"ComfyUI to API"-wizard
We’ve built a tool that takes your ComfyUI workflow (the full export via Comfy → File → Export — not the API export) and turns it into a ready-to-use serverless endpoint. It automates the complicated bits for you.
How it works
1. Upload
workflow.json exported from Comfy → File → Export (not the API export)
2. Analyze your workflow: click Analyze to run our comfy-agent. It detects custom nodes, locates required models and prepares a Dockerfile
3. Generate Dockerfile: review the generated Dockerfile and tweak if you like or just skip ahead
4. Create GitHub repo: we create a repo containing your Dockerfile. You’ll use this in the next step.
5. Deploy on Runpod: create a new endpoint, select the repo (after connecting your GitHub account once) and then we build the docker image for you and deploy it into your endpoint. once that is done you can use the example API request from step 5 to test your endpoint
We’d love your feedback to improve the tool and uncover any edge cases we haven’t covered yet.
https://comfy.getrunpod.io
17 Replies
@angelarose regarding your issue from https://discord.com/channels/912829806415085598/948767517332107274/1425478757665472573
-> we switched auth to only support the https://comfy.getrunpod.io domain. but we added premanent redirects now to the old domains, so the auth problem should be gone
but please give it another try and tell us if you are still running into problems
Hey @NERDDISCO , thank you so much! Im a beginner dev and this is exactly what i have been looking for. I am trying to convert some workflows into a serverless endpoint iwth 0 knowledge, i WILL try these out otnight and let you know how it goes.
May I PM you with any issues I have ?
Happy to help document my steps too and you can reuse the documentation as 'an idiots guide' sort if thing
@Zimzam of course, we can either discuss them here or via dm! looking forward to your feedback
Unknown User•2w ago
Message Not Public
Sign In & Join Server To View
Definitely need some improvement, it detects random models that i don't even have downloaded
Also, is it supposed to be analysing that long?
Or will it just won't stop until it figures out the model?
@Tim
I noticed that the front end doesnt output errors if any. check browser console logs and you'll see (its probably failed) but still says 'analysing' on front end
Thank you, very helpful
Hey,do you or anybody else know whats going on with the 5090s?
Why do these 5090 have different amounts of Ram,when im paying for each one the same?
No idea my friend, I am very new to comfy + runpod + cloud 😄
My guess is that that isnt the ram on the gpu but the ram on the worker
but i could be chatting complete shit
Unknown User•2w ago
Message Not Public
Sign In & Join Server To View
yes sure!
would you mind DMing me your workflow? then i can take a look
yeah, something i will work on next to improve
what you always get is exculisve access to the GPU resources, but RAM / CPU can be different from datacenter to datacenter, so you always get a different amount
yeah exactly, what is fixed is the VRAM of the GPU, but the RAM / CPU for the worker can be different
Sent you a message along with the workflow!
@NERDDISCO just wondering there a YT video for this that going to get released to add to landing page right? haha
yes yes yes 😄