prateek1771/jev_lab

Jev Lab compares Jev, LLM, and plain-code approaches across 24 independent AI decision experiments. Users select an experiment, submit one example or run its labeled dataset, then inspect outcomes and reports in the Next.js or Streamlit UI. The API exposes experiment metadata, live and streamed dataset runs, saved-run reports, and downloads. The experiments span classification and scoring, routing and tool safety, retrieval and verification, and agent and business workflows. Their implementations vary; the sampled clients show Jev and OpenRouter calls, while unsampled project details are represented only at the documented experiment-suite level.