Professional Documents
Culture Documents
Vss Cheat Sheet
Vss Cheat Sheet
Vss Cheat Sheet
Users expect search functionality in every application and website they encounter. Yet more than 80% of business
data is unstructured, stored as text, images, audio, video, or other formats. AI-powered search capabilities are built
into Redis Enterprise to enable Vector Similarity Search.
This is a very
This is a happy dog Vector Similarity Search happy person
That is a
happy person
Today is a sunny day
Embeddings Redis Store Relevant Results
3
1 Transform raw data into embeddings.
Redis supports two vector
Embeddings are numeric representations 2 Store in Redis indexing methods. 4
Type FLOATT32
of unstructured data that capture semantic Both have
information, typically created by an AI model. FLAT
the same
A brute force approach that mandatory DIM 128
Image, text, video or even an audio. searches through all possible parameters.
Where each can be converted into their vectors. Distance
respective embeddings with the help of Cosine
metric
Huggingface model Hub. HNSW
An approximate search that Indexes only need to be created once.
To learn more about embeddings 5
yields faster results with They automatically re-index as new
Huggingface model Hub hashes are stored in Redis.
lower accuracy.
Redis exposes search functionality, combining full text, tag and numeric Achieve unmatched performance and scale Use cases
pre-filters with K Nearest Neighbors (KNN) vector search.
Document
6 7 8 9 12x more retrieval
queries 13x lower
Pre-filters KNN Retrieved Question
per second latency
results answering
(QPS)
Text Numeric
Recommendation
Tag 99.999% systems
L2 Cosine IP
available
Euclidean Cosine Internal Visual search
distance similarity product And much more
User, profile or
contextual parameters
Text, tag,
KNN
Product images numeric
search
filters
Descriptions Recommended
products
Embeddings
OpenAI embeddings
Find relevant documents Answer question
Contstuct prompt