For small projects pip is great, but for larger projects like this whose (flattened) dependency list is long and will surely keep growing more, it is better to move to poetry.

Conda can be a contender but is a very heavy option.

As the main target audience profile for this project is an AI Engineers and poetry is more suitable and it is perfect for robust dependency management, as well as environment management.

And poetry also helps with publishing of your package - I really hope you do that soon.

• I added a request on that line : Feature Request: Publish Cognita as a package to pypi #191

References of why Poetry over Pip

• https://realpython.com/dependency-management-python-poetry/
• https://itheo.tech/pip-vs-poetry-a-comprehensive-comparison
• https://unbiased-coder.com/python-poetry-vs-pip/
• https://www.exxactcorp.com/blog/Deep-Learning/managing-python-dependencies-with-poetry-vs-conda-pip