Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

$vectorSearch

New from version 8.0

Not supported by Elastic cluster.

The $vectorSearch operator in Amazon DocumentDB allows you to perform vector search, a method used in machine learning to find similar data points by comparing their vector representations using distance or similarity metrics. This capability combines the flexibility and rich querying of a JSON-based document database with the power of vector search, enabling you to build machine learning and generative AI use cases such as semantic search, product recommendation, and more.

Parameters

If omitted or set to false, numCandidates is required.

- `<index>` : Name of the Vector Search index to use.
- `<limit>` : Number of documents to return in the results.
- `<numCandidates>` (optional): This field is required if 'exact' is false or omitted. Number of nearest neighbors to use during the search. Value must be less than or equal to (<=) 10000. You can't specify a number less than the number of documents to return ('limit').
- `<path>` : Indexed vector type field to search.
- `<queryVector>` : Array of numbers that represent the query vector.

Example (MongoDB Shell)

The following example demonstrates how to use the $vectorSearch operator to find similar product descriptions based on their vector representations.

Create sample documents

db.products.insertMany([
  {
    _id: 1,
    name: "Product A",
    description: "A high-quality, eco-friendly product for your home.",
    description_vector: [ 0.2, 0.5, 0.8 ]
  },
  {
    _id: 2,
    name: "Product B",
    description: "An innovative and modern kitchen appliance.",
    description_vector: [0.7, 0.3, 0.9]
  },
  {
    _id: 3,
    name: "Product C",
    description: "A comfortable and stylish piece of furniture.",
    description_vector: [0.1, 0.2, 0.4]
  }
]);

Create vector search index

db.runCommand(
    {
        createIndexes: "products",
        indexes: [{
            key: {
                "description_vector": "vector"
            },
            vectorOptions: {
                type: "hnsw",
                dimensions: 3,
                similarity: "cosine",
                m: 16,
                efConstruction: 64
            },
            name: "description_index"
        }]
    }
);

Query example

db.products.aggregate([
  { $vectorSearch: {
      index: "description_index",
      limit: 2,
      numCandidates: 10,
      path: "description_vector",
      queryVector: [0.1, 0.2, 0.3]
    }
  }
]);

Output

[
  {
    "_id": 1,
    "name": "Product A",
    "description": "A high-quality, eco-friendly product for your home.",
    "description_vector": [ 0.2, 0.5, 0.8 ]
  },
  {
    "_id": 3,
    "name": "Product C",
    "description": "A comfortable and stylish piece of furniture.",
    "description_vector": [ 0.1, 0.2, 0.4 ]
  }
]

Code examples

To view a code example for using the $vectorSearch command, choose the tab for the language that you want to use:

Node.js

const { MongoClient } = require('mongodb');

async function findSimilarProducts(queryVector) {
  const client = await MongoClient.connect('mongodb://<username>:<password>@<cluster-endpoint>:27017/?tls=true&tlsCAFile=global-bundle.pem&replicaSet=rs0&readPreference=secondaryPreferred&retryWrites=false');
  const db = client.db('test');
  const collection = db.collection('products');

  const result = await collection.aggregate([
    { $vectorSearch: {
        index: "description_index",
        limit: 2,
        numCandidates: 10,
        path: "description_vector",
        queryVector: queryVector
      }
    }
  ]).toArray();

  console.log(result);
  client.close();
}

findSimilarProducts([0.1, 0.2, 0.3]);

Python

from pymongo import MongoClient


def find_similar_products(query_vector):
    client = MongoClient('mongodb://<username>:<password>@<cluster-endpoint>:27017/?tls=true&tlsCAFile=global-bundle.pem&replicaSet=rs0&readPreference=secondaryPreferred&retryWrites=false')
    db = client.test
    collection = db.products

    result = list(collection.aggregate([
        {
            '$vectorSearch': {
                'index': "description_index",
                'limit': 2,
                'numCandidates': 10,
                'path': "description_vector",
                'queryVector': query_vector
            }
        }
    ]))

    print(result)
    client.close()

find_similar_products([0.1, 0.2, 0.3])