Installation
Welcome to the installation guide for dtx, your AI red teaming and evaluation framework.
This document helps you choose the right installation method based on your environment and needs.
We will only cover installation steps here — for usage instructions, refer to the main user guide.
Installation Options
1. Local Python Installation
Use this if you want to run dtx directly in your Python environment.
- Recommended for development environments.
- Supports local models with
torchandtransformers.
Guide: Local Installation
2. Docker Installation (ddtx)
Use Docker if you want an isolated environment with no local Python dependencies.
- No need to install
torchlocally. - Automatically manages dependencies and environment variables.
Guide: Docker Installation
3. Fast Installation Using uv
Use uv if you want a fast, clean install in new environments.
- Faster dependency resolution.
- Optional support for local models.
Guide: Install Using uv
4. Set Up Additional Dependencies
Includes:
- Ollama for running local models like
qwen2:0.5b - Environment variables for API keys (OpenAI, Hugging Face)
Guide: Set Up Additional Dependencies
Optional Installations
Install Agents Scope Generator
Optional: Install additional tools for automated scope generation.
- Uses AI models to generate scopes interactively.
- Requires model access.
Guide: Install Agents Scope Generator
Recommendations
| Use Case | Recommended Method |
|---|---|
| Development / Local Testing | Local Installation |
| Production, Isolated Environment | Docker Installation |
| Fast Setup in New Environment | Install Using uv |
| Local Models + API Setup | Set Up Additional Dependencies |