Conventions
Shared working agreements.
Conventions knowledge
- What do words mean? Vocabulary tree.
- How are narratives classified? Narrative convention.
- How are UI surfaces and copy named? Frontend convention and Brand copy.
- How do interface locale and Domain content language stay separate? Copy and localisation.
- Which models and call categories are allowed? AI models, then the generated model-call ledger.
Ai Models
8 articles- Adding work — the rulePublic
A new model call site first picks its system (two systems): a typed decision goes through callDecide(state, questions, caller, decisionModelFor('<category ')); text goes through callGenerate(..., modelFor('<category ')). Never both in one call.
- AI Models — tiers, categories, and allocation profilesPublic
The single reference for how the engine picks a model for each kind of work. Source of truth: src/lib/core/ai profiles.ts (systems, tiers, categories, profiles) and src/lib/core/model resolver.ts (modelFor(category) for generation, decisionModelFor(category) for decisions). This doc mirrors that...
- CategoriesPublic
Every modality of LLM work. Categories are intentionally orthogonal: each names one kind of work rather than a model tier, fallback, or miscellaneous bucket. impact is how much dropping the tier costs you — substrate/reasoning work compounds downstream (protect it), volume/interactive work does not.
- Profiles — the allocation matrixPublic
Each column is one profile; read it top to bottom to see every category's tier. B=Budget, Bal=Balanced, Perf=Performance, Front=Frontier. Balanced is the shipped default (DEFAULT PROFILE ID = "balanced smart") and is the original hand tuned allocation.
- The three layersPublic
| Layer | What it is | Where | | | | | | System | Which boundary the call crosses: decide (System 1, typed probabilistic answers) or generate (System 2, text). Two of them; orthogonal to the three below. | AI SYSTEMS, DECISION MODEL | | Tier | A capability/cost rung, mapped to one concrete...
- The two systemsPublic
Every model call in the engine is one of two kinds of thinking, and the boundary it crosses declares which. The split is the first question to ask of any new call — before category, before tier.
- TiersPublic
Low → high. frontier is the most capable/expensive rung; balanced is the fast and capable middle (formerly the "quick" tier).
- Where it's controlledPublic
Director only , fleet wide (like the Pilot). Gate: can.manageAiProfiles (role === "director"). AI Profiles editor — home sidebar → AI Profiles . Full editor: pick a profile, read the category→tier grid, override per category. (src/features/settings/AiProfiles.tsx) Quick switch — a compact...
Narrative
11 articles- Beat ClassesPublic
Beats are the atomic units of scene structure — individual moments that advance story, reveal character, or build world.
- Domain PositionPublic
Position classifies the local delivery state at a given point in the story — where you are in the current rhythm.
- Force grading systemPublic
A Domain is held by five peer forces: World, System, Causality, Fate, and Theme. The scorecard is a realisation grade over a selected record interval. Every force owns an equal 20 point share of the 100. A force cannot earn more than its own share to reimburse another force's shortfall. Because...
- Game Theory TaggingPublic
Every scene is additively decomposed into a sequence of 2×2 style strategic games , one per strategic beat (written to scene.gameAnalysis, never mutating deltas). Each game names the shape of a decision — the full space of choices each party could have made and the consequence of every pairing —...
- Narrative Convention — scene, beat, and game taxonomyPublic
Comprehensive reference for all domain classification systems, scene modes, beat types, structural archetypes, and game theory tagging used throughout Meridians.
- Narrative ShapesPublic
Shapes classify the macro structure of delivery curves — how intensity rises and falls across the full story.
- Pentad archetypesPublic
An archetype names the composition of a whole Domain . It is not the Sign of one Scene and it is not the grade. The grade says how fully the record realises all five forces; the archetype says exactly which forces have formed strongly enough to carry the work.
- Scene cadence (Pentad signs)Public
The pentad sign maps a scene into five force space. Each sign is the high/low configuration of Fate, World, System, Theme, and Causality at that moment. The 32 signs, not the older three force cube corners, are the states of scene cadence.
- Story ScalesPublic
Scales classify domains by scene count — the fundamental measure of scope.
- Usage in MeridiansPublic
These definitions are used throughout the platform:
- World DensityPublic
Density measures the growth of declared World and System structure relative to story length .
Root
4 articles- AI modelsPublic
This maintained index routes to the smallest topical leaf. Update the leaf that owns the claim; do not rebuild a parallel summary here.
- Copy and localisationPublic
Meridians has two language axes. Keep them separate.
- Frontend Convention — UI naming, composition, and visual vocabularyPublic
How code names are formed in this codebase (companion to LANGUAGE.md, which defines the domain vocabulary). This is the present convention — match it when adding or renaming code.
- Narrative conventionPublic
This maintained index routes to the smallest topical leaf. Update the leaf that owns the claim; do not rebuild a parallel summary here.