Prompt Engineering & LLM Eval Kit

Stop guessing whether your prompts work. This toolkit gives you a systematic framework for evaluating and improving LLM outputs. Includes: prompt A/B testing harness (compare two prompts head-to-head across test cases), regression detection (catch when model updates break your prompts), output quality scoring (automated rubrics for accuracy, tone, format compliance), cost-per-query tracking (monitor token usage and estimate spend), prompt versioning system (track what changed and why), and a red-team checklist for adversarial testing. Every component works standalone with Python 3.8+, no external dependencies. Built for teams shipping AI features who need confidence their prompts actually work before they hit production.