Collections

A collection is a named set of points (vectors with a payload) among which you can search. The vector of each point within the same collection must have the same dimensionality and be compared by a single metric. Named vectors can be used to have multiple vectors in a single point, each of which can have their own dimensionality and metric requirements.

Distance metrics are used to measure similarities among vectors. The choice of metric depends on the way vectors obtaining and, in particular, on the method of neural network encoder training.

Qdrant supports these most popular types of metrics:

  • Dot product: Dot - [wiki]
  • Cosine similarity: Cosine - [wiki]
  • Euclidean distance: Euclid - [wiki]
  • Manhattan distance: Manhattan - [wiki]

In addition to metrics and vector size, each collection uses its own set of parameters that controls collection optimization, index construction, and vacuum. These settings can be changed at any time by a corresponding request.

Setting up multitenancy

How many collections should you create? In most cases, you should only use a single collection with payload-based partitioning. This approach is called multitenancy. It is efficient for most of users, but it requires additional configuration. Learn how to set it up

When should you create multiple collections? When you have a limited number of users and you need isolation. This approach is flexible, but it may be more costly, since creating numerous collections may result in resource overhead. Also, you need to ensure that they do not affect each other in any way, including performance-wise.

Note: If you’re running curl from the command line, the following commands assume that you have a running instance of Qdrant on http://localhost:6333. If needed, you can set one up as described in our Quickstart guide. For convenience, these commands specify collections named test_collection1 through test_collection4.

Create a collection

PUT /collections/{collection_name}
{
    "vectors": {
      "size": 300,
      "distance": "Cosine"
    }
}

curl -X PUT http://localhost:6333/collections/test_collection1 \
  -H 'Content-Type: application/json' \
  --data-raw '{
    "vectors": {
      "size": 300,
      "distance": "Cosine"
    } 
  }'

from qdrant_client import QdrantClient, models

client = QdrantClient(url="http://localhost:6333")

client.create_collection(
    collection_name="{collection_name}",
    vectors_config=models.VectorParams(size=100, distance=models.Distance.COSINE),
)

import { QdrantClient } from "@qdrant/js-client-rest";

const client = new QdrantClient({ host: "localhost", port: 6333 });

client.createCollection("{collection_name}", {
  vectors: { size: 100, distance: "Cosine" },
});

use qdrant_client::{
    client::QdrantClient,
    qdrant::{vectors_config::Config, CreateCollection, Distance, VectorParams, VectorsConfig},
};

//The Rust client uses Qdrant's GRPC interface
let client = QdrantClient::from_url("http://localhost:6334").build()?;

client
    .create_collection(&CreateCollection {
        collection_name: "{collection_name}".to_string(),
        vectors_config: Some(VectorsConfig {
            config: Some(Config::Params(VectorParams {
                size: 100,
                distance: Distance::Cosine.into(),
                ..Default::default()
            })),
        }),
        ..Default::default()
    })
    .await?;

import io.qdrant.client.grpc.Collections.Distance;
import io.qdrant.client.grpc.Collections.VectorParams;
import io.qdrant.client.QdrantClient;
import io.qdrant.client.QdrantGrpcClient;

QdrantClient client = new QdrantClient(
    QdrantGrpcClient.newBuilder("localhost", 6334, false).build());

client.createCollectionAsync("{collection_name}",
        VectorParams.newBuilder().setDistance(Distance.Cosine).setSize(100).build()).get();

using Qdrant.Client;
using Qdrant.Client.Grpc;

var client = new QdrantClient("localhost", 6334);

await client.CreateCollectionAsync(
	collectionName: "{collection_name}",
	vectorsConfig: new VectorParams { Size = 100, Distance = Distance.Cosine }
);

In addition to the required options, you can also specify custom values for the following collection options:

  • hnsw_config - see indexing for details.
  • wal_config - Write-Ahead-Log related configuration. See more details about WAL
  • optimizers_config - see optimizer for details.
  • shard_number - which defines how many shards the collection should have. See distributed deployment section for details.
  • on_disk_payload - defines where to store payload data. If true - payload will be stored on disk only. Might be useful for limiting the RAM usage in case of large payload.
  • quantization_config - see quantization for details.

Default parameters for the optional collection parameters are defined in configuration file.

See schema definitions and a configuration file for more information about collection and vector parameters.

Available as of v1.2.0

Vectors all live in RAM for very quick access. The on_disk parameter can be set in the vector configuration. If true, all vectors will live on disk. This will enable the use of memmaps, which is suitable for ingesting a large amount of data.

Create collection from another collection

Available as of v1.0.0

It is possible to initialize a collection from another existing collection.

This might be useful for experimenting quickly with different configurations for the same data set.

Make sure the vectors have the same size and distance function when setting up the vectors configuration in the new collection. If you used the previous sample code, "size": 300 and "distance": "Cosine".

PUT /collections/{collection_name}
{
    "vectors": {
      "size": 100,
      "distance": "Cosine"
    },
    "init_from": {
       "collection": "{from_collection_name}"
    }
}

curl -X PUT http://localhost:6333/collections/test_collection2 \
  -H 'Content-Type: application/json' \
  --data-raw '{
    "vectors": {
      "size": 300,
      "distance": "Cosine"
    },
    "init_from": {
       "collection": "test_collection1"
    }
  }'

from qdrant_client import QdrantClient

client = QdrantClient(url="http://localhost:6333")

client.create_collection(
    collection_name="{collection_name}",
    vectors_config=models.VectorParams(size=100, distance=models.Distance.COSINE),
    init_from=models.InitFrom(collection="{from_collection_name}"),
)

import { QdrantClient } from "@qdrant/js-client-rest";

const client = new QdrantClient({ host: "localhost", port: 6333 });

client.createCollection("{collection_name}", {
  vectors: { size: 100, distance: "Cosine" },
  init_from: { collection: "{from_collection_name}" },
});

use qdrant_client::{
    client::QdrantClient,
    qdrant::{vectors_config::Config, CreateCollection, Distance, VectorParams, VectorsConfig},
};

let client = QdrantClient::from_url("http://localhost:6334").build()?;

client
    .create_collection(&CreateCollection {
        collection_name: "{collection_name}".to_string(),
        vectors_config: Some(VectorsConfig {
            config: Some(Config::Params(VectorParams {
                size: 100,
                distance: Distance::Cosine.into(),
                ..Default::default()
            })),
        }),
        init_from_collection: Some("{from_collection_name}".to_string()),
        ..Default::default()
    })
    .await?;

import io.qdrant.client.QdrantClient;
import io.qdrant.client.QdrantGrpcClient;
import io.qdrant.client.grpc.Collections.CreateCollection;
import io.qdrant.client.grpc.Collections.Distance;
import io.qdrant.client.grpc.Collections.VectorParams;
import io.qdrant.client.grpc.Collections.VectorsConfig;

QdrantClient client =
    new QdrantClient(QdrantGrpcClient.newBuilder("localhost", 6334, false).build());

client
    .createCollectionAsync(
        CreateCollection.newBuilder()
            .setCollectionName("{collection_name}")
            .setVectorsConfig(
                VectorsConfig.newBuilder()
                    .setParams(
                        VectorParams.newBuilder()
                            .setSize(100)
                            .setDistance(Distance.Cosine)
                            .build()))
            .setInitFromCollection("{from_collection_name}")
            .build())
    .get();

using Qdrant.Client;
using Qdrant.Client.Grpc;

var client = new QdrantClient("localhost", 6334);

await client.CreateCollectionAsync(
	collectionName: "{collection_name}",
	vectorsConfig: new VectorParams { Size = 100, Distance = Distance.Cosine },
	initFromCollection: "{from_collection_name}"
);

Collection with multiple vectors

Available as of v0.10.0

It is possible to have multiple vectors per record. This feature allows for multiple vector storages per collection. To distinguish vectors in one record, they should have a unique name defined when creating the collection. Each named vector in this mode has its distance and size:

PUT /collections/{collection_name}
{
    "vectors": {
        "image": {
            "size": 4,
            "distance": "Dot"
        },
        "text": {
            "size": 8,
            "distance": "Cosine"
        }
    }
}

curl -X PUT http://localhost:6333/collections/test_collection3 \
  -H 'Content-Type: application/json' \
  --data-raw '{
    "vectors": {
        "image": {
            "size": 4,
            "distance": "Dot"
        },
        "text": {
            "size": 8,
            "distance": "Cosine"
        }
      }
    }'

from qdrant_client import QdrantClient, models


client = QdrantClient(url="http://localhost:6333")

client.create_collection(
    collection_name="{collection_name}",
    vectors_config={
        "image": models.VectorParams(size=4, distance=models.Distance.DOT),
        "text": models.VectorParams(size=8, distance=models.Distance.COSINE),
    },
)

import { QdrantClient } from "@qdrant/js-client-rest";

const client = new QdrantClient({ host: "localhost", port: 6333 });

client.createCollection("{collection_name}", {
  vectors: {
    image: { size: 4, distance: "Dot" },
    text: { size: 8, distance: "Cosine" },
  },
});

use qdrant_client::{
    client::QdrantClient,
    qdrant::{
        vectors_config::Config, CreateCollection, Distance, VectorParams, VectorParamsMap,
        VectorsConfig,
    },
};

let client = QdrantClient::from_url("http://localhost:6334").build()?;

client
    .create_collection(&CreateCollection {
        collection_name: "{collection_name}".to_string(),
        vectors_config: Some(VectorsConfig {
            config: Some(Config::ParamsMap(VectorParamsMap {
                map: [
                    (
                        "image".to_string(),
                        VectorParams {
                            size: 4,
                            distance: Distance::Dot.into(),
                            ..Default::default()
                        },
                    ),
                    (
                        "text".to_string(),
                        VectorParams {
                            size: 8,
                            distance: Distance::Cosine.into(),
                            ..Default::default()
                        },
                    ),
                ]
                .into(),
            })),
        }),
        ..Default::default()
    })
    .await?;

import java.util.Map;

import io.qdrant.client.QdrantClient;
import io.qdrant.client.QdrantGrpcClient;
import io.qdrant.client.grpc.Collections.Distance;
import io.qdrant.client.grpc.Collections.VectorParams;

QdrantClient client =
    new QdrantClient(QdrantGrpcClient.newBuilder("localhost", 6334, false).build());

client
    .createCollectionAsync(
        "{collection_name}",
        Map.of(
            "image", VectorParams.newBuilder().setSize(4).setDistance(Distance.Dot).build(),
            "text",
                VectorParams.newBuilder().setSize(8).setDistance(Distance.Cosine).build()))
    .get();

using Qdrant.Client;
using Qdrant.Client.Grpc;

var client = new QdrantClient("localhost", 6334);

await client.CreateCollectionAsync(
	collectionName: "{collection_name}",
	vectorsConfig: new VectorParamsMap
	{
		Map =
		{
			["image"] = new VectorParams { Size = 4, Distance = Distance.Dot },
			["text"] = new VectorParams { Size = 8, Distance = Distance.Cosine },
		}
	}
);

For rare use cases, it is possible to create a collection without any vector storage.

Available as of v1.1.1

For each named vector you can optionally specify hnsw_config or quantization_config to deviate from the collection configuration. This can be useful to fine-tune search performance on a vector level.

Available as of v1.2.0

Vectors all live in RAM for very quick access. On a per-vector basis you can set on_disk to true to store all vectors on disk at all times. This will enable the use of memmaps, which is suitable for ingesting a large amount of data.

Vector datatypes

Available as of v1.9.0

Some embedding providers may provide embeddings in a pre-quantized format. One of the most notable examples is the Cohere int8 & binary embeddings. Qdrant has direct support for uint8 embeddings, which you can also use in combination with binary quantization.

To create a collection with uint8 embeddings, you can use the following configuration:

PUT /collections/{collection_name}
{
    "vectors": {
      "size": 1024,
      "distance": "Cosine",
      "datatype": "uint8"
    }
}

curl -X PUT http://localhost:6333/collections/test_collection1 \
  -H 'Content-Type: application/json' \
  --data-raw '{
    "vectors": {
      "size": 1024,
      "distance": "Cosine",
      "datatype": "uint8"
    }
  }'

from qdrant_client import QdrantClient, models

client = QdrantClient(url="http://localhost:6333")

client.create_collection(
    collection_name="{collection_name}",
    vectors_config=models.VectorParams(
        size=1024,
        distance=models.Distance.COSINE,
        datatype=models.Datatype.UINT8,
    ),
)

import { QdrantClient } from "@qdrant/js-client-rest";

const client = new QdrantClient({ host: "localhost", port: 6333 });

client.createCollection("{collection_name}", {
  vectors: {
    image: { size: 1024, distance: "Cosine", datatype: "uint8" },
  },
});

use qdrant_client::{
    client::QdrantClient,
    qdrant::{vectors_config::Config, CreateCollection, Datatype, Distance, VectorParams, VectorsConfig},
};

let client = QdrantClient::from_url("http://localhost:6334").build()?;

qdrant_client
     .create_collection(&CreateCollection {
         collection_name: "{collection_name}".into(),
         vectors_config: Some(VectorsConfig {
             config: Some(Config::Params(VectorParams {
                 size: 1024,
                 distance: Distance::Cosine.into(),
                 datatype: Some(Datatype::Uint8.into()),
                 ..Default::default()
             })),
         }),
         ..Default::default()
     })
     .await?;

import io.qdrant.client.QdrantClient;
import io.qdrant.client.grpc.Collections.Datatype;
import io.qdrant.client.grpc.Collections.Distance;
import io.qdrant.client.grpc.Collections.VectorParams;

QdrantClient client = new QdrantClient(
    QdrantGrpcClient.newBuilder("localhost", 6334, false).build());

client
    .createCollectionAsync("{collection_name}",
        VectorParams.newBuilder()
            .setSize(1024)
            .setDistance(Distance.Cosine)
            .setDatatype(Datatype.Uint8)
            .build())
    .get();

using Qdrant.Client;
using Qdrant.Client.Grpc;

var client = new QdrantClient("localhost", 6334);

await client.CreateCollectionAsync(
  collectionName: "{collection_name}",
  vectorsConfig: new VectorParams {
    Size = 1024, Distance = Distance.Cosine, Datatype = Datatype.Uint8
  }
);

Vectors with uint8 datatype are stored in a more compact format, which can save memory and improve search speed at the cost of some precision. If you choose to use the uint8 datatype, elements of the vector will be stored as unsigned 8-bit integers, which can take values from 0 to 255.

Collection with sparse vectors

Available as of v1.7.0

Qdrant supports sparse vectors as a first-class citizen.

Sparse vectors are useful for text search, where each word is represented as a separate dimension.

Collections can contain sparse vectors as additional named vectors along side regular dense vectors in a single point.

Unlike dense vectors, sparse vectors must be named. And additionally, sparse vectors and dense vectors must have different names within a collection.

PUT /collections/{collection_name}
{
    "sparse_vectors": {
        "text": { },
    }
}

curl -X PUT http://localhost:6333/collections/test_collection4 \
  -H 'Content-Type: application/json' \
  --data-raw '{
    "sparse_vectors": {
        "text": { }
    }
  }'

from qdrant_client import QdrantClient, models

client = QdrantClient(url="http://localhost:6333")

client.create_collection(
    collection_name="{collection_name}",
    sparse_vectors_config={
        "text": models.SparseVectorParams(),
    },
)

import { QdrantClient } from "@qdrant/js-client-rest";

const client = new QdrantClient({ host: "localhost", port: 6333 });

client.createCollection("{collection_name}", {
  sparse_vectors: {
    text: { },
  },
});

use qdrant_client::{
    client::QdrantClient,
    qdrant::{
        vectors_config::Config, CreateCollection, Distance, SparseVectorParams, VectorParamsMap,
        VectorsConfig,
    },
};

let client = QdrantClient::from_url("http://localhost:6334").build()?;

client
    .create_collection(&CreateCollection {
        collection_name: "{collection_name}".to_string(),
        sparse_vectors_config: Some(SparseVectorsConfig {
            map: [
                    (
                        "text".to_string(),
                        SparseVectorParams {},
                    ),
                ]
                .into(),
            }),
        }),
        ..Default::default()
    })
    .await?;

import io.qdrant.client.QdrantClient;
import io.qdrant.client.QdrantGrpcClient;
import io.qdrant.client.grpc.Collections.CreateCollection;
import io.qdrant.client.grpc.Collections.SparseVectorConfig;
import io.qdrant.client.grpc.Collections.SparseVectorParams;

QdrantClient client =
    new QdrantClient(QdrantGrpcClient.newBuilder("localhost", 6334, false).build());

client
    .createCollectionAsync(
        CreateCollection.newBuilder()
            .setCollectionName("{collection_name}")
            .setSparseVectorsConfig(
                SparseVectorConfig.newBuilder()
                    .putMap("text", SparseVectorParams.getDefaultInstance()))
            .build())
    .get();

using Qdrant.Client;
using Qdrant.Client.Grpc;

var client = new QdrantClient("localhost", 6334);

await client.CreateCollectionAsync(
	collectionName: "{collection_name}",
	sparseVectorsConfig: ("text", new SparseVectorParams())
);

Outside of a unique name, there are no required configuration parameters for sparse vectors.

The distance function for sparse vectors is always Dot and does not need to be specified.

However, there are optional parameters to tune the underlying sparse vector index.

Delete collection

DELETE http://localhost:6333/collections/test_collection4 

curl -X DELETE http://localhost:6333/collections/test_collection4 

client.delete_collection(collection_name="{collection_name}")

client.deleteCollection("{collection_name}");

client.delete_collection("{collection_name}").await?;

import io.qdrant.client.QdrantClient;
import io.qdrant.client.QdrantGrpcClient;

QdrantClient client =
    new QdrantClient(QdrantGrpcClient.newBuilder("localhost", 6334, false).build());

client.deleteCollectionAsync("{collection_name}").get();

using Qdrant.Client;

var client = new QdrantClient("localhost", 6334);

await client.DeleteCollectionAsync("{collection_name}");

Update collection parameters

Dynamic parameter updates may be helpful, for example, for more efficient initial loading of vectors. For example, you can disable indexing during the upload process, and enable it immediately after the upload is finished. As a result, you will not waste extra computation resources on rebuilding the index.

The following command enables indexing for segments that have more than 10000 kB of vectors stored:

PATCH /collections/{collection_name}
{
    "optimizers_config": {
        "indexing_threshold": 10000
    }
}

curl -X PATCH http://localhost:6333/collections/test_collection1 \
  -H 'Content-Type: application/json' \
  --data-raw '{
    "optimizers_config": {
        "indexing_threshold": 10000
    }
  }'

client.update_collection(
    collection_name="{collection_name}",
    optimizer_config=models.OptimizersConfigDiff(indexing_threshold=10000),
)

client.updateCollection("{collection_name}", {
  optimizers_config: {
    indexing_threshold: 10000,
  },
});

use qdrant_client::qdrant::OptimizersConfigDiff;

client
    .update_collection(
        "{collection_name}",
        &OptimizersConfigDiff {
            indexing_threshold: Some(10000),
            ..Default::default()
        },
        None,
        None,
        None,
        None,
        None,
    )
    .await?;

import io.qdrant.client.grpc.Collections.OptimizersConfigDiff;
import io.qdrant.client.grpc.Collections.UpdateCollection;

client.updateCollectionAsync(
    UpdateCollection.newBuilder()
        .setCollectionName("{collection_name}")
        .setOptimizersConfig(
            OptimizersConfigDiff.newBuilder().setIndexingThreshold(10000).build())
        .build());

using Qdrant.Client;
using Qdrant.Client.Grpc;

var client = new QdrantClient("localhost", 6334);

await client.UpdateCollectionAsync(
	collectionName: "{collection_name}",
	optimizersConfig: new OptimizersConfigDiff { IndexingThreshold = 10000 }
);

The following parameters can be updated:

  • optimizers_config - see optimizer for details.
  • hnsw_config - see indexing for details.
  • quantization_config - see quantization for details.
  • vectors - vector-specific configuration, including individual hnsw_config, quantization_config and on_disk settings.
  • params - other collection parameters, including write_consistency_factor and on_disk_payload.

Full API specification is available in schema definitions.

Calls to this endpoint may be blocking as it waits for existing optimizers to finish. We recommended against using this in a production database as it may introduce huge overhead due to the rebuilding of the index.

Update vector parameters

Available as of v1.4.0

Qdrant 1.4 adds support for updating more collection parameters at runtime. HNSW index, quantization and disk configurations can now be changed without recreating a collection. Segments (with index and quantized data) will automatically be rebuilt in the background to match updated parameters.

To put vector data on disk for a collection that does not have named vectors, use "" as name:

PATCH /collections/{collection_name}
{
    "vectors": {
        "": {
            "on_disk": true
        }
    }
}

curl -X PATCH http://localhost:6333/collections/test_collection1 \
  -H 'Content-Type: application/json' \
  --data-raw '{
    "vectors": {
        "": { 
            "on_disk": true 
      }
    }
  }'

To put vector data on disk for a collection that does have named vectors:

Note: To create a vector name, follow the procedure from our Points.

PATCH /collections/{collection_name}
{
    "vectors": {
        "my_vector": {
            "on_disk": true
        }
    }
}

curl -X PATCH http://localhost:6333/collections/test_collection1 \
  -H 'Content-Type: application/json' \
  --data-raw '{
    "vectors": {
        "my_vector": { 
           "on_disk": true 
      }
    }
  }'

In the following example the HNSW index and quantization parameters are updated, both for the whole collection, and for my_vector specifically:

PATCH /collections/{collection_name}
{
    "vectors": {
        "my_vector": {
            "hnsw_config": {
                "m": 32,
                "ef_construct": 123
            },
            "quantization_config": {
                "product": {
                    "compression": "x32",
                    "always_ram": true
                }
            },
            "on_disk": true
        }
    },
    "hnsw_config": {
        "ef_construct": 123
    },
    "quantization_config": {
        "scalar": {
            "type": "int8",
            "quantile": 0.8,
            "always_ram": false
        }
    }
}

curl -X PATCH http://localhost:6333/collections/test_collection1 \
  -H 'Content-Type: application/json' \
  --data-raw '{
    "vectors": {
        "my_vector": {
            "hnsw_config": {
                "m": 32,
                "ef_construct": 123
            },
            "quantization_config": {
                "product": {
                    "compression": "x32",
                    "always_ram": true
                }
            },
            "on_disk": true
        }
    },
    "hnsw_config": {
        "ef_construct": 123
    },
    "quantization_config": {
        "scalar": {
            "type": "int8",
            "quantile": 0.8,
            "always_ram": false
        }
    }
}'

client.update_collection(
    collection_name="{collection_name}",
    vectors_config={
        "my_vector": models.VectorParamsDiff(
            hnsw_config=models.HnswConfigDiff(
                m=32,
                ef_construct=123,
            ),
            quantization_config=models.ProductQuantization(
                product=models.ProductQuantizationConfig(
                    compression=models.CompressionRatio.X32,
                    always_ram=True,
                ),
            ),
            on_disk=True,
        ),
    },
    hnsw_config=models.HnswConfigDiff(
        ef_construct=123,
    ),
    quantization_config=models.ScalarQuantization(
        scalar=models.ScalarQuantizationConfig(
            type=models.ScalarType.INT8,
            quantile=0.8,
            always_ram=False,
        ),
    ),
)

client.updateCollection("{collection_name}", {
  vectors: {
    my_vector: {
      hnsw_config: {
        m: 32,
        ef_construct: 123,
      },
      quantization_config: {
        product: {
          compression: "x32",
          always_ram: true,
        },
      },
      on_disk: true,
    },
  },
  hnsw_config: {
    ef_construct: 123,
  },
  quantization_config: {
    scalar: {
      type: "int8",
      quantile: 0.8,
      always_ram: true,
    },
  },
});

use qdrant_client::client::QdrantClient;
use qdrant_client::qdrant::{
    quantization_config_diff::Quantization, vectors_config_diff::Config, HnswConfigDiff,
    QuantizationConfigDiff, QuantizationType, ScalarQuantization, VectorParamsDiff,
    VectorsConfigDiff,
};

client
    .update_collection(
        "{collection_name}",
        None,
        None,
        None,
        Some(&HnswConfigDiff {
            ef_construct: Some(123),
            ..Default::default()
        }),
        Some(&VectorsConfigDiff {
            config: Some(Config::ParamsMap(
                qdrant_client::qdrant::VectorParamsDiffMap {
                    map: HashMap::from([(
                        ("my_vector".into()),
                        VectorParamsDiff {
                            hnsw_config: Some(HnswConfigDiff {
                                m: Some(32),
                                ef_construct: Some(123),
                                ..Default::default()
                            }),
                            ..Default::default()
                        },
                    )]),
                },
            )),
        }),
        Some(&QuantizationConfigDiff {
            quantization: Some(Quantization::Scalar(ScalarQuantization {
                r#type: QuantizationType::Int8 as i32,
                quantile: Some(0.8),
                always_ram: Some(true),
                ..Default::default()
            })),
        }),
    )
    .await?;

import io.qdrant.client.grpc.Collections.HnswConfigDiff;
import io.qdrant.client.grpc.Collections.QuantizationConfigDiff;
import io.qdrant.client.grpc.Collections.QuantizationType;
import io.qdrant.client.grpc.Collections.ScalarQuantization;
import io.qdrant.client.grpc.Collections.UpdateCollection;
import io.qdrant.client.grpc.Collections.VectorParamsDiff;
import io.qdrant.client.grpc.Collections.VectorParamsDiffMap;
import io.qdrant.client.grpc.Collections.VectorsConfigDiff;

client
    .updateCollectionAsync(
        UpdateCollection.newBuilder()
            .setCollectionName("{collection_name}")
            .setHnswConfig(HnswConfigDiff.newBuilder().setEfConstruct(123).build())
            .setVectorsConfig(
                VectorsConfigDiff.newBuilder()
                    .setParamsMap(
                        VectorParamsDiffMap.newBuilder()
                            .putMap(
                                "my_vector",
                                VectorParamsDiff.newBuilder()
                                    .setHnswConfig(
                                        HnswConfigDiff.newBuilder()
                                            .setM(3)
                                            .setEfConstruct(123)
                                            .build())
                                    .build())))
            .setQuantizationConfig(
                QuantizationConfigDiff.newBuilder()
                    .setScalar(
                        ScalarQuantization.newBuilder()
                            .setType(QuantizationType.Int8)
                            .setQuantile(0.8f)
                            .setAlwaysRam(true)
                            .build()))
            .build())
    .get();

using Qdrant.Client;
using Qdrant.Client.Grpc;

var client = new QdrantClient("localhost", 6334);

await client.UpdateCollectionAsync(
	collectionName: "{collection_name}",
	hnswConfig: new HnswConfigDiff { EfConstruct = 123 },
	vectorsConfig: new VectorParamsDiffMap
	{
		Map =
		{
			{
				"my_vector",
				new VectorParamsDiff
				{
					HnswConfig = new HnswConfigDiff { M = 3, EfConstruct = 123 }
				}
			}
		}
	},
	quantizationConfig: new QuantizationConfigDiff
	{
		Scalar = new ScalarQuantization
		{
			Type = QuantizationType.Int8,
			Quantile = 0.8f,
			AlwaysRam = true
		}
	}
);

Collection info

Qdrant allows determining the configuration parameters of an existing collection to better understand how the points are distributed and indexed.

GET /collections/test_collection1

curl -X GET http://localhost:6333/collections/test_collection1

client.get_collection(collection_name="{collection_name}")

client.getCollection("{collection_name}");

client.collection_info("{collection_name}").await?;

client.getCollectionInfoAsync("{collection_name}").get();

await client.GetCollectionInfoAsync("{collection_name}");
Expected result
{
    "result": {
        "status": "green",
        "optimizer_status": "ok",
        "vectors_count": 1068786,
        "indexed_vectors_count": 1024232,
        "points_count": 1068786,
        "segments_count": 31,
        "config": {
            "params": {
                "vectors": {
                    "size": 384,
                    "distance": "Cosine"
                },
                "shard_number": 1,
                "replication_factor": 1,
                "write_consistency_factor": 1,
                "on_disk_payload": false
            },
            "hnsw_config": {
                "m": 16,
                "ef_construct": 100,
                "full_scan_threshold": 10000,
                "max_indexing_threads": 0
            },
            "optimizer_config": {
                "deleted_threshold": 0.2,
                "vacuum_min_vector_number": 1000,
                "default_segment_number": 0,
                "max_segment_size": null,
                "memmap_threshold": null,
                "indexing_threshold": 20000,
                "flush_interval_sec": 5,
                "max_optimization_threads": 1
            },
            "wal_config": {
                "wal_capacity_mb": 32,
                "wal_segments_ahead": 0
            }
        },
        "payload_schema": {}
    },
    "status": "ok",
    "time": 0.00010143
}

If you insert the vectors into the collection, the status field may become yellow whilst it is optimizing. It will become green once all the points are successfully processed.

The following color statuses are possible:

  • 🟢 green: collection is ready
  • 🟡 yellow: collection is optimizing
  • ⚫ grey: collection is pending optimization (help)
  • 🔴 red: an error occurred which the engine could not recover from

Grey collection status

Available as of v1.9.0

A collection may have the grey ⚫ status or show “optimizations pending, awaiting update operation” as optimization status. This state is normally caused by restarting a Qdrant instance while optimizations were ongoing.

It means the collection has optimizations pending, but they are paused. You must send any update operation to trigger and start the optimizations again.

For example:

PATCH /collections/{collection_name}
{
    "optimizers_config": {}
}

curl -X PATCH http://localhost:6333/collections/test_collection1 \
  -H 'Content-Type: application/json' \
  --data-raw '{
    "optimizers_config": {}
  }'

client.update_collection(
    collection_name="{collection_name}",
    optimizer_config=models.OptimizersConfigDiff(),
)

client.updateCollection("{collection_name}", {
  optimizers_config: {},
});

use qdrant_client::qdrant::OptimizersConfigDiff;

client
    .update_collection(
        "{collection_name}",
        &OptimizersConfigDiff::default(),
        None,
        None,
        None,
        None,
        None,
    )
    .await?;

import io.qdrant.client.grpc.Collections.OptimizersConfigDiff;
import io.qdrant.client.grpc.Collections.UpdateCollection;

client.updateCollectionAsync(
    UpdateCollection.newBuilder()
        .setCollectionName("{collection_name}")
        .setOptimizersConfig(
            OptimizersConfigDiff.getDefaultInstance())
        .build());

using Qdrant.Client;
using Qdrant.Client.Grpc;

var client = new QdrantClient("localhost", 6334);

await client.UpdateCollectionAsync(
	collectionName: "{collection_name}",
	optimizersConfig: new OptimizersConfigDiff { }
);

Approximate point and vector counts

You may be interested in the count attributes:

  • points_count - total number of objects (vectors and their payloads) stored in the collection
  • vectors_count - total number of vectors in a collection, useful if you have multiple vectors per point
  • indexed_vectors_count - total number of vectors stored in the HNSW or sparse index. Qdrant does not store all the vectors in the index, but only if an index segment might be created for a given configuration.

The above counts are not exact, but should be considered approximate. Depending on how you use Qdrant these may give very different numbers than what you may expect. It’s therefore important not to rely on them.

More specifically, these numbers represent the count of points and vectors in Qdrant’s internal storage. Internally, Qdrant may temporarily duplicate points as part of automatic optimizations. It may keep changed or deleted points for a bit. And it may delay indexing of new points. All of that is for optimization reasons.

Updates you do are therefore not directly reflected in these numbers. If you see a wildly different count of points, it will likely resolve itself once a new round of automatic optimizations has completed.

To clarify: these numbers don’t represent the exact amount of points or vectors you have inserted, nor does it represent the exact number of distinguishable points or vectors you can query. If you want to know exact counts, refer to the count API.

Note: these numbers may be removed in a future version of Qdrant.

Indexing vectors in HNSW

In some cases, you might be surprised the value of indexed_vectors_count is lower than vectors_count. This is an intended behaviour and depends on the optimizer configuration. A new index segment is built if the size of non-indexed vectors is higher than the value of indexing_threshold(in kB). If your collection is very small or the dimensionality of the vectors is low, there might be no HNSW segment created and indexed_vectors_count might be equal to 0.

It is possible to reduce the indexing_threshold for an existing collection by updating collection parameters.

Collection aliases

In a production environment, it is sometimes necessary to switch different versions of vectors seamlessly. For example, when upgrading to a new version of the neural network.

There is no way to stop the service and rebuild the collection with new vectors in these situations. Aliases are additional names for existing collections. All queries to the collection can also be done identically, using an alias instead of the collection name.

Thus, it is possible to build a second collection in the background and then switch alias from the old to the new collection. Since all changes of aliases happen atomically, no concurrent requests will be affected during the switch.

Create alias

POST /collections/aliases
{
    "actions": [
        {
            "create_alias": {
                "collection_name": "test_collection1",
                "alias_name": "production_collection"
            }
        }
    ]
}

curl -X POST http://localhost:6333/collections/aliases \
  -H 'Content-Type: application/json' \
  --data-raw '{
    "actions": [
        {
            "create_alias": {
                "collection_name": "test_collection1",
                "alias_name": "production_collection"
            }
        }
    ]
}'

client.update_collection_aliases(
    change_aliases_operations=[
        models.CreateAliasOperation(
            create_alias=models.CreateAlias(
                collection_name="example_collection", alias_name="production_collection"
            )
        )
    ]
)

client.updateCollectionAliases({
  actions: [
    {
      create_alias: {
        collection_name: "example_collection",
        alias_name: "production_collection",
      },
    },
  ],
});

client.create_alias("example_collection", "production_collection").await?;

client.createAliasAsync("production_collection", "example_collection").get();

await client.CreateAliasAsync(aliasName: "production_collection", collectionName: "example_collection");

Remove alias

curl -X POST http://localhost:6333/collections/aliases \
  -H 'Content-Type: application/json' \
  --data-raw '{
    "actions": [
        {
            "delete_alias": {
                "collection_name": "test_collection1",
                "alias_name": "production_collection"
            }
        }
    ]
}'

POST /collections/aliases
{
    "actions": [
        {
            "delete_alias": {
                "alias_name": "production_collection"
            }
        }
    ]
}

client.update_collection_aliases(
    change_aliases_operations=[
        models.DeleteAliasOperation(
            delete_alias=models.DeleteAlias(alias_name="production_collection")
        ),
    ]
)

client.updateCollectionAliases({
  actions: [
    {
      delete_alias: {
        alias_name: "production_collection",
      },
    },
  ],
});

client.delete_alias("production_collection").await?;

client.deleteAliasAsync("production_collection").get();

await client.DeleteAliasAsync("production_collection");

Switch collection

Multiple alias actions are performed atomically. For example, you can switch underlying collection with the following command:

POST /collections/aliases
{
    "actions": [
        {
            "delete_alias": {
                "alias_name": "production_collection"
            }
        },
        {
            "create_alias": {
                "collection_name": "test_collection2",
                "alias_name": "production_collection"
            }
        }
    ]
}

curl -X POST http://localhost:6333/collections/aliases \
  -H 'Content-Type: application/json' \
  --data-raw '{
    "actions": [
        {
            "delete_alias": {
                "alias_name": "production_collection"
            }
        },
        {
            "create_alias": {
                "collection_name": "test_collection2",
                "alias_name": "production_collection"
            }
        }
    ]
}'

client.update_collection_aliases(
    change_aliases_operations=[
        models.DeleteAliasOperation(
            delete_alias=models.DeleteAlias(alias_name="production_collection")
        ),
        models.CreateAliasOperation(
            create_alias=models.CreateAlias(
                collection_name="example_collection", alias_name="production_collection"
            )
        ),
    ]
)

client.updateCollectionAliases({
  actions: [
    {
      delete_alias: {
        alias_name: "production_collection",
      },
    },
    {
      create_alias: {
        collection_name: "example_collection",
        alias_name: "production_collection",
      },
    },
  ],
});

client.delete_alias("production_collection").await?;
client.create_alias("example_collection", "production_collection").await?;

client.deleteAliasAsync("production_collection").get();
client.createAliasAsync("production_collection", "example_collection").get();

await client.DeleteAliasAsync("production_collection");
await client.CreateAliasAsync(aliasName: "production_collection", collectionName: "example_collection");

List collection aliases

GET /collections/test_collection2/aliases

curl -X GET http://localhost:6333/collections/test_collection2/aliases

from qdrant_client import QdrantClient

client = QdrantClient(url="http://localhost:6333")

client.get_collection_aliases(collection_name="{collection_name}")

import { QdrantClient } from "@qdrant/js-client-rest";

const client = new QdrantClient({ host: "localhost", port: 6333 });

client.getCollectionAliases("{collection_name}");

use qdrant_client::client::QdrantClient;

let client = QdrantClient::from_url("http://localhost:6334").build()?;

client.list_collection_aliases("{collection_name}").await?;

import io.qdrant.client.QdrantClient;
import io.qdrant.client.QdrantGrpcClient;

QdrantClient client =
    new QdrantClient(QdrantGrpcClient.newBuilder("localhost", 6334, false).build());

client.listCollectionAliasesAsync("{collection_name}").get();

using Qdrant.Client;

var client = new QdrantClient("localhost", 6334);

await client.ListCollectionAliasesAsync("{collection_name}");

List all aliases

GET /aliases

curl -X GET http://localhost:6333/aliases

from qdrant_client import QdrantClient

client = QdrantClient(url="http://localhost:6333")

client.get_aliases()

import { QdrantClient } from "@qdrant/js-client-rest";

const client = new QdrantClient({ host: "localhost", port: 6333 });

client.getAliases();

use qdrant_client::client::QdrantClient;

let client = QdrantClient::from_url("http://localhost:6334").build()?;

client.list_aliases().await?;

import io.qdrant.client.QdrantClient;
import io.qdrant.client.QdrantGrpcClient;

QdrantClient client =
    new QdrantClient(QdrantGrpcClient.newBuilder("localhost", 6334, false).build());

client.listAliasesAsync().get();

using Qdrant.Client;

var client = new QdrantClient("localhost", 6334);

await client.ListAliasesAsync();

List all collections

GET /collections

curl -X GET http://localhost:6333/collections

from qdrant_client import QdrantClient

client = QdrantClient(url="http://localhost:6333")

client.get_collections()

import { QdrantClient } from "@qdrant/js-client-rest";

const client = new QdrantClient({ host: "localhost", port: 6333 });

client.getCollections();

use qdrant_client::client::QdrantClient;

let client = QdrantClient::from_url("http://localhost:6334").build()?;

client.list_collections().await?;

import io.qdrant.client.QdrantClient;
import io.qdrant.client.QdrantGrpcClient;

QdrantClient client =
    new QdrantClient(QdrantGrpcClient.newBuilder("localhost", 6334, false).build());

client.listCollectionsAsync().get();

using Qdrant.Client;

var client = new QdrantClient("localhost", 6334);

await client.ListCollectionsAsync();