Redis Search

Search is often used to get the most similar data in Redis by best semantic match or keyword match. Full-text search, geospatial search, and vector search over vector embeddings are subtypes of search.

Redis Query Engine has search built in natively with a powerful autocomplete engine that can do fuzzy matching and with multiple scoring models and sorting by values. Concurrent searches can be run on Redis Query Engine and allows long-running searches without blocking Redis.

Key full-text search capabilities for Redis Query Engine include: prefix-based searches, exact phrase search and slop-based search, field weights, auto-complete and fuzzy prefix suggestions, and stemming-based query expansion for many languages using Snowball.

It also supports custom functions for query expansion and scoring (see extensions) and document ranking.

Key geospatial search capabilities for Redis Query Engine include search for locations within a specific radius, search based on geometric shapes, such as polygons (representing building layout, lake, etc.), and search on points within a geometric shape.

Key vector search capabilities for Redis Query Engine include running vector search queries with the FT.SEARCH or FT.AGGREGATE commands, semantic searches over vector embeddings, augmented searches with filtering over text/ numerical/geospatial/tag metadata, and advanced querying strategies with vector fields including k-nearest neighbor (KNN), vector range queries, and metadata filters.

Running a search on the ‘index:product’ index to return a list of all those that mention words similar to ‘RAM & SSD’ such as ‘ram and ssd’ is straightforward.