Changelog

🔮 Hex Magic Public Beta​

We’re very excited to announce that Hex Magic – our suite of AI augmentation tools – is moving into public beta. This gives all Hex users a powerful set of AI-driven features to help you do more with data.

Magic lets you use AI to generate, edit, debug, and document code right from your Hex project, with a deep contextual understanding of your code and data.

Opt-in to Hex Magic by visiting your Hex user preferences (Settings → Preferences) and toggling on the Magic option. Users on Pro, Teams, and Enterprise plans get unlimited Magic usage, and users on Community have 100 Magic requests per month.

As a reminder, this is a beta. Magic – like all AI features – is subject to errors, hallucinations, and mistakes. Please use it with care!

Hex Magic is not yet available for workspaces on EU or HIPAA compliant stacks.

🧙 New Magical Powers​

Hex Magic gives you access to 5 Magic tools: Generate, Edit, Explain, Fix, and Explode. You’ll find the Magic menu in the top right corner of SQL, Python, Markdown, and R cells.

For tips, tricks, and a detailed list of examples of how to use Magic, check out the documentation.

🪄 Generate​

ex: show me all orders placed in the last 30 days for > average order price.

Magic Generate lets you ask for code and queries with natural language. Describe what you want to generate in plain English, being as specific as possible, and hit Enter. As Magic thinks, you'll see the generated code begin to fill in below.

Magic has context on your tables and schemas, so you really can just ask natural language questions and Magic will map to the most relevant tables!

✏️ Edit​

ex: change this query to look at unusually small orders.

Magic Edit works just like Generate, but for modifying existing code and queries. If you trigger Magic in a cell that already has code, it'll default to edit mode.

You can ask Magic Edit to do anything to your code, from answering a totally different question to performing manual tasks like renaming every table reference. As Magic thinks, you'll see edited code begin to fill in alongside the previous code, with a line-by-line diff of the changes made.

📝 Explain​

Magic explain automatically documents code – whether it’s something you wrote, or something unfamiliar you’re seeing for the first time. It works similar to a Magic edit, so you'll see the documented code with comments stream back word-by-word.

🩹 Fix​

Magic fix automatically detects and fixes errors in your code. Whenever a cell returns an error, Magic begins preparing a fix, which can be applied with the "Magic fix" button in the error output.

Here, Magic's just fixing a missing comma— but it can solve complex issues, and resolve multiple problems at once rather than step-debugging through them all.

💥 Explode​

Magic explode takes CTEs and "explodes" them into individual chained SQL cells. This can make complex queries more easy to debug, understand, and iterate on.

Explode makes use of query mode, so your exploded queries will not add any additional runtime over the original CTE.

🏷️ Bonus: Rename​

Naming things is hard easy with Magic. You can ask Magic to rename a cell based on the code within it, to keep your project neatly organized.

Other improvements​

• When you click the New Project button, you’ll be greeted with a much friendlier empty state — it now take exactly two clicks to go from the Hex homepage to a working SQL cell. For workspaces with multiple data connections, Admins can also now designate a Default data connection that (you guessed it) the first SQL cell will default to.

• You can now sign up for Hex with your personal email, and spin up a free-forever workspace for any side projects or portfolio work. Or if you’re a data consultant, you can create separate workspaces to hand off to each of your clients.

• We also made improvements to our workspace switcher so you can quickly see which email(s) you're authenticated with, and transition between different workspaces.

• When the output of a SQL query has caching enabled, paging through the results will now be significantly snappier. 🫰

• We fixed a bug that was making dataframe SQL on w i d e tables far slower than it needed to be.

• We’ve made improvements to engine that aggregates data for charts, speeding up chart runtimes along the way. (You might be sensing a theme here…)

• Finally, we also made some improvements to how we write the results of a project to our database, making projects and apps much faster to render results.

🍱 Collections​

There’s a new way to organize and curate your work in Hex: Collections!

Collections are groups of projects or components that can be created, populated, and shared by users in a workspace.

Unlike a traditional file directory, Hex projects and components can live in multiple collections, increasing flexibility and decreasing the potential for data silos.

When adding something to a collection, you have the option to also share, view, or edit permissions with all collection members, ensuring that your team can stay in sync with the latest and greatest insights.

We’ve been using this internally for a little while now, and it’s felt really great. We’d love to hear your thoughts, ideas, and feature requests for collections after trying them out!

Learn more in the docs.

⚙️ Admin panel polish​

The admin settings pages have been overhauled to make it more intuitive to find what you’re looking for. We’ve put everything important at the top level in the sidebar, so you don’t need to remember what section data connections or SSO are nested within.

We’ve also done some general UI polish to spruce up each individual settings page.

🔗 Invite links​

Admins can now generate invite links to easily add new users to Hex, individually or in bulk. You can choose the permissions an invite link grants, set the expiration date of a link, or limit it to just a one-time use.

Whether you’re using Hex to teach a class, run a workshop, or just have a lot of new folks to bring on board, this should drastically improve the manual process of adding users.

Other improvements​

• We fixed an interesting bug that caused Chart cells to aggregate incorrectly when the column on the x axis had a dot "." character in the name.
• We fixed a bug that caused scheduled runs to still process for recently deleted projects
• The openai package is now included in our base package library 🤖
• We are knee-deep in some performance improvements that hopefully we will be sharing with you soon!

🔍 All-new workspace search, including full cell search…​

It always feels harder than it should to find the right report or dashboard. You just want to find the latest MRR numbers. Is that in MRR DASH [sales]? Or “Board MRR update Dash 2”?

Syke, it’s in [official] Company KPI Dashboard, but you’ll never find that by searching “mrr” or “monthly revenue”.

So we’ve rebuilt search to look through metadata, titles, descriptions, text, and even the code in your projects. Search for a table name to find all the projects querying it, or hunt down a report by the name of one of its sections.

Results show a preview of the code or text that matched your query, and you can click into them to be taken directly to that part of the project.

…and advanced filtering​

Too many results? With advanced filtering, you can easily narrow results by creator, date (created or last edited), category, and status.

You can also choose which fields to search by, so you can get more accurate results if you know the string you’re searching for is in a project description or title.

🌹 The perfect date​

We went through every square inch of the Notebook & App view to make sure that anywhere a user inputs a date, or we render a date, the perfect yyyy-mm-dd format is supported.

If you find a spot we missed, let us know and we’ll take you out on a perfect date. Sushi, pizza, ice cream, you name it. Totally serious.

☁️ Custom SQL Formatting with SQLFluff​

You can now upload a custom SQLFluff config that Hex will use when formatting SQL cells. Set up formatting rules for a data connection, and they’ll be available workspace-wide so the entire team’s code can be styled consistently.

Upload .sqlfluff files directly to any project or workspace data connection, in that data connection’s settings. SQLFluff has a ton of rules you can customize, from capitalization to leading commas, or even finicky things like indentation and new lines.

🤏 Compact number formatting​

Number and currency formatting now supports “Compact format”, which provides a more polished view for reports. This is available everywhere number formatting is, from pivot cells to charts and table displays.

Sometimes it really is the small compact things!

Other improvements​

• If an app-only user gets a link to the Notebook view of a project, they now get a much friendlier request access/redirect page.
• We’ve added support for two new databases: AlloyDB and Presto.
• The data browser should load your schemas and tables faster!
• We fixed a bug that caused issues when duplicating a project that had images in text cells.

🍿 Custom Kernels​

Part of the magic of Hex is that anyone can start doing analysis without having to worry about pip installs, virtualenvs, or —heaven forbid— dockerfiles and homebrew commands.

But as companies scale and workloads grow increasingly unique, specific requirements arise. Before you know it, you’re back on the command line tinkering with Python environments, wondering if that pip warning is actually something to worry about or not.

Well, now you can add custom kernels to Hex, making it a one-click process to spin up the tailored environments your company needs. Just give us a Dockerfile, and we’ll do the rest.

This feature is only available for single tenant customers.

If you’re interested in custom kernels, reach out to your Hex point of contact or [email protected].

Faster apps? Simpler settings? Delicious pie? Just your average Thursday for the Hex team. We hope you enjoy these new features hot off the press!

🏎️ Rebuilt query caching engine​

Ever tried to comment out a SQL cell to avoid accidentally rerunning an expensive query? Or taken a peek into your database history and wondered why Hex was running the same query three times in a row?

We’re putting an end to these shenanigans, once and for all: We’ve completely rebuilt our query caching engine to avoid running repeated queries. Now, when you run a SQL query in Hex, we’ll check if the same query has been run recently, and if it has, default to returning results from the cache.

The result: faster projects and apps, and fewer queries sent to your database.

We’ve also made major improvements to the caching UI, so it’s easier to understand, override, and tweak. Instead of finding cache settings on individual SQL cells, you’ll find project-wide controls in the Environment sidebar.

Information for existing caching users

Our new query caching system is a complete rebuild of our old system. If you were using our old caching system, read on for an explanation of how they differ. If you were not using the old SQL caching system, you can head straight to the updated docs to find out how the new system works.

What's the TLDR of the new caching system?​

With our new caching system, any time a SQL cell is run, Hex checks if the same query has been run recently. If it has, Hex pulls results from its cache, rather than sending the query to your warehouse.

When developing logic: Results from the cache are clearly shown in on the SQL cell, with a button to rerun the query without cached results, should you need to refresh the results. There’s also a button to rerun an entire project without using cached results.

In published apps: When an app uses results from the cache, it’s indicated on the published app. An app user can choose to rerun the app without cached results to get fresher data.

By default, results from the cache will be used if the same query has been run within the last 60 minutes. Admins can adjust this default for their workspace, with separate timeouts for developing logic, and published apps.

Further, editors can choose to set their own timeouts per project.

For more information, head to the docs.

What are the main differences between this and the old caching system?​

Old caching system	New caching system
Had to be turned on per cell	On by default, with a workspace cache timeout set by admins, which can be further configured per project by editors.
Cache entries were valid until the next scheduled run, or next app load (depending on configuration).	Configurable cache timeout per project (e.g. 60 minutes) with separate settings for logic and app sessions. Additionally can use the next scheduled run as the expiry.
Cache entries could not be shared between projects.	We now check entries based on query text and data connection details, allowing us to share cache entries between projects. This is useful when projects have been duplicated, or you are using components.
Confusing UI	More intuitive UI

What happened to the old SQL caching system?​

The previous SQL caching system has been fully deprecated: our new system is more performant, and easier to understand.

Any projects that leveraged SQL caching in combination with scheduled runs have been migrated to use the equivalent settings in our new system:

My published apps that loaded fresh data now have results up to an hour old. How can I change this?​

The default cache timeout for workspaces is 60 minutes. This means that any published app that is set to run when an user opens the app may pull results from the cache that are up to 60 minutes old, whereas previously they would have always queried the warehouse.

If you have individual apps that should show fresher data, you can update the cache timeouts from the Environments panel of your project. Note: we recommend that you at least keep a timeout of a few minutes, rather than fully disabling the caching. You will need to republish the app for this change to take effect.

An admin can also adjust the workspace default to a shorter timeout, which will update any project that does not have a custom timeout configured.

How do these changes interact with the "Cache default state" settings?​

Previously, we used the word "cache" to describe two distinct Hex features:

• SQL caching: The ability to fetch results of a SQL query from a cache, rather than send the query to the warehouse.
• App caching (or Cache default state): The ability to view the output of a previous run when opening a published app. In this context, the "output" is limited to the front-end output (e.g. a chart visual, but not the data powering it)

While both of these features were built to make Hex faster, they are somewhat distinct in how they work, and the shared name made this confusing.

So, we redesigned the UI for cache default state (see below), but have not changed how this feature works.

🤔 Understandable app settings​

As we rebuilt query caching, we realized we ought to also tackle the other area where we use the word “cache”: the app run options menu.

If you were ever confused about terms like "cache default state" and "auto rerun", this UI overhaul is for you. There's no new functionality here, just much clearer language that makes it easy to get these settings right.

Hot tip: If you expect viewers to interact with your app via input parameters or visual filtering: Use “Show results from a previous run” and “Pre-run cells in background” together for a big speed boost.

You can read the docs on these settings here.

🔪 Query cancelling​

Caching is one way to save money on expensive queries, but what’s even better than only running a query once? Not running a query at all!

Now if you interrupt a SQL cell while it’s running, the actual query running on the database will also be cancelled.

The place I find myself most grateful for this? When I constantly forget to add a LIMIT 100 to a test query on a large table, and have to scramble to cancel the query before it finishes parsing terabytes of data.

🥧 Pie charts 2.0​

In January, we asked you, our users, if we should add Pie charts to Hex, and the results were striking:

Well strap in everyone, cause now it’s February, the past is over, there’s a delicious warm fragrance wafting through the kitchen office, and we just shipped some expertly crafted pie that even your bubbe will be jealous of— directly to your browser. Also, Barry has lost his twitter privileges.

But really— there are times when pie charts are genuinely good visualizations! And there are times when they aren’t… but some people will want to use them anyway. Support for pie charts felt like a missing feature in the new chart cell, so we went ahead and added them.

And if you hate pie charts? You’re the master of your own domain — just choose not to use them.

Other improvements​

• Previously, if you had a SQL cell with no query in it (maybe you just hadn’t gotten to writing your query yet), or a commented out query (maybe you’re saving it for later), Hex would unhelpfully error out, interrupting your flow. Now, the cell will return an empty dataframe instead.
• Chart cells now have built in data labels — head to the Style tab of the Chart Cell to enable them for your chart. We’ve also shipped a bunch of small improvements and bug fixes to charts.
• We fixed a bug where dragging-and-dropping an image into a text cell took a few seconds to render. It should feel snappy again now!
• We shipped some improvements to the scheduled run manager, including filter options for last run status (success, failed) and run cadence (hourly, daily, etc.), and column-level sorting
• Previously, we made editors visit the App builder before they could start adding elements to the App. That was silly of us, so we changed it so that you can start adding elements to the App at any point.
• Single Value cells now have a new comparison type: absolute change. Thanks Nate Sooter for the suggestion.