• #6

I’m developing a browser-based, chat-like platform for playing adventure games, RPGs, immersive sims and the likes. It’s “modding-first,” meaning that all game content comes in mods. That includes actions, events, components, rules, entities, etc. My goal is that eventually, you could define in mod files the characters, locations, actions, rules, etc. for any existing RPG campaign and you would be able to play through it, with other characters being large language models or GOAP-based artificial intelligences.

From early on, it became clear that the platform was going to be able to support thousands of actions for its actors (which may be human or controlled by a large language model). The code shouldn’t be aware of the specifics of any action, which wasn’t easy to do; I had to extract all the logic for going from place to place from the engine, to the extent that I had to create a domain-specific language for determining target scopes.

When the turn of any actor starts, the system looks at all the actions registered, and determines which are available. I registered some actions like waiting, moving from place to place, following other people, dismissing followers, to some more niche ones (in an “intimacy” mod) like getting close to others and fondling them. Yes, I’m gearing toward erotica in the future. But as I was implementing the actions, it became clear that the availability of some actions wouldn’t be easily discerned for the impossible-to-predict breath of possible entities. For example, if you wanted to implement an “slap {target}” action, you can write a scope that includes actors in the location, but what determines that the actor actually has a head that could be slapped? In addition, what ensures that the acting actor has a hand to slap with?

So I had to create an anatomy system. Fully moddable. The following is a report that I had Claude Code prepare on the first version of the anatomy system.

The Anatomy System: A Deep Dive into Dynamic Entity Body Generation

Executive Summary

The anatomy system is a sophisticated framework for dynamically generating and managing complex anatomical structures for entities in the Living Narrative Engine. It transforms simple blueprint definitions and recipes into fully-realized, interconnected body part graphs with rich descriptions, validation, and runtime management capabilities.

At its core, the system addresses a fundamental challenge in narrative gaming: how to create diverse, detailed, and consistent physical descriptions for entities without manual authoring of every possible combination. The solution is an elegant blend of data-driven design, graph theory, and natural language generation.

System Architecture

The anatomy system follows a modular, service-oriented architecture with clear separation of concerns. The design emphasizes:

• Orchestration Pattern: A central orchestrator coordinates multiple specialized workflows
• Unit of Work Pattern: Ensures transactional consistency during anatomy generation
• Chain of Responsibility: Validation rules are processed in a configurable chain
• Strategy Pattern: Description formatting uses pluggable strategies for different part configurations
• Factory Pattern: Blueprint factory creates anatomy graphs from data definitions

Core Service Layers

1. Orchestration Layer (AnatomyOrchestrator)

• Coordinates the entire generation process
• Manages transactional boundaries
• Handles error recovery and rollback

1. Workflow Layer

• AnatomyGenerationWorkflow: Creates the entity graph structure
• DescriptionGenerationWorkflow: Generates natural language descriptions
• GraphBuildingWorkflow: Builds efficient traversal caches

1. Service Layer

• BodyBlueprintFactory: Transforms blueprints + recipes into entity graphs
• AnatomyDescriptionService: Manages description generation
• BodyGraphService: Provides graph operations and traversal

1. Infrastructure Layer

• EntityGraphBuilder: Low-level entity creation
• SocketManager: Manages connection points between parts
• RecipeProcessor: Processes and expands recipe patterns

Information Flow

The anatomy generation process follows a carefully orchestrated flow:

1. Initialization Phase

When an entity with an anatomy:body component is created, the AnatomyInitializationService detects it and triggers generation if the entity has a recipeId.

2. Blueprint Selection

The system loads two key data structures:

• Blueprint: Defines the structural skeleton (slots, sockets, parent-child relationships)
• Recipe: Provides specific customizations, constraints, and part selections

3. Graph Construction

The BodyBlueprintFactory orchestrates the complex process of building the anatomy:

Blueprint + Recipe → Graph Construction → Entity Creation → Validation → Description Generation

Each step involves:

• Slot Resolution: Blueprint slots are processed in dependency order
• Part Selection: The system selects appropriate parts based on requirements
• Socket Management: Parts are connected via sockets with occupancy tracking
• Constraint Validation: Recipe constraints are continuously checked

4. Description Generation

Once the physical structure exists, the description system creates human-readable text:

• Individual part descriptions are generated using context-aware builders
• Descriptions are composed into a complete body description
• Formatting strategies handle single parts, paired parts, and multiple parts differently

5. Runtime Management

The generated anatomy becomes a living system:

• Parts can be detached (with cascade options)
• The graph can be traversed efficiently via cached adjacency lists
• Events are dispatched for anatomy changes

Core Capabilities

1. Dynamic Entity Generation

• Creates complete anatomical structures from data definitions
• Supports unlimited variety through recipe combinations
• Generates unique entities while maintaining consistency

2. Hierarchical Part Management

• Parts are organized in a parent-child graph structure
• Each part can have multiple sockets for child attachments
• Supports complex anatomies (e.g., creatures with multiple limbs, wings, tails)

3. Intelligent Part Selection

• Matches parts based on multiple criteria (type, tags, properties)
• Supports preferences and fallbacks
• Handles optional vs. required parts gracefully

4. Natural Language Descriptions

• Generates contextual descriptions for individual parts
• Composes full-body descriptions with proper formatting
• Handles pluralization, grouping, and special cases

5. Constraint System

• Enforces recipe-defined constraints (requires/excludes)
• Validates socket compatibility
• Ensures graph integrity (no cycles, orphans, or invalid connections)

6. Runtime Operations

• Part detachment with cascade support
• Efficient graph traversal via cached adjacency lists
• Path finding between parts
• Event-driven notifications for changes

Key Components Deep Dive

AnatomyOrchestrator

The maestro of the system, ensuring all workflows execute in the correct order with proper error handling and rollback capabilities. It implements a Unit of Work pattern to maintain consistency.

BodyBlueprintFactory

The factory transforms static data (blueprints and recipes) into living entity graphs. It handles:

• Dependency resolution for slots
• Socket availability validation
• Part selection and creation
• Name generation from templates

Validation System

A sophisticated chain of validation rules ensures anatomical correctness:

• CycleDetectionRule: Prevents circular parent-child relationships
• OrphanDetectionRule: Ensures all parts are connected
• SocketLimitRule: Validates socket occupancy
• RecipeConstraintRule: Enforces recipe-specific rules
• JointConsistencyRule: Ensures joint data integrity

Description Generation Pipeline

The description system is remarkably sophisticated:

1. BodyPartDescriptionBuilder: Creates individual part descriptions
2. DescriptionTemplate: Applies formatting strategies
3. PartGroupingStrategies: Handles different grouping scenarios
4. TextFormatter: Provides consistent text formatting
5. BodyDescriptionComposer: Orchestrates the complete description

Strengths of the System

1. Modularity and Extensibility

Each component has a single, well-defined responsibility. New features can be added without modifying existing code.

2. Data-Driven Design

Anatomies are defined entirely in data, making it easy to add new creature types without code changes.

3. Robustness

Comprehensive validation, error handling, and rollback mechanisms ensure system reliability.

4. Performance Optimization

• Cached adjacency lists for efficient traversal
• Lazy description generation
• Batched entity operations

5. Developer Experience

• Clear service boundaries
• Extensive logging and debugging support
• Consistent error handling patterns

Expansion Opportunities

1. Dynamic Modification System

• Runtime part growth/shrinkage: Allow parts to change size dynamically
• Transformation support: Enable parts to transform into different types
• Damage modeling: Track part health and visual damage states

2. Advanced Constraints

• Symmetry requirements: Ensure paired parts match when needed
• Resource-based constraints: Limit total mass, magical capacity, etc.
• Environmental adaptations: Parts that change based on environment

3. Procedural Enhancement

• Mutation system: Random variations within constraints
• Evolutionary algorithms: Breed new anatomies from existing ones
• Machine learning integration: Learn optimal configurations

4. Visual Integration

• 3D model mapping: Connect anatomy graph to visual representations
• Animation constraints: Define movement limitations based on anatomy
• Procedural texturing: Generate textures based on part properties

5. Gameplay Systems

• Ability derivation: Generate abilities from anatomy (wings = flight)
• Weakness detection: Identify vulnerable points in anatomy
• Part-specific interactions: Different interactions per body part

6. Description Enhancement

• Contextual descriptions: Change based on observer perspective
• Emotional coloring: Descriptions that reflect entity state
• Cultural variations: Different description styles for different cultures

7. Performance Scaling

• Anatomy LOD (Level of Detail): Simplified anatomies for distant entities
• Streaming support: Load/unload anatomy data dynamically
• Parallel generation: Generate multiple anatomies concurrently

8. Tool Support

• Visual anatomy editor: GUI for creating blueprints and recipes
• Validation sandbox: Test recipes before deployment
• Analytics dashboard: Track anatomy generation patterns

Technical Implementation Details

Design Patterns in Action

The codebase demonstrates excellent use of software design patterns:

• Service Locator: Services are injected via constructor dependencies
• Facade: AnatomyGenerationService provides a simple interface to complex subsystems
• Template Method: Validation rules follow a consistent pattern
• Composite: The anatomy graph itself is a composite structure
• Observer: Event system notifies interested parties of anatomy changes

Error Handling Philosophy

The system follows a “fail-fast” approach with comprehensive error information:

• Validation errors prevent invalid states
• Detailed error messages aid debugging
• Rollback mechanisms prevent partial states
• Event dispatching for error tracking

Extensibility Points

Key extension points for customization:

• Custom validation rules via the ValidationRule base class
• New part grouping strategies via PartGroupingStrategy
• Custom formatters via the formatting service interface
• Additional graph algorithms via the graph service

Conclusion

The anatomy system represents a significant achievement in dynamic content generation. It successfully balances flexibility with consistency, performance with functionality, and simplicity with power. The modular architecture ensures the system can grow and adapt to new requirements while maintaining its core elegance.

The combination of graph-based structural representation, constraint-driven generation, and sophisticated natural language processing creates a system capable of generating virtually unlimited variety while maintaining narrative coherence and technical correctness.

As the Living Narrative Engine evolves, the anatomy system provides a solid foundation for increasingly complex and engaging entity interactions, opening doors to emergent gameplay and storytelling possibilities limited only by imagination.