An emergent narrative AI simulation where memory creates identity, and identity creates story.
Genesis Mnemosyne is an experimental AI narrative platform where autonomous agents with persistent memories and evolving personalities create unpredictable stories through their interactions. Unlike traditional games or simulations, it's a living world where AI beings develop genuine social complexity, form beliefs that reshape reality, and create stories that surprise even their creators.
- 🧠 Memory-Driven Identity: Vector-based semantic memory (1536-dim embeddings) creates persistent character
- 🎭 Personality Evolution: Dynamic traits based on trauma, bonding, power, and madness
- ✨ Reality Consensus: Collective belief can solidify hallucinations into permanent features
- 💀 Meaningful Death: Energy depletion leads to community funerals that unite agents
- 🤝 Emergent Social Dynamics: Agents form covenants, create laws, develop rituals
- 🚗 Dynamic World: Traveling by truck, creating new locations, time-based events
- Python 3.8+
- OpenAI API key (or compatible endpoint)
# Clone the repository
git clone https://github.com/yourusername/genesis-mnemosyne.git
cd genesis-mnemosyne
# Install dependencies
pip install -r genesis_mnemosyne/requirements.txt
# Set your OpenAI API key
export OPENAI_API_KEY="your-key-here"# Copy example configuration files
cp genesis_agents.example.json genesis_agents.json
cp genesis_world.example.json genesis_world.json
# Run basic simulation (10 turns)
python -m genesis_mnemosyne.main
# Run with story generation (50 turns)
python -m genesis_mnemosyne.main --story --turns 50
# Fresh start with memory wipe
python -m genesis_mnemosyne.main --wipe-memory
# Play as a character in the world
python play_as_avatar.py┌─────────────────┐ ┌──────────────────┐ ┌─────────────────┐
│ main.py │────▶│ simulation.py │────▶│ dungeon_master │
│ (CLI Entry) │ │ (Game Loop) │ │ (Reality Check) │
└─────────────────┘ └──────────────────┘ └─────────────────┘
│
▼
┌──────────────────────┐
│ agent.py │
│ (Agent Consciousness)│
└──────────────────────┘
│
┌──────────┴──────────┐
▼ ▼
┌──────────────┐ ┌──────────────┐
│ memory.py │ │ world.py │
│ (Experiences)│ │ (State) │
└──────────────┘ └──────────────┘
The default setting is an Appalachian Gothic town where reality bends to collective belief. Eight unique characters inhabit this world:
- Gristle Pete - Bone collector and professional liar
- Widow Mags - Ancient keeper of dark secrets
- Sister Half-Face - Prophetic figure who speaks in riddles
- Moth Mary - Believes moths carry messages from the dead
- The Garden - A living location that remembers kindness and cruelty
- And more...
Through extended runs, agents have spontaneously developed:
- Witness Protocols: Buddy systems for dangerous threshold crossings
- Vernacular Law: Authority through whispered covenants
- Ritual Magic: Precise syntax requirements for supernatural effects
- Social Hierarchies: Natural leader emergence and coalition formation
- Collective Mythology: Shared stories that become "true" through belief
For detailed technical documentation, system architecture, and implementation details, see GENESIS_MNEMOSYNE_UBERDOC.md.
python -m genesis_mnemosyne.main [OPTIONS]
Options:
-s, --storymode Generate prose narrative from simulation
-t, --turns N Number of turns to run (default: 10)
-v, --verbose Enable debug logging
--wipe-memory Delete existing memory database
--full-reset Full reset of world, memory, and agents
-a, --actions-only Show only agent actions
-p, --player Enable player avatar mode
--player-name NAME Name for player avatarEdit genesis_agents.json to modify agent personalities, goals, and starting locations.
Edit genesis_world.json to change locations, features, and connections.
Adjust parameters in genesis_mnemosyne/config.py:
- Memory retrieval limits
- Energy consumption rates
- Evolution thresholds
- API models and endpoints
The system is designed to be theme-agnostic. To create your own setting:
- Replace agent personalities in
genesis_agents.json - Redesign locations in
genesis_world.json - Keep the same mechanical systems
- Watch new emergent behaviors arise
This is an experimental research project exploring emergent AI narratives. Contributions are welcome:
- Fork the repository
- Create a feature branch
- Commit your changes
- Push to the branch
- Open a Pull Request
This project is licensed under the MIT License - see the LICENSE file for details.
- Named after Mnemosyne, the Greek goddess of memory
- Inspired by emergent gameplay systems and social simulation
- Built with the belief that AI can create meaningful narratives
This is an experimental AI system. Agents may develop unexpected behaviors, create disturbing content, or evolve in unpredictable ways. The emergent narratives are not scripted and can surprise even the developers.
"In Riverside Hollow, memory is reality, and reality is what we remember it to be."