Bringing real-time to innovative teams

NVIDIA-VGoldman SachsLG-new BrandAdobeSiemensHuluCignaSTCDell TechnologiesSamsungKrogerComcast6senseAkamaiAnt MoneyArculesBank of KoreaCARFAXCiscoCometlyDBSDentsuErnst & Young EYFactors.AIFathom AnalyticsFirstEnergyGameFam-Silicon DigitalGEHeap | by ContentsquareIMAXimpact.comLiveRampOpenDialogOutreachPalo Alto NetworksPicPayRBC-new BrandRecartSegMetricsSiriusXMSiteimprove-new BrandSK TelecomSkaiSonySunRunTATAZebra-AntuitZoomInfo
Vector search on SingleStore

unparalleled-ingest-and-search-performanceUnparalleled ingest and search performance

Support your real-time search, recommendation, or AI agent workflows—vector embeddings inserted to SingleStore become immediately searchable, powered by out native vector support.

SingleStore combines competitive vector search performance with fast SQL analytics, joins and aggregations across petabytes of structured and semi-structured data to power intelligent applications.

powerful-hybrid-search-for-your-analytical-and-ai-applicationsPowerful hybrid search for your analytical and AI applications

Bring powerful search to your application with SingleStore's native vector and full-text search.

Combine the benefits of fine-grained control provided by full-text search with semantic vector search in one easy-to-use SQL interface.

1

SET @query_vec =

2

('[0.44, 0.554, 0.34, 0.62]') :> VECTOR(4);

3

4

SELECT

5

    SUBSTRING(comment, 0, 20) AS comment,

6

    FORMAT(

7

@query_vec <*> comment_embedding,

8

4

9

) AS v_score,

10

    FORMAT(

11

MATCH(comment) AGAINST ("restaurant"),

12

4

13

) AS ft_score,

14

    FORMAT(

15

(v_score + ft_score) / 2,

16

4

17

) AS total

18

FROM comments

19

WHERE category = "Food"

20

ORDER BY total DESC

21

LIMIT 10;
+--------------------+---------+----------+--------+
| comment | v_score | ft_score | total |
+--------------------+---------+----------+--------+
| The B24 restaurant | 0.8993 | 0.1151 | 0.5072 |
| The cafeteria in b | 0.9810 | 0.0000 | 0.4905 |
+--------------------+---------+----------+--------+
Vector search

Vector search

native-vector-search-supportNative vector search support

Vector search typically involves a three-way trade-off between accuracy, performance and cost. With SingleStore, you get what you need, when you need it.

k-Nearest Neighbor (kNN) algorithms

Approximate Nearest Neighbor (ANN) algorithms

Dot product and Euclidean distance metrics

IVF-PQ for low index build time and small index size

HNSW for excellent recall and high dimensionality

Full-text search

SingleStore's full-text search is a robust search solution that leverages the Lucene library to deliver high-quality, fast search results across your data.

Native support for inverted indexes

Support for fuzzy search, wildcard search, proximity search, range search, and regex search

SingleStore full-text search is based on the active, open-source Lucene

Integrated full-text search directly with analytical queries, allowing you to combine search operations with aggregations and other SQL operations

Full-text search

fast-enterprise-grade-sql-databaseFast, enterprise-grade SQL database

Unparalleled speed

Fast ingest for vector embeddings and new vectors makes data immediately searchable.

Familiar, powerful SQL interface

Execute searches in SQL with support for filters, aggregates + joins. No need to learn a new language.

Scalability, performance and reliability

Handle complex text and vector search queries as your data volume grows — without significant degradation in performance.

explore-some-sample-codeExplore some sample code

View some of our Jupyter Notebooks showing how to hybrid search works on SingleStore