Hacker News Reader

created something similar but specific to llama-server. it takes advantage of logprob output and completes decisions in a single forward pass https://github.com/khimaros/verdict

we also have several jev type models available in seconds built on top of models on our router

mev (based on mercury) gemmev (based on gemma)

also these take images input

zev, lev etc

https://trustedrouter.com/docs/decide

But… why???

as an alternative option to Jev, and to take advantage of LLM's reasoning for the classification.

Show HN: Jevper – the Jev interface on top of any OpenAI-compatible model