This AI-programmed Obsidian plugin can automatically extract highlighted text from notes, allowing users to add comments, generate AI comments, and interact with the highlighted text. Users can highlight text in various formats and export it as knowledge card images, while enjoying more extended features in the main view.
1 Like
You can view all notes and highlight cards containing highlights in the knowledge base, and you can also have a conversation with AI and highlight cards.
1 Like
Looks nice.
Question:
What models can you load?
Is there plans to support text not in highlights, but say, within backticks or ranges defined by regular expressions?
Thanks 
Does the model refer to the AI model? Currently, AI services such as OpenAI, Gemini, Claude, and Ollma are supported.
At present, there are only three formats highlighted ==, mark, and span. Looking at the follow-up user feedback may take the function of getting the format out and customize it in the settings.
1 Like
Maybe LM Studio could be added too?
Settings in config with customizable stuff sounds good.
I know regex is not usually something the average end user needs or should be confused with, but other users welcome it immensely.
How are the comments stored?
Obsidian comments are in the form %%comment%%.
If the comments are in the json, does that exclude them from core search?
It will be interesting to see how Obsidian implements annotations.
I like where you’re going with this.
1 Like
Thank you , that’s a good suggestion. Indeed, if it’s stored in a JSON file, it can’t be searched. But I feel like inserting comments into the MD document might disrupt the original document structure of the users. This issue needs further deep thinking.
LM is not being considered for the time being. I need to observe more user needs to make a decision.
Agreed, this is a challenge, but … keep in mind that we’re trying to maintain markdown as the storage mechanism: highlights and comments are standard markdown format.
Or, at least, I wish to maintain markdown integrity:)
So, entering comments into json and using html to style the highlight would be a deal breaker for me. TBH, most users don’t care or won’t until they migrate to a new system and have to deal with it.
The thing I like about your work so far is the creation of cards and a way to traverse the highlights and comments outside the note.
There is another plugin that does some of this - Enhanced Annotations. The interesting thing to me with that plugin is the little number in the sidebar beside each highlight and comment. It knows what the note is, searches for each occurance of == highlight == and %% comment %% in the note and uniquely identifies it. Have not seen that anywhere else.
My workflow is to make standard highlights (no extra colors) and place comments at the end of a block; this way it leaves the original content unfettered. Both are searchable at the block level.
I’d recommend leaving the decision to the user - maybe a toggle in settings that let’s them make the call - leave highlights and comments embedded or offer to keep them in json and inform them of the risk.
It may behoove you to continue with the card display work and wait to see what Obsidian does with annotations.
Best wishes
Thank you for your feedback. The solution of inserting the comment content at the bottom of the document and leaving the choice of using JSON or %% to the user is feasible.
The current issue that may need to be addressed is how to bind highlighted text and comments in different formats and display them in the form of cards in the right sidebar.