This was my bachelor's thesis project, completed in 2024 at the University of Calabria, Italy.
A Python desktop application that creates and manages Non‐Player Characters (NPCs) endowed with coherent, Big‐Five‐based personalities and dynamic backstories.
The system couples an easy‐to‐use Tkinter GUI with local Large Language Models (LLMs), allowing real‐time, memory‐aware dialogue generation for role‐playing games and narrative prototypes.

• Big‐Five personality synthesis: generate or edit the five traits numerically and preview them instantly.
• Backstory & dialogue auto‐generation: 3‐shot prompting crafts up‐to‐25‐line origin stories and in‐character replies.
• Persistent conversational memory: chat summaries are stored in SQLite and re‐injected into future prompts for long‐term coherence.
• Model‐agnostic: works with any LLM that exposes an OpenAI‐compatible API and supports ≥ 4 000 tokens context length.
• Statistical evaluation: built‐in IPIP‐50 test runner, CSV logging and Matplotlib/Seaborn plots to benchmark accuracy vs. inference time.