Database / Qdrant Vector DB Interview questions
What is a payload in Qdrant?
A payload is the structured, JSON-formatted metadata attached to a point — separate from its vector — used to store descriptive information like text, tags, categories, numeric values, or nested objects that a query can filter on alongside vector similarity.
payload = { "title": "Getting Started with Vector Search", "category": "tutorial", "views": 1520, "tags": ["ai", "search", "embeddings"], }
Payloads support common JSON types — strings, numbers, booleans, arrays, and nested objects — and they can be filtered using rich conditions like exact match, range comparisons, full-text match, geographic radius, and nested-object conditions, which is what enables combined queries like "vectors similar to this one, published in the last 30 days, tagged 'ai'."
Payloads can also be updated independently of a point's vector — setting, overwriting, or deleting specific payload fields without touching the vector at all — which is useful for keeping metadata like view counts or status flags current without needing to re-embed and re-upsert the entire point.
More Related questions...
