What is the simplest way to isolate data for multiple tenants in a single Qdrant collection?

Question

Accepted Answer

The simplest pattern is one shared collection with a tenant_id payload field on every point, plus a payload index on tenant_id. Every tenant-scoped search, scroll, read, update, and delete should include the tenant filter. The index improves efficient tenant-aware filtering, but it is not an authorization boundary. I would derive tenant_id from authenticated server-side context rather than trusting a client-supplied value. Separate collections or deployments are alternatives when stronger administrative or resource isolation is required. from qdrant_client import models

client.create_payload_index(
    collection_name="documents",
    field_name="tenant_id",
    field_schema=models.PayloadSchemaType.KEYWORD,
)
results = client.query_points(
    collection_name="documents",
    query=[0.01]*768,
    query_filter=models.Filter(must=[
        models.FieldCondition(
            key="tenant_id",
            match=models.MatchValue(value="tenant-42"),
        )
    ]),
    limit=20,
).points tenant_id should exist on every tenant-owned point and use a consistent type and format. Trade-off: one collection is operationally efficient at large tenant counts; separate collections provide a stronger isolation boundary but increase lifecycle and index overhead. Having an indexed tenant_id field does not make the data secure by itself; the application must actually apply the filter. Centralizing Qdrant access in a repository or service layer makes accidental unscoped queries much harder to introduce.

What is the simplest way to isolate data for multiple tenants in a single Qdrant collection?

Use an indexed tenant_id payload field and scope every operation

What is the simplest way to isolate data for multiple tenants in a single Qdrant collection?

Use an indexed tenant_id payload field and scope every operation

Scenario Questions

Follow-up Questions