Press Kit

A deterministic, single-player narrative RPG with persistent simulation and system-driven gameplay.

"A simulated world where dialogue describes the state - it does not create it."

Quick Facts

Title	Ghelmyon
Genre	Single-player narrative RPG / Text-based simulation
Mode	Single-player (play in-browser or self-host)
Platform	Browser-based (CLI also available)
Engine	Custom Python engine (~1.3M lines, ~42% of it the test suite)
Development	Solo developer
Status	Active development - playtest/demo phase
Website	maxwellcranston.com

Description (Short)

Ghelmyon is a deterministic, text-based simulation where player actions are resolved by game systems rather than scripted outcomes. Dialogue is generated from the current world state, but all mechanics - combat, reputation, inventory, and NPC behavior - are fixed and persistent.

Description (Long)

Ghelmyon is a single-player narrative RPG built around a deterministic simulation engine.

Every action produces a typed event that mutates world state: combat resolves through dice-based mechanics, NPCs follow schedules, relationships evolve over time, and factions react to player behavior.

A language model is used only for narration and dialogue. It receives structured state and produces descriptive text, but cannot alter outcomes or introduce new game data.

The result is a system where stories emerge from interacting mechanics rather than authored scripts.

Key Features

Deterministic event-driven simulation
Persistent world state with no resets
1637 named NPCs with schedules, relationships, secrets, and memory
Tactical combat - targeting, fatigue, positioning, injuries
Reputation and faction systems (11 factions, consequences propagate)
289 crafting recipes across 7 stations
41 skill trees learned through use, not XP
13 character classes with distinct playstyles
Emergent gameplay from interacting systems
Generative dialogue grounded in real game state

How It Works

The game is split into two layers:

Simulation layer

Resolves all actions deterministically
Tracks state (HP, inventory, time, reputation, NPC behavior)
Produces events through a central event system

Dialogue layer

Receives structured state after events resolve
Generates scene description and conversation
Cannot modify or override the simulation

This separation ensures consistency and replayability.

Example Gameplay

you> stab wolf high [Attack: Roll 72 vs 45 - Hit! 14 damage to head] [Wolf HP: 16/30]

you> steal bread [Reputation -2 (merchant)] [Guard becomes suspicious]

The mechanics are resolved first; the narration reflects the results.

By the Numbers

1637 Named NPCs

1557 Locations

~62k Automated Tests

~1.3M Lines of Python (~42% tests)

303 Crafting Recipes

368 Quests

41 Skill Trees

13 Classes

97 Spells

125 Creatures

11 Factions

222 Combat Moves

Development

Built using a multi-agent workflow (Overseer + task-specific workers)
~55,000 automated tests covering core systems
Event-driven architecture with strict separation of state and narration
Designed for long-term extensibility and emergent gameplay

Media

Screenshots and gameplay recordings coming soon.
For early access or press copies, use the contact details below.

Contact

Website	maxwellcranston.com
Demo	maxwellcranston.com/intro
Wiki	docs.maxwellcranston.com