mirror of
https://github.com/open-webui/open-webui.git
synced 2026-03-17 12:31:06 -05:00
feat: InvokeAI Image Generation Support #489
No Assignees
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: github-starred/open-webui#489
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 @alfi4000 on GitHub (Mar 17, 2024).
Hi i personally like invokeai more then A1111 because it is healthier handeling gpu resource for consumer hardware is there any plan to support their api for image generation?
Here there repo:
https://github.com/invoke-ai/InvokeAI
It supports very small GPUs too not like A1111 when you get in trouble because of not enough vram!
@justinh-rahb commented on GitHub (Mar 17, 2024):
Unclear whether they even have an API. Did you find any documentation that I've been unable to find so far?
https://github.com/invoke-ai/InvokeAI/discussions/2600
@alfi4000 commented on GitHub (Mar 17, 2024):
There is api when running invokeai just behind access ip adress or localhost and port /docs! ill show I screen recording!
@justinh-rahb commented on GitHub (Mar 17, 2024):
As noted here: https://github.com/invoke-ai/InvokeAI/discussions/2600#discussioncomment-7573198
This doesn't seem to cover generation, although that was some months ago. If the current status is changed, let us know.
@alfi4000 commented on GitHub (Mar 17, 2024):
Oh ok i will check that! And ask Invokeai community because they were talking about It will happen soon and I will check on my instance the api and reach back!
@alfi4000 commented on GitHub (Mar 17, 2024):
Someone Figured out how to do it over endpoint: https://github.com/invoke-ai/InvokeAI/discussions/2600#discussioncomment-8816493
@alfi4000 commented on GitHub (Mar 17, 2024):
It appears to me to be possible this user explained it a bit could you may take a look at it. :) for me it is like a different language I am not a dev or so just a consumer!
@alfi4000 commented on GitHub (Mar 17, 2024):
The Invokeai contributors haven’t answerd me yet maybe the next days!
@psychedelicious commented on GitHub (Mar 18, 2024):
Invoke has had a REST API for quite a while. A maintainer provided the URL to the autodoc for it about a year ago. I've just replied to the linked discussion again: https://github.com/invoke-ai/InvokeAI/discussions/2600#discussioncomment-8821697
@justinh-rahb commented on GitHub (Mar 18, 2024):
Ya... none of that new discussion is giving me any additional confidence that this is something we could easily add, looks like it would indeed be a significant amount of work and possibly require reverse-engineering the client->backend communications. The spec is complete, but that's just a map... it provides no directions. Not saying this is impossible or never gonna happen, just laying out the reality that it's not simple.
@psychedelicious commented on GitHub (Mar 18, 2024):
Sorry, to be clear, I am also a maintainer for Invoke.
Indeed, integrating it into this project would be rather complicated. Besides constructing a valid graph, you'd need a socket.io client to listen for events. You'd also probably want to provide some queue and model install interaction capabilities.
While we have a REST API, it's built for the application and not particularly well suited to run as a headless server for an arbitrary frontend. It certainly can be done but would be a substantial effort.
@justinh-rahb commented on GitHub (Mar 18, 2024):
Totally get it, same for us.. our API exists, but only to serve the needs of our frontend. If someone finds a novel use for that, cool, but that wasn't in our plans 😜
@alfi4000 commented on GitHub (Mar 18, 2024):
Oh so it will most likely not happen yet!?
@justinh-rahb commented on GitHub (Mar 18, 2024):
Unless an unexpected pull request emerges, it appears unlikely that our current contributors will undertake this project, given our focus on numerous other unimplemented features. We already support two image generation backends, including a local one that utilizes the same models as Invoke. Therefore, the justification for developing extensive new code to achieve the same outcome seems limited.
Don't get me wrong, Invoke is a great project you should definitely use, but the thing that makes it great is their interface and we just couldn't hope to replicate even a portion of that within the scope of the type of UI we're developing right now. @psychedelicious thanks for weighing in here, and thanks also for your contributions to the project which I have happily used myself.
@alfi4000 commented on GitHub (Mar 18, 2024):
Ok got it
@linux-universe commented on GitHub (Nov 23, 2024):
Would love to see this because of the great Docker support from both sides