Possible to access ComfyUI interface in serverless to fix custom nodes requirements?
Hi RunPod addicts!
I have a functional ComfyUI install running in a Pod that I want to replicate serverless.
My comfyUI install is made for a specific workflow requiring 18 custom nodes.
I'm trying to deploy this workflow in a new Serverless instance using runpod-worker-comfy project.
My Dockerfile deploys all required custom-nodes but I still have 4/5 nodes that are (IMPORT FAILED) because missing requirements.
I'm trying to troubleshoot the worker env to fix these custom nodes issues.
From the logs I can see that some requirements are missing like deepdiff or evalidate... I feel that the order I install my custom nodes is important because the last one installed could break dependencies required from previous ones, am I right? I can SSH access but I was wondering if there is a way to access ComfyUI interface and use the Manager to try to fix my issues manually? I'm open to any suggestion to help me setup my Serverless worker properly! Thank you!
From the logs I can see that some requirements are missing like deepdiff or evalidate... I feel that the order I install my custom nodes is important because the last one installed could break dependencies required from previous ones, am I right? I can SSH access but I was wondering if there is a way to access ComfyUI interface and use the Manager to try to fix my issues manually? I'm open to any suggestion to help me setup my Serverless worker properly! Thank you!
GitHub
GitHub - blib-la/runpod-worker-comfy: ComfyUI as a serverless API o...
ComfyUI as a serverless API on RunPod. Contribute to blib-la/runpod-worker-comfy development by creating an account on GitHub.
31 Replies
comfy --skip-prompt --no-enable-telemetry set-default /workspace/ComfyUI/
comfy node restore-dependencies
you need adjust path to your ComfyUI install
though first activate venv and install
pip install comfy-cli
i am creating alternate versions of comfy. i am using comfy cli commands to switch the default version to run in Better.
The usecase, is that sometimes I need a very large comfy install with lots of custom nodes, and sometimes more specialised light-weight comfy, for specific use case.
i've created the Bvenv (bcomfy_ultimate) for instance.
I am now running Better "Comfy" but it still runs the original
Trying things inside Better, in order to switch to new "ultimate" install
How do i make Better switch to my "ultimate" comfy install?
@Madiator2011 (Work)
why not just use normal with network storage
i am. I am on net volume. but id like to have multiple comfyui installations
and switch before running Better>comfy
you would need to edit pre_start script then
can i? isnt that in the container?
or in the template?
GitHub
madiator-docker-runpod/official-templates/better-comfyui/pre_start....
RunPod Docker Containers for RunPod. Contribute to kodxana/madiator-docker-runpod development by creating an account on GitHub.
Container Start Command?nope
ok but i am using your Better template. how do i edit it?
Dev2 remember
you cant unless you build own docker image. You can always use container start command but it will break things
ok, so if i use your template, i should only use one comfy install right
i dont see a github of your dev2 to build my own image
so thats a no go
If that’s the case I will just stick one comfy version no prob
I might be able to change it for you if I get more time
ok, that would be cooll no rush. i will stick to one build for now.
Whats the best way to start over the current comfyui.
Delete Application then reinstall, from the Better comfy UI interface?
nevermind, i see it in the UI
i ll proceed
thanks
i ll move my models first
by the way @Madiator2011 (Work) the newest version of comfyui is Python12 Cuda2.5 but yours shows Python11Cuda 2.4
doesnt it update to latest?
there will be compatibility issues with some nodes because of this
I am using 12 locally with big custom nodes without issues. 3.11 your current version causes some issues with some utils from different nodesI need more time for update
ok, having a little issue. Trying to recreate Symlinks
I have my models in /workspace/shared_models now but when i try recreate symlinks button in Better UI I get this
the symblinks are automatic and no need to recreate every time you add new model
ok fixed it
but now i have the same issue in loras
Error: [Errno 40] Too many levels of symbolic links: '/workspace/shared_models/loras/flux/model_metadata.ldb'
it might be because i am using share_folder for SwarmUi as well
hence the model_metadata.ldb file which is from swarmUI
I suspect cause folder named flux
though i cannot see it in explorer
which is weird
oh
dont think that should be an issue. its a random word.
comfy doesnt care if you use 'flux'
is it related to Better symlinking this issue with 'flux'?
but symblinsk might have issues as its going to start making more symblinks in symblinks
but model_metadata.ldb is not a symlink
its a database file
what does it interpret as a symlink inside symlink?
there are no symlinks inside flux path
I'm not sure I did not added swarm ui yet so I cant know what has been broken
Ok I’ll try and create the symlink manually to see if it’s something else
I would say just set the swarm ui path to ComfyUI directly rather than pointing to shared folder
good idea
Yes but the actual models are in shared_models. This is what I did. Was that wrong? From what I understand Better is creating Symlinks in Comfyui/models—so I link SwarmUi to Comfyui/modelsiy will be looking inside su links
Is that what you are saying?
alright all working. just keeping models in comfyui