Roadmap and Workflow

This is the system's central intent document. Future sessions in builder or assistant mode should read this chapter first to understand what this system is for, how work flows through it, and where it is going. Detail lives in the other chapters; this is the map.

What this system is

An efficient, structured, fully-featured system for hosting a simulated world for gameplay. It is a system-agnostic ontology-based knowledge engine: agnostic to narrative, dice, sandbox, or campaign settings, and usable with a startup wizard for configuration. It is a module, not a world — it is loaded into a world project, asked what world to host, and then stores that world as linked, retrievable facts.

Everything is designed for retrieval-first knowledge work: a GM (human or agent) asks a question, and the engine hands back exactly the knowledge that question needs — nothing more. Every convention in this manual exists to serve that one requirement.

Design pillars

1. Explicit ontology for concepts. Every concept is a typed file — one file, one fact cluster, small enough to embed and retrieve as a unit. The Ontology Map chapter is the index of types and their relational rules; the Startup Checklist chapter is the coverage layer that makes sure a booted world decides on every common concept domain.
2. Tagging for recall. File names carry a recall tag after the subject (Thornhollow-place vs Thornhollow-lore), and every file carries tags vocabulary. The tag example marks placeholder content that is template, not world. The tag source-ontology-project marks provenance: files this engine created or reworked while layering onto an existing project.
3. Relational concept linkage. References over prose: location, characters, relatedLore, species, faction. Never restate a fact a linked file owns. canonLevel (canon / established / rumor / myth / theory / retconned) keeps retrieval from presenting rumor as fact; era anchors every claim in time.
4. Anchored memory. Memories are never recalled by pure similarity — they carry where (location), who (characters involved), and when (timestamp/era) from the start. Event files already carry these three anchors; the Memory type keeps all three.
5. Structured people. NPCs and characters map across psychology matrices (alignment, Big Five, dark triad, integrity/corruptibility, stability, beliefs) and body-level appearance (height, build, skin, eyes, hair, dimorphism cues) — appearance is the being, never worn equipment. Dimorphism is a species property. Both are mutable: they track how a person is actually played, and are updated when play contradicts the sheet.
6. Configurable startup. The system boots by capturing the vision first (scope, intent, themes, narratives, flavor, genres) into a writable World Vision scaffold, then asking (genre, magic level, world name, setting, era, tone, game system), branching follow-up questions, and executing the answers: writes a World Profile, scaffolds genre-appropriate lore folders and concept domains, generates mechanic rule files per the game-system summary, and generates the founding lore chunk. The scaffold makes setup resumable — an interrupted boot loses nothing, and a later session (or the user, editing the scaffold directly) finishes from it. The Startup Checklist decides or declares every domain so nothing is missed.
7. Examples plus write-your-own. Every startup question presents example picks, each with backend direction (what choosing it sets and scaffolds), and a free-text field for the user's own answer — custom words always override examples. The per-step map is portable in the Startup Examples chapter.

Architecture at a glance

The type set forms one graph — space (Celestial Body → Cosmology → Landmass → Region → Location, Polity, Weather, Terrain, Architecture, Calendar), people (Species/Background/Class/Subclass ground NPC/Character; Culture, Language, Background, Class, Subclass), belief (Deity, Religion, Magic Tradition), things (one Equipment type, subfoldered; Vehicle, Companion, Flora, Currency, Technology, Affliction), and knowledge (Lore as the connective tissue; World Vision scaffolds record intent; Weather, Calendar, and Prophecy serve time; Memory and Event record play). The full graph lives in the Ontology Map chapter; the domain-by-domain coverage list lives in the Startup Checklist chapter.

Workflow

Boot. Module loads into a project → Startup Protocol runs the vision pass into the World Vision scaffold, then the questionnaire → World Profile written → lore tree and concept domains scaffolded, a starter file per domain → mechanic rule files generated per the game-system summary → founding lore chunk generated → setupStatus: complete. Examples stay as tagged templates; build the new world alongside them. In a clone, assistant instructions may be absent — the Builder's Manual authority is the contract the Startup Protocol encodes.

Layer. When the module loads onto an existing campaign or world (built in another system or another project), the engine runs revamp on demand: it inspects the base project's content, then rebuilds each piece of the earned world into engine format (typed, chunked, named, tagged, anchored, linked — see the ingestion pipeline). Every file the engine creates or reworks in that merge carries the provenance tag source-ontology-project, so the engine's footprint is askable and differentiable from the base system's own files. Merging is always on demand — the tag makes work findable; it is never an automatic trigger.

Play. The GM asks questions; the engine recalls the single chunk that owns each fact, follows references exactly one hop, respects truth levels, keeps state files current (retconned marks overturns, never deleted), and records memories with where/who/when.

Evolve. Profiles and sheets are mutable; startup is rerunnable; roadmap items below land as the engine grows.

Roadmap

Built and working:

• Ontology engine: naming tags, chunking, canon levels, era anchoring, relational linkage; typed graph over 48 registered types (world, geography, people, beliefs, things, adventure, knowledge, play) with the debug sample world (Thornhollow) as example-tagged templates
• World bootstrap: vision pass → questionnaire → World Profile + World Vision scaffold (setupStatus resumable); concept opt-in checklist; startup contract and checklist chapters; example content throughout; startup flows and creation flows
• Examples plus write-your-own everywhere: every wizard step (scope, themes, narratives, flavor, genres, genre, magic system, era, setting, game system) presents example picks with per-option backend direction and a free-text field that overrides examples; the portable Startup Examples chapter maps every step; custom answers always win
• Game-system pinning: startup asks which mechanic family the world runs — d20, d100, narrative, wow-derived, or a written-in system — records gameSystem + gameSystemSummary on the World Profile, and the Game Systems chapter holds the internal summaries (resolution, stats, combat, progression, what to generate) that drive rule-file generation
• Ingestion pipeline: lists→typed chunks, maps→locations (pins linked), prose→lore clusters, foreign schemas migrated, scope scaffolding; Builder's Manual chapters Ingesting External Content, Migration and Lifecycle Conventions, Ontology Map, Naming, Chunking, and Folders, Relational Linkage, Truth and Time
• Package overlay + provenance: when the module layers onto an existing campaign, merge handling is defined in Content Ingestion (assistant instruction) and mirrored in Ingesting External Content and Conventions at a Glance — every file created or rebuilt during a merge carries the source-ontology-project provenance tag, the engine's footprint remains differentiable from base-system files, and rework is callable on demand (never automatic); source-ontology-project is the takeaway query for merge progress
• Migration conventions: retire-then-relocate; slug verification; no-draft refs; single-writer index and handbooks; subagent-drafts-main-commits; type lifecycle registry; index self-verification
• Anchored Memory type: live — .memory.json files in /Memories/ carry the three recall anchors (location, people, when) from creation, plus canonLevel, worldChanged, and relatedEvent; Barge-Burning-memory ships as the example template
• Player's Handbook: chapters drafted but currently builder-facing; awaits the player-facing rewrite once world generation is completed

Planned:

• Player's Handbook rewrite
• World maps, gallery, prelude
• More file-type schemas and computed cross-checks

How to use this document

A new session should: read this chapter to ground scope and intent; read the Startup Protocol when booting a new project; read Knowledge Engine Conventions for naming/chunking/truth; read the Ontology Map for the type graph; read the Startup Checklist for concept coverage; read the Startup Examples chapter for the per-step example map and the Game Systems chapter when pinning or generating mechanics; read the Terrain Taxonomy chapter when classifying geography; read the Migration and Lifecycle Conventions chapter before any conversion or type retirement; and read Ingesting External Content whenever content is imported, merged, or — including as a package — layered onto this engine from another system or project. Update this chapter when the direction of the system changes.