Scale vector search in the cloud

Scale applications powered by large language models (LLMs) in the cloud. Create vector embeddings using any machine learning model (Hugging Face, OpenAI, Cohere, etc.) and index them in Rockset for fast similarity search.

Get Started FreeExplore the tutorial ->

Scale AI applications

Scale personalization, semantic search and chatbots in the cloud. Indexing and updating vector embeddings happen in full isolation of vector search queries for predictable performance at scale. Scale up and down on demand with an architecture that separates compute-storage and compute-compute.
Compute-compute diagram

Get fast, efficient results

Search across vector embeddings using metadata filtering and ANN algorithms. Rockset selects the most cost-effective indexing strategy to deliver millisecond-latency results, saving on compute and memory resources. Rockset indexes minimize the amount of data stored in-memory, enabling efficient search across billions of vectors.
Indexed database

Embrace flexibility and simplicity

Query across nested objects, vector embeddings, geospatial and time-series data using SQL. Rockset has a flexible data model and supports in-place updates, so you can make quick, efficient changes to your vector embeddings. Its fully-managed service simplifies application development and frees teams from indexing, servers, clusters and nodes.
SQL Console

How vector search works

Vector search diagramVector search diagram

Core features

Vector database

Run exact and approximate nearest neighbor searches (ANN) using Rockset for semantic search, personalization and chatbots. 

Real-time updates

Rockset avoids full re-indexing when there are changes to your data, saving on compute and operational costs.

Metadata filtering

Rockset supports filtering of metadata using its cost-based optimizer. Search across vector embeddings, nested objects and geo and time series data for robust filtering capabilities.

Fully-managed

Rockset frees teams from managing indexes, servers, clusters and shards so they can focus on building vector search applications. Rockset is secure and reliable with SOC 2 Type II, CCPA, GDPR and HIPAA compliance and is trusted by thousands of enterprises. 

Converged Index™

Rockset's vector search is fully integrated into its Converged Index with a search index, ANN index, columnar store and row store. As a result, Rockset chooses the most efficient approach to vector search without any manual index management.

Compute-compute separation

Rockset separates the compute for ingestion and indexing vector embeddings from the compute used for vector search queries for performance at scale.

Large Language Models (LLMs)

Create vector embeddings using machine learning models of your choice (OpenAI, Hugging Face, Cohere and more) and store and index them in Rockset. Rockset integrates with LangChain for developing applications powered by language models.

SQL APIs & SDKs

Build vector search applications using the Rockset consoles or SDKs- Python, Node.js, and Go. Run search, aggregation and joins using SQL and turn a parameterized query into a REST endpoint.