Canonical scene state
Every battle is modeled as a single scene document containing rooms, openings, objects, actors, round state, and optional narration fields. The renderer follows that state rather than inventing its own.
Project
Legacy design notes described a GPT-assisted tabletop system where the battlemat is the truth display and the engine behind it owns adjudication, narration, and durable state.
Every battle is modeled as a single scene document containing rooms, openings, objects, actors, round state, and optional narration fields. The renderer follows that state rather than inventing its own.
The legacy architecture notes emphasized durable records and versioned updates. That matters because campaigns need memory, and memory needs a source of truth.
Artificial intelligence is treated as an instrument, not an oracle: useful for continuity, adjudication, and structured narration when it operates inside explicit boundaries.
Why It Matters
The world engine exists so maps, inventories, character state, and tactical scenes can agree with each other over time. The public site does not ship that runtime, but it can explain the design goal clearly: a campaign where systems remain visible and fiction is earned through interaction with them.