OpenAI Python Package

Hi,
Any chance of getting the OpenAI package added to the bundled environment? I wanted to share a workflow with a co-worker without having to go through the headache of having them create a custom conda environment.

Alternatively, is it possible to point the KNIME Python scripting node at the knime_python_llm environment that’s bundled with AI extension?

Not sure if there are any plans to expand the bundled environment with the packages you request. As far as I know the only way for now is to use Conda Propagation Node to make the env available to anyone who uses your workflow. So the prep work is more on you and as long as your colleague has conda installed and configured in their KAP it should be smooth sailing.

I wrote about how to create your own env and use it with Conda environment propagation here

2 Likes

Trying to avoid having them install conda…may just go down the path of learning python node development

The custom node approach ended up working really well. My colleague was able to install it on a clean computer (no Conda / standard KNIME install) and execute a workflow incorporating the custom node successfully.

I used the python node development tutorial to learn the basic mechanics of writing a new node and how to edit the configuration files, @VAGR_ISK’s Python-KNIME node development GPT (Custom GPT to help you building Python-Based KNIME Extensions from ISK) to help port over an existing KNIME python script, and then I poked around the Geospatial Analytics extension files on Github (GitHub - spatial-data-lab/knime-geospatial-extension: This repository is built for KNIME-CGA Geospatial Project, and the goal is to build Python-based nodes for geospatial analysis in KNIME Analytic Platform.) to understand what a properly developed python node looks like.

With those three resources, it took me the better part of a day to get to a functional proof of concept, and I think it’s a pretty accessible approach to solving the problem

3 Likes

wow - kudos :raised_hand: for making that work! It is still on my list to also learn this (have started a few times, but never finished) so thanks for sharing your experience and also the additional resources!

I definitely think it’s worth the effort. I spend most of my development time figuring out how to fully package KNIME workflows for my colleagues to use as push-button tools and this is going to unlock some really interesting solutions.

And honestly, you’re technical skills are so far ahead of where I am at that I am sure you’ll get it all knocked out in an afternoon

2 Likes

Thanks for the encouragement and resources again - actually managed to bundle my first extension on the weekend! A node to interact with local LLM hosted via Ollama :-).

2 Likes