I have searched the existing issues and discussions.
Problem Description
Thank you for developing such a convenient frontend. I mainly enjoy role-playing interactions with models, and although I've used SillyTavern before—which is quite suitable for role-play—I personally prefer the feel of OpenWebUI. After using it extensively, I'd like to share a few suggestions for feature improvements.
Desired Solution you'd like
Feature 1: I’d like to introduce a new chat mode specifically designed for role-playing. When the user remains silent, pressing Enter or clicking the send button should prompt the current model to automatically generate a reply. For instance, if I'm chatting with a model named "Jack," normally I have to speak first for Jack to respond. If I stay silent and hope for Jack’s next response, I currently need to send something like “And then?” or “What happens next?”—but these prompts often feel redundant. I’d prefer that the character continues speaking naturally until I choose to participate.
Feature 2: Enable multi-model group chats. While the @ mention feature can simulate this, it doesn’t quite feel as natural or intuitive.
Feature 3: Allow users to invoke and temporarily embody other models during a conversation, enabling direct interaction between models. This would work like a group chat—imagine I (the user) have been chatting with Jack for several messages, but then suddenly decide to act as Mike (another model), engaging Jack in conversation from Mike’s perspective, using Mike's avatar and personality—not as the user sending messages.
I know my suggestions might be quite numerous, and I apologize. I understand that development isn't easy and not everything can be implemented immediately. Still, I truly hope these features could someday be included. Thank you so much for your hard work!
Alternatives Considered
No response
Additional Context
No response
Originally created by @leetraman822 on GitHub (Aug 22, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/16828
### Check Existing Issues
- [x] I have searched the existing issues and discussions.
### Problem Description
Thank you for developing such a convenient frontend. I mainly enjoy role-playing interactions with models, and although I've used SillyTavern before—which is quite suitable for role-play—I personally prefer the feel of OpenWebUI. After using it extensively, I'd like to share a few suggestions for feature improvements.
### Desired Solution you'd like
Feature 1: I’d like to introduce a new chat mode specifically designed for role-playing. When the user remains silent, pressing Enter or clicking the send button should prompt the current model to automatically generate a reply. For instance, if I'm chatting with a model named "Jack," normally I have to speak first for Jack to respond. If I stay silent and hope for Jack’s next response, I currently need to send something like “And then?” or “What happens next?”—but these prompts often feel redundant. I’d prefer that the character continues speaking naturally until I choose to participate.
Feature 2: Enable multi-model group chats. While the @ mention feature can simulate this, it doesn’t quite feel as natural or intuitive.
Feature 3: Allow users to invoke and temporarily embody other models during a conversation, enabling direct interaction between models. This would work like a group chat—imagine I (the user) have been chatting with Jack for several messages, but then suddenly decide to act as Mike (another model), engaging Jack in conversation from Mike’s perspective, using Mike's avatar and personality—not as the user sending messages.
I know my suggestions might be quite numerous, and I apologize. I understand that development isn't easy and not everything can be implemented immediately. Still, I truly hope these features could someday be included. Thank you so much for your hard work!
### Alternatives Considered
_No response_
### Additional Context
_No response_
It's truly fascinating to engage different models, each embodying a role from a distinct professional field, in a group discussion where they debate the same issue on a shared topic.
<!-- gh-comment-id:3214917007 -->
@i-iooi-i commented on GitHub (Aug 22, 2025):
It's truly fascinating to engage different models, each embodying a role from a distinct professional field, in a group discussion where they debate the same issue on a shared topic.
<!-- gh-comment-id:3215014020 -->
@rgaricano commented on GitHub (Aug 22, 2025):
You can try @pkeffect Agent-Hot Swap and Multimodal functions for the features you're targeting.
There are on openwebui.com but I link the repo (most updated):
https://github.com/pkeffect/functions/tree/main/functions/filters
You can try @pkeffect Agent-Hot Swap and Multimodal functions for the features you're targeting.
Thanks, I've tried out both of these filters. While they can simulate group chats to some extent, the control isn't very good—especially prone to overfitting in character traits. For example, Jack is supposed to speak in a calm and peaceful tone, while Mike speaks enthusiastically and energetically. But when Jack speaks first, Mike's responses tend to unconsciously adopt Jack’s calmer style.
Currently, my workaround for simulating group chats goes like this: I use @mentions to simulate interactions. For instance, I start a new chat with Jack. I say, "Good morning, everyone." Then Jack replies, "Good morning, user! How's your day going?" At that point, I manually edit his message to read: "Jack: Good morning, user! How's your day going?"—adding the character name as a prefix. Then I continue with, "Doing pretty well, is Mike around?" When typing this, I use @Mike to trigger Mike’s response, which comes back as, "Here I am, dude!" At that moment, I manually add the prefix: "Mike:" before his message.
By following this method, I can effectively avoid cross-contamination of speaking styles between characters. It's a bit unconventional, haha—but it works. The key difference from using filters is that each character generates their own messages independently, rather than merging all voices into one unified text stream. This gives me much better control over individual speech patterns.
So, if we could achieve this kind of group chat simulation—where characters speak separately with clear identity and consistent style—it would be ideal!
<!-- gh-comment-id:3216196458 -->
@leetraman822 commented on GitHub (Aug 23, 2025):
> You can try [@pkeffect](https://github.com/pkeffect) Agent-Hot Swap and Multimodal functions for the features you're targeting.
Thanks, I've tried out both of these filters. While they can simulate group chats to some extent, the control isn't very good—especially prone to overfitting in character traits. For example, Jack is supposed to speak in a calm and peaceful tone, while Mike speaks enthusiastically and energetically. But when Jack speaks first, Mike's responses tend to unconsciously adopt Jack’s calmer style.
Currently, my workaround for simulating group chats goes like this: I use @mentions to simulate interactions. For instance, I start a new chat with Jack. I say, "Good morning, everyone." Then Jack replies, "Good morning, user! How's your day going?" At that point, I manually edit his message to read: "Jack: Good morning, user! How's your day going?"—adding the character name as a prefix. Then I continue with, "Doing pretty well, is Mike around?" When typing this, I use @Mike to trigger Mike’s response, which comes back as, "Here I am, dude!" At that moment, I manually add the prefix: "Mike:" before his message.
By following this method, I can effectively avoid cross-contamination of speaking styles between characters. It's a bit unconventional, haha—but it works. The key difference from using filters is that each character generates their own messages independently, rather than merging all voices into one unified text stream. This gives me much better control over individual speech patterns.
So, if we could achieve this kind of group chat simulation—where characters speak separately with clear identity and consistent style—it would be ideal!
<!-- gh-comment-id:3216219723 -->
@pkeffect commented on GitHub (Aug 23, 2025):
You nay want to check your settings on my filters then because they work as they should for me and others who have tested them.
Edited to say, welcome suggestions and PRs for those as well
https://github.com/pkeffect/functions
@pkeffect Hi, could you please unify the UI icon colors with those of OPEN WEBUI to ensure a consistent and cohesive interface aesthetic?
There are color discrepancies between the UI icons and OPEN WEBUI's icons, evident in both their active and inactive states.
<!-- gh-comment-id:3218775898 -->
@i-iooi-i commented on GitHub (Aug 25, 2025):
@pkeffect Hi, could you please unify the UI icon colors with those of OPEN WEBUI to ensure a consistent and cohesive interface aesthetic?
There are color discrepancies between the UI icons and OPEN WEBUI's icons, evident in both their active and inactive states.
Is it possible to simulate this segmented group chat display to enhance immersion? Please note, the accompanying image is merely a Photoshop rendition, not an actual feature.
<!-- gh-comment-id:3218795836 -->
@i-iooi-i commented on GitHub (Aug 25, 2025):
<img width="1731" height="1757" alt="Image" src="https://github.com/user-attachments/assets/11356e71-d7fd-4680-b6a3-5cb4a77dece6" />
Is it possible to simulate this segmented group chat display to enhance immersion? Please note, the accompanying image is merely a Photoshop rendition, not an actual feature.
@leetraman822
Could this be the project you're seeking: https://github.com/maojindao55/botgroup.chat , which offers a group chat experience featuring multiple AI entities interacting?
<!-- gh-comment-id:3218799710 -->
@i-iooi-i commented on GitHub (Aug 25, 2025):
@leetraman822
Could this be the project you're seeking: [https://github.com/maojindao55/botgroup.chat ](https://github.com/maojindao55/botgroup.chat), which offers a group chat experience featuring multiple AI entities interacting?
Yes, actually there are many similar group chat features available. The first one I came across was SillyTavern—it has very powerful and well-developed role-playing capabilities. However, possibly due to my own settings, the same prompts, temperature, top-k, top-p values, etc., result in responses that aren't as intelligent as those from LM Studio or Open WebUI. I'm not sure what's causing this discrepancy. So I really hope that Open WebUI could eventually support group chats, or at least allow appending character replies with the Enter key.
<!-- gh-comment-id:3219614192 -->
@leetraman822 commented on GitHub (Aug 25, 2025):
> Could this be the project you're seeking: (https://github.com/maojindao55/botgroup.chat), which offers a group chat experience featuring multiple AI entities interacting?
Yes, actually there are many similar group chat features available. The first one I came across was SillyTavern—it has very powerful and well-developed role-playing capabilities. However, possibly due to my own settings, the same prompts, temperature, top-k, top-p values, etc., result in responses that aren't as intelligent as those from LM Studio or Open WebUI. I'm not sure what's causing this discrepancy. So I really hope that Open WebUI could eventually support group chats, or at least allow appending character replies with the Enter key.
@pkeffect Hi, could you please unify the UI icon colors with those of OPEN WEBUI to ensure a consistent and cohesive interface aesthetic?
There are color discrepancies between the UI icons and OPEN WEBUI's icons, evident in both their active and inactive states.
@i-iooi-i
Thanks for that. This has been fixed. Version number has not changed but the image has been updated and tested on light and dark themes.
<!-- gh-comment-id:3221676754 -->
@pkeffect commented on GitHub (Aug 25, 2025):
> [@pkeffect](https://github.com/pkeffect) Hi, could you please unify the UI icon colors with those of OPEN WEBUI to ensure a consistent and cohesive interface aesthetic?
> There are color discrepancies between the UI icons and OPEN WEBUI's icons, evident in both their active and inactive states.
@i-iooi-i
Thanks for that. This has been fixed. Version number has not changed but the image has been updated and tested on light and dark themes.
@pkeffect Hi, thanks a lot. I've just tested it, and it seems there's still a small problem: the icons consistently display in gray, rather than turning blue like the integrated ones.These are just minor issues that don't impact usability. Thank you again! ♥
<!-- gh-comment-id:3221721198 -->
@i-iooi-i commented on GitHub (Aug 25, 2025):
@pkeffect Hi, thanks a lot. I've just tested it, and it seems there's still a small problem: the icons consistently display in gray, rather than turning blue like the integrated ones.These are just minor issues that don't impact usability. Thank you again! ♥
<img width="1345" height="742" alt="Image" src="https://github.com/user-attachments/assets/49f74a85-6e71-40a2-8b9d-6a140a997167" />
The new icons identify the theme and change accordingly. Either light or dark now to be more consistent with OI icons, no colors.
<!-- gh-comment-id:3221726925 -->
@pkeffect commented on GitHub (Aug 25, 2025):
@i-iooi-i
The new icons identify the theme and change accordingly. Either light or dark now to be more consistent with OI icons, no colors.
@pkeffect Understood. My sincere thanks for developing this plugin; it's truly excellent, and I anticipate continuing to use it.
<!-- gh-comment-id:3221733075 -->
@i-iooi-i commented on GitHub (Aug 25, 2025):
@pkeffect Understood. My sincere thanks for developing this plugin; it's truly excellent, and I anticipate continuing to use it.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Originally created by @leetraman822 on GitHub (Aug 22, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/16828
Check Existing Issues
Problem Description
Thank you for developing such a convenient frontend. I mainly enjoy role-playing interactions with models, and although I've used SillyTavern before—which is quite suitable for role-play—I personally prefer the feel of OpenWebUI. After using it extensively, I'd like to share a few suggestions for feature improvements.
Desired Solution you'd like
Feature 1: I’d like to introduce a new chat mode specifically designed for role-playing. When the user remains silent, pressing Enter or clicking the send button should prompt the current model to automatically generate a reply. For instance, if I'm chatting with a model named "Jack," normally I have to speak first for Jack to respond. If I stay silent and hope for Jack’s next response, I currently need to send something like “And then?” or “What happens next?”—but these prompts often feel redundant. I’d prefer that the character continues speaking naturally until I choose to participate.
Feature 2: Enable multi-model group chats. While the @ mention feature can simulate this, it doesn’t quite feel as natural or intuitive.
Feature 3: Allow users to invoke and temporarily embody other models during a conversation, enabling direct interaction between models. This would work like a group chat—imagine I (the user) have been chatting with Jack for several messages, but then suddenly decide to act as Mike (another model), engaging Jack in conversation from Mike’s perspective, using Mike's avatar and personality—not as the user sending messages.
I know my suggestions might be quite numerous, and I apologize. I understand that development isn't easy and not everything can be implemented immediately. Still, I truly hope these features could someday be included. Thank you so much for your hard work!
Alternatives Considered
No response
Additional Context
No response
@i-iooi-i commented on GitHub (Aug 22, 2025):
It's truly fascinating to engage different models, each embodying a role from a distinct professional field, in a group discussion where they debate the same issue on a shared topic.
@rgaricano commented on GitHub (Aug 22, 2025):
You can try @pkeffect Agent-Hot Swap and Multimodal functions for the features you're targeting.
There are on openwebui.com but I link the repo (most updated):
https://github.com/pkeffect/functions/tree/main/functions/filters
@leetraman822 commented on GitHub (Aug 23, 2025):
Thanks, I've tried out both of these filters. While they can simulate group chats to some extent, the control isn't very good—especially prone to overfitting in character traits. For example, Jack is supposed to speak in a calm and peaceful tone, while Mike speaks enthusiastically and energetically. But when Jack speaks first, Mike's responses tend to unconsciously adopt Jack’s calmer style.
Currently, my workaround for simulating group chats goes like this: I use @mentions to simulate interactions. For instance, I start a new chat with Jack. I say, "Good morning, everyone." Then Jack replies, "Good morning, user! How's your day going?" At that point, I manually edit his message to read: "Jack: Good morning, user! How's your day going?"—adding the character name as a prefix. Then I continue with, "Doing pretty well, is Mike around?" When typing this, I use @Mike to trigger Mike’s response, which comes back as, "Here I am, dude!" At that moment, I manually add the prefix: "Mike:" before his message.
By following this method, I can effectively avoid cross-contamination of speaking styles between characters. It's a bit unconventional, haha—but it works. The key difference from using filters is that each character generates their own messages independently, rather than merging all voices into one unified text stream. This gives me much better control over individual speech patterns.
So, if we could achieve this kind of group chat simulation—where characters speak separately with clear identity and consistent style—it would be ideal!
@pkeffect commented on GitHub (Aug 23, 2025):
You nay want to check your settings on my filters then because they work as they should for me and others who have tested them.
Edited to say, welcome suggestions and PRs for those as well
https://github.com/pkeffect/functions
@silentoplayz commented on GitHub (Aug 25, 2025):
Related - https://github.com/open-webui/open-webui/issues/16422
@i-iooi-i commented on GitHub (Aug 25, 2025):
@pkeffect Hi, could you please unify the UI icon colors with those of OPEN WEBUI to ensure a consistent and cohesive interface aesthetic?
There are color discrepancies between the UI icons and OPEN WEBUI's icons, evident in both their active and inactive states.
@i-iooi-i commented on GitHub (Aug 25, 2025):
Is it possible to simulate this segmented group chat display to enhance immersion? Please note, the accompanying image is merely a Photoshop rendition, not an actual feature.
@i-iooi-i commented on GitHub (Aug 25, 2025):
@leetraman822
Could this be the project you're seeking: https://github.com/maojindao55/botgroup.chat , which offers a group chat experience featuring multiple AI entities interacting?
@leetraman822 commented on GitHub (Aug 25, 2025):
Yes, actually there are many similar group chat features available. The first one I came across was SillyTavern—it has very powerful and well-developed role-playing capabilities. However, possibly due to my own settings, the same prompts, temperature, top-k, top-p values, etc., result in responses that aren't as intelligent as those from LM Studio or Open WebUI. I'm not sure what's causing this discrepancy. So I really hope that Open WebUI could eventually support group chats, or at least allow appending character replies with the Enter key.
@leetraman822 commented on GitHub (Aug 25, 2025):
@pkeffect commented on GitHub (Aug 25, 2025):
@i-iooi-i
Thanks for that. This has been fixed. Version number has not changed but the image has been updated and tested on light and dark themes.
@i-iooi-i commented on GitHub (Aug 25, 2025):
@pkeffect Hi, thanks a lot. I've just tested it, and it seems there's still a small problem: the icons consistently display in gray, rather than turning blue like the integrated ones.These are just minor issues that don't impact usability. Thank you again! ♥
@i-iooi-i commented on GitHub (Aug 25, 2025):
看图1的效果，你似乎已经在OpenWebui里实现了模拟多人对话的样式了。这是PS的效果还是使用其他插件实现的呢？总之，看上去模拟的有模有样的。
@pkeffect commented on GitHub (Aug 25, 2025):
@i-iooi-i
The new icons identify the theme and change accordingly. Either light or dark now to be more consistent with OI icons, no colors.
@i-iooi-i commented on GitHub (Aug 25, 2025):
@pkeffect Understood. My sincere thanks for developing this plugin; it's truly excellent, and I anticipate continuing to use it.
@leetraman822 commented on GitHub (Aug 26, 2025):
哈哈，这是使用的openwebui里的@提及功能，虽然看起来是在多人讨论，但其实模型并不是主动的。当我选择模型1的时候，我必须使用@提及模型2才能和模型2聊天，否则模型2是不会主动发言的。
图2是在sillytavern里的群聊中，当我说一句话后，群里的成员都会主动发言而无需提及。
也就是说，想要openwebui里加入的效果是：
1：非一问一答的形式。当用户什么都没有输入时，用户按下回车键或发送按钮可以让模型角色再发送一条消息（不是重新生成 也不是消息扩展）。而不是我问一句，对方回答一句的那种。
2：真正的群聊。就像你在PS渲染的图片一样，那种效果可能就是我们一致想要的，每个角色都有单独的消息气泡，看起来就像是不同的人真真正正的聚在一起讨论一样。我很早就注意到聊天界面左上角的+号按钮，当我们按下+号后可以添加另一个模型，这时可以让两个模型一起回答用户的提问。其实，感觉这个功能比较接近，让新添加的模型以正常消息的形式回复，在模型1的下方，模型2会根据上下文内容，而写出自己的回复，而不是模型1和2同时弹出气泡。
@i-iooi-i commented on GitHub (Aug 26, 2025):
@leetraman822 若是只允许通过单独 @ 才能对话的话，确实脱离了 “群组聊天” 的概念。说起来，通过Prompt设定多个不同职业的角色，在群里针对同一个问题，展开思想辩论，确实是很酷的体验！希望早日能在 openwebui 里找到这类的插件。