Orama
A full-text, vector and hybrid search engine and RAG pipeline that runs in the browser, on a server or at the edge, with typo tolerance, facets, geosearch and a plugin system.
Orama is a search engine and retrieval-augmented-generation pipeline that runs in the browser, on a server, or on an edge network. It supports full-text search, vector search and hybrid search, along with facets, geosearch, filters, pinning rules for merchandising, typo tolerance, exact match, BM25 ranking, and stemming and tokenization in 30 languages.
A database instance is created with an indexing schema built from ten data types, including strings, numbers, booleans, enums, geopoints and vectors, plus array versions of each. Vector and hybrid search are enabled by passing a search mode and providing text embeddings at search time; a plugin can generate embeddings, and a secure proxy plugin can call OpenAI's embedding models from the client side without exposing an API key. Since version 3.0.0, Orama can also build ChatGPT- or Perplexity-style chat experiences over the indexed data.
It installs as an npm package, or can be imported directly in a browser or in Deno from a CDN, and is written in TypeScript.