[GH-ISSUE #1397] suggestion on LLM hint to help reduce number of AI tags #888

Open
opened 2026-03-02 11:53:30 +03:00 by kerem · 3 comments
Owner

Originally created by @akohlsmith on GitHub (May 11, 2025).
Original GitHub issue: https://github.com/karakeep-app/karakeep/issues/1397

Describe the feature you'd like

the AI tag suggestion is nice, but I am wondering if I can modify the prompt for tag generation to do things like

  • eliminate underscores/punctuation from the suggested tags
  • look at the existing tag list and if similar tags are found, use them instead of creating new ones
  • if a human/custom tag is very similar to the AI tag, use that tag instead of creating a new one

Describe the benefits this would bring to existing Karakeep users

This feature would go a long way to reducing the "noise" in the AI generated tag list, although I am not sure if the prompt is able to tell the AI what tags are already in use and their count (and perhaps the website for the bookmark as well, which would go a long way to consolidating tags).

Can the goal of this request already be achieved via other means?

I see that custom prompts are available but I'm not sure quite how to use them and am a little gun shy of mis-prompting and causing a real mess in the tag editor. If there were a sandbox mode to test before going live, or to perhaps mark a tag as "testing" then it'd be easier to clean up any potential mess as well.

Have you searched for an existing open/closed issue?

  • I have searched for existing issues and none cover my fundamental request

Additional context

No response

Originally created by @akohlsmith on GitHub (May 11, 2025). Original GitHub issue: https://github.com/karakeep-app/karakeep/issues/1397 ### Describe the feature you'd like the AI tag suggestion is nice, but I am wondering if I can modify the prompt for tag generation to do things like * eliminate underscores/punctuation from the suggested tags * look at the existing tag list and if similar tags are found, use them instead of creating new ones * if a human/custom tag is very similar to the AI tag, use that tag instead of creating a new one ### Describe the benefits this would bring to existing Karakeep users This feature would go a long way to reducing the "noise" in the AI generated tag list, although I am not sure if the prompt is able to tell the AI what tags are already in use and their count (and perhaps the website for the bookmark as well, which would go a long way to consolidating tags). ### Can the goal of this request already be achieved via other means? I see that custom prompts are available but I'm not sure quite how to use them and am a little gun shy of mis-prompting and causing a real mess in the tag editor. If there were a sandbox mode to test before going live, or to perhaps mark a tag as "testing" then it'd be easier to clean up any potential mess as well. ### Have you searched for an existing open/closed issue? - [x] I have searched for existing issues and none cover my fundamental request ### Additional context _No response_
Author
Owner

@Eragos commented on GitHub (May 12, 2025):

Last days I have posted a comment about this topic https://github.com/karakeep-app/karakeep/discussions/843#discussioncomment-13080110
With an example how you can use it. Feel free to discuss with us over-there ;-)

<!-- gh-comment-id:2872851836 --> @Eragos commented on GitHub (May 12, 2025): Last days I have posted a comment about this topic https://github.com/karakeep-app/karakeep/discussions/843#discussioncomment-13080110 With an example how you can use it. Feel free to discuss with us over-there ;-)
Author
Owner

@ben-kenney commented on GitHub (May 18, 2025):

I think this is a good idea. IMHO AI tagging is currently defeating the value of tags because it doesn't reuse any existing tags.

For example, I can save one page and get tags for "stock market", "investing"
And another page gives me tags of "stocks", "investments"
-> 4 different tags that are basically the same thing

I tried the method that @Eragos has suggested but it doesn't seem to work. I think you also need to pass the ai a list of existing tags so it knows which ones to use.

It would also be nice to be able to rework the default prompt. For example:

Aim for a variety of tags, including broad categories, specific keywords, and potential sub-genres

This might not work towards some people's goals of minimizing the numbers of tags.

<!-- gh-comment-id:2888936413 --> @ben-kenney commented on GitHub (May 18, 2025): I think this is a good idea. IMHO AI tagging is currently defeating the value of tags because it doesn't reuse any existing tags. For example, I can save one page and get tags for "stock market", "investing" And another page gives me tags of "stocks", "investments" -> 4 different tags that are basically the same thing I tried the method that @Eragos [has suggested](https://github.com/karakeep-app/karakeep/discussions/843#discussioncomment-13080110) but it doesn't seem to work. I think you also need to pass the ai a list of existing tags so it knows which ones to use. It would also be nice to be able to rework the default prompt. For example: > Aim for a variety of tags, including broad categories, specific keywords, and potential sub-genres This might not work towards some people's goals of minimizing the numbers of tags.
Author
Owner

@thiswillbeyourgithub commented on GitHub (May 27, 2025):

I think this should be closed in favor of #111

<!-- gh-comment-id:2911887819 --> @thiswillbeyourgithub commented on GitHub (May 27, 2025): I think this should be closed in favor of #111
Sign in to join this conversation.
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
starred/karakeep#888
No description provided.