Jev in production › Search, ranking and matching

Jev RAG

Vector-free BM25 RAG pipeline where Jev reranks the top candidates; author's BEIR NFCorpus benchmark shows nDCG@10 rising to 0.444 in the hybrid setup (author).

Open on GitHub ↗

0.444measured against a baseline, as published by the source
Use
Search, ranking and matching
Industry
AI infrastructure
Form
Open-source tool
Stage
Beta
Listed
2026-09-26
Found via
github
Repository
aifabrice/jev-rag
Stars
10
Forks
2
Last push
2026-10-10
Language
Python
License
MIT
Jev RAG screenshot
docs/assets/demo-ui.png in the aifabrice/jev-rag README, MIT; shown from GitHub.

The README opens with

Jev RAG is an open-source local knowledge search engine with seven selectable pipelines: BM25 + Jev by default, agentic lexical search, embedding hybrid retrieval, multi-round Agentic Hybrid, taxonomy-routed hybrid retrieval, a unified Jev Passage Gate, and hierarchical Jev Line Search.

Use Jev RAG to search a local document folder, rerank candidate passages with Jev, and stream grounded answers with file citations. The default path is a vector-free RAG architecture: it requires neither embeddings nor a vector database. Jev RAG is local-first rather than fully offline because selected passages are sent to the configured Jev and answer-model providers.

Badge

For the project's own README, linking back here:

Listed in Jev in production

Also used for search, ranking and matching