How do you implement multitenancy in Qdrant?

Qdrant's recommended multitenancy pattern is partitioning tenants within a single shared collection using a payload field (like tenant_id), rather than creating a separate collection per tenant, which avoids the operational overhead of managing potentially thousands of small collections.

client.create_payload_index(
    collection_name="documents",
    field_name="tenant_id",
    field_schema=models.KeywordIndexParams(type="keyword", is_tenant=True),
)

results = client.query_points(
    collection_name="documents",
    query=query_vector,
    query_filter=models.Filter(
        must=[models.FieldCondition(key="tenant_id", match=models.MatchValue(value="tenant_42"))]
    ),
    limit=10,
)

Marking the tenant field's payload index with is_tenant=True is a specific optimization hint: Qdrant uses it to organize storage so that a given tenant's data is grouped together more efficiently on disk, improving the performance of tenant-scoped queries compared to treating the field as just another generic keyword filter.

This shared-collection approach scales far better operationally than a collection-per-tenant design, since Qdrant's production guidance specifically warns that having too many shards — which a separate collection per user tends to produce, since every collection has its own shards — leads to performance degradation, making payload-based tenant partitioning the recommended default for most multi-tenant applications.