Production-ready billion scale vector database — Elasticsearch

Elasticsearch's open source vector database offers an efficient way to create, store, and search vector embeddings.

Combine text search and vector search for hybrid retrieval, resulting in the best of both capabilities for greater relevance and accuracy.

Try Elasticsearch today

Discover the latest innovations that make Elasticsearch and Lucene the top choice for vector databases.

Read blog

Learn to use Elasticsearch as a vector database for embeddings, powering search and building use cases like retrieval augmented generation (RAG), summarization, and Q&A.

Discover more on Search Labs

Elastic is the first to offer better binary quantization (BBQ), an optimization for vector databases with faster, more accurate vector search and 95% memory reduction.

Learn more about BBQ

Elasticsearch — the most widely deployed vector database

Copy to try locally in two minutes

curl -fsSL https://elastic.co/start-local | sh

Read docs

Deploy for production

Start free cloud trial

Or, download on-prem

A vector database is your starting point …

You need more than a vector database for a great search experience. Elasticsearch offers multiple retrieval types, flexible machine learning models, and advanced search features like aggregations, filtering, and auto-complete.

Run in the cloud, serverless, on-prem, or air gapped.

Read Elasticsearch Relevance Engine docs

Create inference endpoint

Use a machine learning model and apply it to your data at ingestion time.

Learn more about inference API & E5 model.

PUT _inference/text_embedding/my-e5-endpoint
{
  "service": "elasticsearch",
  "service_settings": {
    "num_allocations": 1,
    "num_threads": 1,
		"model_id": ".multilingual-e5-small"
  }
}

Generate embeddings
Capture meaning, context, and associations by transforming data into data in dense vectors. Create embeddings blog
POST _inference/my-e5-endpoint { "input": "How many adult mallard ducks fit in an american football field?" }

Index embeddings

Elasticsearch's vector storage is based on Lucene's HNSW algorithm which exceeds comparative benchmarks (see 'luceneknn').

Our vector search approach

PUT my-index
{
  "mappings": {
    "properties": {
      "my_vector": {
        "type": "dense_vector"
      },
      "my_text" : {
        "type" : "text"
      },
      "category": {
        "type": "text"
      }
    }
  }
}



# Index a document
PUT my-index/_doc/2
{
  "my_vector" : [
        0.018743202,
        -0.011059246,
   …
        0.05155631,
        0.04044561
      ],
  "my_text" :"57,000 average sized mallard ducks fit on an American football field",
"category" : "weird_stats"
}

Search embeddings

Run kNN search to fit your use case — apply the ANN algorithm for speed and scale using Lucene's HNSW index or exact match.

Explore ANN and KNN.


GET /my-index/_search
{
  "retriever": {
    "knn": { 
      "field": "my_vector", 
      "query_vector_builder":{
      "text_embedding": {
        "model_id": "my-e5-endpoint",
        "model_text": "How many adult mallard ducks fit in an american football field??"
      }
    }, 
      "k": 10, 
      "num_candidates": 10 
    }
  }
}

Use filters and faceting to refine search

Rely on Elasticsearch's rich framework of filtering and faceting capabilities available for vector search for building intuitive end-user applications.

Read about kNN filtering

Learn more about faceting

GET /my-index/_search
{
  "retriever": {
    "knn": {
      "field": "my_vector",
      "query_vector_builder":{
      "text_embedding": {
        "model_id": "my-e5-endpoint",
        "model_text": "How many adult mallard ducks fit in an american football field??"
      }
    }, 
      "k": 10,
      "num_candidates": 10,
      "filter": [
        {
          "term": {
            "category": "weird_stats"
          }
        }
      ]
    }
  }
}

Combine retrieval methods with hybrid search

With hybrid retrieval, pick from retrieval methods that deliver the best results for your use case: BM25, Elastic's trained sparse encoder model (ELSER), or dense vectors.

Watch video about hybrid search


GET /my-index/_search
{
  "retriever": {
    "rrf": {
      "retrievers": [
        {
          "standard": {
            "query": {
              "bool": {
                "filter": [
                  {
                    "term": {
                      "category": "weird_stats"
                    }
                  }
                ]
              }
            }
          }
        },
        {
          "knn": {
            "field": "vector",
            "query_vector_builder": {
              "text_embedding": {
                "model_id": "my-e5-endpoint",
                "model_text": "How many adult mallard ducks fit in an american football field??"
              }
            },
            "k": 10,
            "num_candidates": 10
          }
        }
      ],
      "rank_constant": 1,
      "rank_window_size": 50
    }
  }
}

Apply document-level security and compliance authorization

Assign granular role-based access controls with document- and field-level security with coverage across widely adopted compliance frameworks.

Set up authorization

Set up compliance

POST /_security/role/employee_role {
  "indices": [
    {
      "names": [ "example-index" ],
      "privileges": [ "read" ],
      "field_security" : {
        "grant" : [ "state", "name.first", "name.last" ]
      }
    }
  ]
}


GET example-index/_search {
  "query": {...},
  "knn" = {...},
  "aggs": {...}
}

PUT _inference/text_embedding/my-e5-endpoint
{
  "service": "elasticsearch",
  "service_settings": {
    "num_allocations": 1,
    "num_threads": 1,
		"model_id": ".multilingual-e5-small"
  }
}

Vector Database Integrations

Why use a vector database?

Try it using the AI Playground

Semantic search
Focus search on intent and contextual meaning beyond text matching.
Try hands-on learning
Multi-modal search
Search across all your data: text, vector, image, audio, video, geo, or unstructured data.
Field mapping types
GenAI search
Build retrieval augmented generation workloads for GenAI search experiences using vector and hybrid techniques.
ChatGPT and private data

Vector database superset

Choose a vector database based on the vector search experience you want to build.

Some vector databases

Elasticsearch

Embeddings

Store embeddings

full support

full support (free)

Generate embeddings

some support

full support (paid)

Retrieval

Machine learning

Search experience tools

Elastic tools

Elasticsearch — in action

See how organizations are building AI search applications to improve customer experience and help users find exactly what they're looking for.

See more customer stories

Customer spotlight
Reed, the UK's largest recruiter, brings job searchers and employers together using vector embeddings in Elasticsearch.
Learn more
Customer spotlight
Stack Overflow combines the power of human experts with generative AI to accelerate the retrieval of trusted information from developer knowledge bases.
Learn more
Customer spotlight
Adobe scales, manages multiple use cases, and puts machine learning features to work with Elastic.
Learn more

Get started implementing vector search

Frequently asked questions

What is a vector database and how does it work?

A vector database stores information as vectors, which are numerical representations of data objects, also known as vector embeddings. It uses vector embeddings for multi-modal search across a massive data set of structured, unstructured, and semi-structured data, such as images, text, videos and audio. Vector databases are built to manage vector embeddings and therefore offer a complete solution for data management.

The Search AI Company

Generative AI

Search

Security

Observability

By solution

Industries

Production-ready billion scale vector database — Elasticsearch

Elasticsearch — the most widely deployed vector database

Copy to try locally in two minutes

Deploy for production

A vector database is your starting point …

Vector Database Integrations

Why use a vector database?

Semantic search

Multi-modal search

GenAI search

Vector database superset

Some vector databases

Elasticsearch

Embeddings

Retrieval

Machine learning

Search experience tools

Elastic tools

Elasticsearch — in action

Customer spotlight

Customer spotlight

Customer spotlight

Get started implementing vector search

Blogs

webinars

Demo projects

Frequently asked questions

What is a vector database and how does it work?

What are vector embeddings?

What are the benefits of a vector database?

Is Elasticsearch a vector database?

Why choose Elastic as your vector database?

Follow us

About us

Join us

Partners

Trust & Security

Investor relations

Excellence Awards