Repository navigation
Home
The Agentic Data Generation Toolkit is designed to provide an easy-to-use interface for both human users and Agents. The primary purpose of this toolkit is to provide a framework for experimentation with Agents that generate data.
The framework provides all the automation needed to execute a scenario while providing the user with consistent measurements across scenarios, as well as tracking and reporting of results.
I started this project before MCP was announced. I needed a mechanism for easily automating and duplicating experiments that are both human as well as agent driven. This packaged supported all the experiments I needed to perform for my dissertation. My goal was and remains to fill a gap for researchers so if I built it for myself others can take advantage of this as well.
Looking forward I plan on continuing to main
- Lab Journal: A "lab journal" which can be invoked through an experiment.
- Comprehensive Reports: Reports saved to disk for both preview and results of an experiment.
- Extensible Architecture: The framework is designed to be extensible both on load and during execution.
To install adgtk, please use one of the following methods:
To install the package using pip:
pip install adgtkIf you wish to clone the repository and install the package manually, follow these steps:
- Activate your virtual environment.
- Download the project from https://github.com/fred78108/adgtk.
- From the root folder of
adgtk, run:
(.venv) $ python -m pip install -e .This allows you to modify your copy of adgtk and evaluate the results immediately, which is ideal for custom development.