Search use case
This section documents core Elasticsearch search capabilities. These capabilities are available across all Elastic deployments, solutions, and project types.
Use this section to understand search techniques, query methods, ranking strategies, and data ingestion for search-powered applications.
Using the Elasticsearch solution or serverless project type? The Elasticsearch solution documentation covers additional UI tools included with these options.
Use Elasticsearch search capabilities for use cases such as:
- Website and documentation search
- Ecommerce product catalogs
- Content recommendation systems
- RAG (Retrieval Augmented Generation) systems
- Geospatial search applications
- Question answering systems
- Analytics dashboards and data exploration
- Custom observability or cybersecurity search tools
- Much more!
The following subjects are covered in this section:
| Topic | Description |
|---|---|
| Get started | Create deployments, connect to Elasticsearch, and run your first searches |
| Ingest data | Learn about options for getting data into Elasticsearch for search use cases |
| Search approaches | Compare search techniques available in Elasticsearch, including full-text, vector, semantic, and hybrid search |
| Build your queries | Implement your search approaches using specific query languages |
| Ranking and reranking | Control result ordering and relevance |
| RAG | Learn about tools for retrieval augmented generation with Elasticsearch |
| Building applications | Integrate Elasticsearch into your websites or applications |
For an introduction to core Elasticsearch concepts such as indices, documents, and mappings, refer to The Elasticsearch data store.
To dive more deeply into the building blocks of Elasticsearch clusters, including nodes, shards, primaries, and replicas, refer to Distributed architecture.
Elasticsearch functions as a vector database by storing vector embeddings and retrieving the most similar results to a query vector. Vector embeddings are numerical representations of data, such as text, images, or audio, created by machine learning models. Because similar items are positioned closer together in vector space, Elasticsearch can use these embeddings to perform semantic similarity search and return results based on meaning rather than exact keyword matches.
This capability is the foundation for vector search and related use cases in Elasticsearch. It enables you to work with semantic retrieval on the same data and infrastructure that you already use for full-text search, structured filters, and aggregations. Elasticsearch supports this by storing embeddings in vector field types such as dense_vector and sparse_vector, alongside your other indexed data.
To use Elasticsearch as a vector database, you can use the semantic_text field type. When you index content into a semantic_text field, Elasticsearch automatically generates vector embeddings using a configured machine learning model and stores them in the underlying vector field. These embeddings are indexed for efficient k-nearest neighbor (kNN) search, enabling fast similarity-based retrieval at query time.
For a high-level understanding of vector search concepts and capabilities in Elasticsearch, refer to Vector search. For an overview of use cases and guidance on how to implement them, refer to Vector search use cases.