razorback16/openjev ? reverse-engineered prompt

Reverse engineered prompt

Build me an open source decision server that feels like a fast, reliable “system one” API.

It should take a piece of state and a set of typed questions, like yes or no, multiple choice, and scored answers, then return probabilities and confidence instead of freeform text. I want it to be compatible with TypeSafe Jev so existing SDKs can talk to it without changes, and it should also support the OpenAI Decisions API style. Please include endpoints for decisions and model listing, and keep the responses strict so answers never drift off schema.

It should run with an open model on GPU hardware, and also support Apple silicon. Text only is fine for the first pass, but if it is practical, add image input too. Make it fast, calibrated, and easy to run locally with Docker, with clear defaults and a simple way to test it from curl or a small Python example. If you need to look up current docs online, go ahead.

Are you gonna build this?

make sure you review the code using coderabbit

Try freeSponsored — opens CodeRabbit in a new tab