Meilisearch

Meilisearch is an open source search engine written in Rust, built to be embedded into applications and websites without extensive tuning. It returns results in under 50 milliseconds, tolerates typos by default, and combines full-text with semantic (hybrid) search. It ships with a RESTful API and official SDKs, so most teams can add search without building their own ranking engine, and example applications for movies, ecommerce, and CRM search show the same core engine adapting to different data shapes. The project publishes a public roadmap so users can see and vote on features under consideration before they ship, and demo apps like a hundred-million-image Flickr search show the engine handling large datasets beyond typical product catalogs.

Key features

• Hybrid search: Combines semantic vector search with traditional full-text search for more relevant results.
• Typo tolerance: Matches queries even with misspellings, with no extra configuration required.
• Faceting and filtering: Build faceted search interfaces with custom filters in a few lines of code.
• Sorting: Sort results by price, date, or any other field your users need.
• Geosearch: Filter and sort documents using geographic coordinates.
• Multi-tenancy: Serve isolated search results to different tenants using tenant tokens.
• Conversational search: Lets users ask natural-language questions and get AI-generated answers grounded in indexed data.

Ideal use cases

Meilisearch fits e-commerce product search, in-app search for SaaS products, documentation search, and any application that needs sub-50ms search-as-you-type without operating a heavier search stack like Elasticsearch. Its multi-tenancy and security controls also make it a fit for CRM-style products where different customers need isolated, filtered views of the same underlying index. The hybrid and conversational search features extend that further to natural-language product finders and support-style search, where users phrase a question instead of picking exact keywords. Personalization features let results be re-ranked per user based on prior behavior, which fits catalog and content products where a single ranking doesn't serve every visitor equally well.

Self-hosting and deployment

Meilisearch runs as a single binary or Docker container and scales horizontally through replication and sharding for larger datasets. A managed Meilisearch Cloud option exists alongside the self-hosted path for teams that would rather not operate the infrastructure themselves. Security is handled through API keys with fine-grained permissions, so different parts of an application can hold keys scoped to only the indexes and actions they need. Because search settings, ranking rules, and synonyms are configured through the same REST API used for indexing and querying, infrastructure-as-code setups can version an index's configuration alongside application code instead of managing it through a separate admin console. Dependency status, license, and merge-queue state are all tracked as public badges in the repository, which is useful context for teams evaluating the project's release hygiene before adopting it.