Skip to main content
Search is the core operation in Qdrant. It finds the nearest neighbors to a given query vector based on the configured distance metric.

​
API Endpoint

POST /collections/{collection_name}/points/search
Find points most similar to your query vector. 
curl -X POST http://localhost:6333/collections/my_collection/points/search \
  -H 'Content-Type: application/json' \
  -d '{
    "vector": [0.2, 0.1, 0.9, 0.7],
    "limit": 10
  }'

​
Search Parameters

​
vector
array
required
Query vector to search for. Must match the collection’s vector dimensions.
​
limit
integer
default:"10"
Maximum number of results to return.
​
offset
integer
default:"0"
Skip first N results. Useful for pagination.
​
score_threshold
number
Minimum score threshold. Only return results with score above this value.
​
with_payload
boolean | array
Include payload in the results. Can be true, false, or an array of specific fields.
​
with_vector
boolean | array
Include vectors in the results. Can be true, false, or an array of specific vector names.
​
filter
object
Filter conditions to apply. See the filtering guide for details.

​
Search with Score Threshold

Only return results that meet a minimum similarity score. 
curl -X POST http://localhost:6333/collections/my_collection/points/search \
  -H 'Content-Type: application/json' \
  -d '{
    "vector": [0.2, 0.1, 0.9, 0.7],
    "limit": 10,
    "score_threshold": 0.8
  }'
The score threshold interpretation depends on the distance metric:
  • Cosine: 0 to 1 (higher is more similar)
  • Dot: -∞ to +∞ (higher is more similar)
  • Euclid: 0 to +∞ (lower is more similar)

​
Control Payload and Vector Returns

​
Return Specific Payload Fields

curl -X POST http://localhost:6333/collections/my_collection/points/search \
  -H 'Content-Type: application/json' \
  -d '{
    "vector": [0.2, 0.1, 0.9, 0.7],
    "limit": 5,
    "with_payload": ["city", "country"]
  }'

​
Include/Exclude Vectors

{
  "vector": [0.2, 0.1, 0.9, 0.7],
  "limit": 5,
  "with_vector": true
}

​
Search with Named Vectors

For collections with multiple named vectors, specify which vector to search. 
curl -X POST http://localhost:6333/collections/multi_vector_collection/points/search \
  -H 'Content-Type: application/json' \
  -d '{
    "vector": {
      "name": "text",
      "vector": [0.2, 0.1, 0.9, 0.7]
    },
    "limit": 10
  }'

​
Search with Filters

Combine vector similarity with payload filtering. 
curl -X POST http://localhost:6333/collections/my_collection/points/search \
  -H 'Content-Type: application/json' \
  -d '{
    "vector": [0.2, 0.1, 0.9, 0.7],
    "filter": {
      "must": [
        {
          "key": "country",
          "match": {
            "value": "Germany"
          }
        }
      ]
    },
    "limit": 5
  }'

​
Advanced Search Parameters

​
Search Params

Fine-tune the search algorithm for better performance or accuracy. 
curl -X POST http://localhost:6333/collections/my_collection/points/search \
  -H 'Content-Type: application/json' \
  -d '{
    "vector": [0.2, 0.1, 0.9, 0.7],
    "limit": 10,
    "params": {
      "hnsw_ef": 128,
      "exact": false
    }
  }'
​
params.hnsw_ef
integer
default:"dynamic"
Size of the dynamic candidate list for HNSW index. Higher values improve accuracy but reduce speed.
​
params.exact
boolean
default:"false"
If true, perform exact search (brute force) instead of approximate search.
​
params.indexed_only
boolean
default:"false"
If true, only search indexed vectors, skip unindexed segments.
Perform multiple searches in a single request for better efficiency. 
curl -X POST http://localhost:6333/collections/my_collection/points/search/batch \
  -H 'Content-Type: application/json' \
  -d '{
    "searches": [
      {
        "vector": [0.2, 0.1, 0.9, 0.7],
        "limit": 3
      },
      {
        "vector": [0.5, 0.3, 0.2, 0.8],
        "limit": 3
      }
    ]
  }'

​
Response Format

​
result
array
Array of scored points.
​
result[].id
integer | string
Point ID.
​
result[].version
integer
Point version.
​
result[].score
number
Similarity score.
​
result[].payload
object
Point payload (if requested).
​
result[].vector
array | object
Point vector(s) (if requested).
Response Example
{
  "result": [
    {
      "id": 1,
      "version": 3,
      "score": 0.95,
      "payload": {
        "city": "Berlin",
        "country": "Germany"
      },
      "vector": [0.2, 0.15, 0.88, 0.72]
    },
    {
      "id": 5,
      "version": 2,
      "score": 0.87,
      "payload": {
        "city": "Munich",
        "country": "Germany"
      }
    }
  ],
  "status": "ok",
  "time": 0.001234
}

​
Query Parameters

​
consistency
string
Read consistency level:
  • majority - Wait for majority of replicas
  • quorum - Wait for quorum of replicas
  • all - Wait for all replicas
​
timeout
integer
Operation timeout in seconds.
Large limit values (>1000) may impact performance. Consider pagination for large result sets.

​
Best Practices

  1. Limit Size: Keep limit values reasonable (10-100) for best performance
  2. Score Threshold: Use score thresholds to filter out low-quality results
  3. Payload Selection: Only request needed payload fields to reduce response size
  4. Batch Requests: Use batch search when performing multiple searches
  5. HNSW EF: Increase hnsw_ef for better accuracy at the cost of speed