A namespace is an isolated partition within a vector database that separates one collection of vectors from another, so queries against one namespace only return results from that namespace. It is a lightweight mechanism for keeping different sets of data logically distinct inside a shared index or database.
Namespaces are the standard tool for multi-tenancy. By giving each customer their own namespace, a SaaS application keeps tenants’ data separate and prevents queries from crossing boundaries, while still sharing the underlying infrastructure. Because creating a namespace is cheap, this pattern scales to very large numbers of tenants efficiently.
Namespaces are also useful for organising different kinds of content within one application — keeping product, user, and document embeddings apart so each can be queried independently. The exact terminology and isolation guarantees vary between systems: some implement namespaces as fully separate indexes, others as partitions within a shared index enforced by filtering. Either way, the role is to provide clean logical separation without the overhead of entirely separate databases.