jenna-russell/storyscope ? reverse-engineered prompt

Build me a Python project that recreates the StoryScope pipeline from the paper, so I can run the full workflow from prompts to story analysis.

I want it to take writing prompts, generate stories with a choice of AI provider, turn each story into a structured narrative template, compare the templates across sources, discover a taxonomy of narrative features, apply those features to all stories, and then train models that can tell human writing from AI writing and also identify which author produced a story. Please make it work from the command line in separate steps, with a simple config file for choosing OpenAI, Anthropic, Gemini, or local models.

Use the included sample data and let it save outputs to folders I choose. It should also support feature importance analysis so I can see which narrative traits matter most. If anything needs current library guidance, look it up online and use the best available approach.