08-02-2026, 06:31 PM
there is new comfyui sdk that might help you with tydiffusion.
I hope you look for it and thanks for everything.
?️ **We're making ComfyUI a better backend to build on**
@everyone A lot of you don't just *use* ComfyUI — you build on top of it. A Blender plugin that generates textures. Character art for your favorite LLM roleplaying app. Krita painting straight into a workflow. A consumer AI app handling thousands of jobs a day. A pipeline that runs a workflow over every single frame of a film.
Until now, all of that meant reverse-engineering the same endpoints our web UI uses and hoping we didn't change them out from under you. We're starting to fix that properly.
**Step one: official ComfyUI SDKs.** ?
? Python — `pip install comfy-sdk`
? TypeScript — `npm i @comfyorg/sdk`
? And a documented HTTP API underneath both, for every other language
The same code runs against Comfy Cloud or against your own local ComfyUI — you change the URL. (Local needs a small open-source proxy during the beta; that moves into ComfyUI core once the API settles.)
```python
from comfy_sdk import Comfy
client = Comfy("http://127.0.0.1:8189")
# client = Comfy("https://cloud.comfy.org", api_key="comfyui-...")
wf = client.workflows.from_file("workflow_api.json")
asset = client.assets.from_file("photo.png")
wf.set_input("10", "image", asset)
# Run workflow
# Get outputs using the output node Id as a reference
job = client.run(wf)
for output in job.get_outputs("9"):
output.to_file(output.name)
```
**It's a beta, and that's the entire point.** ? We plan to support these APIs for a long time — which means the shape we settle on is the shape you're going to live with. Right now it's cheap for us to change. Once version 1.0 is released, changing it becomes much more involved.
So come discuss it with us and tell us what you think while your opinion can make a big impact! More details and discussions in: <#1533263616973864980>
I hope you look for it and thanks for everything.
?️ **We're making ComfyUI a better backend to build on**
@everyone A lot of you don't just *use* ComfyUI — you build on top of it. A Blender plugin that generates textures. Character art for your favorite LLM roleplaying app. Krita painting straight into a workflow. A consumer AI app handling thousands of jobs a day. A pipeline that runs a workflow over every single frame of a film.
Until now, all of that meant reverse-engineering the same endpoints our web UI uses and hoping we didn't change them out from under you. We're starting to fix that properly.
**Step one: official ComfyUI SDKs.** ?
? Python — `pip install comfy-sdk`
? TypeScript — `npm i @comfyorg/sdk`
? And a documented HTTP API underneath both, for every other language
The same code runs against Comfy Cloud or against your own local ComfyUI — you change the URL. (Local needs a small open-source proxy during the beta; that moves into ComfyUI core once the API settles.)
```python
from comfy_sdk import Comfy
client = Comfy("http://127.0.0.1:8189")
# client = Comfy("https://cloud.comfy.org", api_key="comfyui-...")
wf = client.workflows.from_file("workflow_api.json")
asset = client.assets.from_file("photo.png")
wf.set_input("10", "image", asset)
# Run workflow
# Get outputs using the output node Id as a reference
job = client.run(wf)
for output in job.get_outputs("9"):
output.to_file(output.name)
```
**It's a beta, and that's the entire point.** ? We plan to support these APIs for a long time — which means the shape we settle on is the shape you're going to live with. Right now it's cheap for us to change. Once version 1.0 is released, changing it becomes much more involved.
So come discuss it with us and tell us what you think while your opinion can make a big impact! More details and discussions in: <#1533263616973864980>