Everyone who has ever tried to tweak a prompt knows that changing the prompt to make it work on one scenario can lead to break another. Up until now there was no easy solution for this.<p>PromptWatch is a LangChain tracing on steroids and now introducing also Unit tests!<p>Based on annotated previous sessions or prompts, it allows you to replay these inputs with new prompt / model.<p>And just 3 lines of code (more/less) to set it up.