In invoke web UI, pressing "Use All" on a txt2img does not change web UI state to indicate change to txt2img

[Perryman]

Describe your environment

• GPU: cuda
• VRAM: 12G
• CPU arch: x64
• OS: Linux
• Python: Anaconda
• Branch: On branch main
• Commit: 194d4c7

Describe the bug
If you're in the invoke web UI and you're doing img2img work, then you click on an image and click "use all," the state of the pending command changes based on whether the image you picked was img2img or txt2img. The web UI does not switch back to txt2img, but the output command and output are txt2img.

To Reproduce
Steps to reproduce the behavior:

1. Run invoke.py --web
2. Have at least 1 txt2img generated (any parameters)
3. Send any image to img2img
4. Select any previously generated txt2img output using the right pane
5. Click "Use All" on the top bar
6. Invoke the image
7. Observe that the output is "txt2img" and the web UI still indicates img2img mode

Expected behavior
I saw the img2img UI, so I expected to be in img2img mode.
If "Use All" is intended to use the mode, then it should change the web UI.
If "Use All" isn't intended to change the mode, then the option slicer logic might be wrong.

Screenshots
If applicable, add screenshots to help explain your problem.

Additional context

InvokeAI/frontend/src/features/options/optionsSlice.ts

Lines 191 to 198 in 194d4c7

	if (type === 'img2img') {
	if (init_image_path) state.initialImagePath = init_image_path;
	if (mask_image_path) state.maskPath = mask_image_path;
	if (strength) state.img2imgStrength = strength;
	if (typeof fit === 'boolean') state.shouldFitToWidthHeight = fit;
	state.shouldUseInitImage = true;
	} else {
	state.shouldUseInitImage = false;

[psychedelicious]

Hmm, I'm not sure we actually want to change this. What if you want to use all parameters in img2img?

@blessedcoolant , related to the bug you just found, what do you think?

[Perryman]

I think the core thing I’m reporting here is that the state of the web UI should match the state. If the UI reflects the state and sends you to txt2img, then you could just click img2img to go back. If it keeps you in img2img mode, but you actually wanted txt2img, then you’d click txt2img to go back. Or, alternatives like “use all with current mode” and some kind of context drop-down could be considered too.

edit: sorry for the accidental close, misclicked on mobile.

[psychedelicious]

Ahh, I misunderstood the issue at first. I see what is happening. An assumption made a while ago is no longer correct. Fixed in #1198.