Databricks Generative AI Engineer Associate Practice Question 3: Delta Sync Change Data Feed
Databricks Certified Generative AI Engineer Associate · Test 1 · question 3 of 10
An engineer writes chunked documents to a Delta table in Unity Catalog. They then try to create a Delta Sync index with Databricks-managed embeddings on a standard Databricks AI Search (formerly Vector Search) endpoint, so that the index updates automatically as rows are added. Index creation fails with a message about the source table.
What should the engineer do so that the index can be created and stay in sync automatically?
Show answer and explanation
Answer:
A Delta Sync index tracks changes to its source Delta table, and for standard endpoints the source table must have Change Data Feed enabled (for example, ALTER TABLE ... SET TBLPROPERTIES (delta.enableChangeDataFeed = true)). Once it is enabled, the index can be created and later syncs pick up only changed rows.
Option B avoids the requirement, but a Direct Vector Access index must be updated manually through the REST API or SDK whenever the data changes. That breaks the requirement for automatic updates.
Option C only changes who computes the embeddings. A self-managed embeddings index is still a Delta Sync index on a Delta table, so the Change Data Feed requirement still applies. It also adds work the engineer does not need.
Option D is wrong. Syncing reads from the table, and the failure is about a missing table property, not a missing write privilege. Granting MODIFY would also give more access than the job needs.
Reference: Databricks AI Search