Hi there,
I am trying to do some analysis that will eventually be deployed on our dashboard. The analysis requires python libraries that are not on condo but can be git cloned. I know can do this via Notebook command, but not sure if it makes sense that this will need to happen every time I run the notebook workflow(probably after each device reporting).
Could you advice the best course for this?
The notebook environment does not have the ability to pull in packages that are not included in the base image you select for security reasons, whether thye’re from Conda, Pip, or Git repos.
That said, we are actually publishing a new base image in our next release, and if you have particular packages in mind, please let us know what they are and we may be able to include them in the image. We typically only consider packages that are published on Pip and that are widely used, maintained, and come from trusted sources.
Thanks for the reply,
The package I want is pretty niche, but it is from NIST
https://pages.nist.gov/pyMCR/
If there is any work around to this let me know.
Thanks
Michelle
Thanks @Andreia_Smith-Moritz . I’ll send this request over to the engineering team and let you know if they can include it in our next notebook image. If so, it will be in our next platform release, which will be a few weeks from now (but should be by year’s end).
Thanks Dylan! Losant has been a huge for us prototyping and piloting our technology.
Is there a way to install packages now? need cantools
Hi @Jie_Li and welcome to Losant!
It is still not possible for users to import their own packages and I don’t see us changing that policy. There are serious information security concerns with allowing arbitrary package imports into our notebook environment.
However, we still can publish new notebook images that expose new packages per user requests such as this one. Are there any packages you need other than cantools?
cantools and python-can. Thank you
One thing I just want to clarify - Losant’s Notebooks feature runs in a sandboxed, offline environment. It is meant for deep analysis of large data sets through Jupyter notebooks. You must explicitly define the inputs to provide to the notebook, which are then downloaded and provided as static files to the execution environment.
I can see a use case for cantools as it exposes several methods for parsing CANbus data, which could be provided to the notebook as a static text file. But python-can seems to be strictly a Python interface for communicating with CANbus hardware - something you cannot do inside of a notebook environment.
Are you instead trying to work with CANbus devices in an Edge Workflow?
thanks, I only need the decode function, so it will work on the notebook directly.