infrastructure//Cloudflare//Vectorize

Vectorize is Cloudflare's edge-native vector database: it stores embeddings and performs cosine similarity search at the edge. It is a separate product from D1, because D1 is SQLite and has no vector extensions natively. pgvector takes the other approach to the same problem: it adds vectors to an existing SQL database (one system, one query), where Vectorize is a standalone store with its own query.

Vectorize is Cloudflare's edge-native vector database: it stores embeddings and performs cosine similarity search at the edge. It is a separate product from D1, because D1 is SQLite and has no vector extensions natively. pgvector takes the other approach to the same problem: it adds vectors to an existing SQL database (one system, one query), where Vectorize is a standalone store with its own query.

An index holds up to 5M vectors, and an account up to 50K namespaces.

It integrates with Workers and Workers AI, so the whole chain runs on Cloudflare:

1Inference2Embed3Store4Search

It has no hybrid search (BM25 and vectors in one query). That takes D1 (FTS5 full-text search) and Vectorize together, two products for what Weaviate does in one.