AI search engine and MCP server for bots and agents. One Rust binary indexes code, docs, logs, JSONL/CSV and PDFs for BM25 full-text, vector and hybrid search, code search without grep, agent memory and RAG. Elasticsearch-compatible API; Corpus Hub serves hundreds of reference memory corpora.
title: "Does XERJ do semantic search?" evidence:
- claim: "The neural embedder is opt-in behind --embed-mode neural" source: "docs/EXPERIMENTAL_ONNX.md"
- claim: "The default embedder is lexical feature hashing" source: "docs/EXPERIMENTAL_ONNX.md"
Does XERJ do semantic search?
XERJ can do semantic search, but only when you opt in. Start the node with
--embed-mode neural and it downloads a MiniLM-class model and embeds your text
with it. The default embedder is lexical feature hashing: it has no model, it
cannot connect synonyms, and it will happily rank an unrelated sentence above an
exact synonym.
So: lexical by default, neural is opt-in. If synonym matching matters to you, turn the neural embedder on and measure it on your own corpus first.