Plainbook -- LaTeX & Table of Contents Implementation

[Uchechi E]

Hi all, I was working on a project with plainbook involving EDA and I thought of a simple QOL addition that I thought we could consider adding.

Similar to how Kaggle has a table of contents for Markdown & how Jupyter Notebook showcases the headings in a notebook, I was thinking that we could implement something similar to help users navigate their project a little bit better.

I also thought about the idea of adding LaTeX support for comment cells, but I am a bit hesitant about this idea as I believe it might be a bit unnecessary. However, I did add that feature in the explanation for cells. Nonetheless, I would like to hear your guys' thoughts. Let me know your thoughts. Thanks!

I have provided a picture reference of what I am thinking below.

[Luca de Alfaro]

I would say, yes to both ideas. 

(btw sorry -- do a git fetch -- I forgot to check in a couple of files -- sorry). 

Adding latex support is really quite easy via mathjax.  I think it will be useful in general.  There is no reason not to do it. 

For the other idea, ... yes, but how precisely? 

I think that when we save the explanation, right now we ask for a 2-3 word summary of the cell to give the name to the cell.  I think it would be ideal to ask, in the same AI call, for a short one-line summary of the cell, and store that independently; we could then use that for the TOC.  Or do you have better ideas on how to give names to cells that we can use in the index?  

Well, another simple idea would be to use the beginning of the description of course. 

Thoughts? 

Luca

--
To post to the group, email plai...@googlegroups.com
---
You received this message because you are subscribed to the Google Groups "Plainbook" group.
To unsubscribe from this group and stop receiving emails from it, send an email to plainbook+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/plainbook/9bf77484-eff7-4662-bde0-63f5d89b2c6an%40googlegroups.com.

[Uchechi E]

I think to lessen token usage, we can make it so that the text in H1 comments is the text that shows for the TOC. And then the text in H2 comments is the ones a bit smaller under the H1 comment.


Here is an example:

Comment cell #1: # Header 1 text: Dataset preparation
Comment cell #2: ## Sub-header 2 text: Cleaning up Data
Comment cell #3: # Header 3 text: Model Architecture

TOC:
| Dataset preparation
|     - Cleaning up Data
| Model Architecture
|

This way, we do not have to use AI to generate descriptions for the cell names & can save token usage.
However, on that note, I did have the idea of having the ability to rename cell names as well.

[Luca de Alfaro]

Yes, we can certainly use the text in comment cells, and their H1/H2/... in markdown. 

The problem I see is that, given that the cells are already in natural language, there seems to be a bit less of an incentive to put even more comment cells in.  I don't use them a lot for instance. 

But maybe we can do a mix: Use markdown headers in bold in the Toc, for #, ##, ###, and then for code cells, simply use the beginning of the description and then dots. 

Where would you put this? One idea would be on the left, normally "folded" (hidden) but you can unfold it in a TOC panel on the left side of the screen... is this what you have in mind? 

To view this discussion visit https://groups.google.com/d/msgid/plainbook/34cf978e-7f34-4bc3-b61f-9b6c560bf399n%40googlegroups.com.