Can't get phoenix traces correctly

Hi all, I’m following the “Nvidia’s NeMo Agent Toolkit: Making Agents Reliable” course on DeepLearning.ai but running everything locally on WSL.

It’s working mostly fine, but I can’t seem to get the phoenix traces right. It’s logging just the top-level workflow calls, but I can’t see LLM calls or tool interactions, like I would see when using langgraph directly or even if I do it from the course’s notebook.

I’m using Python 3.12.13 and have installed:

nvidia-nat

nvidia-nat[langchain]

nvidia-nat[phoenix]

Has anyone had this issue?