LibreChat
-
@NCKNE Can you now invite other private users to use your private task-specific knowledge-primed with WebUI?
From what I have seen:
If you work alone. I'm sure WebUI is fine.
If you work as a team. LibreChat has a very, very valuable feature that I've not see existing or planned for WebUI.
I expect the difficulty in seeing the difference is those that don't work as a team on things, so will never have the need to share what LibreChat calls "agents".
Those working as a team just don't have a team-tool to collaborate with.
AI chat requires a ton of refinement for specific purposes.
If you want basic Q&A AI chat, use anything.
If you need a specific type and style of output, based on additional knowledge input, and for the output to be consistent across a team, the choices I know of are Poe.com with those fees, or LibreChat hosted on a server.
-
@firmansi Yeah, they might be things "better", and there might be things with WebUI that are "good enough".
The reason for LibreChat is specifically for collaboration.
I use Cloudron as a collaboration platform among many users for different things, so I value collaborative tools that save time and repeat conversations or effort.
LibreChat solves that for me, and it would be nicer to have it self-hosted with Cloudron, than to be setting up a separate small server just for this one app, and then all the maintenance involved in that.
There's a need, problem, and solution. We don't need to re-invent anything. Just an ability to use each thing for it's unique advantages.
-
This...
@marcusquinn said in LibreChat:
If you need a specific type and style of output, based on additional knowledge input, and for the output to be consistent across a team, the choices I know of are Poe.com with those fees, or LibreChat hosted on a server.
...does sound an awful lots like this:
you can create a custom model that is enriched with your own knowledge (aka you own documents) and share that model with other to chat with
I've not tried doing it with either so can't really comment on any differences.
But I think everyone is agreed that having LibreChat on Cloudron would be nice.
-
@jdaviescoates Yeah. I can see why people look at the basics of it being a chat app, and think they compare.
There's so many man-hours that are saved with collaborative tools.
When you setup a collaborative tool on Poe.com, you are locking that entire team into it a subscription to it for life to continue to use.
If we had the same with LibreChat on Cloudron, all those processes are now owned and self-hosted, with no per-user pricing, to utilise and get value from.
I don't doubt WebUI is useful, and great it already packaged.
We just missed out on having the collaborative LibreChat, so I have to make a value case for it, so it's not dismissed as similar enough to not be a priority.
If collaboration had no value, I can use LibreChat, locally — but the ability to share identical, maintained, pre-trained knowledge apps/agents with LibreChat is so damn valuable if you work with a team.
The biggest overhead with all teamwork is "explaining". Pre-trained LibreChat agents mostly eliminate that explaining and updating everyone overhead, and gives an interface to consistent input/output for all users.
AI is notoriously inconsistent and lacking in recent or private knowledge.
LibreChat solves all of this, when hosted on a server for multiple users to share.
Web Design https://www.evergreen.je
Development https://brandlight.org
Life https://marcusquinn.com -
@marcusquinn said in LibreChat:
the ability to share identical, maintained, pre-trained knowledge apps/agents with LibreChat is so damn valuable if you work with a team.
But how is that different to this:
you can create a custom model that is enriched with your own knowledge (aka you own documents) and share that model with other to chat with.
?
The way you keep on insisting that it is totally different and like it's a whole other level of collaborative capability makes me think there must be some difference, but it's not at all clear what that difference is
You normally do a better job at explaining than you seem to be managing this time
Is LibreChat collaborative like multiple people can join the very same chat window and chat at the same time? Like people can co-edit docs with OnlyOffice/ Collabora?
Perhaps some screenshots or even screencasts are in order?
Even better, someone package LibreChat and we can all see for ourselves!
-
@jdaviescoates Tired, impatient, and time-sensitive. Sorry!
If you get anything from me here it is always to save time, our most precious resources, and I have a lot less of it left them most.
Doubly impatient when I have to resize screenshots up upload them to this NodeBB setup, but here we are...
- https://www.librechat.ai/docs/features/agents
- https://www.librechat.ai/docs/features/agents#sharing-and-permissions
It's all there in the docs.
This is an "Agent" (what Poe.com calls a "App". You only have to use Poe.com for 30 seconds to get what they are.)
Create these, and have a server-hosted (Cloudron-hosted) instance of LibreChat, and now your pre-trained "Agents" (in LibreChat lingo), can be shared.
This feature just doesn't exist in OpenWebUI, and I can't see it coming any time soon.
The difference to me with having LibreChat on Cloudron is probably $100/month plus compared to doing the same with Poe.com. Plus not having my knowledge training setups locked up in someone else's SaaS.
Plus, frickin hours, and hours, and hours of long-hand wasted time teaching a team how to all setup the same pre-trained chat, and keep updating it as it is refined — when this is an already a solved problem but LibreChat Agents and sharing those Agents among users on the same server. Not conversations. Agents. AKA Apps in Poe.com language.
With the bridge between problem and solution apparently being persuasion that there is a LOT more value to this in LibreChat, and this specific highly valuable time, money and settings-saving thing value is not in OpenWebUI.
If you don't work with a team, you probably don't need it.
If you work with a team, and are going to be using AI, then this has a lot of value that I'm not seeing anywhere else.
That's the use-case and value-proposition.
Is there anyone left in this thread that still doesn't get it?
-
@marcusquinn said in LibreChat:
Doubly impatient when I have to resize screenshots up upload them to this NodeBB setup, but here we are...
ShareX with (on Cloudron) Xbackbone are brilliant for screenshotting. If you regularly use screenshots, it is invaluable.
-
@LoudLemur Thanks. I'm using Shottr, and just found the resize x0.5 option, so that seems to work.
-
@marcusquinn You‘re welcome. We are using openwebui with RAG (our own documents aka knowledge) for the IT support team. We are using it on cloudron together with the private IONOS AI Hub. Great service if you are serving EU customers.
