Loading

Elasticsearch vector and full-text search in 10 minutes

Elasticsearch is a search and vector engine that supports full-text, semantic, and hybrid search over your data. This guide uses official Elasticsearch clients to take you from an empty Vector Database project to real search results in about 10 minutes.

Use it if you're new to search in Elasticsearch or want to experiment with meaning-based retrieval alongside keyword matching. By the end of the tutorial, you:

  • Create a Vector Database project and connect an official Elasticsearch client.
  • Create an index and add sample data.
  • Run semantic and hybrid searches and aggregate the data with ES|QL.
Tip

Are you an agent? Use the elasticsearch-onboarding skill.

Create a free Elasticsearch Vector Database project. It's serverless and built for search and vector workloads, so you don't need to size or manage a cluster.

The project takes about a minute to start. The Getting started page then displays the Project endpoint and a generated API key. Copy both values.

Set the project endpoint and API key as environment variables:

export ES_URL="https://YOUR-PROJECT.es.REGION.aws.elastic.cloud:443"
export ES_API_KEY="YOUR_API_KEY"
		
$Env:ES_URL = "https://YOUR-PROJECT.es.REGION.aws.elastic.cloud:443"
$Env:ES_API_KEY = "YOUR_API_KEY"
		
Important

Never hard-code credentials or commit them to source control. For local development, store the values in a .env file, configure your application to load it, and add .env to your .gitignore file.

Install the client for your language. For a list of available clients, refer to Elasticsearch clients.

pip install elasticsearch
		

For supported Python versions and other requirements, refer to the Python client documentation.

npm init --yes
npm pkg set type=module
npm install @elastic/elasticsearch
npm install --save-dev typescript tsx
		

For supported Node.js versions and other requirements, refer to the JavaScript client documentation.

composer require elasticsearch/elasticsearch
		

For supported PHP versions and other requirements, refer to the PHP client documentation.

gem install elasticsearch
		

For supported Ruby versions and other requirements, refer to the Ruby client installation documentation.

dotnet add package Elastic.Clients.Elasticsearch
		

For supported .NET versions and other requirements, refer to the .NET client documentation.

dependencies {
    implementation "co.elastic.clients:elasticsearch-java:VERSION"
}
		

Replace VERSION with the version from the latest Java client release. For supported Java versions and other requirements, refer to the Java client documentation.

go get github.com/elastic/go-elasticsearch/v9
		

For supported Go versions and other requirements, refer to the Go client installation documentation.

Use the environment variables to create the client and check the connection:

import os
from elasticsearch import Elasticsearch, helpers

es = Elasticsearch(
    os.environ["ES_URL"],
    api_key=os.environ["ES_API_KEY"],
)

print(es.info())
		
import { Client } from "@elastic/elasticsearch";

const es = new Client({
  node: process.env.ES_URL,
  auth: { apiKey: process.env.ES_API_KEY! },
});

console.log(await es.info());
		
<?php

require __DIR__ . "/vendor/autoload.php";

use Elastic\Elasticsearch\ClientBuilder;

$es = ClientBuilder::create()
    ->setHosts([getenv("ES_URL")])
    ->setApiKey(getenv("ES_API_KEY"))
    ->build();

print_r($es->info()->asArray());
		
require "elasticsearch"

es = Elasticsearch::Client.new(
  url: ENV.fetch("ES_URL"),
  api_key: ENV.fetch("ES_API_KEY")
)

puts es.info
		
using System.Text.Json;
using System.Text.Json.Serialization;
using Elastic.Clients.Elasticsearch;
using Elastic.Clients.Elasticsearch.Core.Bulk;
using Elastic.Clients.Elasticsearch.Esql;
using Elastic.Transport;

var url = Environment.GetEnvironmentVariable("ES_URL")
    ?? throw new InvalidOperationException("ES_URL is not set.");
var apiKey = Environment.GetEnvironmentVariable("ES_API_KEY")
    ?? throw new InvalidOperationException("ES_API_KEY is not set.");

var settings = new ElasticsearchClientSettings(new Uri(url))
    .Authentication(new ApiKey(apiKey));
var es = new ElasticsearchClient(settings);

var response = await es.InfoAsync();
Console.WriteLine(response);
		
import java.io.BufferedReader;
import java.io.InputStreamReader;
import java.io.StringReader;

import co.elastic.clients.elasticsearch.ElasticsearchClient;
import co.elastic.clients.elasticsearch._types.Refresh;
import co.elastic.clients.elasticsearch.core.BulkRequest;
import co.elastic.clients.elasticsearch.core.SearchResponse;
import co.elastic.clients.elasticsearch.core.search.Hit;
import co.elastic.clients.elasticsearch.esql.EsqlFormat;
import co.elastic.clients.transport.endpoints.BinaryResponse;

String serverUrl = System.getenv("ES_URL");
String apiKey = System.getenv("ES_API_KEY");

ElasticsearchClient es = ElasticsearchClient.of(b -> b
    .host(serverUrl)
    .apiKey(apiKey)
);

System.out.println(es.info());
		
package main

import (
	"bytes"
	"context"
	"encoding/csv"
	"encoding/json"
	"fmt"
	"io"
	"log"
	"os"
	"strings"

	"github.com/elastic/go-elasticsearch/v9"
	"github.com/elastic/go-elasticsearch/v9/esutil"
	"github.com/elastic/go-elasticsearch/v9/typedapi/types/enums/esqlformat"
)

func main() {
	es, err := elasticsearch.New(
		elasticsearch.WithAddresses(os.Getenv("ES_URL")),
		elasticsearch.WithAPIKey(os.Getenv("ES_API_KEY")),
	)
	if err != nil {
		log.Fatal(err)
	}

	typed, err := elasticsearch.NewTyped(
		elasticsearch.WithAddresses(os.Getenv("ES_URL")),
		elasticsearch.WithAPIKey(os.Getenv("ES_API_KEY")),
	)
	if err != nil {
		log.Fatal(err)
	}

	res, err := es.Info()
	if err != nil {
		log.Fatal(err)
	}
	defer res.Body.Close()
	if res.IsError() {
		log.Fatal(res)
	}

	if _, err := io.Copy(os.Stdout, res.Body); err != nil {
		log.Fatal(err)
	}
}
		

Create the books index, then add sample data.

Elasticsearch uses dynamic mapping to determine field types from the first documents you index. The only field you need to define is description, which you set to semantic_text before indexing so you can search it by meaning.

Note

semantic_text automatically embeds text at ingest time using an inference endpoint. The inference endpoint connects to an embedding model that converts indexed text and search queries into vectors. When you search the field, Elasticsearch embeds your query the same way and matches on meaning. For example, "space rescue" can match "stranded astronaut" with no words in common.

es.indices.create(
    index="books",
    mappings={
        "properties": {
            "description": {
                "type": "semantic_text",
            }
        }
    },
)
		
  1. In this example, semantic_text uses a default inference endpoint. The model used by this endpoint is multilingual, so you can index and search text in multiple languages. To use custom models, refer to Configure inference endpoints for semantic_text.
await es.indices.create({
  index: "books",
  mappings: {
    properties: {
      description: {
        type: "semantic_text",
      },
    },
  },
});
		
  1. In this example, semantic_text uses a default inference endpoint. The model used by this endpoint is multilingual, so you can index and search text in multiple languages. To use custom models, refer to Configure inference endpoints for semantic_text.
$es->indices()->create([
    "index" => "books",
    "body" => [
        "mappings" => [
            "properties" => [
                "description" => [
                    "type" => "semantic_text",
                ],
            ],
        ],
    ],
]);
		
  1. In this example, semantic_text uses a default inference endpoint. The model used by this endpoint is multilingual, so you can index and search text in multiple languages. To use custom models, refer to Configure inference endpoints for semantic_text.
es.indices.create(
  index: "books",
  body: {
    mappings: {
      properties: {
        description: {
          type: "semantic_text"
        }
      }
    }
  }
)
		
  1. In this example, semantic_text uses a default inference endpoint. The model used by this endpoint is multilingual, so you can index and search text in multiple languages. To use custom models, refer to Configure inference endpoints for semantic_text.
await es.Indices.CreateAsync<Book>("books", c => c
    .Mappings(m => m
        .Properties(p => p
            .SemanticText(b => b.Description)
        )
    )
);
		
  1. In this example, semantic_text uses a default inference endpoint. The model used by this endpoint is multilingual, so you can index and search text in multiple languages. To use custom models, refer to Configure inference endpoints for semantic_text.
es.indices().create(c -> c
    .index("books")
    .withJson(new StringReader("""
        {
          "mappings": {
            "properties": {
              "description": {
                "type": "semantic_text"
              }
            }
          }
        }
        """))
);
		
  1. In this example, semantic_text uses a default inference endpoint. The model used by this endpoint is multilingual, so you can index and search text in multiple languages. To use custom models, refer to Configure inference endpoints for semantic_text.
res, err := es.Indices.Create(
	"books",
	es.Indices.Create.WithBody(strings.NewReader(`{
	  "mappings": {
	    "properties": {
	      "description": {
	        "type": "semantic_text"
	      }
	    }
	  }
	}`)),
)
if err != nil {
	panic(err)
}
defer res.Body.Close()
		
  1. In this example, semantic_text uses a default inference endpoint. The model used by this endpoint is multilingual, so you can index and search text in multiple languages. To use custom models, refer to Configure inference endpoints for semantic_text.

Next, index five books in one bulk request:

books = [
    {"title": "The Left Hand of Darkness", "author": "Ursula K. Le Guin", "release_year": 1969,
     "description": "An envoy visits an icy planet whose people have no fixed gender, feeling out politics and friendship across a deep cultural gap."},
    {"title": "Project Hail Mary", "author": "Andy Weir", "release_year": 2021,
     "description": "A lone astronaut wakes with amnesia on a spaceship and has to stop a disaster that threatens all life on Earth."},
    {"title": "The Name of the Wind", "author": "Patrick Rothfuss", "release_year": 2007,
     "description": "A gifted young musician and magician tells the story of his rise from orphan to legend."},
    {"title": "Klara and the Sun", "author": "Kazuo Ishiguro", "release_year": 2021,
     "description": "An artificial friend watches human love and loneliness while hoping a child will pick her."},
    {"title": "Dune", "author": "Frank Herbert", "release_year": 1965,
     "description": "On a desert planet prized for a rare spice, a young heir is pulled into a war over ecology, religion, and power."},
]

helpers.bulk(es, ({"_index": "books", "_source": b} for b in books), refresh="wait_for")
		
const books = [
  {
    title: "The Left Hand of Darkness",
    author: "Ursula K. Le Guin",
    release_year: 1969,
    description: "An envoy visits an icy planet whose people have no fixed gender, feeling out politics and friendship across a deep cultural gap.",
  },
  {
    title: "Project Hail Mary",
    author: "Andy Weir",
    release_year: 2021,
    description: "A lone astronaut wakes with amnesia on a spaceship and has to stop a disaster that threatens all life on Earth.",
  },
  {
    title: "The Name of the Wind",
    author: "Patrick Rothfuss",
    release_year: 2007,
    description: "A gifted young musician and magician tells the story of his rise from orphan to legend.",
  },
  {
    title: "Klara and the Sun",
    author: "Kazuo Ishiguro",
    release_year: 2021,
    description: "An artificial friend watches human love and loneliness while hoping a child will pick her.",
  },
  {
    title: "Dune",
    author: "Frank Herbert",
    release_year: 1965,
    description: "On a desert planet prized for a rare spice, a young heir is pulled into a war over ecology, religion, and power.",
  },
];

await es.bulk({
  refresh: "wait_for",
  operations: books.flatMap((book) => [
    { index: { _index: "books" } },
    book,
  ]),
});
		
$books = [
    [
        "title" => "The Left Hand of Darkness",
        "author" => "Ursula K. Le Guin",
        "release_year" => 1969,
        "description" => "An envoy visits an icy planet whose people have no fixed gender, feeling out politics and friendship across a deep cultural gap.",
    ],
    [
        "title" => "Project Hail Mary",
        "author" => "Andy Weir",
        "release_year" => 2021,
        "description" => "A lone astronaut wakes with amnesia on a spaceship and has to stop a disaster that threatens all life on Earth.",
    ],
    [
        "title" => "The Name of the Wind",
        "author" => "Patrick Rothfuss",
        "release_year" => 2007,
        "description" => "A gifted young musician and magician tells the story of his rise from orphan to legend.",
    ],
    [
        "title" => "Klara and the Sun",
        "author" => "Kazuo Ishiguro",
        "release_year" => 2021,
        "description" => "An artificial friend watches human love and loneliness while hoping a child will pick her.",
    ],
    [
        "title" => "Dune",
        "author" => "Frank Herbert",
        "release_year" => 1965,
        "description" => "On a desert planet prized for a rare spice, a young heir is pulled into a war over ecology, religion, and power.",
    ],
];

$operations = [];
foreach ($books as $book) {
    $operations[] = [
        "index" => [
            "_index" => "books",
        ],
    ];
    $operations[] = $book;
}

$es->bulk([
    "refresh" => "wait_for",
    "body" => $operations,
]);
		
books = [
  {
    title: "The Left Hand of Darkness",
    author: "Ursula K. Le Guin",
    release_year: 1969,
    description: "An envoy visits an icy planet whose people have no fixed gender, feeling out politics and friendship across a deep cultural gap."
  },
  {
    title: "Project Hail Mary",
    author: "Andy Weir",
    release_year: 2021,
    description: "A lone astronaut wakes with amnesia on a spaceship and has to stop a disaster that threatens all life on Earth."
  },
  {
    title: "The Name of the Wind",
    author: "Patrick Rothfuss",
    release_year: 2007,
    description: "A gifted young musician and magician tells the story of his rise from orphan to legend."
  },
  {
    title: "Klara and the Sun",
    author: "Kazuo Ishiguro",
    release_year: 2021,
    description: "An artificial friend watches human love and loneliness while hoping a child will pick her."
  },
  {
    title: "Dune",
    author: "Frank Herbert",
    release_year: 1965,
    description: "On a desert planet prized for a rare spice, a young heir is pulled into a war over ecology, religion, and power."
  }
]

operations = books.flat_map do |book|
  [
    { index: { _index: "books" } },
    book
  ]
end

es.bulk(
  body: operations,
  refresh: "wait_for"
)
		
var books = new[]
{
    new Book(
        "The Left Hand of Darkness",
        "Ursula K. Le Guin",
        1969,
        "An envoy visits an icy planet whose people have no fixed gender, feeling out politics and friendship across a deep cultural gap."
    ),
    new Book(
        "Project Hail Mary",
        "Andy Weir",
        2021,
        "A lone astronaut wakes with amnesia on a spaceship and has to stop a disaster that threatens all life on Earth."
    ),
    new Book(
        "The Name of the Wind",
        "Patrick Rothfuss",
        2007,
        "A gifted young musician and magician tells the story of his rise from orphan to legend."
    ),
    new Book(
        "Klara and the Sun",
        "Kazuo Ishiguro",
        2021,
        "An artificial friend watches human love and loneliness while hoping a child will pick her."
    ),
    new Book(
        "Dune",
        "Frank Herbert",
        1965,
        "On a desert planet prized for a rare spice, a young heir is pulled into a war over ecology, religion, and power."
    )
};

var operations = new BulkOperationsCollection();
foreach (var book in books)
{
    operations.Add(new BulkIndexOperation<Book>(book)
    {
        Index = "books"
    });
}

await es.BulkAsync(new BulkRequest
{
    Refresh = Refresh.WaitFor,
    Operations = operations
});

public record Book(
    string Title,
    string Author,
    [property: JsonPropertyName("release_year")] int ReleaseYear,
    string Description
);
		
record Book(
    String title,
    String author,
    int release_year,
    String description
) {}

Book[] books = {
    new Book(
        "The Left Hand of Darkness",
        "Ursula K. Le Guin",
        1969,
        "An envoy visits an icy planet whose people have no fixed gender, feeling out politics and friendship across a deep cultural gap."
    ),
    new Book(
        "Project Hail Mary",
        "Andy Weir",
        2021,
        "A lone astronaut wakes with amnesia on a spaceship and has to stop a disaster that threatens all life on Earth."
    ),
    new Book(
        "The Name of the Wind",
        "Patrick Rothfuss",
        2007,
        "A gifted young musician and magician tells the story of his rise from orphan to legend."
    ),
    new Book(
        "Klara and the Sun",
        "Kazuo Ishiguro",
        2021,
        "An artificial friend watches human love and loneliness while hoping a child will pick her."
    ),
    new Book(
        "Dune",
        "Frank Herbert",
        1965,
        "On a desert planet prized for a rare spice, a young heir is pulled into a war over ecology, religion, and power."
    )
};

BulkRequest.Builder bulk = new BulkRequest.Builder()
    .refresh(Refresh.WaitFor);

for (Book book : books) {
    bulk.operations(op -> op
        .index(idx -> idx
            .index("books")
            .document(book)
        )
    );
}

es.bulk(bulk.build());
		
type Book struct {
	Title       string `json:"title"`
	Author      string `json:"author"`
	ReleaseYear int    `json:"release_year"`
	Description string `json:"description"`
}

books := []Book{
	{
		Title:       "The Left Hand of Darkness",
		Author:      "Ursula K. Le Guin",
		ReleaseYear: 1969,
		Description: "An envoy visits an icy planet whose people have no fixed gender, feeling out politics and friendship across a deep cultural gap.",
	},
	{
		Title:       "Project Hail Mary",
		Author:      "Andy Weir",
		ReleaseYear: 2021,
		Description: "A lone astronaut wakes with amnesia on a spaceship and has to stop a disaster that threatens all life on Earth.",
	},
	{
		Title:       "The Name of the Wind",
		Author:      "Patrick Rothfuss",
		ReleaseYear: 2007,
		Description: "A gifted young musician and magician tells the story of his rise from orphan to legend.",
	},
	{
		Title:       "Klara and the Sun",
		Author:      "Kazuo Ishiguro",
		ReleaseYear: 2021,
		Description: "An artificial friend watches human love and loneliness while hoping a child will pick her.",
	},
	{
		Title:       "Dune",
		Author:      "Frank Herbert",
		ReleaseYear: 1965,
		Description: "On a desert planet prized for a rare spice, a young heir is pulled into a war over ecology, religion, and power.",
	},
}

indexer, err := esutil.NewBulkIndexer(esutil.BulkIndexerConfig{
	Client:  es,
	Index:   "books",
	Refresh: "wait_for",
})
if err != nil {
	panic(err)
}

ctx := context.Background()
for _, book := range books {
	data, err := json.Marshal(book)
	if err != nil {
		panic(err)
	}

	err = indexer.Add(ctx, esutil.BulkIndexerItem{
		Action: "index",
		Body:   bytes.NewReader(data),
	})
	if err != nil {
		panic(err)
	}
}

if err := indexer.Close(ctx); err != nil {
	panic(err)
}
		
Tip

If you want to experiment with a larger dataset, use the 1,000-book sample dataset.

Use the indexed book data to run semantic and hybrid searches. Semantic search matches meaning, while hybrid search combines semantic and keyword matching.

Run a semantic search against the description field:

resp = es.search(
    index="books",
    query={
        "semantic": {
            "field": "description",
            "query": "surviving alone in space",
        }
    },
)

for hit in resp["hits"]["hits"]:
    print(hit["_score"], hit["_source"]["title"])
		
interface Book {
  title: string;
}

const resp = await es.search<Book>({
  index: "books",
  query: {
    semantic: {
      field: "description",
      query: "surviving alone in space",
    },
  },
});

for (const hit of resp.hits.hits) {
  console.log(hit._score, hit._source?.title);
}
		
$response = $es->search([
    "index" => "books",
    "body" => [
        "query" => [
            "semantic" => [
                "field" => "description",
                "query" => "surviving alone in space",
            ],
        ],
    ],
]);

foreach ($response["hits"]["hits"] as $hit) {
    echo $hit["_score"] . " " . $hit["_source"]["title"] . "\n";
}
		
response = es.search(
  index: "books",
  body: {
    query: {
      semantic: {
        field: "description",
        query: "surviving alone in space"
      }
    }
  }
)

response["hits"]["hits"].each do |hit|
  puts "#{hit["_score"]} #{hit["_source"]["title"]}"
end
		
var response = await es.SearchAsync<SearchBook>(s => s
    .Indices("books")
    .Query(q => q
        .Semantic(semantic => semantic
            .Field("description")
            .Query("surviving alone in space")
        )
    )
);

foreach (var hit in response.Hits)
{
    Console.WriteLine($"{hit.Score} {hit.Source?.Title}");
}

public record SearchBook(
    [property: JsonPropertyName("title")] string Title
);
		
record Book(
    String id,
    String title,
    String author,
    int release_year,
    String description
) {}

String query = """
    {
      "query": {
        "semantic": {
          "field": "description",
          "query": "surviving alone in space"
        }
      }
    }
    """;

SearchResponse<Book> response = es.search(s -> s
        .index("books")
        .withJson(new StringReader(query)),
    Book.class
);

for (Hit<Book> hit : response.hits().hits()) {
    System.out.println(hit.score() + " " + hit.source().title());
}
		
query := strings.NewReader(`{
	  "query": {
	    "semantic": {
	      "field": "description",
	      "query": "surviving alone in space"
	    }
	  }
	}`)

	res, err := es.Search(
		es.Search.WithContext(context.Background()),
		es.Search.WithIndex("books"),
		es.Search.WithBody(query),
	)
	if err != nil {
		log.Fatal(err)
	}
	defer res.Body.Close()
	if res.IsError() {
		log.Fatal(res)
	}

	var response struct {
		Hits struct {
			Hits []struct {
				Score  float64 `json:"_score"`
				Source struct {
					Title string `json:"title"`
				} `json:"_source"`
			} `json:"hits"`
		} `json:"hits"`
	}
	if err := json.NewDecoder(res.Body).Decode(&response); err != nil {
		log.Fatal(err)
	}

for _, hit := range response.Hits.Hits {
	fmt.Println(hit.Score, hit.Source.Title)
}
		

Project Hail Mary ranks first because its description includes "A lone astronaut," which is semantically similar to "surviving alone in space" even though the wording differs.

Run a hybrid search that uses full-text search on the title field and semantic search on the description field:

resp = es.search(
    index="books",
    query={
        "bool": {
            "should": [
                {
                    "match": {
                        "title": "wind"
                    }
                },
                {
                    "semantic": {
                        "field": "description",
                        "query": "young magician coming of age",
                    }
                },
            ]
        }
    },
)

for hit in resp["hits"]["hits"]:
    print(hit["_score"], hit["_source"]["title"])
		
  1. The bool.should clauses are optional, but at least one must match. When both clauses match a document, Elasticsearch adds their scores, which can rank that document higher.
  2. The match clause performs full-text search on title and scores how well the analyzed text matches wind.
  3. The semantic clause searches description by meaning and contributes its semantic similarity score.
interface Book {
  title: string;
}

const resp = await es.search<Book>({
  index: "books",
  query: {
    bool: {
      should: [
        {
          match: {
            title: "wind",
          },
        },
        {
          semantic: {
            field: "description",
            query: "young magician coming of age",
          },
        },
      ],
    },
  },
});

for (const hit of resp.hits.hits) {
  console.log(hit._score, hit._source?.title);
}
		
  1. The bool.should clauses are optional, but at least one must match. When both clauses match a document, Elasticsearch adds their scores, which can rank that document higher.
  2. The match clause performs full-text search on title and scores how well the analyzed text matches wind.
  3. The semantic clause searches description by meaning and contributes its semantic similarity score.
$response = $es->search([
    "index" => "books",
    "body" => [
        "query" => [
            "bool" => [
                "should" => [
                    [
                        "match" => [
                            "title" => "wind",
                        ],
                    ],
                    [
                        "semantic" => [
                            "field" => "description",
                            "query" => "young magician coming of age",
                        ],
                    ],
                ],
            ],
        ],
    ],
]);

foreach ($response["hits"]["hits"] as $hit) {
    echo $hit["_score"] . " " . $hit["_source"]["title"] . "\n";
}
		
  1. The bool.should clauses are optional, but at least one must match. When both clauses match a document, Elasticsearch adds their scores, which can rank that document higher.
  2. The match clause performs full-text search on title and scores how well the analyzed text matches wind.
  3. The semantic clause searches description by meaning and contributes its semantic similarity score.
response = es.search(
  index: "books",
  body: {
    query: {
      bool: {
        should: [
          {
            match: {
              title: "wind"
            }
          },
          {
            semantic: {
              field: "description",
              query: "young magician coming of age"
            }
          }
        ]
      }
    }
  }
)

response["hits"]["hits"].each do |hit|
  puts "#{hit["_score"]} #{hit["_source"]["title"]}"
end
		
  1. The bool.should clauses are optional, but at least one must match. When both clauses match a document, Elasticsearch adds their scores, which can rank that document higher.
  2. The match clause performs full-text search on title and scores how well the analyzed text matches wind.
  3. The semantic clause searches description by meaning and contributes its semantic similarity score.
var response = await es.SearchAsync<SearchBook>(s => s
    .Indices("books")
    .Query(q => q
        .Bool(b => b
            .Should(
                should => should
                    .Match(match => match
                        .Field(book => book.Title)
                        .Query("wind")
                    ),
                should => should
                    .Semantic(semantic => semantic
                        .Field("description")
                        .Query("young magician coming of age")
                    )
            )
        )
    )
);

foreach (var hit in response.Hits)
{
    Console.WriteLine($"{hit.Score} {hit.Source?.Title}");
}

public record SearchBook(
    [property: JsonPropertyName("title")] string Title
);
		
  1. The bool.should clauses are optional, but at least one must match. When both clauses match a document, Elasticsearch adds their scores, which can rank that document higher.
  2. The match clause performs full-text search on title and scores how well the analyzed text matches wind.
  3. The semantic clause searches description by meaning and contributes its semantic similarity score.
record Book(
    String id,
    String title,
    String author,
    int release_year,
    String description
) {}

String query = """
    {
      "query": {
        "bool": {
          "should": [
            {
              "match": {
                "title": "wind"
              }
            },
            {
              "semantic": {
                "field": "description",
                "query": "young magician coming of age"
              }
            }
          ]
        }
      }
    }
    """;

SearchResponse<Book> response = es.search(s -> s
        .index("books")
        .withJson(new StringReader(query)),
    Book.class
);

for (Hit<Book> hit : response.hits().hits()) {
    System.out.println(hit.score() + " " + hit.source().title());
}
		
  1. The bool.should clauses are optional, but at least one must match. When both clauses match a document, Elasticsearch adds their scores, which can rank that document higher.
  2. The match clause performs full-text search on title and scores how well the analyzed text matches wind.
  3. The semantic clause searches description by meaning and contributes its semantic similarity score.
query := strings.NewReader(`{
	  "query": {
	    "bool": {
	      "should": [
	        {
	          "match": {
	            "title": "wind"
	          }
	        },
	        {
	          "semantic": {
	            "field": "description",
	            "query": "young magician coming of age"
	          }
	        }
	      ]
	    }
	  }
	}`)

	res, err := es.Search(
		es.Search.WithContext(context.Background()),
		es.Search.WithIndex("books"),
		es.Search.WithBody(query),
	)
	if err != nil {
		log.Fatal(err)
	}
	defer res.Body.Close()
	if res.IsError() {
		log.Fatal(res)
	}

	var response struct {
		Hits struct {
			Hits []struct {
				Score  float64 `json:"_score"`
				Source struct {
					Title string `json:"title"`
				} `json:"_source"`
			} `json:"hits"`
		} `json:"hits"`
	}
	if err := json.NewDecoder(res.Body).Decode(&response); err != nil {
		log.Fatal(err)
	}

for _, hit := range response.Hits.Hits {
	fmt.Println(hit.Score, hit.Source.Title)
}
		
  1. The bool.should clauses are optional, but at least one must match. When both clauses match a document, Elasticsearch adds their scores, which can rank that document higher.
  2. The match clause performs full-text search on title and scores how well the analyzed text matches wind.
  3. The semantic clause searches description by meaning and contributes its semantic similarity score.

The Name of the Wind ranks first because it matches both clauses, so its full-text and semantic scores combine.

With aggregations, you can summarize groups of data, such as the number of books released in each decade. Use ES|QL, a piped query language, to run this aggregation:

resp = es.esql.query(query="""
    FROM books
    | STATS books = COUNT(*) BY decade = release_year - (release_year % 10)
    | SORT decade ASC
    | LIMIT 10
""")

cols = {c["name"]: i for i, c in enumerate(resp["columns"])}
for row in resp["values"]:
    print(f"{row[cols['decade']]}s: {row[cols['books']]}")
		
const resp = await es.esql.query({
  query: `
    FROM books
    | STATS books = COUNT(*) BY decade = release_year - (release_year % 10)
    | KEEP decade, books
    | SORT decade ASC
  `,
});

for (const [decade, count] of resp.values ?? []) {
  console.log(`${decade}s: ${count}`);
}
		
$response = $es->esql()->query([
    "body" => [
        "query" => <<<'ESQL'
            FROM books
            | STATS books = COUNT(*) BY decade = release_year - (release_year % 10)
            | KEEP decade, books
            | SORT decade ASC
            ESQL,
    ],
]);

foreach ($response["values"] as [$decade, $count]) {
    echo $decade . "s: " . $count . "\n";
}
		
response = es.esql.query(
  body: {
    query: <<~ESQL
      FROM books
      | STATS books = COUNT(*) BY decade = release_year - (release_year % 10)
      | KEEP decade, books
      | SORT decade ASC
    ESQL
  }
)

response["values"].each do |decade, count|
  puts "#{decade}s: #{count}"
end
		
var response = await es.Esql.QueryAsync(r => r
    .Query("""
        FROM books
        | STATS books = COUNT(*) BY decade = release_year - (release_year % 10)
        | KEEP decade, books
        | SORT decade ASC
        """)
    .Format(EsqlFormat.Json)
);

using var result = JsonDocument.Parse(response.Body);
foreach (
    var row in result.RootElement.GetProperty("values").EnumerateArray()
)
{
    Console.WriteLine($"{row[0]}s: {row[1]}");
}
		
String query = """
    FROM books
    | STATS books = COUNT(*) BY decade = release_year - (release_year % 10)
    | KEEP decade, books
    | SORT decade ASC
    """;

BinaryResponse response = es.esql().query(q -> q
    .query(query)
    .format(EsqlFormat.Csv)
);

try (BufferedReader reader = new BufferedReader(
    new InputStreamReader(response.content())
)) {
    reader.lines().skip(1).forEach(line -> {
        String[] values = line.split(",", -1);
        System.out.println(values[0] + "s: " + values[1]);
    });
}
		
query := `FROM books
		| STATS books = COUNT(*) BY decade = release_year - (release_year % 10)
		| KEEP decade, books
		| SORT decade ASC`

response, err := typed.Esql.Query().
		Query(query).
		Format(esqlformat.Csv).
		Do(context.Background())
	if err != nil {
		log.Fatal(err)
	}

	rows, err := csv.NewReader(bytes.NewReader(response)).ReadAll()
	if err != nil {
		log.Fatal(err)
	}

for _, row := range rows[1:] {
	fmt.Printf("%ss: %s\n", row[0], row[1])
}
		

The complete example combines everything in one file. Set ES_URL and ES_API_KEY, then run it.

If you want to continue using your Vector Database project, select a next step based on your goal.

To learn how continued usage is billed, refer to Serverless project billing dimensions. If you don't plan to continue using the project, delete it to avoid additional usage charges.