[GH-ISSUE #4355] Add the ability to delete a single message (only the selected one, instead of deleting all messages after it) #133306

Closed
opened 2026-05-24 22:33:58 -05:00 by GiteaMirror · 6 comments
Owner

Originally created by @CookSleep on GitHub (Aug 4, 2024).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/4355

Is your feature request related to a problem? Please describe.
Many times, I only need to delete a single message in the conversation to remove unnecessary information and continue the discussion. However, the current delete function removes all messages after the selected one, which is very annoying.

Describe the solution you'd like
Add the ability to delete individual messages.

Describe alternatives you've considered
Replace the message that needs to be deleted with a space or something else, but this will make the chat history look very odd.

Originally created by @CookSleep on GitHub (Aug 4, 2024). Original GitHub issue: https://github.com/open-webui/open-webui/issues/4355 **Is your feature request related to a problem? Please describe.** Many times, I only need to delete a single message in the conversation to remove unnecessary information and continue the discussion. However, the current delete function removes all messages after the selected one, which is very annoying. **Describe the solution you'd like** Add the ability to delete individual messages. **Describe alternatives you've considered** Replace the message that needs to be deleted with a space or something else, but this will make the chat history look very odd.
Author
Owner

@GrayXu commented on GitHub (Aug 6, 2024):

Messages can be deleted at the granularity of individual q-and-a pairs.

<!-- gh-comment-id:2271612934 --> @GrayXu commented on GitHub (Aug 6, 2024): Messages can be deleted at the granularity of individual q-and-a pairs.
Author
Owner

@justinh-rahb commented on GitHub (Aug 6, 2024):

Messages can be deleted at the granularity of individual q-and-a pairs.

@GrayXu the safest method is to remove the pair, some models behave weirdly if you feed them a chat context that seems disjointed and out of order (not following user, assistant, user, assistant, etc. format).

<!-- gh-comment-id:2271658869 --> @justinh-rahb commented on GitHub (Aug 6, 2024): > Messages can be deleted at the granularity of individual q-and-a pairs. @GrayXu the safest method is to remove the pair, some models behave weirdly if you feed them a chat context that seems disjointed and out of order (not following user, assistant, user, assistant, etc. format).
Author
Owner

@thiswillbeyourgithub commented on GitHub (Aug 6, 2024):

Messages can be deleted at the granularity of individual q-and-a pairs.

@GrayXu the safest method is to remove the pair, some models behave weirdly if you feed them a chat context that seems disjointed and out of order (not following user, assistant, user, assistant, etc. format).

To me that's the point. If I want to experiment with breaking LLMs I should be able to. Also there's got to be some models finetuned in a way that makes it a legit use case.

<!-- gh-comment-id:2271702008 --> @thiswillbeyourgithub commented on GitHub (Aug 6, 2024): > > Messages can be deleted at the granularity of individual q-and-a pairs. > > @GrayXu the safest method is to remove the pair, some models behave weirdly if you feed them a chat context that seems disjointed and out of order (not following user, assistant, user, assistant, etc. format). To me that's the point. If I want to experiment with breaking LLMs I should be able to. Also there's got to be some models finetuned in a way that makes it a legit use case.
Author
Owner

@justinh-rahb commented on GitHub (Aug 6, 2024):

To me that's the point. If I want to experiment with breaking LLMs I should be able to. Also there's got to be some models finetuned in a way that makes it a legit use case.

This sort of experimentation is more easily done in the Playground area of the application. You can set the message roles arbitrarily at any time and re-run generation.

<!-- gh-comment-id:2272213372 --> @justinh-rahb commented on GitHub (Aug 6, 2024): > To me that's the point. If I want to experiment with breaking LLMs I should be able to. Also there's got to be some models finetuned in a way that makes it a legit use case. This sort of experimentation is more easily done in the Playground area of the application. You can set the message roles arbitrarily at any time and re-run generation.
Author
Owner

@thiswillbeyourgithub commented on GitHub (Aug 7, 2024):

To me that's the point. If I want to experiment with breaking LLMs I should be able to. Also there's got to be some models finetuned in a way that makes it a legit use case.

This sort of experimentation is more easily done in the Playground area of the application. You can set the message roles arbitrarily at any time and re-run generation.

Good point I am now convinced. I had forgotten about even trying the playground because it's not intuitive to go in the bottom left menu. I'll open an issue for that if there isn't any.

<!-- gh-comment-id:2272781678 --> @thiswillbeyourgithub commented on GitHub (Aug 7, 2024): > > To me that's the point. If I want to experiment with breaking LLMs I should be able to. Also there's got to be some models finetuned in a way that makes it a legit use case. > > This sort of experimentation is more easily done in the Playground area of the application. You can set the message roles arbitrarily at any time and re-run generation. Good point I am now convinced. I had forgotten about even trying the playground because it's not intuitive to go in the bottom left menu. I'll open an issue for that if there isn't any.
Author
Owner

@tjbck commented on GitHub (Aug 27, 2024):

Closing in favour of #4923

<!-- gh-comment-id:2312387232 --> @tjbck commented on GitHub (Aug 27, 2024): Closing in favour of #4923
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/open-webui#133306