Get Data Into Databricks - Generative AI Databricks Apps

Summary

• Rapid app development: Databricks Apps provides built-in infrastructure and data security, enabling developers to focus on building generative AI applications.
• Agent framework integration: Use Mosaic AI agent framework within Databricks Apps to build intelligent agents with data governed by Unity Catalog.
• Multiple framework support: Build apps using popular Python frameworks like Dash, Flask, Streamlit, or custom code with pre-built templates.

Get Data Into Databricks - Generative AI Databricks Apps

Try Databricks today: https://dbricks.co/3EAWLK6.
Description: Introducing Databricks Apps: A fast and secure solution for building and deploying web applications directly within Databricks. Enjoy effortless app development with built-in infrastructure and robust data security. Seamlessly integrate with the Databricks Data and AI platform, including Databricks SQL, Model Serving and Unity Catalog. Focus on creating cutting-edge Generative AI applications with ease.

Chapters

• 0:00 Introduction to apps
• 0:43 Setting up the agent
• 1:35 Creating a new app
• 2:52 Managing app infrastructure
• 3:59 Using the physical agent
• 5:10 Redeploying with changes
• 5:57 Conclusion

FAQs

What is a Databricks App?

A Databricks App is a web application built and deployed directly within Databricks using popular Python frameworks like Dash, Flask, Streamlit, or custom code.

What agent framework is used in Databricks Apps?

Databricks Apps uses the Mosaic AI agent framework, allowing developers to build intelligent agents with access to governed data and resources managed by Unity Catalog.

How long does it take to create and deploy a Databricks App?

Initial app creation takes about 5 minutes to initialize the container, install dependencies, and create a service principal. Subsequent redeployments take only seconds.

What Python frameworks are supported for building Databricks Apps?

Databricks Apps supports Dash, Flask, Radiology, Streamlit, and custom source code with pre-built templates for chatbots and data applications.

Full transcript

0:00 Introduction to apps

Hi there uh I'm a Solutions architect at datab bricks and here I'm really excited to introduce datab bricks apps it's a fast and secure way to build and deploy web applications directly in datab bricks and it provides effortless app development experience with built-in infrastructure and data security it also provides seamless integration with datab bricks ecosystem that includ includes services like data brick SQL model serving and unity catalog so what does this all mean for you it really lets you focus on building gen applications with ease to show you how simple it is let's jump into a workspace and go through how I built a agent system in datab BRS app so here's the scenario I want to build a physical therapy coaching agent that has access to Patient profiles their physical history and also to give the agent access to tools I can call external apis and custom python functions all these are set up using Mosaic AI agent framework and all the underlying data and tools are governed by unity catalog once the agent is created then I'll deploy it to a model serving endpoint and then I can invoke that endpoint with the in an app to create a new app we'll go to the compute page and then you'll find that there is an apps tab within it there is an option to create an app look into that so here we have the option of building an app from a template using popular python Frameworks such as Dash flask radio shiny and streamlet and there some pre-built templates like chatbot or data apps or you could build a custom app using custom source codes and packages and that's the option we go with here so once you click into this then you could Define the app name and you'll also see that there are options to Define app resources so this is where you could Grant the app's service principle permissions to various uh services like jobs model serving endpoints and SQL warehouses so after you define name then you create an app so I won't do that here but because I already have one built but that's how you would create an app in the beginning when you first create the app datab braks will initialize a container install dependencies create a service principle and run your app code this process takes about 5 minutes and once the container is up and running subsequent redeployment of the app will only take seconds so from here I'll jump to an app I've already created and show you what that looks like so here you can see the live app URL the deployment source and the deployment history and you could also see the app resources available to the service principal and you could see the logs here so you could understand what's going on behind the scenes and if you want to share this app with other users then you simply change the permission here and you enter the group with the user that you want to Grant it to so let's click into the app and chat with the agents so here in this pretty basic streamlet interface we have the agents running in the model survey endpoint so I can ask questions uh like maybe what is my next appointment today as a physical physical therapist and the agent will invoke uh relevant functions and retrieve the right information in this uh instance it's going to find the date and then the user ID the appointment and I could follow up with additional questions or instructions that give me the full user profile associated with that user ID and now it's going to call another function and to generate uh or to retrieve the user history physical history and personality and preferences and it'll even print out a nice summary of what the patient is and some of the key uh profile informations so after trying this out for a bit maybe I want to make some minor changes to the code and redeploy so this is where I can go to the source code and go to the chatbot app and maybe I'll want to add balloons every time um a chat message was sent so then I made that change I could redeploy it here go ahead and click deploy okay so now it's going through redeployment and after about 7 seconds it's done so if we go back to the app refresh now you see that balloons pop up every time so what I showed here is just the tip of the Iceberg uh with what is possible with datab brakes apps you're not limited to just python Frameworks you could also use JavaScript go or other languages you prefer so you could really build much more complicated sophisticated applications that's tailored to your p uh company's Brands and use cases so I think it's a super cool product and I hope you find it interesting as well you can learn more about it in data bricks documentations so please give it a try and I'm really excited to see what you could build thank you

Learn more about the Databricks Data and AI platform.