Tantivy is a full-text search engine library written in Rust that provides a programmatic foundation for building search applications rather than a ready-made server.
The tool addresses the need for fast, embeddable full-text search by implementing a design inspired by Apache Lucene. It operates as a Rust crate that developers integrate into their applications, handling indexing and querying through an in-process library rather than requiring a separate service. The engine supports configurable tokenization with stemming for multiple languages, BM25 scoring, natural query syntax including phrase queries, and faceted search capabilities. Indexing is multithreaded and incremental, with support for various field types including text, numerics, dates, IP addresses, and hierarchical facets. The tool compresses document storage using LZ4 or Zstd and applies SIMD integer compression on compatible platforms.
Tantivy suits developers building search functionality into Rust applications who want to avoid the operational overhead of running a separate search service. It works well for command-line tools due to its sub-10ms startup time and for applications requiring embedded search without external dependencies. The README positions it as closer to Lucene than to Elasticsearch or Solr, emphasizing that it is a library component rather than a standalone server. For distributed search across multiple nodes, the README directs users to Quickwit, a separate distributed engine built on top of Tantivy.
Development activity shows consistent engagement with the codebase through regular updates and maintenance. The project maintains comprehensive documentation including examples and a command-line interface tool that demonstrates practical usage patterns. Performance remains a focus area with published benchmarks tracking query and indexing speed across different workloads.
