Payload index types
Qdrant provides payload index schemas suited to different data and query semantics, including keyword, integer, float, bool, geo, text, and other specialized types supported by the current release. The index type should match how the field is queried rather than merely its storage representation.
For a tags field where values are matched exactly, a keyword index is appropriate. For a price field queried with numeric ranges such as price < 100, an integer or float index should match the field's numeric type.
Text indexing is different from keyword indexing because full-text search tokenizes text and supports text-oriented matching, while keyword matching treats a value as an exact categorical value. Geo fields have geographic semantics rather than ordinary numeric comparisons.
The common mistake is using a text index for every string field. Product category, tenant ID, status, and similar categorical fields usually have exact-match semantics and should be modeled accordingly. Available index schemas and behavior are version-dependent.
Keyword indexes fit exact categorical matching
Integer and float indexes support numeric filtering
Text indexes support tokenized full-text matching
Geo indexes are designed for geographic conditions
Your product has a status field with values active, paused, and deleted. Which payload index semantics fit exact status filtering?
A price filter needs values between 50 and 100. Why would a keyword index be a poor fit?
A tags field sometimes contains multiple tags per point. How would you verify that the chosen index and filter semantics match the stored payload shape?
A team indexes a numeric price field as text and range queries become awkward. What data-model issue would you identify?
Your payload contains categorical, numeric, geographic, and full-text fields. How would you establish an index schema for each without over-indexing?
A text field has both exact-code queries and natural-language searches. Would you use one index type or separate fields? Explain.
You are designing a large multi-tenant collection with many filter dimensions and strict write throughput requirements. How would you select payload indexes using workload evidence rather than schema completeness?
A new Qdrant release changes or adds payload-index capabilities. How would you validate index compatibility before upgrading a production cluster?