What's your use case?
In other words, what's your pain point?
Ans: Everytime we have some new data, we try to make the same pipeline which somewhat starts off the same, for example forst load then preprocess, etc etc.

Is your request related to a problem, or perhaps a frustration?
Ans: Its related to frustration, the integration of a chat based template creator would help us save a lot of time making the same pipeline over and over again for different projects.

Tell us the story that led you to write this request.
I was trying out orange with a lot of different datasets due to changing use cases, and before building the solution I would try out the models with orange. What I found myself doing is making the same pipeline with 10% changes.

What's your proposed solution?
Basically we can make an LLM train of some sets of predefined pipelines to understand the connections and the steps what would be traditionally followed and then help users chat with orange (lets call it OrangeJuice) to help make some templates for them. The chat would also help the user edit and understand what next steps should be taken based on the data they have. This can also be done with some local llm that runs on CPU/GPU.

Are there any alternative solutions?
Templates would be close, but that requires a very exhaustive list of all the templates possible. Also this would need a lot of manual effort by user to edit if they ever need to edit a template.