Introduction

Apache Lucene is an open-source Java library for full-text search. It gives you the building blocks to add fast, scalable search to any Java application — without running a separate search server.

​

What Lucene is

• A Java library you add as a dependency to your project
• A toolkit for building inverted indexes over arbitrary text and structured data
• An engine that executes queries against those indexes and returns ranked results
• The foundation for most widely-used open-source search systems

​

What Lucene is not

• A database — it does not replace a relational or document store
• A search server — it has no HTTP API or wire protocol out of the box
• A crawler or scraper — you are responsible for feeding documents to it
• A managed service — you operate it as part of your application

​

Key use cases

Application search — embed Lucene to add a search box to a web app, mobile backend, or desktop tool. Index whatever content matters: articles, products, logs, code, messages. Search backends — build a dedicated indexing and query service in Java. Lucene handles relevance ranking, faceting, highlighting, and geospatial queries at scale. Log and event analytics — index structured fields alongside free text. Query by date range, category, and keyword in a single pass. Semantic and vector search — since Lucene 9, the library includes native k-nearest-neighbor (KNN) vector search alongside classic term-based retrieval, enabling hybrid search pipelines.

​

Architecture overview

Every Lucene workflow follows three phases: 1. Analyze — raw text passes through an Analyzer pipeline. The analyzer tokenizes the text, lowercases it, removes stop words, and applies stemming or other filters. The result is a stream of Tokens ready to be indexed. 2. Index — an IndexWriter writes tokens and stored field values into a Directory (either on disk or in memory). Internally, Lucene organises data into immutable segments that are periodically merged. Each segment contains an inverted index mapping terms to the documents that contain them. 3. Search — a DirectoryReader opens the index read-only, and an IndexSearcher executes Query objects against it. Lucene scores matching documents using BM25 by default and returns a TopDocs result containing ranked ScoreDoc entries.

​

Current version

These docs cover Lucene 11.0.0. Lucene 11 requires Java 25 and brings continued improvements to HNSW vector search, automaton handling, and merge policy configuration.

​

Next steps