Meilisearch is an open-source search engine written in Rust, aimed at powering search-as-you-type experiences in web and mobile apps. It prioritises relevance tuning, sub-50ms response times on moderate datasets, and a REST API that an application developer can integrate in an afternoon.
Internally Meilisearch stores documents in an LMDB-backed index and uses a custom inverted index plus a bucket-sort ranking pipeline with configurable ranking rules — words, typo, proximity, attribute, sort, exactness — that can be reordered per index. Typo tolerance is built in via a Levenshtein automaton, and the query pipeline supports filtering, faceting, geo search, synonyms, and stop-word lists. Recent releases add vector search with HNSW indexes and hybrid search that combines lexical BM25-style ranking with vector similarity, which brings LLM retrieval use cases into the same engine. Client SDKs exist for JavaScript, Python, Ruby, PHP, Go, Rust, Java, and .NET, and the project ships a hosted Meilisearch Cloud alongside the self-hostable binary.
Meilisearch competes with Elasticsearch, OpenSearch, Typesense, and Algolia, and its niche is being a single small Rust binary that gives you Algolia-quality relevance defaults without running a JVM cluster.