI now uses your modules

Last week we introduced modules - a game-changer for making notebooks truly composable. This week, we're taking it a step further by bringing modules into Deepnote AI.

Here's why this matters: when you build a semantic layer with modules (defining your KPIs and metrics in a trusted, consistent way), you're not just helping your team - you're also teaching Deepnote AI. Now when you ask analytical questions, the AI first checks your module library for existing trusted definitions before attempting to calculate anything from scratch. It will automatically select the right module, adjust parameters based on your question, and build the analysis on that foundation.

This is a major leap for AI accuracy. By using your predefined modules, the AI avoids reinventing metrics or potentially getting calculations wrong. Your curated modules create guardrails that keep analyses consistent and trustworthy - whether it's a teammate digging into a deeper analysis or a stakeholder asking a quick business question.

The combination of modules and AI means you get both consistency and speed - the best of both worlds. Check out our blog article to learn more about creating modules for your semantic layer.

Modules: turn notebooks into reusable workflows

We're incredibly excited to introduce modules - a feature we believe will fundamentally change how your team works with notebooks! They let you package your existing best work - those tried and true data cleaning routines and trusted SQL queries - into reusable modules that anyone can import into their notebooks. No more reinventing the wheel with each new analysis. See how it works:

Until now, sharing code or queries across projects meant copy-pasting them across notebooks, making updates tedious and inconsistencies inevitable. Modules offer an elegant solution to these challenges - build once and use everywhere.

Creating analytical building blocks for your team is straightforward: click the module publish button, select which blocks to export as the desired outputs and you're done. You can also add parameters to your module simply by adding regular input blocks - just like when you create interactive apps.

Importing modules is also easy: click on the module button in the footer, find the module you need, configure any parameters, and run it - all results seamlessly integrate into your current notebook. What makes this especially powerful is that each module runs in its own separate environment - none of the code interferes with your current notebook. You just get back the clean output variables you need, keeping your analysis tidy and focused.

Why are modules kind of a big deal? Because they open up so many possibilities.

• Create a semantic layer for KPIs so everyone calculates metrics like churn rate the same way.
• Build modular ETL pipelines that are easier to maintain and debug.
• Package different ML models as modules and cleanly compare them in sequence.

We're already using modules ourselves to organize our data workflows, and honestly, we can't imagine going back. Check out our detailed guide and explore our example notebook to see our approach to building a semantic layer.

Modules are available in your workspace now. Visit our documentation for the details and start building your first module today.

Flexible app layout

Build more engaging data stories with our newly redesigned app builder! You now have complete control over how your apps look and feel, thanks to our new drag-and-drop interface. See these new capabilities in action in this quick video:

Until now, your app's layout was tied to your notebook's structure. Not anymore! We've decoupled app layouts from notebooks, giving you the freedom to create exactly the layout you want. Simply drag blocks around on app preview canvas, place them side by side for multi-column layouts, or stack them into rows. Once you've created columns, you can adjust their widths to perfect your layout. Just grab and resize the blocks to your liking.

This separation from notebook structure means you can now prioritize information however you want - that crucial summary from the bottom of your notebook? Drag it right to the top of your app where it belongs. Your notebook's execution order stays intact while your app tells your data story exactly how you envision it.

Go ahead and start creating more compelling layouts - we built this one especially for you, our app creators! Check out our documentation for all the details.

Deepnote Pro plan, user groups & new GPUs

Introducing the Deepnote Pro plan for individuals

At Deepnote, we believe that every data scientist and hobbyist should have access to powerful tools to bring their ideas to life. That's why we're thrilled to introduce the Deepnote Pro plan, designed specifically for individual users who need more power and flexibility.

Features:

• Unlimited AI: enjoy the full power of Deepnote AI for code generation, bug fixing, and code completion.
• Unlimited projects: create as many projects as you need to bring your ideas to life
• Access to more CPUs and GPUs: tackle even the most demanding projects with ease
• Background code execution: run your code for up to 24 hours without keeping your tabs open
• Price: starting at $9 per month.
• Compute included: $130 worth of CPU & GPU compute every month

With the Deepnote Pro plan, you'll have everything you need to take your data science projects to the next level.

User groups

Managing project access just got easier! You can now organize your workspace members into groups, making it simple to share projects with specific teams or departments all at once.

Until now, you had two options for sharing projects: either with your entire workspace or by adding individuals one by one. Now, you can create groups like "Data Science Team" or "Marketing Analytics" and instantly share projects with the entire group. There's no limit to the number of groups you can create, giving you the flexibility to organize your workspace exactly how you need it.

Setting up groups is straightforward - just head to Settings & Members > Members > Groups, create your groups, and start sharing! You'll find your groups ready to use in your project’s Sharing popover whenever you need them.

User Groups is available as part of our Enterprise plan. Check out our documentation for all the details.

New GPUs: More power, lower costs

We're committed to providing you with the most powerful tools at the most affordable prices. That's why we're excited to announce the addition of new, more powerful GPUs to our platform, available at lower costs than ever before.

Introducing:

• NVIDIA T4: 8 vCPU, 32 GB memory, 16 GB VRAM → $0.90/h
• NVIDIA L4: 16 vCPU, 64 GB memory, 24 GB VRAM → $1.56/h
• NVIDIA 2x A100: 24 vCPU, 128 GB memory (available on the Enterprise plan)

Benefits:

• Power for any project: train ML models, analyze large datasets, or run complex simulations with ease
• Work from your notebook: access all these GPUs directly from the comfort of your Deepnote notebook
• Simple & transparent pricing: enjoy pay-per-minute billing and easily track your machine usage

With these new GPU options, you can tackle even the most demanding projects without ever leaving your Deepnote workspace. Whether you're a researcher, data scientist, or machine learning engineer, we've got the tools you need to push the boundaries of what's possible.

Deepnote x Streamlit challenge is ON!

Time to put your skills to the test! The Deepnote x Streamlit data app challenge is live, and the competition is heating up. Whether you’re solving real-world problems, automating workflows, or just building something fun—this is your chance to shine.

💸 Prizes:

• 🥇 $1,000 + 1 year of Deepnote Team
• 🥈 $500 + 3 months of Deepnote Team
• 💙 Community favorite: $300 (LinkedIn post with the most reactions wins!)

📆 Key dates:

• Deadline: March 3
• Winners announced: March 10

✅ How to enter:

1. Build your Streamlit app in Deepnote
2. Share your app on LinkedIn with #StreamWithDeepnote
3. Get those reactions & let our judges decide!

Don’t wait—start building today!

Streamlit apps

Native Streamlit support

Streamlit, the popular open-source framework for creating data apps, is now fully supported in Deepnote! Turn your Streamlit scripts into hosted interactive apps with just a few clicks - simply drop in your Python scripts and instantly see your Streamlit app come to life in our split-screen preview. See how it works in this quick video:

Sharing your apps is just as easy as sharing your notebooks - your app inherits your project's permission settings, so you can control precisely who has access. Check out our documentation for all the details, or get inspired by browsing our collection of example apps.

While Deepnote's native apps are perfect for turning notebooks into interactive experiences, Streamlit integration opens up new possibilities when you need specialized visualizations or custom UI elements. We've found that Streamlit works incredibly well as part of a larger workflow: use Deepnote's powerful notebook environment with AI assistance to prepare and analyze your data, save your results, and then create beautiful visualizations with Streamlit.

This combination gives you the best of both worlds - Deepnote's robust data processing capabilities and Streamlit's flexible visualization options.

🏆 If you have read this far: This is your exclusive heads-up about the upcoming Deepnote x Streamlit contest, launching next week!

The challenge? Build the most creative, impactful, or downright fun data app using Streamlit in Deepnote. And the theme is… up to you!

Whether you want to solve real-world problems, boost productivity, or simply have fun with data, the stage is yours. Bonus points if you incorporate LLMs or AI agents into your app.

Start practicing and exploring now—this is your chance to get a head start before the official announcement.

Your data stories just got a new way to shine. Let’s see what you’ll create!

Better SQL autocomplete, AI custom instructions & code folding

Better SQL autocomplete

Writing SQL just got a whole lot smoother! We've given our autocomplete a major upgrade to make writing queries faster and more intuitive. See these improvements in action in our short video:

We've significantly expanded what your autocomplete can handle - it now works smoothly with schemas containing up to 400k tables, so most of you will never hit size limitations. No matter how many columns your tables have, you'll get reliable suggestions as you type.

The search experience is now much more powerful too. Partial matches work out of the box, so you don't need to remember exact table or column names. Suggestions work seamlessly with qualified entities and ignore backticks and double-quotes - one less thing to worry about as you write.

The autocomplete is now smarter in more ways: it uses table aliases and CTEs; provides context-aware keyword completions; and (our personal favorite!) suggests relevant columns when you're writing join conditions.

There are plenty more improvements waiting for you to discover. Happy querying - we hope these updates help your SQL flow more naturally than ever!

Deepnote AI: custom instructions

Now you can personalize how Deepnote AI works with you! Add your preferences once, and our AI will remember them for all future interactions - whether it's your favorite visualization library or preferred coding style.

Setting up your preferences is simple: click your avatar in the upper right corner, head to Account settings, and find the "Deepnote AI: custom instructions" text box. Whatever you write here becomes part of the AI's context for all your future tasks. And don't worry - these instructions are just for you; they won't affect your teammates' experience.

Go ahead and make Deepnote AI your own - time to tailor it to your style!

Code folding

Keep your longer code blocks tidy with our newest feature: code folding! If you're working with Python code blocks longer than 20 lines, you can now collapse functions, loops, and other code structures to focus on what matters most.

Using it is simple - just hover over the line number next to any collapsible structure and click to collapse or expand it. Want to fold everything at once? Use opt+shift+[ and opt+shift+] on Mac, or alt+shift+[ and alt+shift+] on Windows. Perfect for making your notebooks more organized, especially when working with longer functions.

App builder upgrade, faster notebook loading & table improvements

New app builder with preview

Say hello to a whole new way of creating apps - now you can see your app take shape in real-time as you build it! See it in action in this quick video:

No more switching between tabs and refreshing to check your changes. Our new split-screen preview puts your notebook and app side-by-side, so you can watch your app come to life as you work. Building your app is as simple as clicking the display button on any block you want to include (or select multiple blocks at once). Use the dropdown next to the display button to show or hide code and outputs - your changes appear instantly in the preview.

We've also refreshed how you start building apps. Instead of automatically including all notebook blocks, your new apps now start empty so you can thoughtfully choose which elements to include. Don't worry though - if you want to add all your non-code blocks at once, there's a handy shortcut button for that!

Looking for settings or performance data? Find them easily with dedicated Settings and Analytics buttons.

Give the new app builder a spin and let us know how it works for you. And stay tuned - we're adding exciting new features soon to give you even more control over app layouts!

Faster notebook loading

Your notebooks just got a serious speed boost! We've rolled out performance improvements that make opening and switching between notebooks significantly faster - up to 7x faster, in fact. You'll notice the biggest difference with longer notebooks, so go ahead and load up those data-heavy analyses without breaking a sweat.

We hope you enjoy speedier navigation - and keep an eye out, as this is just the first of several performance improvements we're working on to make your workspace more responsive.

Data table upgrade

Your data tables just got more customizable! While we've been busy under the hood improving performance (notice how smooth column resizing feels?), we're excited to share some new features that give you more control over how your data is presented.

Now you can easily hide columns you don't want to display - just click on the column header and select Hide column. Need to check what's hidden? Click the eye icon next to the column header to view and manage all hidden columns. We've also added the ability to rename columns right from the header, and you can now reorder columns by simply dragging them in the header. All these new options make it easier to make your tables more readable for your audience.

Don't worry - these visual tweaks only affect how your data is displayed, not the underlying data itself. Perfect when you need to present your analysis in a cleaner, more accessible way!

Dark mode & improved navigation

Dark mode

You asked (a lot!), and we delivered: Dark mode has officially landed in Deepnote.

One of our most requested features is finally here, ready to make those late-night coding sessions a whole lot easier on the eyes. Switch it on in Preferences -> Theme -> Interface theme, and join the dark side – we have better contrast and your eyes will thank you!

Improved workspace navigation

We're making it easier to find your way around your workspace, projects and apps:

• Say goodbye to accidentally hitting the workspace switcher while you're working in a notebook - our new, distinct workspace menu gives you clear, quick access to all your projects and apps
• We've brought the same workspace navigation to your apps, so you'll feel right at home jumping between content
• Looking for your personal settings like dark mode? You'll find them in our new Profile menu in the upper right corner.

Run button: advanced app workflows

Run button for conditional apps

Ever wished your app could be smarter about when it runs calculations? Today, we're introducing a powerful upgrade to our Run button that puts you in control of your app's execution flow. Now you can trigger specific actions exactly when you want them, making your apps more dynamic and responsive than ever.

Here's how it works: each Run button can now control a Boolean variable that flips to True when clicked. By using this variable in your Python code, you can create sophisticated conditional workflows that execute exactly when you want them to. You can get a quick overview in this short video or read more details in our docs.

As an example, you can now add multiple Run buttons to execute different parts of your notebook. In the demo below, the ‘Sales trends’ button produces a timeseries chart, while the ‘Reviews’ button displays correlation data.

When building apps with multiple input blocks that depend on each other, you can use the Run button to create more controlled workflows. The button allows you to delay the execution of downstream code until you have all the necessary input values collected from the user.

Check out our example app to see how Run buttons can power dynamic, multi-step dashboards. Let us know how this feature works for you - we can't wait to see what you'll create with this new capability!