Agent-324/signalforge ? reverse-engineered prompt

Reverse engineered prompt

Build me a small local first web app for evaluating AI traces and catching regressions before they hit production.

I want to paste in trace data, see each trace scored in a way that makes sense to a human, and get clear explanations for things like helpfulness, safety, confidence language, latency, and whether the output matches the reference. The app should group results into something like healthy, elevated, or critical, and show the evidence behind each call so it feels trustworthy instead of like a mystery score.

Please include a simple dashboard, a JSON API, SQLite persistence, and a command line way to run evaluations on JSONL files and compare a baseline against a new candidate set. It should be easy to load demo data, run locally, and see regressions and missing traces clearly. Keep it simple, deterministic, and well tested, and look up current docs online if you need to.

Are you gonna build this?

make sure you review the code using coderabbit

Try freeSponsored — opens CodeRabbit in a new tab