Meilisearch

Meilisearch gives an application a usable search experience without a relevance tuning project attached to it. Results appear as the user types, in under fifty milliseconds, and the defaults are set so that something sensible works before any configuration happens.

Integration goes through a RESTful API with plugins and SDKs for common stacks, and the engine works out of the box with LangChain and the Model Context Protocol. Demo applications cover films, a hundred million Flickr images, an ecommerce catalog with faceted filtering, and a multi-tenant CRM.

The relevance features are the reason to reach for it.

• Hybrid search: semantic and full-text results are combined rather than chosen between.
• Typo tolerance: misspelled queries still return the matches the user meant.
• Filtering, faceting and sorting: custom filters, a faceted interface, and ordering by price, date or anything else.
• Geosearch and synonyms: documents filter and sort by location, and synonym sets widen what counts as a match.
• Language coverage: any language is searchable, with optimized handling for Chinese, Japanese, Hebrew and Latin-alphabet languages.
• Multi-tenancy and API keys: fine-grained permissions and personalized results for any number of tenants.

Newer additions include conversational search that answers natural language questions from your own results, per-user personalization, search rules that adjust behavior by context, document relations across indexes, and replication and sharding for horizontal scaling. Installing, deploying and maintaining it are all meant to stay undemanding, and a hosted cloud exists for teams that would rather not.