Jypyter notebooks in CDF is running inside your browser and is based on pyodide. There is a subset of packages that works with pyodide. psutil seems like something wouldn’t work here.

There workaround is to install jupyter / python locally on your computer.

CDF functions runs in a normal python environment, so that’s different.

Hi @Diana Chimnaz Johan, did the above help you?

Hello,

It helped to confirm the limitation of the jupyter notebook and streamlit since both use pyodide. I was wondering if there is a plan to lift that limitation? 

Thanks,

Diana

We don’t have any plans to deploy jupyter or streamlit outside pyodide in the Cognite platform. 

What are you trying to do that does not work in pyodide?

Hi @Diana Chimnaz Johan,

We are following up to see whether you're satisfied with the responses you've received.

Best regards,
Dilini  

Hi Dilni,

Yes, I got all the answers I needed.

Thanks,

Diana

@Murad Sæter 
sklearn, pytorch, tensorflow are all very well known data science libraries. None of them have a pure python wheel.

Unless I am wrong, you can’t use them in CDF Streamlit. Do you agree?

Neither does numba which is a prerequisite for stumpy which I am trying to use for time series segmentation.

It would be nice to have these packages work. For example, by letting users choose and operate inside docker containers like on DigitalOcean. Otherwise, many interesting things will have to be done outside of CDF.

You are right. There are limitations with pydiode, but we have currently no plans of deploying any server side notebook or streamlit functionality. If you need that you can use any cloud provider and access CDF using our SDKs.