Hi @<1717350310768283648:profile|SplendidFlamingo62> , you could also add it as a bash init script inside the container. Basically run a bash script to manually install it or import it in whatever fashion you do so manually today.

WDYT?

Ah sorry. I use pip install -r requirements.txt. I have the environment variables also set.

I mean when you manually setup your environment, how do you install those packages?

Hi @<1523701070390366208:profile|CostlyOstrich36> , thanks for the anwer. This would work but then I would only have one "global" version of the lib and could not install in Task1 lib==1.0 and in Task2 lib==2.0 or?

How do you currently install from your 'internal' repo?

I think then you can use Task.add_requirements - None

And inject any relevant env variables into the container

We let clearml track the repository and have the login data in the agent configuration.

Thank you for your idea. Unfortunately, it does not work. The package will be tracked locally correctly, but when the agent runs this for the first time, it will overwrite the original PIP with PIP (Package @ Nexus). When I clone this task and enqueue it, it will fail because the username and password are no longer in the Nexus link.