积分
点是Qdrant操作的核心实体。 一个点是由一个向量和一个可选的负载组成的记录。
看起来是这样的：
// This is a simple point
{
"id": 129,
"vector": [0.1, 0.2, 0.3, 0.4],
"payload": {"color": "red"},
}
您可以在一个集合中基于向量相似性搜索分组点。 此过程在搜索和过滤部分中有更详细的描述。
本节解释了如何创建和管理向量。
任何点修改操作都是异步的，并且分两个步骤进行。 在第一阶段，操作被写入预写日志。
从这一刻起，即使机器断电，服务也不会丢失数据。
点ID
Qdrant 支持使用 64-bit unsigned integers 和 UUID 作为点的标识符。
UUID 字符串表示的示例：
- 简单:
936DA01F9ABD4d9d80C702AF85C822A8 - 带连字符的：
550e8400-e29b-41d4-a716-446655440000 - urn:
urn:uuid:F9168C5E-CEB2-4faa-B6BF-329BF39FA1E4
这意味着在每个请求中可以使用UUID字符串代替数字ID。 示例：
PUT /collections/{collection_name}/points
{
"points": [
{
"id": "5c56c793-69f3-4fbf-87e6-c4bf54c28c26",
"payload": {"color": "red"},
"vector": [0.9, 0.1, 0.1]
}
]
}
from qdrant_client import QdrantClient, models
client = QdrantClient(url="http://localhost:6333")
client.upsert(
collection_name="{collection_name}",
points=[
models.PointStruct(
id="5c56c793-69f3-4fbf-87e6-c4bf54c28c26",
payload={
"color": "red",
},
vector=[0.9, 0.1, 0.1],
),
],
)
import { QdrantClient } from "@qdrant/js-client-rest";
const client = new QdrantClient({ host: "localhost", port: 6333 });
client.upsert("{collection_name}", {
points: [
{
id: "5c56c793-69f3-4fbf-87e6-c4bf54c28c26",
payload: {
color: "red",
},
vector: [0.9, 0.1, 0.1],
},
],
});
use qdrant_client::qdrant::{PointStruct, UpsertPointsBuilder};
use qdrant_client::Qdrant;
let client = Qdrant::from_url("http://localhost:6334").build()?;
client
.upsert_points(
UpsertPointsBuilder::new(
"{collection_name}",
vec![PointStruct::new(
"5c56c793-69f3-4fbf-87e6-c4bf54c28c26",
vec![0.9, 0.1, 0.1],
[("color", "Red".into())],
)],
)
.wait(true),
)
.await?;
import java.util.List;
import java.util.Map;
import java.util.UUID;
import static io.qdrant.client.PointIdFactory.id;
import static io.qdrant.client.ValueFactory.value;
import static io.qdrant.client.VectorsFactory.vectors;
import io.qdrant.client.QdrantClient;
import io.qdrant.client.QdrantGrpcClient;
import io.qdrant.client.grpc.Points.PointStruct;
QdrantClient client =
new QdrantClient(QdrantGrpcClient.newBuilder("localhost", 6334, false).build());
client
.upsertAsync(
"{collection_name}",
List.of(
PointStruct.newBuilder()
.setId(id(UUID.fromString("5c56c793-69f3-4fbf-87e6-c4bf54c28c26")))
.setVectors(vectors(0.05f, 0.61f, 0.76f, 0.74f))
.putAllPayload(Map.of("color", value("Red")))
.build()))
.get();
using Qdrant.Client;
using Qdrant.Client.Grpc;
var client = new QdrantClient("localhost", 6334);
await client.UpsertAsync(
collectionName: "{collection_name}",
points: new List<PointStruct>
{
new()
{
Id = Guid.Parse("5c56c793-69f3-4fbf-87e6-c4bf54c28c26"),
Vectors = new[] { 0.05f, 0.61f, 0.76f, 0.74f },
Payload = { ["color"] = "Red" }
}
}
);
import (
"context"
"github.com/qdrant/go-client/qdrant"
)
client, err := qdrant.NewClient(&qdrant.Config{
Host: "localhost",
Port: 6334,
})
client.Upsert(context.Background(), &qdrant.UpsertPoints{
CollectionName: "{collection_name}",
Points: []*qdrant.PointStruct{
{
Id: qdrant.NewID("5c56c793-69f3-4fbf-87e6-c4bf54c28c26"),
Vectors: qdrant.NewVectors(0.05, 0.61, 0.76, 0.74),
Payload: qdrant.NewValueMap(map[string]any{"color": "Red"}),
},
},
})
和
PUT /collections/{collection_name}/points
{
"points": [
{
"id": 1,
"payload": {"color": "red"},
"vector": [0.9, 0.1, 0.1]
}
]
}
client.upsert(
collection_name="{collection_name}",
points=[
models.PointStruct(
id=1,
payload={
"color": "red",
},
vector=[0.9, 0.1, 0.1],
),
],
)
client.upsert("{collection_name}", {
points: [
{
id: 1,
payload: {
color: "red",
},
vector: [0.9, 0.1, 0.1],
},
],
});
use qdrant_client::qdrant::{PointStruct, UpsertPointsBuilder};
use qdrant_client::Qdrant;
let client = Qdrant::from_url("http://localhost:6334").build()?;
client
.upsert_points(
UpsertPointsBuilder::new(
"{collection_name}",
vec![PointStruct::new(
1,
vec![0.9, 0.1, 0.1],
[("color", "Red".into())],
)],
)
.wait(true),
)
.await?;
import java.util.List;
import java.util.Map;
import static io.qdrant.client.PointIdFactory.id;
import static io.qdrant.client.ValueFactory.value;
import static io.qdrant.client.VectorsFactory.vectors;
import io.qdrant.client.QdrantClient;
import io.qdrant.client.QdrantGrpcClient;
import io.qdrant.client.grpc.Points.PointStruct;
QdrantClient client =
new QdrantClient(QdrantGrpcClient.newBuilder("localhost", 6334, false).build());
client
.upsertAsync(
"{collection_name}",
List.of(
PointStruct.newBuilder()
.setId(id(1))
.setVectors(vectors(0.05f, 0.61f, 0.76f, 0.74f))
.putAllPayload(Map.of("color", value("Red")))
.build()))
.get();
using Qdrant.Client;
using Qdrant.Client.Grpc;
var client = new QdrantClient("localhost", 6334);
await client.UpsertAsync(
collectionName: "{collection_name}",
points: new List<PointStruct>
{
new()
{
Id = 1,
Vectors = new[] { 0.05f, 0.61f, 0.76f, 0.74f },
Payload = { ["color"] = "Red" }
}
}
);
import (
"context"
"github.com/qdrant/go-client/qdrant"
)
client, err := qdrant.NewClient(&qdrant.Config{
Host: "localhost",
Port: 6334,
})
client.Upsert(context.Background(), &qdrant.UpsertPoints{
CollectionName: "{collection_name}",
Points: []*qdrant.PointStruct{
{
Id: qdrant.NewIDNum(1),
Vectors: qdrant.NewVectors(0.05, 0.61, 0.76, 0.74),
Payload: qdrant.NewValueMap(map[string]any{"color": "Red"}),
},
},
})
两者都是可能的。
向量
qdrant中的每个点可能有一个或多个向量。 向量是Qdrant架构的核心组件， qdrant依赖不同类型的向量来提供不同类型的数据探索和搜索。
以下是支持的向量类型列表：
| 密集向量 | 由大多数嵌入模型生成的常规向量。 |
| 稀疏向量 | 没有固定长度，但只有少量非零元素的向量。 适用于精确的令牌匹配和协同过滤推荐。 |
| MultiVectors | 具有固定长度但可变高度的数字矩阵。 通常从像ColBERT这样的后期交互模型中获得。 |
可以将多种类型的向量附加到单个点上。 在Qdrant中，我们称这些为命名向量。
了解更多关于向量类型的信息，以及它们在向量部分中如何存储和优化的内容。
上传点
为了优化性能，Qdrant 支持批量加载点。也就是说，您可以在一个 API 调用中将多个点加载到服务中。 批处理可以最大限度地减少创建网络连接的开销。
Qdrant API 支持两种创建批次的方式 - 面向记录和面向列。 在内部，这些选项没有区别，只是为了交互的方便而设置。
批量创建点：
PUT /collections/{collection_name}/points
{
"batch": {
"ids": [1, 2, 3],
"payloads": [
{"color": "red"},
{"color": "green"},
{"color": "blue"}
],
"vectors": [
[0.9, 0.1, 0.1],
[0.1, 0.9, 0.1],
[0.1, 0.1, 0.9]
]
}
}
client.upsert(
collection_name="{collection_name}",
points=models.Batch(
ids=[1, 2, 3],
payloads=[
{"color": "red"},
{"color": "green"},
{"color": "blue"},
],
vectors=[
[0.9, 0.1, 0.1],
[0.1, 0.9, 0.1],
[0.1, 0.1, 0.9],
],
),
)
client.upsert("{collection_name}", {
batch: {
ids: [1, 2, 3],
payloads: [{ color: "red" }, { color: "green" }, { color: "blue" }],
vectors: [
[0.9, 0.1, 0.1],
[0.1, 0.9, 0.1],
[0.1, 0.1, 0.9],
],
},
});
或面向记录的等效内容：
PUT /collections/{collection_name}/points
{
"points": [
{
"id": 1,
"payload": {"color": "red"},
"vector": [0.9, 0.1, 0.1]
},
{
"id": 2,
"payload": {"color": "green"},
"vector": [0.1, 0.9, 0.1]
},
{
"id": 3,
"payload": {"color": "blue"},
"vector": [0.1, 0.1, 0.9]
}
]
}
client.upsert(
collection_name="{collection_name}",
points=[
models.PointStruct(
id=1,
payload={
"color": "red",
},
vector=[0.9, 0.1, 0.1],
),
models.PointStruct(
id=2,
payload={
"color": "green",
},
vector=[0.1, 0.9, 0.1],
),
models.PointStruct(
id=3,
payload={
"color": "blue",
},
vector=[0.1, 0.1, 0.9],
),
],
)
client.upsert("{collection_name}", {
points: [
{
id: 1,
payload: { color: "red" },
vector: [0.9, 0.1, 0.1],
},
{
id: 2,
payload: { color: "green" },
vector: [0.1, 0.9, 0.1],
},
{
id: 3,
payload: { color: "blue" },
vector: [0.1, 0.1, 0.9],
},
],
});
use qdrant_client::qdrant::{PointStruct, UpsertPointsBuilder};
client
.upsert_points(
UpsertPointsBuilder::new(
"{collection_name}",
vec![
PointStruct::new(1, vec![0.9, 0.1, 0.1], [("city", "red".into())]),
PointStruct::new(2, vec![0.1, 0.9, 0.1], [("city", "green".into())]),
PointStruct::new(3, vec![0.1, 0.1, 0.9], [("city", "blue".into())]),
],
)
.wait(true),
)
.await?;
import java.util.List;
import java.util.Map;
import static io.qdrant.client.PointIdFactory.id;
import static io.qdrant.client.ValueFactory.value;
import static io.qdrant.client.VectorsFactory.vectors;
import io.qdrant.client.QdrantClient;
import io.qdrant.client.QdrantGrpcClient;
import io.qdrant.client.grpc.Points.PointStruct;
QdrantClient client =
new QdrantClient(QdrantGrpcClient.newBuilder("localhost", 6334, false).build());
client
.upsertAsync(
"{collection_name}",
List.of(
PointStruct.newBuilder()
.setId(id(1))
.setVectors(vectors(0.9f, 0.1f, 0.1f))
.putAllPayload(Map.of("color", value("red")))
.build(),
PointStruct.newBuilder()
.setId(id(2))
.setVectors(vectors(0.1f, 0.9f, 0.1f))
.putAllPayload(Map.of("color", value("green")))
.build(),
PointStruct.newBuilder()
.setId(id(3))
.setVectors(vectors(0.1f, 0.1f, 0.9f))
.putAllPayload(Map.of("color", value("blue")))
.build()))
.get();
using Qdrant.Client;
using Qdrant.Client.Grpc;
var client = new QdrantClient("localhost", 6334);
await client.UpsertAsync(
collectionName: "{collection_name}",
points: new List<PointStruct>
{
new()
{
Id = 1,
Vectors = new[] { 0.9f, 0.1f, 0.1f },
Payload = { ["color"] = "red" }
},
new()
{
Id = 2,
Vectors = new[] { 0.1f, 0.9f, 0.1f },
Payload = { ["color"] = "green" }
},
new()
{
Id = 3,
Vectors = new[] { 0.1f, 0.1f, 0.9f },
Payload = { ["color"] = "blue" }
}
}
);
import (
"context"
"github.com/qdrant/go-client/qdrant"
)
client, err := qdrant.NewClient(&qdrant.Config{
Host: "localhost",
Port: 6334,
})
client.Upsert(context.Background(), &qdrant.UpsertPoints{
CollectionName: "{collection_name}",
Points: []*qdrant.PointStruct{
{
Id: qdrant.NewIDNum(1),
Vectors: qdrant.NewVectors(0.9, 0.1, 0.1),
Payload: qdrant.NewValueMap(map[string]any{"color": "red"}),
},
{
Id: qdrant.NewIDNum(2),
Vectors: qdrant.NewVectors(0.1, 0.9, 0.1),
Payload: qdrant.NewValueMap(map[string]any{"color": "green"}),
},
{
Id: qdrant.NewIDNum(3),
Vectors: qdrant.NewVectors(0.1, 0.1, 0.9),
Payload: qdrant.NewValueMap(map[string]any{"color": "blue"}),
},
},
})
Python客户端具有加载点的附加功能，其中包括：
- 并行化
- 重试机制
- 支持延迟批处理
例如，您可以直接从硬盘读取数据，以避免将所有数据存储在RAM中。您可以使用upload_collection和upload_points方法来实现这些功能。
与基本的upsert API类似，这些方法支持面向记录和面向列的格式。
列式格式：
client.upload_collection(
collection_name="{collection_name}",
ids=[1, 2],
payload=[
{"color": "red"},
{"color": "green"},
],
vectors=[
[0.9, 0.1, 0.1],
[0.1, 0.9, 0.1],
],
parallel=4,
max_retries=3,
)
面向记录的格式：
client.upload_points(
collection_name="{collection_name}",
points=[
models.PointStruct(
id=1,
payload={
"color": "red",
},
vector=[0.9, 0.1, 0.1],
),
models.PointStruct(
id=2,
payload={
"color": "green",
},
vector=[0.1, 0.9, 0.1],
),
],
parallel=4,
max_retries=3,
)
Qdrant中的所有API，包括点加载，都是幂等的。 这意味着连续多次执行相同的方法等同于单次执行。
在这种情况下，意味着具有相同id的点在重新上传时将被覆盖。
幂等性属性非常有用，例如，如果您使用不提供恰好一次保证的消息队列。 即使在这样的系统中，Qdrant也能确保数据一致性。
如果集合是使用多个向量创建的，每个向量数据可以使用向量的名称提供：
PUT /collections/{collection_name}/points
{
"points": [
{
"id": 1,
"vector": {
"image": [0.9, 0.1, 0.1, 0.2],
"text": [0.4, 0.7, 0.1, 0.8, 0.1, 0.1, 0.9, 0.2]
}
},
{
"id": 2,
"vector": {
"image": [0.2, 0.1, 0.3, 0.9],
"text": [0.5, 0.2, 0.7, 0.4, 0.7, 0.2, 0.3, 0.9]
}
}
]
}
client.upsert(
collection_name="{collection_name}",
points=[
models.PointStruct(
id=1,
vector={
"image": [0.9, 0.1, 0.1, 0.2],
"text": [0.4, 0.7, 0.1, 0.8, 0.1, 0.1, 0.9, 0.2],
},
),
models.PointStruct(
id=2,
vector={
"image": [0.2, 0.1, 0.3, 0.9],
"text": [0.5, 0.2, 0.7, 0.4, 0.7, 0.2, 0.3, 0.9],
},
),
],
)
client.upsert("{collection_name}", {
points: [
{
id: 1,
vector: {
image: [0.9, 0.1, 0.1, 0.2],
text: [0.4, 0.7, 0.1, 0.8, 0.1, 0.1, 0.9, 0.2],
},
},
{
id: 2,
vector: {
image: [0.2, 0.1, 0.3, 0.9],
text: [0.5, 0.2, 0.7, 0.4, 0.7, 0.2, 0.3, 0.9],
},
},
],
});
use std::collections::HashMap;
use qdrant_client::qdrant::{PointStruct, UpsertPointsBuilder};
use qdrant_client::Payload;
client
.upsert_points(
UpsertPointsBuilder::new(
"{collection_name}",
vec![
PointStruct::new(
1,
HashMap::from([
("image".to_string(), vec![0.9, 0.1, 0.1, 0.2]),
(
"text".to_string(),
vec![0.4, 0.7, 0.1, 0.8, 0.1, 0.1, 0.9, 0.2],
),
]),
Payload::default(),
),
PointStruct::new(
2,
HashMap::from([
("image".to_string(), vec![0.2, 0.1, 0.3, 0.9]),
(
"text".to_string(),
vec![0.5, 0.2, 0.7, 0.4, 0.7, 0.2, 0.3, 0.9],
),
]),
Payload::default(),
),
],
)
.wait(true),
)
.await?;
import java.util.List;
import java.util.Map;
import static io.qdrant.client.PointIdFactory.id;
import static io.qdrant.client.VectorFactory.vector;
import static io.qdrant.client.VectorsFactory.namedVectors;
import io.qdrant.client.grpc.Points.PointStruct;
client
.upsertAsync(
"{collection_name}",
List.of(
PointStruct.newBuilder()
.setId(id(1))
.setVectors(
namedVectors(
Map.of(
"image",
vector(List.of(0.9f, 0.1f, 0.1f, 0.2f)),
"text",
vector(List.of(0.4f, 0.7f, 0.1f, 0.8f, 0.1f, 0.1f, 0.9f, 0.2f)))))
.build(),
PointStruct.newBuilder()
.setId(id(2))
.setVectors(
namedVectors(
Map.of(
"image",
List.of(0.2f, 0.1f, 0.3f, 0.9f),
"text",
List.of(0.5f, 0.2f, 0.7f, 0.4f, 0.7f, 0.2f, 0.3f, 0.9f))))
.build()))
.get();
using Qdrant.Client;
using Qdrant.Client.Grpc;
var client = new QdrantClient("localhost", 6334);
await client.UpsertAsync(
collectionName: "{collection_name}",
points: new List<PointStruct>
{
new()
{
Id = 1,
Vectors = new Dictionary<string, float[]>
{
["image"] = [0.9f, 0.1f, 0.1f, 0.2f],
["text"] = [0.4f, 0.7f, 0.1f, 0.8f, 0.1f, 0.1f, 0.9f, 0.2f]
}
},
new()
{
Id = 2,
Vectors = new Dictionary<string, float[]>
{
["image"] = [0.2f, 0.1f, 0.3f, 0.9f],
["text"] = [0.5f, 0.2f, 0.7f, 0.4f, 0.7f, 0.2f, 0.3f, 0.9f]
}
}
}
);
import (
"context"
"github.com/qdrant/go-client/qdrant"
)
client, err := qdrant.NewClient(&qdrant.Config{
Host: "localhost",
Port: 6334,
})
client.Upsert(context.Background(), &qdrant.UpsertPoints{
CollectionName: "{collection_name}",
Points: []*qdrant.PointStruct{
{
Id: qdrant.NewIDNum(1),
Vectors: qdrant.NewVectorsMap(map[string]*qdrant.Vector{
"image": qdrant.NewVector(0.9, 0.1, 0.1, 0.2),
"text": qdrant.NewVector(0.4, 0.7, 0.1, 0.8, 0.1, 0.1, 0.9, 0.2),
}),
},
{
Id: qdrant.NewIDNum(2),
Vectors: qdrant.NewVectorsMap(map[string]*qdrant.Vector{
"image": qdrant.NewVector(0.2, 0.1, 0.3, 0.9),
"text": qdrant.NewVector(0.5, 0.2, 0.7, 0.4, 0.7, 0.2, 0.3, 0.9),
}),
},
},
})
自 v1.2.0 版本起可用
命名向量是可选的。在上传点时，可以省略一些向量。
例如，你可以上传一个只包含image向量的点，以及另一个只包含text向量的点。
当上传一个具有现有ID的点时，首先删除现有点，然后仅插入指定的向量。换句话说，整个点被替换，任何未指定的向量都被设置为null。要保留现有向量不变并仅更新指定的向量，请参阅更新向量。
自 v1.7.0 版本起可用
点可以包含密集和稀疏向量。
稀疏向量是一个数组，其中大多数元素的值为零。
可以利用这一特性来实现优化的表示，因此它们的形状与密集向量不同。
它们被表示为(index, value)对的列表，其中index是一个整数，value是一个浮点数。index是向量中非零值的位置。values是非零元素的值。
例如，以下向量：
[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 2.0, 0.0, 0.0]
可以表示为一个稀疏向量：
[(6, 1.0), (7, 2.0)]
Qdrant 在其所有 API 中使用以下 JSON 表示。
{
"indices": [6, 7],
"values": [1.0, 2.0]
}
indices 和 values 数组的长度必须相同。并且 indices 必须是唯一的。
如果indices没有排序，Qdrant 会在内部对它们进行排序，因此您可能无法依赖元素的顺序。
稀疏向量必须命名，并且可以像密集向量一样上传。
PUT /collections/{collection_name}/points
{
"points": [
{
"id": 1,
"vector": {
"text": {
"indices": [6, 7],
"values": [1.0, 2.0]
}
}
},
{
"id": 2,
"vector": {
"text": {
"indices": [1, 1, 2, 3, 4, 5],
"values": [0.1, 0.2, 0.3, 0.4, 0.5]
}
}
}
]
}
client.upsert(
collection_name="{collection_name}",
points=[
models.PointStruct(
id=1,
vector={
"text": models.SparseVector(
indices=[6, 7],
values=[1.0, 2.0],
)
},
),
models.PointStruct(
id=2,
vector={
"text": models.SparseVector(
indices=[1, 2, 3, 4, 5],
values=[0.1, 0.2, 0.3, 0.4, 0.5],
)
},
),
],
)
client.upsert("{collection_name}", {
points: [
{
id: 1,
vector: {
text: {
indices: [6, 7],
values: [1.0, 2.0],
},
},
},
{
id: 2,
vector: {
text: {
indices: [1, 2, 3, 4, 5],
values: [0.1, 0.2, 0.3, 0.4, 0.5],
},
},
},
],
});
use std::collections::HashMap;
use qdrant_client::qdrant::{PointStruct, UpsertPointsBuilder, Vector};
use qdrant_client::Payload;
client
.upsert_points(
UpsertPointsBuilder::new(
"{collection_name}",
vec![
PointStruct::new(
1,
HashMap::from([("text".to_string(), vec![(6, 1.0), (7, 2.0)])]),
Payload::default(),
),
PointStruct::new(
2,
HashMap::from([(
"text".to_string(),
vec![(1, 0.1), (2, 0.2), (3, 0.3), (4, 0.4), (5, 0.5)],
)]),
Payload::default(),
),
],
)
.wait(true),
)
.await?;
import java.util.List;
import java.util.Map;
import static io.qdrant.client.PointIdFactory.id;
import static io.qdrant.client.VectorFactory.vector;
import io.qdrant.client.grpc.Points.NamedVectors;
import io.qdrant.client.grpc.Points.PointStruct;
import io.qdrant.client.grpc.Points.Vectors;
client
.upsertAsync(
"{collection_name}",
List.of(
PointStruct.newBuilder()
.setId(id(1))
.setVectors(
Vectors.newBuilder()
.setVectors(
NamedVectors.newBuilder()
.putAllVectors(
Map.of(
"text", vector(List.of(1.0f, 2.0f), List.of(6, 7))))
.build())
.build())
.build(),
PointStruct.newBuilder()
.setId(id(2))
.setVectors(
Vectors.newBuilder()
.setVectors(
NamedVectors.newBuilder()
.putAllVectors(
Map.of(
"text",
vector(
List.of(0.1f, 0.2f, 0.3f, 0.4f, 0.5f),
List.of(1, 2, 3, 4, 5))))
.build())
.build())
.build()))
.get();
using Qdrant.Client;
using Qdrant.Client.Grpc;
var client = new QdrantClient("localhost", 6334);
await client.UpsertAsync(
collectionName: "{collection_name}",
points: new List<PointStruct>
{
new()
{
Id = 1,
Vectors = new Dictionary<string, Vector> { ["text"] = ([1.0f, 2.0f], [6, 7]) }
},
new()
{
Id = 2,
Vectors = new Dictionary<string, Vector>
{
["text"] = ([0.1f, 0.2f, 0.3f, 0.4f, 0.5f], [1, 2, 3, 4, 5])
}
}
}
);
import (
"context"
"github.com/qdrant/go-client/qdrant"
)
client, err := qdrant.NewClient(&qdrant.Config{
Host: "localhost",
Port: 6334,
})
client.Upsert(context.Background(), &qdrant.UpsertPoints{
CollectionName: "{collection_name}",
Points: []*qdrant.PointStruct{
{
Id: qdrant.NewIDNum(1),
Vectors: qdrant.NewVectorsMap(map[string]*qdrant.Vector{
"text": qdrant.NewVectorSparse(
[]uint32{6, 7},
[]float32{1.0, 2.0}),
}),
},
{
Id: qdrant.NewIDNum(2),
Vectors: qdrant.NewVectorsMap(map[string]*qdrant.Vector{
"text": qdrant.NewVectorSparse(
[]uint32{1, 2, 3, 4, 5},
[]float32{0.1, 0.2, 0.3, 0.4, 0.5}),
}),
},
},
})
修改点
要更改一个点，您可以修改其向量或其有效载荷。有几种方法可以做到这一点。
更新向量
自 v1.2.0 版本起可用
此方法更新给定点上的指定向量。未指定的向量保持不变。所有给定的点必须存在。
REST API (模式):
PUT /collections/{collection_name}/points/vectors
{
"points": [
{
"id": 1,
"vector": {
"image": [0.1, 0.2, 0.3, 0.4]
}
},
{
"id": 2,
"vector": {
"text": [0.9, 0.8, 0.7, 0.6, 0.5, 0.4, 0.3, 0.2]
}
}
]
}
client.update_vectors(
collection_name="{collection_name}",
points=[
models.PointVectors(
id=1,
vector={
"image": [0.1, 0.2, 0.3, 0.4],
},
),
models.PointVectors(
id=2,
vector={
"text": [0.9, 0.8, 0.7, 0.6, 0.5, 0.4, 0.3, 0.2],
},
),
],
)
client.updateVectors("{collection_name}", {
points: [
{
id: 1,
vector: {
image: [0.1, 0.2, 0.3, 0.4],
},
},
{
id: 2,
vector: {
text: [0.9, 0.8, 0.7, 0.6, 0.5, 0.4, 0.3, 0.2],
},
},
],
});
use std::collections::HashMap;
use qdrant_client::qdrant::{
PointVectors, UpdatePointVectorsBuilder,
};
client
.update_vectors(
UpdatePointVectorsBuilder::new(
"{collection_name}",
vec![
PointVectors {
id: Some(1.into()),
vectors: Some(
HashMap::from([("image".to_string(), vec![0.1, 0.2, 0.3, 0.4])]).into(),
),
},
PointVectors {
id: Some(2.into()),
vectors: Some(
HashMap::from([(
"text".to_string(),
vec![0.9, 0.8, 0.7, 0.6, 0.5, 0.4, 0.3, 0.2],
)])
.into(),
),
},
],
)
.wait(true),
)
.await?;
import java.util.List;
import java.util.Map;
import static io.qdrant.client.PointIdFactory.id;
import static io.qdrant.client.VectorFactory.vector;
import static io.qdrant.client.VectorsFactory.namedVectors;
client
.updateVectorsAsync(
"{collection_name}",
List.of(
PointVectors.newBuilder()
.setId(id(1))
.setVectors(namedVectors(Map.of("image", vector(List.of(0.1f, 0.2f, 0.3f, 0.4f)))))
.build(),
PointVectors.newBuilder()
.setId(id(2))
.setVectors(
namedVectors(
Map.of(
"text", vector(List.of(0.9f, 0.8f, 0.7f, 0.6f, 0.5f, 0.4f, 0.3f, 0.2f)))))
.build()))
.get();
using Qdrant.Client;
using Qdrant.Client.Grpc;
var client = new QdrantClient("localhost", 6334);
await client.UpdateVectorsAsync(
collectionName: "{collection_name}",
points: new List<PointVectors>
{
new() { Id = 1, Vectors = ("image", new float[] { 0.1f, 0.2f, 0.3f, 0.4f }) },
new()
{
Id = 2,
Vectors = ("text", new float[] { 0.9f, 0.8f, 0.7f, 0.6f, 0.5f, 0.4f, 0.3f, 0.2f })
}
}
);
import (
"context"
"github.com/qdrant/go-client/qdrant"
)
client, err := qdrant.NewClient(&qdrant.Config{
Host: "localhost",
Port: 6334,
})
client.UpdateVectors(context.Background(), &qdrant.UpdatePointVectors{
CollectionName: "{collection_name}",
Points: []*qdrant.PointVectors{
{
Id: qdrant.NewIDNum(1),
Vectors: qdrant.NewVectorsMap(map[string]*qdrant.Vector{
"image": qdrant.NewVector(0.1, 0.2, 0.3, 0.4),
}),
},
{
Id: qdrant.NewIDNum(2),
Vectors: qdrant.NewVectorsMap(map[string]*qdrant.Vector{
"text": qdrant.NewVector(0.9, 0.8, 0.7, 0.6, 0.5, 0.4, 0.3, 0.2),
}),
},
},
})
要更新点并替换其所有向量，请参阅上传点。
删除向量
自 v1.2.0 版本起可用
此方法仅从给定点中删除指定的向量。其他向量保持不变。点永远不会被删除。
REST API (模式):
POST /collections/{collection_name}/points/vectors/delete
{
"points": [0, 3, 100],
"vectors": ["text", "image"]
}
client.delete_vectors(
collection_name="{collection_name}",
points=[0, 3, 100],
vectors=["text", "image"],
)
client.deleteVectors("{collection_name}", {
points: [0, 3, 10],
vector: ["text", "image"],
});
use qdrant_client::qdrant::{
DeletePointVectorsBuilder, PointsIdsList,
};
client
.delete_vectors(
DeletePointVectorsBuilder::new("{collection_name}")
.points_selector(PointsIdsList {
ids: vec![0.into(), 3.into(), 10.into()],
})
.vectors(vec!["text".into(), "image".into()])
.wait(true),
)
.await?;
import java.util.List;
import static io.qdrant.client.PointIdFactory.id;
client
.deleteVectorsAsync(
"{collection_name}", List.of("text", "image"), List.of(id(0), id(3), id(10)))
.get();
await client.DeleteVectorsAsync("{collection_name}", ["text", "image"], [0, 3, 10]);
import (
"context"
"github.com/qdrant/go-client/qdrant"
)
client.DeleteVectors(context.Background(), &qdrant.DeletePointVectors{
CollectionName: "{collection_name}",
PointsSelector: qdrant.NewPointsSelector(
qdrant.NewIDNum(0), qdrant.NewIDNum(3), qdrant.NewIDNum(10)),
Vectors: &qdrant.VectorsSelector{
Names: []string{"text", "image"},
},
})
要删除整个点，请参阅删除点。
更新有效载荷
学习如何在负载部分修改点的负载。
删除点
REST API (模式):
POST /collections/{collection_name}/points/delete
{
"points": [0, 3, 100]
}
client.delete(
collection_name="{collection_name}",
points_selector=models.PointIdsList(
points=[0, 3, 100],
),
)
client.delete("{collection_name}", {
points: [0, 3, 100],
});
use qdrant_client::qdrant::{DeletePointsBuilder, PointsIdsList};
client
.delete_points(
DeletePointsBuilder::new("{collection_name}")
.points(PointsIdsList {
ids: vec![0.into(), 3.into(), 100.into()],
})
.wait(true),
)
.await?;
import java.util.List;
import static io.qdrant.client.PointIdFactory.id;
client.deleteAsync("{collection_name}", List.of(id(0), id(3), id(100)));
using Qdrant.Client;
var client = new QdrantClient("localhost", 6334);
await client.DeleteAsync(collectionName: "{collection_name}", ids: [0, 3, 100]);
import (
"context"
"github.com/qdrant/go-client/qdrant"
)
client, err := qdrant.NewClient(&qdrant.Config{
Host: "localhost",
Port: 6334,
})
client.Delete(context.Background(), &qdrant.DeletePoints{
CollectionName: "{collection_name}",
Points: qdrant.NewPointsSelector(
qdrant.NewIDNum(0), qdrant.NewIDNum(3), qdrant.NewIDNum(100),
),
})
指定要移除的点的另一种方法是使用过滤器。
POST /collections/{collection_name}/points/delete
{
"filter": {
"must": [
{
"key": "color",
"match": {
"value": "red"
}
}
]
}
}
client.delete(
collection_name="{collection_name}",
points_selector=models.FilterSelector(
filter=models.Filter(
must=[
models.FieldCondition(
key="color",
match=models.MatchValue(value="red"),
),
],
)
),
)
client.delete("{collection_name}", {
filter: {
must: [
{
key: "color",
match: {
value: "red",
},
},
],
},
});
use qdrant_client::qdrant::{Condition, DeletePointsBuilder, Filter};
client
.delete_points(
DeletePointsBuilder::new("{collection_name}")
.points(Filter::must([Condition::matches(
"color",
"red".to_string(),
)]))
.wait(true),
)
.await?;
import static io.qdrant.client.ConditionFactory.matchKeyword;
import io.qdrant.client.grpc.Points.Filter;
client
.deleteAsync(
"{collection_name}",
Filter.newBuilder().addMust(matchKeyword("color", "red")).build())
.get();
using Qdrant.Client;
using static Qdrant.Client.Grpc.Conditions;
var client = new QdrantClient("localhost", 6334);
await client.DeleteAsync(collectionName: "{collection_name}", filter: MatchKeyword("color", "red"));
import (
"context"
"github.com/qdrant/go-client/qdrant"
)
client, err := qdrant.NewClient(&qdrant.Config{
Host: "localhost",
Port: 6334,
})
client.Delete(context.Background(), &qdrant.DeletePoints{
CollectionName: "{collection_name}",
Points: qdrant.NewPointsSelectorFilter(
&qdrant.Filter{
Must: []*qdrant.Condition{
qdrant.NewMatch("color", "red"),
},
},
),
})
此示例从集合中移除所有具有{ "color": "red" }的点。
获取积分
有一个通过ID检索点的方法。
REST API (模式):
POST /collections/{collection_name}/points
{
"ids": [0, 3, 100]
}
client.retrieve(
collection_name="{collection_name}",
ids=[0, 3, 100],
)
client.retrieve("{collection_name}", {
ids: [0, 3, 100],
});
use qdrant_client::qdrant::GetPointsBuilder;
client
.get_points(GetPointsBuilder::new(
"{collection_name}",
vec![0.into(), 30.into(), 100.into()],
))
.await?;
import java.util.List;
import static io.qdrant.client.PointIdFactory.id;
client
.retrieveAsync("{collection_name}", List.of(id(0), id(30), id(100)), false, false, null)
.get();
using Qdrant.Client;
var client = new QdrantClient("localhost", 6334);
await client.RetrieveAsync(
collectionName: "{collection_name}",
ids: [0, 30, 100],
withPayload: false,
withVectors: false
);
import (
"context"
"github.com/qdrant/go-client/qdrant"
)
client, err := qdrant.NewClient(&qdrant.Config{
Host: "localhost",
Port: 6334,
})
client.Get(context.Background(), &qdrant.GetPoints{
CollectionName: "{collection_name}",
Ids: []*qdrant.PointId{
qdrant.NewIDNum(0), qdrant.NewIDNum(3), qdrant.NewIDNum(100),
},
})
此方法有额外的参数 with_vectors 和 with_payload。
使用这些参数，您可以选择您想要的部分点作为结果。
排除这些部分可以帮助您不浪费流量传输无用的数据。
单个点也可以通过API获取：
REST API (模式):
GET /collections/{collection_name}/points/{point_id}
滚动点
有时可能需要在不了解id的情况下获取所有存储的点，或者遍历符合过滤器的点。
REST API (模式):
POST /collections/{collection_name}/points/scroll
{
"filter": {
"must": [
{
"key": "color",
"match": {
"value": "red"
}
}
]
},
"limit": 1,
"with_payload": true,
"with_vector": false
}
client.scroll(
collection_name="{collection_name}",
scroll_filter=models.Filter(
must=[
models.FieldCondition(key="color", match=models.MatchValue(value="red")),
]
),
limit=1,
with_payload=True,
with_vectors=False,
)
client.scroll("{collection_name}", {
filter: {
must: [
{
key: "color",
match: {
value: "red",
},
},
],
},
limit: 1,
with_payload: true,
with_vector: false,
});
use qdrant_client::qdrant::{Condition, Filter, ScrollPointsBuilder};
client
.scroll(
ScrollPointsBuilder::new("{collection_name}")
.filter(Filter::must([Condition::matches(
"color",
"red".to_string(),
)]))
.limit(1)
.with_payload(true)
.with_vectors(false),
)
.await?;
import static io.qdrant.client.ConditionFactory.matchKeyword;
import static io.qdrant.client.WithPayloadSelectorFactory.enable;
import io.qdrant.client.grpc.Points.Filter;
import io.qdrant.client.grpc.Points.ScrollPoints;
client
.scrollAsync(
ScrollPoints.newBuilder()
.setCollectionName("{collection_name}")
.setFilter(Filter.newBuilder().addMust(matchKeyword("color", "red")).build())
.setLimit(1)
.setWithPayload(enable(true))
.build())
.get();
using Qdrant.Client;
using static Qdrant.Client.Grpc.Conditions;
var client = new QdrantClient("localhost", 6334);
await client.ScrollAsync(
collectionName: "{collection_name}",
filter: MatchKeyword("color", "red"),
limit: 1,
payloadSelector: true
);
import (
"context"
"github.com/qdrant/go-client/qdrant"
)
client, err := qdrant.NewClient(&qdrant.Config{
Host: "localhost",
Port: 6334,
})
client.Scroll(context.Background(), &qdrant.ScrollPoints{
CollectionName: "{collection_name}",
Filter: &qdrant.Filter{
Must: []*qdrant.Condition{
qdrant.NewMatch("color", "red"),
},
},
Limit: qdrant.PtrOf(uint32(1)),
WithPayload: qdrant.NewWithPayload(true),
})
返回所有color = red的点。
{
"result": {
"next_page_offset": 1,
"points": [
{
"id": 0,
"payload": {
"color": "red"
}
}
]
},
"status": "ok",
"time": 0.0001
}
Scroll API 将以逐页方式返回所有匹配过滤器的点。
所有结果点按ID排序。要查询下一页，需要在offset字段中指定已看到的最大ID。
为了方便，这个ID也会在next_page_offset字段中返回。
如果next_page_offset字段的值为null - 则表示已到达最后一页。
按有效载荷键排序点
自 v1.8.0 版本起可用
使用scroll API时，您可以按有效载荷键对结果进行排序。例如，如果您的有效载荷具有"timestamp"字段，您可以按时间顺序检索点，如下例所示：
POST /collections/{collection_name}/points/scroll
{
"limit": 15,
"order_by": "timestamp", // <-- this!
}
client.scroll(
collection_name="{collection_name}",
limit=15,
order_by="timestamp", # <-- this!
)
client.scroll("{collection_name}", {
limit: 15,
order_by: "timestamp", // <-- this!
});
use qdrant_client::qdrant::{OrderByBuilder, ScrollPointsBuilder};
client
.scroll(
ScrollPointsBuilder::new("{collection_name}")
.limit(15)
.order_by(OrderByBuilder::new("timestamp")),
)
.await?;
import io.qdrant.client.grpc.Points.OrderBy;
import io.qdrant.client.grpc.Points.ScrollPoints;
client.scrollAsync(ScrollPoints.newBuilder()
.setCollectionName("{collection_name}")
.setLimit(15)
.setOrderBy(OrderBy.newBuilder().setKey("timestamp").build())
.build()).get();
await client.ScrollAsync("{collection_name}", limit: 15, orderBy: "timestamp");
import (
"context"
"github.com/qdrant/go-client/qdrant"
)
client, err := qdrant.NewClient(&qdrant.Config{
Host: "localhost",
Port: 6334,
})
client.Scroll(context.Background(), &qdrant.ScrollPoints{
CollectionName: "{collection_name}",
Limit: qdrant.PtrOf(uint32(15)),
OrderBy: &qdrant.OrderBy{
Key: "timestamp",
},
})
你需要使用order_by key参数来指定有效载荷键。然后你可以添加其他字段来控制排序，例如direction和start_from：
"order_by": {
"key": "timestamp",
"direction": "desc" // default is "asc"
"start_from": 123, // start from this value
}
order_by=models.OrderBy(
key="timestamp",
direction="desc", # default is "asc"
start_from=123, # start from this value
)
order_by: {
key: "timestamp",
direction: "desc", // default is "asc"
start_from: 123, // start from this value
}
use qdrant_client::qdrant::{start_from::Value, Direction, OrderByBuilder};
OrderByBuilder::new("timestamp")
.direction(Direction::Desc.into())
.start_from(Value::Integer(123))
.build();
import io.qdrant.client.grpc.Points.Direction;
import io.qdrant.client.grpc.Points.OrderBy;
import io.qdrant.client.grpc.Points.StartFrom;
OrderBy.newBuilder()
.setKey("timestamp")
.setDirection(Direction.Desc)
.setStartFrom(StartFrom.newBuilder()
.setInteger(123)
.build())
.build();
using Qdrant.Client.Grpc;
new OrderBy
{
Key = "timestamp",
Direction = Direction.Desc,
StartFrom = 123
};
import "github.com/qdrant/go-client/qdrant"
qdrant.OrderBy{
Key: "timestamp",
Direction: qdrant.Direction_Desc.Enum(),
StartFrom: qdrant.NewStartFromInt(123),
}
当排序基于非唯一值时，无法依赖ID偏移量。因此，响应中不会返回next_page_offset。然而，您仍然可以通过将"order_by": { "start_from": ... }与{ "must_not": [{ "has_id": [...] }] }过滤器结合来进行分页。
计算点数
自 v0.8.4 版本起可用
有时，了解有多少点符合过滤条件而不进行实际搜索可能很有用。
例如，我们可以强调以下场景：
- 分面搜索结果大小的评估
- 确定分页的页数
- 调试查询执行速度
REST API (模式):
POST /collections/{collection_name}/points/count
{
"filter": {
"must": [
{
"key": "color",
"match": {
"value": "red"
}
}
]
},
"exact": true
}
client.count(
collection_name="{collection_name}",
count_filter=models.Filter(
must=[
models.FieldCondition(key="color", match=models.MatchValue(value="red")),
]
),
exact=True,
)
client.count("{collection_name}", {
filter: {
must: [
{
key: "color",
match: {
value: "red",
},
},
],
},
exact: true,
});
use qdrant_client::qdrant::{Condition, CountPointsBuilder, Filter};
client
.count(
CountPointsBuilder::new("{collection_name}")
.filter(Filter::must([Condition::matches(
"color",
"red".to_string(),
)]))
.exact(true),
)
.await?;
import static io.qdrant.client.ConditionFactory.matchKeyword;
import io.qdrant.client.grpc.Points.Filter;
client
.countAsync(
"{collection_name}",
Filter.newBuilder().addMust(matchKeyword("color", "red")).build(),
true)
.get();
using Qdrant.Client;
using static Qdrant.Client.Grpc.Conditions;
var client = new QdrantClient("localhost", 6334);
await client.CountAsync(
collectionName: "{collection_name}",
filter: MatchKeyword("color", "red"),
exact: true
);
import (
"context"
"github.com/qdrant/go-client/qdrant"
)
client, err := qdrant.NewClient(&qdrant.Config{
Host: "localhost",
Port: 6334,
})
client.Count(context.Background(), &qdrant.CountPoints{
CollectionName: "midlib",
Filter: &qdrant.Filter{
Must: []*qdrant.Condition{
qdrant.NewMatch("color", "red"),
},
},
})
返回符合给定过滤条件的计数数量：
{
"count": 3811
}
批量更新
自 v1.5.0 版本起可用
您可以批量执行多个点更新操作。这包括插入、更新和删除点、向量和有效载荷。
批量更新请求由一系列操作组成。这些操作按顺序执行。这些操作可以批量处理：
- 更新插入分数:
upsert或UpsertOperation - 删除点:
delete_points或DeleteOperation - 更新向量:
update_vectors或UpdateVectorsOperation - 删除向量:
delete_vectors或DeleteVectorsOperation - 设置有效载荷:
set_payload或SetPayloadOperation - 覆盖有效载荷:
overwrite_payload或OverwritePayload - 删除有效载荷:
delete_payload或DeletePayloadOperation - 清除有效载荷:
clear_payload或ClearPayloadOperation
以下示例片段使用了所有操作。
REST API (模式):
POST /collections/{collection_name}/points/batch
{
"operations": [
{
"upsert": {
"points": [
{
"id": 1,
"vector": [1.0, 2.0, 3.0, 4.0],
"payload": {}
}
]
}
},
{
"update_vectors": {
"points": [
{
"id": 1,
"vector": [1.0, 2.0, 3.0, 4.0]
}
]
}
},
{
"delete_vectors": {
"points": [1],
"vector": [""]
}
},
{
"overwrite_payload": {
"payload": {
"test_payload": "1"
},
"points": [1]
}
},
{
"set_payload": {
"payload": {
"test_payload_2": "2",
"test_payload_3": "3"
},
"points": [1]
}
},
{
"delete_payload": {
"keys": ["test_payload_2"],
"points": [1]
}
},
{
"clear_payload": {
"points": [1]
}
},
{"delete": {"points": [1]}}
]
}
client.batch_update_points(
collection_name="{collection_name}",
update_operations=[
models.UpsertOperation(
upsert=models.PointsList(
points=[
models.PointStruct(
id=1,
vector=[1.0, 2.0, 3.0, 4.0],
payload={},
),
]
)
),
models.UpdateVectorsOperation(
update_vectors=models.UpdateVectors(
points=[
models.PointVectors(
id=1,
vector=[1.0, 2.0, 3.0, 4.0],
)
]
)
),
models.DeleteVectorsOperation(
delete_vectors=models.DeleteVectors(points=[1], vector=[""])
),
models.OverwritePayloadOperation(
overwrite_payload=models.SetPayload(
payload={"test_payload": 1},
points=[1],
)
),
models.SetPayloadOperation(
set_payload=models.SetPayload(
payload={
"test_payload_2": 2,
"test_payload_3": 3,
},
points=[1],
)
),
models.DeletePayloadOperation(
delete_payload=models.DeletePayload(keys=["test_payload_2"], points=[1])
),
models.ClearPayloadOperation(clear_payload=models.PointIdsList(points=[1])),
models.DeleteOperation(delete=models.PointIdsList(points=[1])),
],
)
client.batchUpdate("{collection_name}", {
operations: [
{
upsert: {
points: [
{
id: 1,
vector: [1.0, 2.0, 3.0, 4.0],
payload: {},
},
],
},
},
{
update_vectors: {
points: [
{
id: 1,
vector: [1.0, 2.0, 3.0, 4.0],
},
],
},
},
{
delete_vectors: {
points: [1],
vector: [""],
},
},
{
overwrite_payload: {
payload: {
test_payload: 1,
},
points: [1],
},
},
{
set_payload: {
payload: {
test_payload_2: 2,
test_payload_3: 3,
},
points: [1],
},
},
{
delete_payload: {
keys: ["test_payload_2"],
points: [1],
},
},
{
clear_payload: {
points: [1],
},
},
{
delete: {
points: [1],
},
},
],
});
use std::collections::HashMap;
use qdrant_client::qdrant::{
points_update_operation::{
ClearPayload, DeletePayload, DeletePoints, DeleteVectors, Operation, OverwritePayload,
PointStructList, SetPayload, UpdateVectors,
},
PointStruct, PointVectors, PointsUpdateOperation, UpdateBatchPointsBuilder, VectorsSelector,
};
use qdrant_client::Payload;
client
.update_points_batch(
UpdateBatchPointsBuilder::new(
"{collection_name}",
vec![
PointsUpdateOperation {
operation: Some(Operation::Upsert(PointStructList {
points: vec![PointStruct::new(
1,
vec![1.0, 2.0, 3.0, 4.0],
Payload::default(),
)],
..Default::default()
})),
},
PointsUpdateOperation {
operation: Some(Operation::UpdateVectors(UpdateVectors {
points: vec![PointVectors {
id: Some(1.into()),
vectors: Some(vec![1.0, 2.0, 3.0, 4.0].into()),
}],
..Default::default()
})),
},
PointsUpdateOperation {
operation: Some(Operation::DeleteVectors(DeleteVectors {
points_selector: Some(vec![1.into()].into()),
vectors: Some(VectorsSelector {
names: vec!["".into()],
}),
..Default::default()
})),
},
PointsUpdateOperation {
operation: Some(Operation::OverwritePayload(OverwritePayload {
points_selector: Some(vec![1.into()].into()),
payload: HashMap::from([("test_payload".to_string(), 1.into())]),
..Default::default()
})),
},
PointsUpdateOperation {
operation: Some(Operation::SetPayload(SetPayload {
points_selector: Some(vec![1.into()].into()),
payload: HashMap::from([
("test_payload_2".to_string(), 2.into()),
("test_payload_3".to_string(), 3.into()),
]),
..Default::default()
})),
},
PointsUpdateOperation {
operation: Some(Operation::DeletePayload(DeletePayload {
points_selector: Some(vec![1.into()].into()),
keys: vec!["test_payload_2".to_string()],
..Default::default()
})),
},
PointsUpdateOperation {
operation: Some(Operation::ClearPayload(ClearPayload {
points: Some(vec![1.into()].into()),
..Default::default()
})),
},
PointsUpdateOperation {
operation: Some(Operation::DeletePoints(DeletePoints {
points: Some(vec![1.into()].into()),
..Default::default()
})),
},
],
)
.wait(true),
)
.await?;
import java.util.List;
import java.util.Map;
import static io.qdrant.client.PointIdFactory.id;
import static io.qdrant.client.ValueFactory.value;
import static io.qdrant.client.VectorsFactory.vectors;
import io.qdrant.client.grpc.Points.PointStruct;
import io.qdrant.client.grpc.Points.PointVectors;
import io.qdrant.client.grpc.Points.PointsIdsList;
import io.qdrant.client.grpc.Points.PointsSelector;
import io.qdrant.client.grpc.Points.PointsUpdateOperation;
import io.qdrant.client.grpc.Points.PointsUpdateOperation.ClearPayload;
import io.qdrant.client.grpc.Points.PointsUpdateOperation.DeletePayload;
import io.qdrant.client.grpc.Points.PointsUpdateOperation.DeletePoints;
import io.qdrant.client.grpc.Points.PointsUpdateOperation.DeleteVectors;
import io.qdrant.client.grpc.Points.PointsUpdateOperation.PointStructList;
import io.qdrant.client.grpc.Points.PointsUpdateOperation.SetPayload;
import io.qdrant.client.grpc.Points.PointsUpdateOperation.UpdateVectors;
import io.qdrant.client.grpc.Points.VectorsSelector;
client
.batchUpdateAsync(
"{collection_name}",
List.of(
PointsUpdateOperation.newBuilder()
.setUpsert(
PointStructList.newBuilder()
.addPoints(
PointStruct.newBuilder()
.setId(id(1))
.setVectors(vectors(1.0f, 2.0f, 3.0f, 4.0f))
.build())
.build())
.build(),
PointsUpdateOperation.newBuilder()
.setUpdateVectors(
UpdateVectors.newBuilder()
.addPoints(
PointVectors.newBuilder()
.setId(id(1))
.setVectors(vectors(1.0f, 2.0f, 3.0f, 4.0f))
.build())
.build())
.build(),
PointsUpdateOperation.newBuilder()
.setDeleteVectors(
DeleteVectors.newBuilder()
.setPointsSelector(
PointsSelector.newBuilder()
.setPoints(PointsIdsList.newBuilder().addIds(id(1)).build())
.build())
.setVectors(VectorsSelector.newBuilder().addNames("").build())
.build())
.build(),
PointsUpdateOperation.newBuilder()
.setOverwritePayload(
SetPayload.newBuilder()
.setPointsSelector(
PointsSelector.newBuilder()
.setPoints(PointsIdsList.newBuilder().addIds(id(1)).build())
.build())
.putAllPayload(Map.of("test_payload", value(1)))
.build())
.build(),
PointsUpdateOperation.newBuilder()
.setSetPayload(
SetPayload.newBuilder()
.setPointsSelector(
PointsSelector.newBuilder()
.setPoints(PointsIdsList.newBuilder().addIds(id(1)).build())
.build())
.putAllPayload(
Map.of("test_payload_2", value(2), "test_payload_3", value(3)))
.build())
.build(),
PointsUpdateOperation.newBuilder()
.setDeletePayload(
DeletePayload.newBuilder()
.setPointsSelector(
PointsSelector.newBuilder()
.setPoints(PointsIdsList.newBuilder().addIds(id(1)).build())
.build())
.addKeys("test_payload_2")
.build())
.build(),
PointsUpdateOperation.newBuilder()
.setClearPayload(
ClearPayload.newBuilder()
.setPoints(
PointsSelector.newBuilder()
.setPoints(PointsIdsList.newBuilder().addIds(id(1)).build())
.build())
.build())
.build(),
PointsUpdateOperation.newBuilder()
.setDeletePoints(
DeletePoints.newBuilder()
.setPoints(
PointsSelector.newBuilder()
.setPoints(PointsIdsList.newBuilder().addIds(id(1)).build())
.build())
.build())
.build()))
.get();
要使用单一操作类型批量处理多个点，请直接在该操作中使用批处理功能。
等待结果
如果API调用时使用了&wait=false参数，或者没有明确指定该参数，客户端将收到接收数据的确认：
{
"result": {
"operation_id": 123,
"status": "acknowledged"
},
"status": "ok",
"time": 0.000206061
}
此响应并不意味着数据已经可以检索。这使用了一种最终一致性形式。在实际处理之前，可能需要很短的时间，因为更新集合是在后台进行的。事实上，这样的请求最终可能会失败。如果要插入大量向量，我们还建议使用异步请求以利用流水线。
如果您的应用程序逻辑要求保证在API响应后立即可以搜索向量，则使用标志?wait=true。
在这种情况下，API只会在操作完成后返回结果：
{
"result": {
"operation_id": 0,
"status": "completed"
},
"status": "ok",
"time": 0.000206061
}