Skip to content

Building a user-facing interface for an agent, such as an app, Slack, or Teams

For a Generative AI agent to be usable, it needs a front-end that end users can interact with, and Databricks supports multiple deployment paths for exposing an agent beyond notebooks. The choice of interface depends on where the users already work (a custom web UI, a chat platform like Slack/Teams, or an embedded Databricks-native experience) and how the backend agent is exposed (typically via Model Serving endpoints). Regardless of the interface chosen, the underlying agent logic and serving endpoint remain the same, so only the presentation layer changes.

1 · Learn the must-know

  • Databricks Apps allow you to build and host lightweight, interactive web UIs (e.g., using frameworks like Streamlit, Gradio, or Dash) directly within the Databricks workspace, calling a Model Serving endpoint on the backend.
  • Agents deployed to Model Serving expose a REST API endpoint, which is the common integration point for any external interface, including Slack bots, Microsoft Teams apps, or custom web/mobile clients.
  • Integrating with Slack or Teams typically requires building a middleware/bot layer (e.g., using each platform's SDK/webhook framework) that receives user messages, calls the Model Serving endpoint, and returns the formatted response.
  • A key gotcha is authentication: interfaces calling a Model Serving endpoint need a valid Databricks personal access token or service principal token, and this must be managed securely (e.g., via secrets) rather than hardcoded in the app.
  • Choice of interface should match the usage scenario: internal experimentation or demos often use Databricks Apps or notebooks, while production business workflows favor embedding into existing collaboration tools like Slack/Teams where users already operate.
  • Streaming responses (token-by-token) from the serving endpoint can improve perceived responsiveness in chat-style interfaces, but the interface must be built to handle streaming output if the endpoint supports it.

2 · Check your understanding

Check this objectiveFree · always available

A Generative AI Engineer is building an internal HR policy assistant that employees will access through a Slack channel. The Databricks App backend calls a Mosaic AI Agent Framework serving endpoint and needs both a Slack bot token and a Databricks personal access token to operate. The engineer wants these credentials inaccessible to anyone browsing the app's source code repository. How should the engineer store these credentials?

Your objective map0 tried · 0 answered correctly · 56 untouched

What you have tried across Databricks GenAI Engineer's objectives, not a readiness score.

Design Applications10.71% of the exam*0 of 6 tried
Data Preparation14.29% of the exam*0 of 8 tried
Application Development23.21% of the exam*0 of 13 tried
Assembling and Deploying Applications26.79% of the exam*0 of 15 tried
Governance7.14% of the exam*0 of 4 tried
Evaluation and Monitoring17.86% of the exam*0 of 10 tried

* Our estimate. Databricks publishes no section weights.

3 · Keep going