AI - Tantivy
Empowering developers with fast, configurable full-text search capabilities through Tantivy's Rust library, featuring BM25 scoring, natural query language support, and various data types.
- Name
- Tantivy -https://github.com/quickwit-oss/tantivy
- Last Audited At
About Tantivy
Tantivy is a search engine library developed in Rust. It offers full-text search capabilities with configurable tokenizers for various languages, including stemming support for 17 Latin languages and third-party offerings for Chinese, Japanese, and Korean. Tantivy boasts fast performance, with a tiny startup time under 10ms, ideal for command-line tools. The library employs BM25 scoring similar to Lucene, supports natural query language, phrase queries, and incremental indexing. Additionally, it features multithreaded indexing, an mmap directory, SIMD integer compression, fast fields (single valued and multivalued for u64, i64, f64), text fields, and various data types such as dates, ip, bool, and hierarchical facets. Tantivy offers compressed document stores with options like LZ4, Zstd, or None. It supports range queries, faceted search, and configurable indexing with optional term frequency and position indexing. Furthermore, it includes a JSON field, an aggregation collector for histograms, range buckets, averages, and stats metrics, log merge policy with deletes, and a Searcher Warmer API.