3/ Deploying our frontend App with Lakehouse Applications

Mosaic AI Agent Evaluation review app is used for collecting stakeholder feedback during your development process.

You still need to deploy your own front end application!

Let's leverage Databricks Lakehouse Applications to build and deploy our first, simple chatbot frontend app.

Add your application configuration

Lakehouse apps let you work with any python framework. For our small demo, we will create a small configuration file containing the model serving endpoint name used for our demo and save it in the `chatbot_app/app.yaml` file.

Deploying our application

Our application is made of 2 files under the `chatbot_app` folder:
- `main.py` containing our python code
- `app.yaml` containing our configuration

All we now have to do is call the API to create a new app and deploy using the `chatbot_app` path:

Lakehouse apps come with an auto-provisioned Service Principal. Let's grant this Service Principal access to our model endpoint before deploying...

Once the app is created, we can (re)deploy the code as following:

Your Lakehouse app is ready and deployed!

Open the UI to start requesting your chatbot.

As improvement, we could improve our chatbot UI to provide feedback and send it to Mosaic AI Quality Labs, so that bad answers can be reviewed and improved.

Conclusion

We saw how Databricks provides an end to end platform:
- Building and deploying an endpoint
- Buit-in solution to review, analyze and improve our chatbot
- Deploy front-end genAI application with lakehouse apps!

Next: ready to take it to a next level?

Open the [03-advanced-app/01-PDF-Advanced-Data-Preparation]($../03-advanced-app/01-PDF-Advanced-Data-Preparation) notebook series to learn more about unstructured data, advanced chain, model evaluation and monitoring.