LOB BrainLOB Brain
Back to Documentation

Hybrid RAG Engine

High-performance RAG combining FTS5 keyword indexing and local semantic search.

Typical RAG pipelines rely entirely on slow cloud APIs for embeddings. LOB leverages a local ONNX runtime embedded in the Rust binary, processing embeddings in under 4ms.

When calling brain_search, LOB fuses FTS5 keyword metrics (BM25) with semantic vector similarities, generating highly accurate search matches locally.