Browser RAG

N0X indexes documents in a Web Worker, combines vector retrieval with BM25 keyword matching, and uses MMR reranking to return focused excerpts for chat.

• Small files are injected directly for fast answers.
• Large files are chunked, embedded, cached, and searched locally.
• PDF, DOCX, TXT, Markdown, CSV, HTML, and JSON are supported.