You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
For text as well as for code workers it would be great to have AI text revisions where changes e.g. in form of proofreading or prompt-based specific rewrites can be tracked and interactively approved or rejected. With the help of AI, I have spun up a working prototype that creates an odt file(libreoffice text format that can also be opened in MS Word) where the AI's edits are presented in track changes mode. The output looks like this:
(these edits have been generated by Gemini with a prompt requesting light edits to improve readability by younger readers - as well as explanations for the changes which are not shown here)
I am trying to add comments as something that can be rendered as well, this is where the AIs reasoning behind certain changes could live (which might be helpful in the case of reasoning models #109 . I am experimenting with <cmnt> </cmnt> tags for that (see the last two iterations on lovable). Once this is done I can have it refactor and clean up the code and push it to GitHub - but do have a look and see if you d like to remix it on lovable to steer the implementation in a way that is easier to integrate into writing tools.
If this could become part of writing-tools, it would make it a complete text revisions suite and take away the pain point of having to parse the entire generated output to spot changes. This also goes a long way to ensure transparency and interpretability - important for team work or just longer texts.
The text was updated successfully, but these errors were encountered:
Hello! Your idea is great, and it's something that's certainly on the table to add! #25 is about this.
Thank you for sharing your implementation—it looks cool! However, for Writing Tools, it would be more efficient for me to directly write something in Python + PySide when I'm working on it.
Great to know its on your todo list, sorry I missed #25 I'd have posted my idea there if I had found it during my pre-posting search.-Feel free to merge the above there if you prefer, though I d say my implementation idea is a bit different. Here is a lovable created implementation of the same functionality, but based on python and PySide
Tbh this code goes beyond what i can evaluate, so I hope it is a helpful starting point. The basic functionality works, what doesn't work yet in the exported odt file is displaying not just track changes but comments as well.
Maybe you can have a look and are able to get a bit further with lovable before manual finishing touches. I hope it saves time and makes it possible for the feature to arrive sooner. If the lovable Ai isn't there yet, just disregard the code.
For text as well as for code workers it would be great to have AI text revisions where changes e.g. in form of proofreading or prompt-based specific rewrites can be tracked and interactively approved or rejected. With the help of AI, I have spun up a working prototype that creates an odt file(libreoffice text format that can also be opened in MS Word) where the AI's edits are presented in track changes mode. The output looks like this:
(these edits have been generated by Gemini with a prompt requesting light edits to improve readability by younger readers - as well as explanations for the changes which are not shown here)
find it here https://text-changes-tracker.lovable.app/
code is here https://lovable.dev/projects/209231b2-569d-48a5-905c-eede5748e301
I am trying to add comments as something that can be rendered as well, this is where the AIs reasoning behind certain changes could live (which might be helpful in the case of reasoning models #109 . I am experimenting with
<cmnt> </cmnt>
tags for that (see the last two iterations on lovable). Once this is done I can have it refactor and clean up the code and push it to GitHub - but do have a look and see if you d like to remix it on lovable to steer the implementation in a way that is easier to integrate into writing tools.If this could become part of writing-tools, it would make it a complete text revisions suite and take away the pain point of having to parse the entire generated output to spot changes. This also goes a long way to ensure transparency and interpretability - important for team work or just longer texts.
The text was updated successfully, but these errors were encountered: