Typesense

Typesense is an open-source, typo-tolerant search engine written in C++. It is purpose-built for instant search experiences in websites and applications and is commonly used as a simpler, cheaper alternative to Algolia and Elasticsearch. It runs as a single binary, exposes a REST API, and supports faceting, filtering, sorting, synonyms, curation, geo search, JOINs, vector search and clustering.

Yes. Typesense is designed to be self-hosted: download the single binary or run the official Docker image, start the server, and point your application at its REST API. There are no external runtime dependencies. For production, Raft-based clustering provides high availability with automatic leader election and failover. Typesense Cloud is optional if you prefer a managed service.

Yes. Typesense indexes embeddings from any machine learning model and performs nearest-neighbor (KNN) search for semantic search, recommendations, visual search and RAG. It can also generate embeddings automatically using built-in models or providers such as OpenAI, and combine keyword and semantic results through hybrid search with rank fusion. Version 30 also adds MMR-based result diversification.

Typesense offers similar instant-search capabilities — typo tolerance, faceting, filtering, sorting, synonyms and merchandising — while being open source and self-hostable, which avoids per-search pricing and vendor lock-in. Its JavaScript instant-search adapter is compatible with Algolia InstantSearch.js, so you can reuse existing UI components. Typesense Cloud is available as a managed alternative if you do not want to operate the cluster yourself.

Typesense Cloud is the official managed hosting option for Typesense. It adds a visual dashboard, self-service scaling, geo-distributed search delivery, team access, SSO, role-based access control and support plans on top of the open-source engine. The underlying search engine remains the same, so you can move between self-hosted and cloud deployments.