StorySpark: Module-wise Evolutionary Search for Story Premise Generation

StorySpark implements a module-wise evolutionary search framework that treats narrative components like persona and event as dynamic local search spaces. The system uses feedback-driven mutation and Pareto-guided selection to refine individual modules conditioned on the existing premise, balancing branch coverage with high-potential narrative directions.