Introducing Vector Buckets
Supabase is introducing Vector Buckets, a new storage option that provides durable and cost-efficient vector storage with built-in similarity search capabilities.
Why it matters
Vector Buckets provide a scalable and cost-effective solution for storing and querying large volumes of high-dimensional vectors, enabling a wide range of AI-powered applications.
Key Points
- 1Vector Buckets are a new bucket type in Supabase Storage that store high-dimensional vectors and optional metadata
- 2Vector Buckets provide scalable vector storage for large datasets, with capacity for tens of millions of vectors per index
- 3Vector Buckets offer built-in similarity search capabilities, allowing users to find conceptually related items based on vector representations
- 4Vector Buckets are designed to provide sub-second similarity search performance suitable for backend workflows, AI agents, and many user-facing features
Details
Vector Buckets are a new storage option from Supabase that combines the durability and cost-efficiency of Amazon S3 with built-in similarity search capabilities. This allows users to store large volumes of high-dimensional vectors, such as those used for semantic search, recommendations, and media similarity, without the need to manage a separate vector database. Vector Buckets store vectors in S3-backed object storage, providing capacity for tens of millions of vectors per index, while still allowing users to query the vectors through Postgres. The similarity search capabilities of Vector Buckets enable a variety of use cases, including semantic document search, product and content recommendations, and de-duplication of media libraries. Vector Buckets are designed to provide sub-second search performance, making them suitable for backend workflows, AI agents, and many user-facing features, though they may not be the best choice for ultra-low latency, high-throughput applications.
No comments yet
Be the first to comment