1. What is the primary role of a vector database in an AI system?
2. Why is approximate nearest neighbor (ANN) search used instead of exact search at scale?
3. What data structure does HNSW (Hierarchical Navigable Small World) use?
4. What is 'hybrid search' in vector database terminology?
5. What does 'index ef_search' (or 'ef' at query time) control in HNSW?
6. Why can cosine similarity be more appropriate than Euclidean distance for text embeddings?
7. What is a 'metadata filter' in a vector database query?