mirror of
https://github.com/open-webui/open-webui.git
synced 2026-05-06 02:48:13 -05:00
[GH-ISSUE #8050] enh: channels #53644
No Assignees
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: github-starred/open-webui#53644
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @tjbck on GitHub (Dec 25, 2024).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/8050
Originally assigned to: @tjbck on GitHub.
@ajikmr commented on GitHub (Dec 26, 2024):
Currently everybody sees the channel on the left. Group users can send/read the message.
There should be an option so that only members of the group for any channel could see the channel name on the left sidebar. Others should not see channel (name) on the side bar. This will be good for privacy reasons.
@ecker00 commented on GitHub (Dec 28, 2024):
This is shaping up. 👍 I guess there is no way to interact with AI in channels until @ mention is implemented?
Will there be a way to control scope/context sent to the AI? Is it current message only, or entire chat history?
@srajangarg commented on GitHub (Jan 2, 2025):
@tjbck this is amazing! thanks for working on this
essentially this led me to imagine a hybrid llm chat / channel space where users can talk to each other while "looming" different shared chats and responses with each other and build on each others chats. would it be possible to give this more definition? i would definitely want to pick this up!
edit: i feel like the first step would be to allow multiple users to build on the same chat
@tjbck commented on GitHub (Jan 2, 2025):
@ecker00 You technically can with https://github.com/open-webui/bot , do keep in mind this boilerplate and the channels feature overall is highly experimental, and the underlying API/functions powering this feature might change drastically overtime.
@flefevre commented on GitHub (Jan 13, 2025):
As we are working with generative ai, it could be great to clic on a channel and ask for a basic resume for the following messages. It will summit the day messages, week messages. But I do not know where the resume go
Nb I stil do not understand, if you prefer I create new discussion ticket or if it is fine I add here additional suggestions. Please feel free to tell the process you prefer.
@tcgumus commented on GitHub (Jan 17, 2025):
@tjbck how can i call the bot from channel?
@42atomys commented on GitHub (Jan 28, 2025):
Hi @tjbck,
I'm interested to improve/work this part, as I have work currently on a standalone implementation of multi-users discussion with an IA agent.
We are at the beginning of channels so discussion about the "epic" are currently the state right ?
@Notysoty commented on GitHub (Mar 19, 2025):
@tcgumus Please check this out https://github.com/open-webui/bot. Change the token in env.py with your token from openwebui>>Settings>>Account>>API keys> JWT Token and run python ./examples/ai.py
@flefevre commented on GitHub (Jun 2, 2025):
Additional ideas: https://github.com/open-webui/open-webui/discussions/14588
Expand "Channels" to Enable Multi-Agent Conversations with Assistant Models and Shared Memor
@tarunwadhwa-paypay commented on GitHub (Jul 7, 2025):
Hi !! Thank you for the good work. Just wanted to ask if adding emojis in chat messages (not reaction but message content) under consideration? I can even raise a pull request. We do use channels for a lot of internal communications and feels like emojis will add to the experience.
@ldpg-dev commented on GitHub (Sep 22, 2025):
@tjbck Any information about https://github.com/open-webui/open-webui/issues/17314? You marked as completed but I don't see it in v0.6.30
The channel/chat that originated a notification should at least be in bold for that feature to be usable. Else a notification arrives and nobody knows where it comes from.
@Classic298 commented on GitHub (Sep 22, 2025):
@ldpg-dev it was marked as completed because channel related feature proposals should be posted HERE and not in separate new issues to avoid duplication and uncontrolled spread of new issues for essentially the same topic. (That's why tim responses to #17314 with the ID of this issue to reference to it.)
And it was not included in 0.6.30 or any version before that (otherwise it'd be in the changelog)
@ldpg-dev commented on GitHub (Sep 22, 2025):
It's the single most critical feature imo. It's the only thing that keeps channels from being usable. Because all other functionality is already there
@flefevre commented on GitHub (Sep 22, 2025):
I will close my ticket to reference this one
🚀 Feature Request: Autonomous Bot Participation in Channels #17620
@N0YU0 commented on GitHub (Oct 2, 2025):
When creating a prompt message in channel if the model has a thinking phase that prints out and it should be a collapsed element just like in a regular chat.
Channel's thinking tags:
Regular chat thinking tags:
@frenzybiscuit commented on GitHub (Oct 5, 2025):
Channels does not appear to support streaming or regeneration's. Could we get those two features added?
@Hekas1 commented on GitHub (Oct 11, 2025):
Let me also add whether this is possible to implement in the channel system.
Global search across channels and channel content, and ideally, across the content of processed submitted files.
And is it possible to reference knowledge using #
@impr3ssi0n commented on GitHub (Nov 7, 2025):
Channels have the ability to upload files, but in my channel, llm can't access the file context. Tell them it can't find it.
@silentoplayz commented on GitHub (Dec 3, 2025):
It would be a welcome improvement if any URL that references a specific note inside a channel within Open WebUI could behave like an ordinary in-app hyperlink: when the user clicks it, the application would simply navigate to the Notes page in the same browser tab instead of launching a fresh tab or window. This would keep the workflow seamless, avoid cluttering the tab bar, and make cross-referencing notes feel as native as jumping between channels or conversations.
@jqqin commented on GitHub (Feb 19, 2026):
We've encountered the same issue. It appears that channels currently only process image files, while text files and PDFs are not being passed to the models. Would really appreciate a fix for this!
@daniporr commented on GitHub (Feb 20, 2026):
From #21581
@nrmjeremy commented on GitHub (Mar 7, 2026):
Closed #22284 as a duplicate of this. Adding KB/RAG injection as a missing item — full diagnosis with Docker logs here: [link]. Happy to contribute a PR if helpful.
@srprca commented on GitHub (Mar 13, 2026):
Hello!
I would like to request several features for the Channels functionality. Is this the right place to do so? If not, please let me know where to re-submit my request.
I use Channels to chat with multiple models in such a way that they see each other's responses and react to them, as opposed to the Chat mode, where the participating models don't "see" each other's responses.
I believe the following UI elements could improve the workflow a lot:
Thank you so much for developing this awesome piece of software!
@gdzikowskitt commented on GitHub (Mar 27, 2026):
Hello!
Good job on the feature!
I'm wondering, as it is not on a checklist above and the only PR was closed with redirection here - will tool calling be a feature for Channels?
Keep up the good work! Thank you!
@FxMorin commented on GitHub (Apr 16, 2026):
Channels (groups/shared) currently leak all registered users through the API and when mentioning using
@.There should be a setting to limit user visibility between the following:
Having a setting for read-only channels to prevent seeing users could also help prevent leaking unrelated users.
While doing this, you could also add the feature from slack. Where if you ping a user that isn't in the channel, and you have perms to add users. It tells you that this user isn't in the channel and asks if you want to add them.
Users should be gated like models currently are.
@silenceroom commented on GitHub (Apr 20, 2026):
Echo on this requset to limit the user selection scope in the private channel. Currently the available user list for mention in a channel is all registered users in the system which make the communication too hard to use. It should be the members only plus the models (this part does a good job to limit the members could only access the models they have access to).
@martin-grimm commented on GitHub (Apr 22, 2026):
Are tool calling and MCP integration something that is being considered for channels/threads? I don't see it in the check list above and @tjbck closed #23823 as completed.
I understand that this is a challenge, as outside of channels tool and MCP availability settings are scoped to individual models and must be enabled by the user for individual conversations.
However, it would be a very powerful tool in any discussion / meeting:
Imagine a thread about an issue from GitHub, Jira, you name it:
A similar workflow could happen with a meeting agenda point pulled from a note, Confluence, etc.
One could even have skills to enable a model to function as a moderator/facilitator, opening additional threads for agenda points (given additional channel / thread tools for that).
Visibility to tools / notes / skills should be scoped to the user currently sending the request to the model.