Update ideas - confirmation for developing updates

[Uchechi E]

Good afternoon, everyone! I apologize if it seems I have been inactive; I was in San Francisco for the past couple of days. I just wanted to cover a few things in this email.

1. Am I in the clear to start working on the small QOL LaTeX & TOC update? Just wanted to know before I pushed out any updates.

2. I have noticed a small detail about cell generation: when a small change is created to the cell description, despite the change, plainbook will automatically regenerate code for the cell. (eg. Cell #1 "In this cell, create a Fibbonaci Sequence", then ran, but then the user changes it to "Create a Fibbonaci Sequence" and runs the code again). This could pose a problem for users who have limited usage rates for their AI's or for users trying to save on token usage (along with this, for each cell & for the entire project, a user's token usage count would be a great addition as well).

Let me know your thoughts on this. If it is an issue that we'd need to program a solution for, I do have some ideas & I'd be happy to help develop it!

I also wanted to let you all know that I will be flying back home to KC on Aug. 6th, in case if anything is brought up.

For recently pushed updates, I tried out candidate 0.4.0, and everything seemed great.

That's all I have. I hope everyone is doing well!!

[Luca de Alfaro]

Hi Uchechi! 

Yes, you can start work on latex and TOC, with the following notes: 

0.  Take the develop branch as starting point.  I am making a few commits to it, but my commits are in a very separate area, so they should not interfere with what you plan to do and we should be able to easily merge. 

1. For latex, you don't have to do anything server-side, we should simply include mathjax on the client, and it will take care of the rendering.  Please include a static mathjax .js among the packages, just as I have done; it's all static, so one can view a notebook even if the internet connection is down. 

2. For the content bar, it's as we discussed, make it foldable out from the left part of the screen.  If you go to ... overleaf.com, note how you can have a side bar that can be folded in (as in image below) then fold out.  Try to do something similar, including ability to manually resize.

I think a good idea would be to put the code for the sidebar in a Vue component SideIndex.js or some such name (up to you).  

For the "small changes": we have done already a lot of work on trying to ensure that a small change to a cell does not require the _subsequent_ cells to be rebuilt.  There is more to do and you can join the discussions, with Ranjit and Dhyan, on the topic if interested.  But the point is, for the same cell, there is no easy way other than asking an LLM (that is, an LLM call) to know whether a change is small or not.  How would you do otherwise?  A small (in n. of characters) change can sometimes entirely change the semantics.  Do you have ideas on this?  I don't see a good solution... ? 

Again THANKS for contributing!!

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/0494d875-2245-4aea-bc0e-89aa69f7b59fn%40googlegroups.com.

[Uchechi E]

Alright! I will make sure to keep those notes in mind while working on the update.

That is the problem I was thinking of as well. I have two solutions that might help solve this problem: 

1. Because plainbook runs linearly, we could add an if statement to check if the cell description has changed. If it has, then we use a small LLM model to check if the new description is similar to the old one (such as Ollama). However, this one seems like quite a hassle to implement

2. We could also use a regular LLM. After the if statement, if the description is not the same, then the cell goes through an AI prompt asking if the text is the same. It will then output 0 if it is virtually the same and 1 if it is a completely different task. It will save its response to a variable in the code, and before running the code, we check if that variable is 0 or 1.

These are the two ideas I have, but I'm not necessarily too confident in them, and I feel like they have already been proposed.

For Ranjit and Dhyan, what are your ideas on solving this problem? Please let me know what you have, thank you!

[Luca de Alfaro]

Let me respond in a new thread to make it easier to find. 

Luca

To view this discussion visit https://groups.google.com/d/msgid/plainbook/4fc07655-ae5f-4a72-a3d2-2394e83330aen%40googlegroups.com.