Query points

Nearest vectors

For an existing collection with four-dimensional unnamed vectors:

curl 'http://localhost:6333/collections/notes/points/query' \
  -H 'Content-Type: application/json' \
  --data '{"query":[0.2,0.1,0.9,0.7],"limit":3,"with_payload":true}'

Set using to the vector name for a collection with named vectors. Match the query vector's dimension to the collection configuration.

Filter results

{
  "query": [0.2, 0.1, 0.9, 0.7],
  "filter": {
    "must": [{"key": "city", "match": {"value": "London"}}]
  },
  "limit": 3,
  "with_payload": true
}

Send this body to the same query endpoint. Use limit for the result count.

Scroll points

curl 'http://localhost:6333/collections/notes/points/scroll' \
  -H 'Content-Type: application/json' \
  --data '{"limit":10,"with_payload":true,"with_vector":false}'

Pass result.next_page_offset as the next request's offset. A null offset marks the final page. Scrolling enumerates points; it does not rank by vector similarity.

Source

These request bodies use QueryRequest and ScrollRequest from the pinned specification. They require an existing collection and are not a record of requests executed against a live server.