We are excited to announce that creating vector search indexes on views is now Generally Available.
Creating indexes on compatible views enables you to optimize your data for search without disrupting other consumers or duplicating data. Currently, Atlas Vector Search supports views only with the following stages:
- $match with the $expr operator for partially index a collection (e.g. only documents which contain vectors)
- $addFields to derive new fields to search on (e.g. splicing vectors to test different dimensionalities)
- $set to remodel existing data into compatible types (e.g. Decimal128 to Double for filtering)
You can read more on the announcement blog and the docs.