Load comments from a CSV or Excel file and watch two models label them at once, by default Jev against DeepSeek, or any other OpenAI-compatible chat model. It records the run for replay and produces an offline report.
The author tested it on 10,000 real comments, with a separate review of both models' labels. It also includes a plain-language explainer of how Jev works.
The app and docs are in Chinese.
What you get
- Full JavaScript source code, MIT licensed
- Head-to-head comparison app with replay and reports
- Demo data and a usage guide
Jev
TypeScript
JavaScript
OpenRouter
CodeRabbit participant
Python
Node.js