devwithjevRoutes Graph RAG Extraction with Jev
Joseph used Jev with GLiNER 2.5 base to speed up entity and relationship extraction, reporting a ~25x performance boost. Jev helps determine which extractions are correct and sends the rest to a local 35B model.
First real use case with Jev got up and running last night.
Used it with GLiNER 2.5 base to help dramatically speed up entity / relationship extraction.
GLiNER by itself for Graph RAG over a huge document corpus is fairly unreliable. Using Jev to help determine what it got right and then send everything else to my local 35B model.
~25x performance boost.
GLiNER alone was fairly unreliable for Graph RAG over a huge document corpus.
Also filed under Research & data
- Jev Structured Decision Model Demo
Jev is TypeSafe AI's structured decision model. The repository contains research notes and an interactive Cloudflare Worker demo for it.
- Japanese Print Metadata Extraction
John Resig explored using Jev with his existing Japanese print metadata extraction pipeline, which uses gpt-5.6 luna. He found Jev could replace luna in some cases to save money, and augment it in others for higher quality.