-
Notifications
You must be signed in to change notification settings - Fork 1.5k
Description
In trying out this project with locally-run llm's (qwen3:8b), I found that the documentation isn't always completely correct, so it'd be great if there was an easy way to tell it to make edits afterwards. The expectation being that the edits happen with the same RAG enhancement as initial creation, but now with additional hints given by the user. For that it might also be helpful if the user could manually add additional relevant source files.
I believe the simplest way of achieving this would be by adding a button at the bottom of each page "Request an Edit", which will give an input prompt where the user can tell the LLM what needs to be changed. From there, the page is regenerated just like during the initial generation, but this time with the context added by the user.
A more involved way would be to have an annotations section like in Microsoft Word or Google Docs, where people can add their annotations to single paragraphs/ elements. A user could add all of their comments, and then that would be sent to the model to regenerate the page if necessary, or just the specific section(s).