A team is building semantic search over a 5-million-document knowledge base using a vector index in Azure AI Search connected to a Foundry-hosted model. Storage and query costs for the index have grown far larger than expected, and the team wants to reduce those costs while keeping the index able to find semantically related passages, not just exact keyword matches. (Select TWO.) Which two changes address this?
Select all correct answers, then click Submit.
Short Explanation
When a vector index gets expensive at five million documents, the two levers that actually shrink the bill without giving up meaning-based search are about the vectors themselves: make each one smaller to begin with, and compress the ones you keep. Picking an embedding model that outputs fewer dimensions per document means less to store and less to compare on every search. Turning on quantization squeezes each of those stored numbers into fewer bits, shrinking the footprint further for a small, usually tolerable, accuracy cost. Neither of those gives up the ability to match on meaning rather than exact words. Swapping to a plain keyword index would cut cost too, but it throws away semantic matching entirely, which defeats the whole point of building this search in the first place. And pulling back more results per query only adds more computation and data movement, pushing the cost in the wrong direction.
Full Explanation
The correct answers are B and C. Choosing an embedding model with a smaller output dimension means every one of the 5 million document vectors takes less storage and less compute to compare at query time, directly cutting cost while the index still performs semantic similarity search. Enabling quantization compresses each stored vector's dimensions into fewer bits, shrinking the index's memory and storage footprint further with only a small, usually acceptable, loss in precision, again while preserving semantic matching. Option D is incorrect because dropping to a keyword-only full-text index abandons semantic search entirely; it would find exact term matches but miss the related-meaning passages the team explicitly wants to keep finding. Option A is incorrect because retrieving more chunks per query increases, rather than decreases, the compute and data transferred on every search, working directly against the team's cost goal.