nohn3043-arch/anthropomorphic-agent-engine

This project is a small Python engine that simulates the shape of an emotional personality without knowing anything about the meaning of events. It takes in three simple numbers representing threat, belonging, and fatigue, and turns them into changing emotional states, memories that fade over time, and lasting trauma markers, purely through math. The README describes the core as the body, and says the meaning behind events, such as deciding that criticism feels threatening, is supplied separately by a small piece of code called a mapper. Inside the core, seven emotional dimensions such as joy, anger, fear, and trust drift toward changing targets over time, at a pace the code calls viscosity. If threat crosses a certain level, a trauma marker forms and biases future emotional reactions. Memories fade using a mathematical decay curve, with a limit of 64 stored memory traces, and older or less important ones get replaced. Suppressed negative emotions can build up and eventually burst out after a cooldown period, and a character's ability to trust can shrink after repeated negative experiences, recovering only slowly afterward. Developers customize a character's personality by writing their own mapper, a small class that decides what each named event, like praise or criticism, means in terms of threat and belonging. Swapping mappers changes how a character reacts while the underlying engine stays the same. The engine has no external dependencies beyond plain Python, works the same way every time given the same input, and is meant to be used for things like game characters with long term memory, virtual assistants, interactive storytelling, and psychological simulation research.