Consider a Retrieval-Augmented Generation (RAG) system tasked with finding product specifications using specific part numbers (SKUs). When employing a pure vector similarity approach for precise SKU searches (e.g., 'XR-4892-B'), the results are often unsatisfactory because SKUs inherently lack semantic content. Users type free-text questions that mention a part number alongside other words (e.g., 'max operating temperature for XR-4892-B'), and the application passes the raw question to the retriever without parsing out the SKU. What capability within Mosaic AI Vector Search most effectively overcomes this challenge?
Answer & explanation
Correct answer: C
WHY C is correct: Hybrid keyword-similarity search combines vector-based embedding search with traditional keyword-based BM25 search. The Databricks documentation explicitly calls out that this is 'particularly useful in RAG applications where source data has unique keywords such as SKUs or identifiers that are not well suited to pure similarity search.' Results from both methods are fused using Reciprocal Rank Fusion (RRF).
WHY NOT A: Increasing embedding dimension improves the model's ability to represent semantic concepts in dense vector space, but SKUs are arbitrary alphanumeric sequences that don't carry distributional semantic signal — higher dimensions don't help with exact-match retrieval of opaque identifiers.
WHY NOT B: Direct Vector Access Index changes how embeddings are updated (manual vs. auto-sync) but does not add keyword search capability. It does not improve exact-match retrieval of SKU queries.
WHY NOT D: A metadata filter needs the exact SKU value extracted from the query first. The application passes the raw question, so there is no filter value to apply. Hybrid search matches the SKU token directly from the query text and handles the semantic part of the question at the same time.