An independent game development collective building games, gameplay systems, and experimental interactive technology.
4Fates Studios brings together small development teams working on independently initiated game projects. Teams vary between projects, with shared collaborators, tools, and development spaces connecting work across the collective.
Our projects span Unity and Unreal Engine and explore gameplay systems, procedural generation, developer tools, game AI, and AI-driven character interaction.
Turn-Based RPG · Unity · C# · Procedural Generation · Local AI
Project Warden is a turn-based RPG set in a dystopian world contained within a massive tower inspired by the mythical Tower of Babel.
As players ascend the tower, they encounter characters who can join and leave their party throughout the journey. These recurring characters are also the focus of an experimental AI-driven dialogue system intended to support more dynamic conversations and persistent character context.
The game is currently being developed by an 8-person team.
Development includes:
- Turn-based RPG gameplay systems
- Procedural labyrinth and environment generation
- Modular interaction and inventory systems
- Custom Unity Editor tools and content pipelines
- Locally hosted language model experimentation
- Retrieval-based character and world knowledge
- Persistent context derived from previous conversations
- Internal dialogue evaluation and data-collection tooling
Alongside the game, we are developing an experimental Python backend for AI-driven character interaction. Character and world information can be retrieved from project documentation, while previous conversations can be condensed into new context and stored for future retrieval.
The system is being designed around small, locally hosted language models with the goal of running AI-driven character dialogue entirely on the player's machine, without requiring internet connectivity, cloud inference, external servers, or paid API tokens. A major focus of the project is balancing dialogue quality with model size, inference performance, and accessibility across consumer hardware.
Project Warden is actively developed across public and private repositories. Some production assets and experimental backend infrastructure are not currently public.
Unity · C# · Python · FastAPI · RAG
Boba Stop is a senior capstone project developed by a three-person team exploring dynamic NPC interaction through AI-generated dialogue.
The project combines a Unity game client with a Python FastAPI backend and Retrieval-Augmented Generation (RAG), allowing NPC dialogue to reference character-specific information and contextual knowledge.
The project served as an early exploration of AI-driven NPC dialogue and influenced several of the ideas now being developed further for Project Warden.
Unreal Engine 5 · C++ · Blueprints
Echoes of Hubris is a Metroidvania developed by a six-person team in Unreal Engine 5.
The project features custom systems for inventory, combat, character movement, enemy behavior, and multi-phase boss encounters, combining C++ and Blueprint development across a multidisciplinary team.
While each project is built around its own game and team, 4Fates also serves as a space for experimenting with the technology and development practices behind those games.
Areas of interest include:
- Gameplay systems and game architecture
- Procedural generation
- Developer tools and content pipelines
- Game and enemy AI
- AI-driven character dialogue
- Local language model inference
- Retrieval-Augmented Generation (RAG)
- Persistent character context and memory
- Offline AI inference on consumer hardware
- Small-model optimization and quality/performance tradeoffs
| Project | Technologies | Status |
|---|---|---|
| Project Warden | Unity, C#, Python, Local LLMs | In Development |
| Boba Stop | Unity, C#, Python, FastAPI, RAG | Completed |
| Echoes of Hubris | Unreal Engine 5, C++, Blueprints | Completed |
Teams and contributors vary between projects. Individual repositories contain additional source code, technical information, and project documentation where publicly available.