Converting a script into a full-blown extension -- what is involved and how much work is it?

[marcsyp]

Hi all --

I just got my hands dirty recently modifying a publicly-available script for my own purposes. I would now like to convert it to a full blown extension so that it appears in the UI and allows me to use X/Y/Z Grid and other scripts alongside it (instead of taking up the only available script slot).

The script is fully functional with a usable UI. I have taken a look at the documentation for Scripts and Extensions but I'm having a hard time understanding the sum total of the differences and how much refactoring would be required to convert the script into an extension with its own UI. Can someone give me an idea of how big of a project this would be, and what the key steps of the process would be?

Thank you!

[marcsyp]

Any thoughts? Sorry to bother you @AUTOMATIC1111, but I'm really passionate about this project and continue to develop it, and would really love some guidance, even if minimal. Thanks!!

[missionfloyd]

Here's how to create extensions.
https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Developing-extensions

And here's some examples.
https://github.com/udon-universe/stable-diffusion-webui-extension-templates

[marcsyp]

Thank you, I guess I'll just dig in to this and see if I can get something going!

[marcsyp]

So... I did a little reading and playing around with this, and I've made some progress. For instance, I was able to get my script to be shown in the UI as its own accordion by changing the following lines in the python:

def show(self, is_img2img):
# return is_img2img
return scripts.AlwaysVisible

However, the script no longer triggers in img2img at all, and I imagine that's because I need to hook to an event somewhere. I've dug through other extensions but they are usually significantly more complicated, and so I haven't been able to decipher where the trigger point is. Two questions:

1. I have an "Enabled" checkbox in my UI. Anyone have any suggestions on how to trigger the script to run when the enabled checkbox is active?
2. Anyone have a good suggestion on an Extension that has an AlwaysOn script paradigm that is relatively simple that I can use as a starter example? Things like ControlNet, Dynamic Prompts, etc are pretty complex and difficult to navigate as I am starting on this project. Thank you!

[tostangs]

@marcsyp If you're still looking for some assistance on how to create a basic extension by inheriting the Script class, I've created a very simple example that allows you to upload your generated content to your choice of an aws s3 bucket assuming the the IAM user has proper permissions.

Heres the code, it was my first go at creating a primitive extension, but you can build the complexity as much as you want from this basic format.

All I did was reference the scripts.py --> Script class definition defined here to determine which types of methods I should overwrite to perform my desired functionalities.

Hope that helped if you're not already far along in dev now

[w-e-w]

@tostangs if you wish to have your extension listed on the index, make a PR at https://github.com/AUTOMATIC1111/stable-diffusion-webui-extensions

[marcsyp]

@tostangs I'm returning to this thread after some time and though I didn't see it at the time, the pointer to the scripts.py file is actually very helpful. I have since developed my original script pretty far and then built my first AlwaysOn script from scratch and am now much more confident in working with A1111 code. The big piece missing for me in the beginning is that I didn't realize that run() was for regular scripts and process() is triggered for AlwaysOn scripts, which is documented nicely in scripts.py but I managed to figure out at some point through persistent digging and experimentation.

My biggest step forward happened when I finally got VS Code to attach to my webui process and was able to start setting and hitting breakpoints and visualize the code flow as well as locals, globals, and stack traces. That was a game changer. :)

[j-beniwal]

I have one doubt, is there a way we can direclty make api call to the script and get the results.

[w-e-w]

I'm not entirely sure what you're asking

---

it is possible to use any extension that adds the image generation pipeline from the API
see wiki API https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/API#example-script

and these extension typically makes no sense to call them directly

---

other types of extensions that are not family imagination pipeline like you say adding a tab to do extra stuff
you cannot call thes extensions using the image generation API
the extension developer will have to implement some sort of API endpoint or the user can use gradio client to use any UI as an API but this can get a bit complicated

[JeoiReqi]

I Do Not See Any Specific Language You Are Using, But If It Is Python (Like All Scripts Should Be) It Is A 1 Line Command In The Terminal.

Replace 'your_script.py' With Your Filename & Wait A Minute For Everything To Complete. Now You Have A Fully Working .EXE As Long As The Original Script Is Functional.

pyinstaller --onefile your_script.py