Project 89 - ArgOS

A sophisticated agent simulation system built on BitECS, featuring autonomous agents with advanced cognitive architectures, capable of dynamic interactions, self-spawning, and emergent narrative generation.

Overview

ArgOS is an experimental platform for creating and running autonomous agent simulations. It uses a custom cognitive architecture that enables agents to:

• Process sensory input and context
• Maintain different types of memory (working, episodic, semantic, procedural)
• Execute complex cognitive functions (planning, reasoning, decision making)
• Generate contextual responses and actions
• Maintain emotional states and belief systems
• Interact with other agents and their environment

Current Features

• Entity Component System (BitECS)

  • Efficient agent state management
  • Component-based architecture
  • Fast query system

• Agent Systems

  • Thinking System (cognitive processing)
  • Room System (environment management)
  • Action System (behavior execution)
  • Perception System (stimuli processing)

• Memory Management

  • Thought history
  • Experience tracking
  • Context awareness

• Action Framework

  • Speech capabilities
  • Environment interaction
  • Tool usage system

Getting Started

1. Install dependencies:

npm install

2. Run the basic conversation example:

npm run start

This will start a simulation with two agents in a room, demonstrating basic interaction capabilities.

Architecture

The system is built on several core components:

1. World State

   • Resource management
   • Narrative state tracking
   • Population management

2. Agent Components

   • Core agent properties
   • Memory systems
   • Action capabilities
   • Relationship tracking

3. Systems

   • Cognitive processing
   • Environmental interaction
   • Action execution
   • State management

For a detailed architectural overview, see DESIGN_DOC.md.

Development Status

Currently implemented:

• Basic agent interactions and conversations
• Simple Thought generation with LLM integration
• Basic Environment awareness and room system
• Basic action and perception system
• Simple Memory tracking (thoughts and experiences)
• Speech and examination tools

In progress:

• Physical actions and body awareness
• Enhanced agent perception (sight, sound, smell)
• Long term vector memory
• Goal setting and planning system
• Multi-agent coordination
• Core memory systems (childhood, significant experiences)
• Relationship formation and tracking

Planned features:

• Self-spawning capabilities (agent reproduction)
• Dynamic narrative generation
• World generation from text prompts
• World state as entity relationships
• Meta-agent for narrative control
• Tool system for world modification
• Long-term persistence and database integration
• Advanced memory hierarchies
  • Working memory
  • Episodic memory
  • Semantic memory
  • Procedural memory

Running Examples

The project includes several example scenarios:

• basic-conversation.ts: Two agents engaging in basic interaction

Contributing

This is an experimental project in active development. Feel free to explore and experiment with the codebase.

License

MIT