Orama

Orama is an open-source, zero-dependency search engine written in TypeScript and developed by OramaSearch Inc. It runs in any JavaScript runtime and supports full-text, vector, and hybrid search, with an answer engine for chat-style experiences. The project describes itself as a complete search engine and RAG pipeline in less than 2kb.

What It Is

Orama is a JavaScript library installed through a package manager (for example npm install @orama/orama) or imported directly as a browser module from a CDN. Deno is supported via npm specifiers. A JavaScript runtime is the only requirement.

Developers create a database instance with a typed schema, insert documents, and query them with search. Schemas support string, number, boolean, enum, geopoint, array types, and fixed-size vectors for embeddings.

Search Capabilities

The documentation lists full-text search with BM25, typo tolerance, exact match, filters, facets, field boosting, geosearch, and results pinning (merchandising). Vector and hybrid search are selected through a mode option, and the caller supplies embeddings at search time. Stemming and tokenization are available in 30 languages, with dedicated guides for Chinese and Japanese.

RAG, Chat, and Plugins

Since v3.0.0, Orama provides AnswerSession for building ChatGPT-style answer experiences over indexed data. A Secure Proxy plugin lets client-side code call OpenAI securely. An embeddings plugin can generate vectors at insert and search time using TensorFlow.js.

Official plugins also cover analytics, data persistence, QPS, PT15, and documentation-site integrations for Vitepress, Docusaurus, Astro, and Nextra. A plugin system allows custom plugins.

Setup Path

Orama ships ESM modules by default and still supports CommonJS. TypeScript projects should set moduleResolution to Node16 or NodeNext. The repository is licensed under Apache 2.0 according to the README.