AuraGen Documentation

Welcome to AuraGen, a sophisticated data generation engine designed to produce diverse, high-quality risky trajectories for Agentic Systems.

Python Version License

Overview

AuraGen operates in two distinct stages:

  1. Generate Harmless Trajectories: Create clean, task-oriented agent action/response traces across many scenarios

  2. Inject Risk: Programmatically mutate the harmless trajectories to introduce realistic risks while maintaining coherence and plausibility

Key Features

  • Flexible API Integration (OpenAI, external APIs, custom providers)

  • Dynamic Configuration via YAML (including custom API key types)

  • Comprehensive Risk Injection Framework

  • Scenario-based Design with Constraints

Quick Start

# Clone the repository
git clone <repository-url>
cd Agentic-Guardian

conda create -n aura python=3.11 -y
conda activate aura

# Install dependencies
pip install -r requirements.txt --upgrade

# Configure API keys
python config/configure_api_keys.py

# Generate and inject risks
python generate_and_inject.py

Table of Contents

User Guide

Indices and Tables