Introducing: Postgres Best Practices
We are releasing Agent Skills for Postgres Best Practices to help AI coding agents write high quality, correct Postgres code.
Cloud Infrastructure
Summary
Supabase introduces Vector Buckets, an S3-backed storage layer for durable, large-scale vector storage with built-in similarity search. This tool complements pgvector by handling tens of millions of vectors efficiently, enabling semantic search and RAG workflows via k-NN queries.
Highlights
- Vector Buckets provide S3-backed durability for storing up to tens of millions of vectors per index.
- Built-in similarity search supports cosine, euclidean, and L2 metrics for k-NN queries.
- Vectors can be queried via Supabase clients or directly from Postgres using a foreign data wrapper.
- Ideal for large datasets where pgvector might cause table bloat or performance issues.
- Supports semantic search, recommendations, and media similarity use cases with optional metadata.
auto-generated
Context
Audience
Software Developers
DomainDatabase Engineering
Formatblog post
Accessfree online
Topics
Related
Supabase StoragepgvectorAmazon S3Vector EmbeddingsSimilarity SearchRAG