PRODUCTION SCENARIO
A Generative AI Engineer built a search feature on a Direct Vector Access index and computes the embeddings inside the application. Rows edited in the application's own Delta table keep returning their old text in search results.
What is the application responsible for in this setup?
Answering here is anonymous. Nothing is saved unless you sign in.
Show answer and explanation
Answer: Writing updated vectors and metadata into the index through the SDK or REST API
A Direct Vector Access index supports direct read and write of vectors and metadata, and the user is responsible for updating it through the REST API or the Python SDK, so an edit reaches search only when the application upserts it. The other three are Delta Sync responsibilities: Delta Sync is the index type that runs a sync pipeline on a schedule, reads the source table on your behalf, and calls an embedding endpoint for you. Direct Vector Access means taking that work on instead of delegating it.
Official Source: https://docs.databricks.com/aws/en/ai-search/create-ai-search ↗