Node.js built-in SQLite now enables hybrid search with zero external dependencies

Node.js 22.5 introduced a native SQLite module that eliminates the need for third-party packages like better-sqlite3 or native build tools such as node-gyp. Developers can now build hybrid search systems combining SQLite's built-in FTS5 full-text search and vector similarity using only standard library imports. FTS5 supports Porter stemming and Unicode-aware tokenisation, allowing lexical queries to match word variants and diacritics across languages. Vector embeddings are stored as binary data in regular tables, making a separate vector database unnecessary for datasets in the tens of thousands. The two ranking signals — BM25 relevance and cosine similarity — are merged using Reciprocal Rank Fusion, a position-based method that avoids score normalisation issues and requires no retuning when models change.