I built a playground of interative A/B testing for RAG

To iteratively improve RAG performance, current evaluation solutions still take lots of manually work or lots of coding. And it requires close collaboration between AI engineers and domain experts (who may not know how to code).

So I built this playground to show a smoothier workflow that enables continuously improvement of RAG, it can 1) run RAGs and get evaluation results quickly 2) generate insight both tech & non-tech can understand 3) provide an UI for domain experts to review and update flagged entries easier.