You have an Azure Databricks workspace named Workspace1.

The correct answer is B. The %pip install command (or pip3 in a terminal context) creates an isolated, per-session library environment in notebooks, which is the Unity Catalog-compatible approach. Unity Catalog workspaces require cluster access mode set to ' Shared ' or ' Single User, ' and %pip installs work seamlessly within those modes without requiring cluster restarts.

Option A (custom dependency script) introduces extra maintenance work for every environment change — exactly what the question says to avoid. Option C installs libraries at the cluster level and requires a manual restart, which disrupts other users sharing the cluster and bypasses the per-notebook isolation model that Unity Catalog recommends. Option D uploads libraries to the Workspace file system (legacy DBFS approach), which is being deprecated in favour of Unity Catalog Volumes for library storage.