004 / SYSTEM DESIGN

Paths of Wonder

A data-driven interactive fiction engine built to grow — three stories in, an AI generation layer scoped next.

RoleProduct design & engineering
ForRavensbreath Lab
Year2026
ToolsPython · Interactive Fiction · AI Generation Layer

01 — THE PROBLEM

Interactive fiction engines are either too rigid to write in or too complex to build on.

02 — CONSTRAINTS

What the system had to respect.

01Story data, engine logic, and player state needed to stay fully separated
02Scenes had to be data-driven so new stories don't require new code
03Player progress had to autosave without manual intervention
04Foundation had to support an AI layer later without a rebuild

03 — THE SYSTEM

A workflow people can follow.

01

Architect

A clean three-file structure keeps story data, engine logic, and player state fully separated.

02

Author

Scenes are data-driven and variables substitute dynamically, so new stories don't mean new code.

03

Save

Player progress autosaves to JSON without any manual intervention.

04

Test

A developer test mode makes it fast to debug new stories without replaying from the beginning.

05

Expand

Three stories are in progress now, with an AI scene-generation layer already scoped for what comes next.

04 — DECISIONS

The choices that made it practical.

Decision 01

Separate the story from the code.

Data, engine logic, and player state live apart, so the architecture doesn't fight the writing.

Decision 02

Build the foundation before the AI layer.

Getting the base architecture right now means the generation layer won't mean rebuilding from scratch later.

05 — RESULT

A data-driven interactive fiction engine built to grow — three stories in, an AI generation layer scoped next.

3Stories in active development
LIVEEngine published and running on GitHub
NEXTAI scene-generation layer already scoped

Too rigid, or too complex

Most interactive fiction engines are either too rigid for real storytelling or too complex to build on top of. This one was built to grow — a foundation solid enough that adding an AI generation layer later wouldn’t mean rebuilding from scratch.

A Python interactive fiction engine, built on a clean three-file architecture, keeps story data, engine logic, and player state fully separated. Scenes are data-driven, variables substitute dynamically, and player progress autosaves to JSON without any manual intervention. A developer test mode makes it fast to build and debug new stories without replaying from the beginning every time.

Three stories, and what comes next

Three stories are currently in progress: The Summoned Mage, Space Walker, and Tech Hero. The engine is published and live on GitHub, with the next phase already scoped — an AI scene generation layer sitting on top of the existing architecture: structured output handling, stateful conversation, cost-aware generation triggers. The same instinct that shows up in the client work — build the foundation right so the interesting problem is easier to solve.