nexgencoder3/maya-persona-architecture

Maya Persona Architecture is a collection of carefully structured text files and a Python loader designed to give a local AI language model the personality of a specific fictional character named Maya. The goal is not to write a simple prompt that tells an AI how to act, but to build a persistent, multi-file framework that defines who Maya is, how she speaks, what she remembers, and how she reacts emotionally. The framework is organized into three broad categories. The persona layer covers Maya's backstory, family history, personal values, and areas of vulnerability. The voice layer defines her linguistic patterns, word choices, and conversational rhythm. A large collection of behavioral framework files handles things like emotional responses, memory selection, attachment patterns, anti-sycophancy rules, and safeguards that prevent her from sounding like a generic AI assistant. Loading Maya into a model takes a few minutes. You can use Ollama with a model like Mistral Nemo by running two commands: one to pull the base model and one to create a new model that bakes in Maya's personality. Alternatively, you can paste the lean system prompt directly into LM Studio's system prompt field. A Python loader script handles more advanced setups, including backends that use an OpenAI-compatible API. The framework includes conversation examples that show how Maya talks, a prompt bank for image generation, and a Python script that compiles all the modular files into one combined system prompt. A development handoff document explains the decisions made during development, which is useful if you want to understand why specific behavioral rules exist. The project is licensed under MIT and is designed as a fork-friendly starting point. You modify the files to create your own persona rather than using Maya as-is.