Help with resources for using LLMs as fictional characters

Hey ya'll,

I'm an ex-cognitive scientist turned NLP Data Scientist by day, and science fiction author by night.

I want to bring fictional characters in my prose to life with Local LLMs, and I'm looking for the best resources out there for doing this kind of work (datasets, models, libraries, common patterns, etc.). Could you help me out?

For context, I recently got a 64GB Mac Mini for this and other Local LLM side-projects, and my work pays for about $750 USD of LLM API tokens for personal use per year I could use to create my own training data. I work with BERT & GPT-style models at work, and I've done some Local LLM work on my MacBook with >8B models (mostly just basic vector-database-based RAG for question-answering and summarization over PDFs). I also have detailed character notes for persona prompting as well as world-building notes for RAG-based pipelines.

I would like to go beyond persona prompting and RAG, though. I've been reading mechanistic interpretability / steering research for the last few months, and am very interested in using these methods to more precisely control character behavior and personality. So anything in this space specifically would be very appreciated.

Cheers!

TL;DR - Looking for for resources on using LLMs in fiction, specifically using LLMs as fictional characters/NPCs. Particularly interested in applying mechanistic interpretability / steering methods on top of persona prompting and RAG.