Hand it a PDF, Word or PowerPoint file and a few category rules in plain English. The text is read on your own machine, then Jev decides which category the document belongs to, or where one document ends and the next begins inside a bundle.
Each answer comes with probabilities and a flag when it needs a human look. The demo uses real IRS, Treasury and SEC documents, and a visual report lets you browse all 40 test documents and the timings.
Use it as a Python library, a command-line tool or a local app. Needs Python 3.11 or newer and a TypeSafe API key. Built by Jerry Liu, independent of LlamaIndex's hosted products.
What you get
- Full Python source code, Apache 2.0 licensed
- Python library, command-line tool and local app
- Real-document demo and benchmark report
Jev
Python
PostgreSQL
Rust
OpenRouter
