Multitenancy and namespaces

Vector Databases for Embeddings with Pinecone

James Chapman

Curriculum Manager, DataCamp

Multitenancy

Serve multiple tenants in isolation
Separate different customers' data
- Security and privacy
Reduce query latency

Multitenancy strategies

Namespaces
- Advantages: Reduces the need for additional indexes
- Disadvantages: Tenants share resources, complex data
Metadata Filtering
- Advantages: Allows querying across multiple tenants
- Disadvantages: Shared resources, challenging cost tracking
Separate Indexes
- Advantages: Physically separates tenants, allocates individual resources
- Disadvantages: Requires more effort and cost

Namespaces

Created implicitly during upsertion if they don't exist

index.upsert(
  vectors=vector_set1, namespace="namespace1"
)

index.upsert(
  vectors=vector_set2, namespace="namespace2"
)

Inspecting namespaces

index.describe_index_stats()

{'dimension': 1536,
 'index_fullness': 0.0,
 'namespaces': {'namespace1': {'vector_count': 5},
                'namespace2': {'vector_count': 5}},
 'total_vector_count': 10}

Querying vectors from namespaces

query_result = index.query(
    vector=vector,

    namespace='namespace1',

    top_k=3
)

Deleting vectors from namespaces

index.delete(
    ids=["1", "2"],
    namespace='namespace1'
)

Let's practice!

Vector Databases for Embeddings with Pinecone