s[TOR-Y]telling

s[TOR-Y]telling utilizes a local Large Language Model (Llama) to power its unique s[TOR-Y] Sandbox. This AI generates live text-based interactions, allowing players to probe and analyze sentient narrative entities in real-time.Technical Implementation & Privacy: The AI inference runs entirely on the player’s local hardware. No data is sent to the cloud, ensuring total player privacy and allowing for full offline playability.Safety & Moderation: We have implemented a multi-layered safety strategy to ensure a safe experience:Model-Level Guardrails: We utilize the Instruct version of the model, which features built-in safety weights designed to refuse harmful content.System-Level Directives: All prompts include an additional safety layer to prevent sexual violence, pedophilia, minor abuse/exploitation, explicit sexual behavior, or hate speech.Gameplay Integration: While the primary Visual Novel is hand-authored, the results of your sandbox evaluations influence the story's trajectory. The AI serves as a tool for narrative exploration, bridging the gap between open-ended analysis and cinematic storytelling.Generative AI was also used in the process of creating the game static assets (images, text, and music).