Vector Search

Kind: mechanism

Layer: Correctness Core

Record: architecture:vector-search

Severity: contextual

Scope: retrieval, search, RAG

Canonical: Ontology

A mechanism that finds records whose embeddings lie close to a query's embedding.

Listed in Architecture principles, after Retrieval-Augmented Generation (RAG) and before Knowledge Graphs.

Requires

• Embeddings
• Vector Index

Reinforces

• Retrieval-Augmented Generation (RAG)
• Semantic Search

Enables

• Similarity Retrieval

Conflicts with

• Exact Keyword Search Only

In tension with

• Explainability/Recall

Tensions

• Vector Search / Explainability/Recall

Violated by

• Exact Keyword Search Only

Refactored by

• Evaluation Suite

Severity

• contextual

Category

• Model Architecture

Linked from

• Systems built around a model
• Model Architecture
• Model Architecture
• Severity levels
• The resolutions