cannot connect to host.

[xSK9x]

hey! i am having an issue with connecting the extension to comfyUI and I am banging my head in frustration trying to fix it using AI to help me as I just don't know what is wrong.

I have attached a couple screenshots with all the information I think is relevant to the issue.

[xSK9x]

since my report, I have changed my Comfyui to the portable version and it is now 127.0.0.1:8188.

the native Image gen extension on sillytavern seems to also work perfectly fine.

but still comfyinject is having trouble recognizing that Comfyui is running.

got the workflow JSON to be recognized, but the checkpoint box still is failing to recognize comfyui.

in the end, i end up with image generation failed.

[Spadic21]

Thanks for the screenshots, that helps a lot.

One quick thing to clarify: the arrow next to the Checkpoint field only tries to fetch checkpoint models from ComfyUI. It’s just a convenience list. If that request fails, it can show “Could not reach ComfyUI,” but that message is a bit misleading because ComfyUI itself might still be reachable.

The important part is the text box. Whatever filename you type there is what ComfyInject will insert into the workflow’s {{CHECKPOINT}} placeholder. So even if the dropdown doesn’t populate, generation can still work as long as the filename is typed exactly as it appears in the ComfyUI models folder.

Since you mentioned that SillyTavern’s built-in image generation works, another thing that might be worth checking is the workflow itself. Some workflows made for ST’s native image generation use a different placeholder format than ComfyInject. If that’s the case, ComfyInject wouldn’t be able to inject its values into the graph.

A custom workflow also doesn’t need every placeholder, only the ones relevant to the nodes it uses. The important ones are usually things like the checkpoint and prompts.

Would you be able to attach or paste your Anima.json workflow file here? That will help check whether the placeholders line up with what ComfyInject expects, or if something else might be going wrong.

If possible, it would also help to check the browser console after a generation fails and see if there are any [ComfyInject] errors there.

Thanks again for bringing this up and for the detailed report and screenshots. That stuff really does help a lot when tracking down issues like this.

[xSK9x]

sure! it is a custom workflow and admittedly i am a noob with these things, especially extensions.

Anima (1).json

i think it may help to have a video tutorial on youtube because these things can be pretty complicated!

anyways here is the error i got from the console:

[Spadic21]

Thanks, this helps a lot. I found the immediate issue.

Your workflow file has invalid JSON syntax because it includes trailing commas in at least two places, so ComfyInject can’t parse it before generation even starts. That matches the console error exactly.

The two spots are the easy negative node and the easy positive node, where the inputs objects end with a trailing comma after the placeholder value.

There’s also one other thing to fix after that: your workflow currently has the model filename hardcoded in the UNETLoader node, so the Checkpoint field in ComfyInject won’t affect it yet. If you want ComfyInject’s Checkpoint box to control the model, that field should use {{CHECKPOINT}} instead.

So the fixes are:

• Remove the trailing commas from the workflow JSON

• Replace the hardcoded unet_name value with {{CHECKPOINT}} if you want the Checkpoint field to work with this workflow

Once those are fixed, it should get much further than it is now.

One other quick thing to double-check: the filename in the Workflow field in ComfyInject has to match the workflow file in the workflows folder exactly, including capitalization and the .json extension.

I’m assuming the (1) in the file you sent was just from duplicating it to share, but it’s worth making sure the filename in the extension settings matches the actual file name in the folder.

[xSK9x]

Hey! I appreciate you helping me, through this. these fixes indeed put us closer to fixing this, but I now get this error.

Here is my comfyinject settings.

here is the version of the workflow with your fixes in it!

Anima.json

Do you have any idea what this could mean?

Thanks

[Spadic21]

Hey, thanks for sending the updated workflow and screenshots, that helps a lot.

The new error you're seeing (POST /prompt 400) means ComfyInject is successfully reading the workflow and sending it to ComfyUI, but ComfyUI itself is rejecting the submitted workflow during validation.

A couple things stood out that are worth checking.

Sampler value

In your ComfyInject settings the sampler is set to euler a.
ComfyUI’s API usually expects the internal sampler identifier instead of the UI label, so try changing it to:

euler_ancestral

and test again.

Quick check with the workflow itself

Before digging further into the extension side, it would help to confirm whether the workflow itself runs normally in ComfyUI.

Could you try this:

• Open the same Anima.json workflow directly in ComfyUI
• Select the same model (animayume_v02.safetensors)
• Try running the workflow manually

If it runs successfully in ComfyUI, then we know the workflow itself is valid and we can focus on what values ComfyInject is sending.

If it doesn't run in ComfyUI either, then the issue is likely inside the workflow itself.

One thing to check specifically: in the workflow, click the Load Diffusion Model / UNETLoader node and look at the dropdown for unet_name.

If animayume_v02.safetensors does not appear in that dropdown, that usually means the workflow is using a loader that does not match the model type or folder.

Also, if the ComfyUI terminal prints any error when you try generating, that message would be very helpful to see.

[xSK9x]

did all of that, unfortunately still does not work!

i did get a bit more information from it though;

this is from the sillytavern side. something i missed the first time and seems to be on startup.

and this is on comfyui's side. it looks to me like the prompts aren't filling in the gaps i think.

let me know what you think, since im stuck on this.

Alternatively, i will use your comfy default workflow and see if it works out of the box; if it does then i will modify it with the basics to see if that will work instead.

[xSK9x]

turns out it really was my workflow the whole time! i finally got it to work. thanks for your help :D