Databricks Apps 2026: Git Deployment & Lakebase Autoscaling

0:04Hey folks, today I have the pleasure to host for the second time Pascal. Hi Pascal. Glad to be back. I'm looking forward to talk about Databricks Yeah, exactly. Few months ago we recorded an intro video to help you get started quickly with Databricks app through the cookbook he and his team built. And

0:27today he's back because we have some exciting features to show you and one of them is possibility to deploy apps directly from GitHub for example. But Pascal is here to walk us through those different steps and why this is a game changer and also we have other features as we will show you during the demo. Exactly. So yeah, a

0:50very useful new feature that we have are get backed apps for Databricks apps. So that makes it super easy to deploy an app directly from a get repository. So I think most people when they develop an app, they develop it in their IDE in Casa with cloud code on their local machine and then you push the code into get repository, GitHub, GitLab and so on. And then you need to deploy it to

1:12Databricks apps somehow, right? And in the past you had to get your code somehow into the workspace file system first and then from there you could deploy it. But now with these get backed apps, I can just directly deploy from my GitHub repo for example. And that's exactly what can show you also. So let me share my screen here. So I have here in GitHub a super simple

1:36dash app that runs on or it can run on Databricks apps. And if I now want to deploy this, I can use this new get backed apps feature. So as you can see here, this is a private repository. So this will become also important later. And if I want to deploy this, I will just already copy here my URL and then I will switch into Databricks. And here I

1:59can create a new app. So, that's the same way as before. I just go to create app. Uh and then I will choose a custom app now. And I will give it a name. So, Git Baked App as for example. And now, what is new here? I have this configure Git step in the deployment process. So, here I will

2:21just put my uh URL for my Git repository. Uh but we automatically select kind of the right uh Git provider. So, this supports all of the Git providers that we were already supporting for Git folders. Exactly. And now I can create my app. And while the app compute starts, we can now take care of uh setting up the

2:44credentials, right? Because, as you remember, the repository was private. So, we need to somehow give our app uh permissions to also um get the code from our private Git repo. Um this is super simple. So, it's different between the different Git providers. For GitLab, uh sorry, for GitHub, I can just here link my account. So, I can just click here on link. I can

3:06select my user. And then uh this is basically all I have to do. So, it's now authenticated. I can see here this Git credential is now connected um or added to my app. So, my app now also has permission to get the app code uh from GitHub. There it is. Yeah, now that the app deployment is completed, I can run

3:28app. So, I will go here to my app URL. And then we will see my fantastic uh two-minute bike coordinates uh dash sales dashboard here uh in action. And that's how easy it is to do a deployment directly from GitHub. I can also, of course, do this via the Databricks API or CLI, right? So, if I have a uh GitHub action, for example, uh

3:51I could trigger a redeployment um also from there. So this makes it really easy to automate also deployments directly from Git. And yeah, that's how easy it is now to deploy apps. Yeah, and during the last couple months, Databricks announced so many new features and enhancements. What would be your top three announcements?

4:13Oh yeah, there's so much happened around apps. It's really hard to pick like top three. I think what comes to mind first is lake base integration. So we already had integration with the provisioned lake lake base. But now Databricks apps also supports lake base auto scaling, which is recently added as an app resource. So if I add a resource here to

4:35my app, I can also now find under the databases the auto scaling databases. I can add those. And that makes it really easy to experiment with a Postgres database with an app. And they're also super fast and super cost-efficient. Yeah, so that's a big new. And specifically like reverse ETL use cases where for example where you

4:58have KPIs that you're computing maybe on real time. Or maybe it's it's for example sales data, real-time sales data. You can check the app and see the like number of items you're selling being updated in the real time with the with the very low latency and a very efficient way. And as you mentioned, it's very cost-efficient. Why? Because

5:22first of all, it's like the app is on with in less than 100 milliseconds. And you have all the features that you used to have in Neon, which is the let's say the acquisition that Databricks, the company that Databricks acquired to have this lake base auto scaling feature. And you have other things like

5:45branching with zero copy. We have those like permissions, time frameworks, this easy way to sync data like Delta tables to lakehouse. It's really we have the platform has now complete ecosystem from the ingestion till the serving of the KPIs through

6:07the apps like this. Exactly. Lakehouse auto scaling really is a super easy way to add a fast auto key database to your Databricks app. So I use it basically with all of my new apps now. That's one big change that we made. And then another great feature is that we now support more extensive logging

6:29observability capabilities for apps. So we always had for Databricks apps here these real-time logs which show me what's happening inside my app compute, right? But when I stop my app or if I delete it, then these logs they also go away, right? And I sometimes need like an audit trail or I need to search across logs of multiple

6:52apps. And for that I can now enable automatic real logging for my Databricks app. And that way all of the logs of my app and also metrics and also other hotel elements can be logged into a Unity Catalog table. And then I can really just write a SQL query to analyze my logs. I can place alerts on my logs. I have a never-ending

7:14audit trail if I want that. So that's a very powerful new feature that we also have for Databricks apps. And maybe number three feature Exactly. Number three, I think that's the most ubiquitous feature that everyone uses now, right? It's the byte coding. I think apps are very

7:36well suited to be byte coded, right? Because you have direct like visual feedback. And what we have there is what we tools that we can actually recommend. So very popular tool to use is the AI DevKit, right? So this is here available on GitHub. This is a set of skills that I can set up for my code code or my class arc. And this has skills for all kinds

7:59of features of Databricks including also for Databricks apps. So I can set this up in my code code. I can tackle code, build an app that is the UI for I don't know a training a machine learning model where I uh label some images or something like that. And AI DevKit will be able to build the app but also in the background machine learning model on Databricks.

8:22So it will be able to help me build like an end-to-end AI training AI solution on Databricks. So that's super useful. And I know you are also are a connoisseur of AI DevKits. Are using it extensively probably. So I think that's super useful if you have tried it once. And I think you already spoiled the next announcement that's going to be AppKit that we can

8:44see on your browser. Exactly. So we are also thinking about like specifically just focus on Databricks apps. What is the easiest way to write code these Databricks apps to build them in an AI assistant way. So we now also have our own toolkit that which is AppKit that is built by our Databricks engineering team around Databricks apps.

9:07Which really has specific features and plugins that make it super easy and efficient and safe from errors basically to write code Databricks apps. And this is also something to keep an eye on over the next months because we are investing a lot of resources here. It's going to become much more powerful. And yes, easy to try out. We have all documentation here where you can find

9:30all of the available integrations of Databricks services. And yeah, this is also what I use from time to time to build new database apps. And what I can recommend as a web coding tool and we can also put a link to that in the description of the video. Yeah, I mean, this this is amazing like if you if you come back maybe in the past maybe 2 years ago, we had only

9:55Python Python apps like Python frameworks to build apps. Then we added Node.js apps. Now we're adding the more features. Extended all the let's say features we have on Databricks to the apps. I'll have a bunch of additional things that's going to come be our be announced by the Data and AI Summit. So,

10:17stay tuned. Make sure to register for the data for the online version of the Data and AI Summit if you can't make it to San Francisco. And we're more likely going to be interviewing Pascal again for the third time right after summit so we can review and show you how you can get started with all the latest features

10:39that's going to be announced. Thank you Pascal so much. >> Thanks for having me. Looking forward to be back soon. Yeah, have fun building Databricks apps. Thanks, buddy. Bye. Bye-bye.