An entry in a Cosmos DB indexing policy, of type flat, quantizedFlat or diskANN, that makes VectorDistance searches faster. It can't be changed once created and needs the vector search feature turned on.
Read more: Microsoft Learn
In the Ultra Transcenders books
Each book explains Vector index in context, with comparison tables and the common traps.
Terms in this definition
- Cosmos DB
NoSQL database distributed globally, offering writes in multiple regions, automatic indexing and latency below 10 ms.
- Indexing policy
A Cosmos DB container setting that decides which paths get indexed and by which index types. Every property is indexed by default, so leaving out paths that are never queried cuts the RU cost of writes.
- FLAT
A RediSearch vector index that compares every vector exactly, by brute force. It fits smaller data sets or cases needing complete results, whereas HNSW is approximate.
- DiskANN
A vector index from Microsoft Research for approximate nearest-neighbour search. Azure Cosmos DB offers it as diskANN (vectors of up to 4,096 dimensions) and Azure Database for PostgreSQL as pg_diskann, and it works best once a collection holds more than roughly 50,000 vectors.
- Vector search
Searching embeddings for similar meaning, so paraphrases are found via nearest-neighbour algorithms like HNSW; exact identifiers, however, may be missed.
Related terms
- QPS
How many queries a system handles each second, a figure used to compare different kinds of vector index.
- Semantic index
An automatically built vector index over Microsoft Graph data that helps Copilot match information by meaning and related words instead of exact keywords, always within what each person may see. It can't be switched off, though sites hidden from search are left out of it too.