milvus-logo
LFAI
< Docs
  • Node

search()

This method performs a vector similarity search.

You must load the collection before searching or querying.

Example

import { MilvusClient, DataType, MetricType } from "@zilliz/milvus2-sdk-node";

// simple search example
new milvusClient(MILUVS_ADDRESS).search({
  collection_name: "my_collection",
  vector: [1, 2, 3, 4],
});

// complex search example
new milvusClient(MILUVS_ADDRESS).search({
  collection_name: "my_collection",
  vector: [1, 2, 3, 4],
  filter: "count > 5",
  limit: 10,
  offset: 2,
  metric_type: MetricType.L2,
  param: { nprobe: 1024 },
  consistency_level: ConsistencyLevelEnum.Strong,
  output_fields: ["your field defined in schema"],
});

// batch search example
new milvusClient(MILUVS_ADDRESS).search({
  collection_name: "my_collection",
  vectors: [
    [1, 2, 3, 4],
    [5, 6, 7, 8],
  ],
});

Response

Single search response

{
  status: { error_code: 'Success', reason: '' },
  results: [
    { score: 22, age: '434848878802251176' },
    { score: 22, age: '434848878802251181' },
    { score: 23, age: '434848878802251173' },
    { score: 25, age: '434848878802251179' }
  ]
}

Batch search response

{
  status: { error_code: 'Success', reason: '' },
  results: [
    [
    { score: 22, age: '434848878802251176' },
    { score: 22, age: '434848878802251181' },
    { score: 23, age: '434848878802251173' },
    { score: 25, age: '434848878802251179' }
   ],
   [
    { score: 22, age: '434848878802251176' },
    { score: 22, age: '434848878802251181' },
    { score: 23, age: '434848878802251173' },
    { score: 25, age: '434848878802251179' }
    ],
  ]
}

Parameters

ParametersDescriptionType
collection_nameName of the collection to search onString
vector? or vectors?vector or vectors to searchNumber[] or Number[][]
limit? or topk?Topk, by default: 10Number
offset?offset, by default: 0Number
output_fields?Vector or scalar field to be returned, by default, we will output all fields in the collection,String[]
partitions_names?An array of the names of the partitions to search onString[]
metric_type?similarity metric, by default, it is 'L2'String
filter? or expr?Boolean expression to filter the dataString
params?Optional search param, it depends on vector index, for example {nprobe: 1024} ,Please refer to https://milvus.io/docs/index.md.Object
consistency_level?Consistency in a distributed database specifically refers to the property that ensures every node or replica has the same view of data when writing or reading data at a given time.Enum
timeout?This parameter is used to specify the length of time, in milliseconds, that the RPC (Remote Procedure Call) is allowed to run. If no value is provided, the default is undefined.Number
Feedback

Was this page helpful?