[GH-ISSUE #10602] When generating images, the thinking process is also sent to ComfyUI #15952

New Issue

GiteaMirror · 2026-04-19T22:00:59-05:00

GiteaMirror commented

2026-04-19 22:00:59 -05:00

Originally created by @lzf-tech on GitHub (Feb 23, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/10602

When generating images, the thinking process is also sent to ComfyUI. I don't think the thinking process should be sent. Can this be fixed?

Originally created by @lzf-tech on GitHub (Feb 23, 2025). Original GitHub issue: https://github.com/open-webui/open-webui/issues/10602 When generating images, the thinking process is also sent to ComfyUI. I don't think the thinking process should be sent. Can this be fixed?

GiteaMirror closed this issue

2026-04-19 22:01:00 -05:00

GiteaMirror commented

2026-04-19 22:01:01 -05:00

@sanjibnarzary commented on GitHub (Feb 23, 2025):

As per as comfyui is concerned I think without knowing your workflow json it is difficult to identify.

@sanjibnarzary commented on GitHub (Feb 23, 2025): As per as comfyui is concerned I think without knowing your workflow json it is difficult to identify.

GiteaMirror commented

2026-04-19 22:01:02 -05:00

@lzf-tech commented on GitHub (Feb 23, 2025):

As per as comfyui is concerned I think without knowing your workflow json it is difficult to identify.

I don't think it's necessary to handle this from the perspective of ComfyUI. Instead, it should be handled on the WebUI side. When users directly click on the generated images in the reply based on the generated content, the WebUI doesn't need to send the content of the thinking label to ComfyUI.

@lzf-tech commented on GitHub (Feb 23, 2025): > As per as comfyui is concerned I think without knowing your workflow json it is difficult to identify. I don't think it's necessary to handle this from the perspective of ComfyUI. Instead, it should be handled on the WebUI side. When users directly click on the generated images in the reply based on the generated content, the WebUI doesn't need to send the content of the thinking label to ComfyUI.

GiteaMirror commented

2026-04-19 22:01:02 -05:00

@rgaricano commented on GitHub (Feb 23, 2025):

As "thinking" process/tags are adopted by diferent system a easy way to deal with this is apply a input filter to comfy workflow:
https://github.com/stavsap/comfyui-ollama/issues/84

(maybe in some ways this "thinking info" could be usefull for comfy specific workflows)

@rgaricano commented on GitHub (Feb 23, 2025): As "thinking" process/tags are adopted by diferent system a easy way to deal with this is apply a input filter to comfy workflow: https://github.com/stavsap/comfyui-ollama/issues/84 (maybe in some ways this "thinking info" could be usefull for comfy specific workflows)

GiteaMirror commented

2026-04-19 22:01:03 -05:00

@michaelmarziani commented on GitHub (Feb 24, 2025):

When generating images, the thinking process is also sent to ComfyUI. I don't think the thinking process should be sent. Can this be fixed?

I'm not able to reproduce this with my setup. I followed the instructions to set up the ComfyUI integration and in the terminal I see what looks like the JSON being sent to ComfyUI and there's nothing that looks like a thinking process in the prompt or other JSON fields.

Could you please post your logs that show this?

@michaelmarziani commented on GitHub (Feb 24, 2025): > When generating images, the thinking process is also sent to ComfyUI. I don't think the thinking process should be sent. Can this be fixed? I'm not able to reproduce this with my setup. I followed the instructions to set up the ComfyUI integration and in the terminal I see what looks like the JSON being sent to ComfyUI and there's nothing that looks like a thinking process in the prompt or other JSON fields. Could you please post your logs that show this?

GiteaMirror commented

2026-04-19 22:01:04 -05:00

@lzf-tech commented on GitHub (Feb 24, 2025):

When generating images, the thinking process is also sent to ComfyUI. I don't think the thinking process should be sent. Can this be fixed?

I'm not able to reproduce this with my setup. I followed the instructions to set up the ComfyUI integration and in the terminal I see what looks like the JSON being sent to ComfyUI and there's nothing that looks like a thinking process in the prompt or other JSON fields.

Could you please post your logs that show this?
It should be that my previous description was incorrect. It was not generated by open-webui, but by the model. Some models, such as deepseekR1, will display the thinking process when outputting, with special tags. open-webui should also have been adapted accordingly. The screenshot is as follows. When I directly generate an image in this reply, open-webui should carry the
tag.

@lzf-tech commented on GitHub (Feb 24, 2025): > > When generating images, the thinking process is also sent to ComfyUI. I don't think the thinking process should be sent. Can this be fixed? > > I'm not able to reproduce this with my setup. I followed the instructions to set up the ComfyUI integration and in the terminal I see what looks like the JSON being sent to ComfyUI and there's nothing that looks like a thinking process in the prompt or other JSON fields. > > Could you please post your logs that show this? It should be that my previous description was incorrect. It was not generated by open-webui, but by the model. Some models, such as deepseekR1, will display the thinking process when outputting, with special tags. open-webui should also have been adapted accordingly. The screenshot is as follows. When I directly generate an image in this reply, open-webui should carry the <details> tag. ![Image](https://github.com/user-attachments/assets/6a1ac630-429e-4d8b-bde2-85a0c1e9788f)

GiteaMirror commented

2026-04-19 22:01:05 -05:00

@michaelmarziani commented on GitHub (Feb 27, 2025):

Hi @lzf-tech, thanks for the screenshot but I was looking for console or terminal logs. What we see in your screenshot is the users view of the thought process in the Web UI interface, but that doesn't mean this is what is sent to the image API. To see that, you'd need to access the console logs. I don't know what your setup is; I installed open-webui directly, without docker, so I can see the console logs in my terminal after running open-webui serve.

Do you have any terminal logs you can post when you run an image generation?

@michaelmarziani commented on GitHub (Feb 27, 2025): Hi @lzf-tech, thanks for the screenshot but I was looking for console or terminal logs. What we see in your screenshot is the users view of the thought process in the Web UI interface, but that doesn't mean this is what is sent to the image API. To see that, you'd need to access the console logs. I don't know what your setup is; I installed open-webui directly, without docker, so I can see the console logs in my terminal after running `open-webui serve`. Do you have any terminal logs you can post when you run an image generation?

GiteaMirror referenced this issue

2026-04-19 23:36:02 -05:00

[GH-ISSUE #15952] issue: Docling Not work, tesseract_ocr Invalid resolution 0 dpi. Using 70 instead. #17730

GiteaMirror referenced this issue

2026-04-25 07:10:38 -05:00

[GH-ISSUE #15952] issue: Docling Not work, tesseract_ocr Invalid resolution 0 dpi. Using 70 instead. #33259

GiteaMirror referenced this issue

2026-05-05 19:20:00 -05:00

[GH-ISSUE #15952] issue: Docling Not work, tesseract_ocr Invalid resolution 0 dpi. Using 70 instead. #56396

GiteaMirror referenced this issue

2026-05-13 04:21:23 -05:00

[GH-ISSUE #15952] issue: Docling Not work, tesseract_ocr Invalid resolution 0 dpi. Using 70 instead. #72334