Definition
The architectural capability of a vector database to store, index, and query structured attributes (e.g., timestamps, document IDs, or source categories) alongside high-dimensional embeddings. While it significantly increases retrieval precision by enabling pre-filtering, it introduces trade-offs regarding storage overhead and indexing latency.
Focuses on chunk-level attributes in a vector store, not generic website SEO tags.
"Color-coded labels on individual folders in a massive archive, allowing a librarian to ignore all folders except the 'Red' ones before looking at the content inside."
- Vector Search(Prerequisite)
- Hybrid Search(Component)
- Self-Querying(Advanced Implementation)
- Pre-filtering(Execution Strategy)
Conceptual Overview
The architectural capability of a vector database to store, index, and query structured attributes (e.g., timestamps, document IDs, or source categories) alongside high-dimensional embeddings. While it significantly increases retrieval precision by enabling pre-filtering, it introduces trade-offs regarding storage overhead and indexing latency.
Disambiguation
Focuses on chunk-level attributes in a vector store, not generic website SEO tags.
Visual Analog
Color-coded labels on individual folders in a massive archive, allowing a librarian to ignore all folders except the 'Red' ones before looking at the content inside.