What is payload filtering in Qdrant?

Payload filtering lets a query narrow results to points whose payload matches specified conditions, combined directly with vector similarity search rather than as a separate post-processing step, so a query can express something like "find similar vectors, but only where category equals 'news' and price is under $50."

from qdrant_client import models

results = client.query_points(
    collection_name="documents",
    query=query_vector,
    query_filter=models.Filter(
        must=[
            models.FieldCondition(key="category", match=models.MatchValue(value="news")),
            models.FieldCondition(key="price", range=models.Range(lt=50)),
        ]
    ),
    limit=10,
)

Filters are built from conditions combined with boolean clauses — must (AND), should (OR), and must_not (NOT) — and individual conditions support exact value matching, numeric or date ranges, full-text matching, geographic radius, and checks against array membership, covering most common filtering needs without needing to drop down into a separate query language.

Filtering can run without any dedicated payload index by scanning payloads directly, but for frequently filtered fields, creating a payload index (covered separately) makes filtered search dramatically faster, especially as a collection grows into millions of points.