Agient: Psychologically-Grounded Agent Simulation Framework

Agient is a Python framework for creating psychologically realistic agent-based simulations that model human behavior, relationships, and decision-making processes. It combines psychological models with LLM capabilities to create more human-like agents.

Features

• 🧠 Psychological Modeling

  • Five Factor Model (Big Five) personality traits
  • Emotional states and memory
  • Beliefs and values systems
  • Goal-oriented behavior

• 🤝 Social Dynamics

  • Agent-to-agent communication
  • Relationship tracking
  • Social network modeling
  • Influence mechanics

• 💭 LLM Integration

  • OpenAI GPT-4/3.5
  • Anthropic Claude
  • Extensible provider system

• 📊 Memory Management

  • SQLite-based persistent storage
  • Importance-based memory filtering
  • Emotional context tracking
  • Memory cleanup utilities

File Structure

The project has been organized into a more structured format:

├── examples/
│   ├── senator_example.py
│   ├── teacher_example.py
│   └── student_example.py
├── llm/
│   ├── __init__.py
│   ├── base.py
│   ├── config.py
│   ├── providers/
│   │   ├── __init__.py
│   │   ├── openai.py
│   │   └── anthropic.py
├── main.py
├── models/
│   ├── __init__.py
│   ├── base.py
│   ├── demographics.py
│   ├── memory.py
│   └── personality.py
├── storage.py
├── tests/
│   ├── __init__.py
│   └── test_storage.py
├── pyproject.toml
└── README.md

Getting Started

Prerequisites

• Python 3.11 or later
• An OpenAI API key (for LLM integration)
• An Anthropic API key (for LLM integration)

Installation

1. Clone the repository:

   git clone https://github.com/danielmerja/agient.git
   cd agient

2. Install dependencies:

   poetry install

3. Set up environment variables for API keys:

   export OPENAI_API_KEY="your_openai_api_key"
   export ANTHROPIC_API_KEY="your_anthropic_api_key"

Running Examples

To run the provided examples, execute:

python examples/senator_example.py
python examples/teacher_example.py
python examples/student_example.py

Running Tests

To run the tests, execute:

pytest