From perception to Explain this

designed · Phase 1 (#31, #35–#38, #41) · PLAN §7.2–§7.4.2, §7.7, §13.5

The path one decision takes, with the component that owns each step and what it leaves behind for Explain this.

 tick N, Agency phase
 ┌──────────────────┐   perceived_entities, perceived_danger, perceived_warmth (attributes)
 │ PerceptionSystem │──▶ knows edges: {food_value: {value, observed_at, provenance}}            ─┐
 └──────────────────┘                                                                            │ graph
 ┌──────────────────┐   hunger += PerTick(hunger_rate_per_day)                                   │
 │ HungerSystem     │──▶                                                                        ─┘
 └──────────────────┘
 tick N+1, Agency phase (reads what the above wrote)
 ┌──────────────────┐   Option(actor, "move", target, Need:"safety", Score, Reason, Inputs{…})
 │ BaselineOptions  │──▶ Option(actor, "eat", bush,    Need:"hunger", …)                      ──▶ ctx.Options (bus, transient)
 └──────────────────┘   Option(actor, "fetch_water", bucket, Need:"safety", …)
 ┌──────────────────┐   per actor: × personality weight → argmax, StableHash tiebreak
 │ DecisionSystem   │──▶ SetAttribute(current_action, _target, _reason)                        ──▶ graph
 └──────────────────┘   decision record {winner, runners-up[3], inputs, weights}               ──▶ trace
 tick N+2, Agency phase
 ┌──────────────────┐   move:  RemoveEdge(contains old) + AddEdge(contains new)
 │ ActionSystem     │──▶ eat:   Delta(hunger, −meal_value) + Delta(bush.food_value, −bite_size) ──▶ batch
 └──────────────────┘   intents carry Inputs where a reader would ask why
 ┌──────────────────┐   sums deltas, priority for sets, one container per node, bounds
 │ IntentResolver   │──▶ intent_applied / intent_conflict events                               ──▶ graph + trace
 └──────────────────┘
 ┌──────────────────┐   Explain(eventId): event → intent (ordinal) → source (system:Decision)
 │ Causal trace     │     → decision record → its Inputs (perceived_danger=0.8, caution=0.8)
 └──────────────────┘     → WhyChanged(kira, perceived_danger, gen) → PerceptionSystem intent
                            → its Inputs (inn.temperature=310) → WhyChanged(inn, temperature, gen)
                            → FireSystem propagation → ignition intent → … → player:harness action
 ┌──────────────────┐
 │ Viewer           │──▶ "Kira left the square because the inn was burning and she is cautious.
 └──────────────────┘     The inn caught fire because the smithy's heat crossed 250°. The smithy was set alight by you."

The contracts along the way

Step	Owner	Writes	Reads	PLAN
Sense	PerceptionSystem	perceived_*, knows edges	containment, permeability.light/sound, salient attributes	§7.2, §7.3
Need	HungerSystem	hunger	hunger_rate_per_day	§7.4.1
Offer	BaselineOptions	options (bus)	perceived_*, hunger, energy, view.Facts/Known — never a target's live attributes	§7.4.1, §7.4.2
Choose	DecisionSystem	current_action*; decision record	ctx.Options, personality_*	§7.4.2, §7.6
Do	ActionSystem	intents per §7.7	current_action*	§7.7
Apply	Resolver	graph; events	batch	§18.1, §18.6
Remember	Causal trace	records	applied intents, decisions, events	§13.5
Tell	Viewer (Explain this)	prose	Explain, History	§13.4

What makes the chain possible

Three small things, each cheap, each required: every event has an id and the ordinal of the intent that produced it; every decision record keeps the runners-up and the inputs; and intents may carry Inputs — the snapshot values they were computed from — so that a physics link ("the inn caught fire because…") has a next hop. A system that omits Inputs ends the chain at itself, which is acceptable but shallower.

See also

The tick · Option bus · Causal trace · Player: Explain this