LivingSim

A world that remembers and continues.

A deterministic social-consequence engine for games. NPCs remember typed events and later change their goals and actions because of them — running offline, with every reaction traceable back to its cause.

Get started with the SDK LivingSimSdk reference The engine Author a world

Recipes: one player action to memory, feeling, why and next move, in the getting-started page.

Run the demo: dotnet run --project tools/LivingSim.ProductProof -- --scenario village-council-hearing

What's inside the core

Each card is one capability the engine ships today.

Memory

NPCs form lasting memories of typed events — each with confidence, a source, and possible distortion.

Emotion

Decaying, multi-axis emotional state: anger, fear, joy, shame, pride, grief, hope, and more.

Relationships

Pairwise affinity, trust, respect, intimacy, and conflict that evolve with every interaction.

Reputation

Standing on axes like helpful, honest, dangerous, and untrustworthy — shaped by what others witness and hear.

Needs & Goals

Six social needs (safety, belonging, status, resources, truth, autonomy) generate the goals a character actually pursues.

Planning

A goal-oriented planner turns a chosen goal into concrete typed actions — move, exchange, wait.

Utility scoring

Goals are ranked by multiplicatively-compensated utility, so the character does what matters most right now.

Social exchanges

Multi-phase interactions like gossip propagate news between characters, with confidence decaying each hop.

Trauma

Durable trauma tied to people and places drives avoidance for days or weeks of game time.

Beliefs & expectations

Characters expect outcomes; when reality differs, surprise, vindication, or betrayal follow.

Identity

Self-image anchors — violating one produces shame, upholding it pride.

Norms & roles

Place- and world-scoped social norms that actions uphold or violate, registered by everyone present.

Secrets & concealment

Hidden affiliations and secrets that can be suspected, discovered, revealed, or confessed.

Theory of mind

Characters model what others know — and stay quiet when they believe you already know.

Habits

Repeated choices become earned tendencies that bias future decisions.

Causality graph

Every later action traces back through an append-only graph of the events that caused it.

Determinism & save/load

Same seed and inputs produce a byte-identical run — including after saving and reloading.

Multi-tier population

Dozens of full-fidelity NPCs, thousands more lightly tracked, and a hundred thousand in cohorts — all deterministic.

Material state

Health, money, items, employment, and legal status as real, durable consequences.

Explainability

A read-only inspector projects memory, goals, relationships, and the "why" behind any committed plan.

How it works

You feed the engine typed events — "Mara accused Brann", "Brann apologized" — and it updates each character's memory, emotion, relationships, reputation, and goals. Characters then act on that accumulated state. Nothing is scripted line by line; consequences emerge and stay explainable.

typed event → memory → emotion → relationship → reputation → goal → plan → action → next event
                                                                              ↓
                                                            causality graph (the "why")

• Get started — what it is and what to copy-paste, in five minutes.
• Author a world — declare a starting world as data (optional; the engine works from C# alone).
• Roadmap — what's shipped (M0–M4: internal checkpoint + The House Share playable proof), the Unity import proof (M5), and the store arcs after it (M6–M7). Which milestone is live is recorded in docs/product/milestone-status.md (in the repository).