active_genie

https://github.com/Roriz/active_genie

ActiveGenie is an enabler for creating reliable GenAI features, offering powerful, model-agnostic tools across any provider. It allows you to settle subjective comparisons with a `ActibeGenie::Comparator` module that stages a political debate, get accurate scores from an AI jury using `ActiveGenie::Scorer`, and rank large datasets using `ActiveGenie::Ranker`'s tournament-style system. This reliability is built on three core pillars: - Custom Benchmarking: Testing for consistency with every new version and model update. - Reasoning Prompting: Utilizing human reasoning techniques (like debate and jury review) to control a model's reasoning. - Overfitting Prompts: Highly specialized, and potentially model-specific, prompt for each module's purpose.