Skip to main content

Requirements

Workspace​

  • Unity Catalog.
  • Databricks Apps with user authorization, so the app can act with the signed-in user's permissions.
  • A SQL warehouse for the app's queries. Each user can choose another one under Settings.
  • Serverless jobs. Every report runs as a serverless notebook job.
  • Foundation Model API chat endpoints for the assistant. The app uses the first available Claude Sonnet or GPT endpoint; each user can choose another chat model under Settings.
  • Access from serverless compute to GitHub, where report jobs download the pinned Impulse wheel. Without it, see Serverless without GitHub access.

Data​

A silver schema in the Impulse layout: container_metrics, channel_metrics and channels, optionally container_tags and channel_tags, as described in Silver Layer Schema. The Impulse demo data works unchanged. Other layouts work through the app's data model.

Permissions​

Users need:

  • USE CATALOG and USE SCHEMA on the silver and gold catalogs and schemas, and SELECT on the silver tables.
  • CREATE TABLE and MODIFY on the gold schema.
  • CAN_USE on the app's SQL warehouse, or on the one they choose under Settings, and CAN_QUERY on the chat model endpoint.

Report jobs run as the app's service principal. It needs USE CATALOG, USE SCHEMA and SELECT on the silver data, and additionally MODIFY and CREATE TABLE on the gold schema; see Grant the service principal access.

Tools for the installation​

  • Git.
  • The Databricks CLI, authenticated against the workspace. The installation was tested with v1.21.0.
  • Node.js 22.12 or later, to build the frontend.

Python isn't needed for the installation: the Databricks Apps runtime installs the app's Python dependencies.