Knowledge & Data · Software component
Vector Index Builder
Software componentKnowledge & DataKnowledge & Dataarc:VectorIndexBuilder
A load-stage component that builds the configured approximate-nearest-neighbour index over a vector collection with the similarity metric matching the embedding model, then loads it into query-node memory.
Responsibility. Builds and loads the similarity-search index for a vector collection.
Also known as: Vector indexing stage
Relationships
is configured by structural
- Vector Index Build Configuration abstract Ch6.3B
writes dependency
- Vector Index Store abstract Ch6.3B
is orchestrated by control
Design guidance
- MUST use the similarity metric (cosine, L2 or inner product) that matches the embedding model's training objective.
- SHOULD load the built index into memory and block until resident before admitting queries.
Quantitative guidance
As stated by the sources; verify before use.
- Wrong similarity metric can degrade retrieval quality by 40% even with good embeddings (Ch6.3B).
- Without indexing, queries over a million chunks can time out after 30 seconds instead of returning in milliseconds (Ch6.3B).
Classification
- Patterns
- Index then load into memory before serving
- Risks mitigated
- Linear-scan query timeoutsRetrieval degradation from similarity-metric mismatchCollection-not-loaded query failures
Sources
- Ch6.3B: T. Nguyen, "ETL Worked Example - Load Phase & Pipeline Integration," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 6.3B. ISBN: 9798244538229.