Git is cool to save revisions of workflows.

Is it possible :

1. Auto save current workflow as "working-workflow.json" to the repository.
2. A option to allow auto sync with remote repository every N minutes.

This extension is an incredible time-saver and it's getting more polished at light speed. Thank you for sharing it with the community.

A great addition to its current features, assuming it's in scope with the goal of the project, would be the capability to display the image metadata in the gallery.

Today, I save the metadata via the SD Prompt Saver node, part of the SD Prompt Reader custom node suite, because it's the most reliable and comprehensive node around for the job:

When I have to check that image metadata, I have to go the file explorer, reach the ComfyUI output image, and finally open the image inside the standalone app SD Prompt Reader (which is great, from the same author of the nodes).

It's fine, but this workflow could be really sped up if your ComfyUI Browser could show the metadata right inside the Outputs view.

Just an idea.

This is a great tool!

On windows, I have experienced an error:
[comfyui-browser] failed
[comfyui-browser] 'grep' is not recognized as an internal or external command,
operable program or batch file.

I will have to modify the code changing "grep -w" to "findstr" since windows platform does not have grep by default.

Maybe add a check of OS and use findstr for windows instead

Options to sort the browser display by newest/oldest, file type, file name, etc would be nice.

My workflow repository link:

Description:

My workflow repository link:
Description:

一直在加载，但是找不到错误日志

This node is great for me. Awesome work!

It's good to add support of git sync via token. In gitlab , we can use https://oauth2:<your-access-token>@gitlab.com/your-username/your-repo.git to clone and sync files.the <your-access-token> is personalized for user from settings.

In github it looks like https://<token>@github.com/your-username/your-repo.git

If input this single line in the comfyui brower's git repository , can save workflows from cloud server like colab.

I have ~15,000 files the browser is processing which is a lot but does not enough to reasonably cause such a slowdown. Submitting a PR soon to remedy this issue!

Hi. Magnific extensions, I cannot use with windows 10 ( I am not expert but ever tell me error connection, and save local ).

I need some node (as separate extension) that take the value from another node, similar to your node "select node imput", in this form I can use the values to incorporate to korkgroups.

Sample. I use "global seed.seed" to use into text to watermark over the image. (It is a sample).
I need this node :) to copy "internal" value to make "public" value in another node.

Thanks.
PD.: Sorry my bad english.

Refused to connect in comfyui workflows page, but works in new tab page.

如果是中文命名,点击加载会提示"No workflow found here", 英文+数字就没问题.

Finally, a comfortable ComfyUI Browser is here, thanks!
However, I've got a suggestion regarding the saved workflows. Can you add an option to save the png to workflows, or just set a custom image for each? It'll be more comfortable to go through with images representing workflows, rather than names only. Other than that, all's great.
EDIT: Nevermind, I'm blind and didn't get that I can save to workflows through image browser itself.

Hello, thank you for developing this great tool.
I wanted to include your node in my workflow, but I only get empty grids when trying to alternate some primitive INT inputs.
It does work if I select inputs from the Ksampler, but that's what a normal xy plot already does.
Any idea on what could be the cause?
Thank you and keep up the great work!!

Here is the workflow:
xyz_INT_FAIL.json

Hi, thanks for the work on the project!

XYZPlot doesn't work with prompts like I'd expect them to. In my head, if I've already got a prompt, then the axis values for a prompt should be appended, not replaced. That said, I appreciate the generalized utility of XYZPlot and how it can be tied to any field. In most cases, a replace operation would be preferred (which XYZPlot does right now).

For prompts, I'd like to avoid human error and excessively large prompts in the XYZPlot node for doing replacing. In this way I'll have a sort of base prompt and the axis for that prompt will add its own values. I think this is what the screenshot is showing in the example of XYZPlot, but I haven't been able to reproduce it myself. When I inspected the data being sent to queue up the new prompt, it looks like my prompt was being overridden by one of the axis values entirely.

I have a branch at https://github.com/LoganBarnett/comfyui-browser/tree/debug-xyz-node which provides one means of doing this for XYZPlot by introducing an update_mode variable for each axis (I only cover X and Y at the moment). While it does function, I'm not convinced this is the ideal way to perform this task. I believe this requires each item in each prompt axis to have a comma to separate the prompt elements. Otherwise it's just a straight concatenation. For example, if you have a prompt of sunset on a beach and the axis values are 1girl; 1boy then you would get sunset on a beach1girl and sunset on a beach1boy. You could fix it with sunset on a beach, (note the added comma) and 1girl,; 1boy, (more commas, assuming you might use two or more axes for your prompt modifier.

I'm thinking maybe we should lean to have a substituter variable for each axis instead. You can leave it blank to replace the value. Then you would write a very semantic prompt like sunset on a beach, {person}, {time}, and your X axis could be 1girl; 1boy, and the Y axis could be day; night, and these would work very well together. You'd just have to give XYZPlot the additional {person} and {time} as x_substituter and y_substituter inputs, respectively. This approach will make the node a little larger / more noisy, but I think it gives us the power we are accustomed to seeing with ComfyUI (and XYZPlot) without adding magic that will later create a higher maintenance burden.

Thoughts?

I plan on implementing this, but I won't take offense if you'd rather see this done a different way and I've already arrived at a pull request before you could weigh in :)

Currently for model download we have the choice of Checkpoint, LoRA, VAE, Embedding, and ControlNet. By choosing the right model type, the downloaded model is place in the right directory.

What is missing is the Unet model type, which will allow the placement of Unet model such as Stablity Cascade into the models/unet directory.

As I said on Reddit a few weeks ago, your XYZ Plot node is the elegant and powerful way to plot complex parameter variations in a complex workflow. Thanks for sharing it with the community.

I'm preparing to include this node in the upcoming AP Workflow 9.0 (it's already in the Early Access 3 prerelease), and I wanted to ask if you could add a way to show arbitrary axis labels.

It would make the plot easier to read, especially if you are showing the plot to people that didn't prepare the node configuration.

Thank you for considering my request.

Close和Browse in new tab这一行都没有，就离谱

I have this setup:

The intention is to generate an XYZ plot that modifies the checkpoint name in the SDParameterGenerator node, part of the SD Prompt Reader suite. But I also tried doing the same with an Efficient Loader node.

I receive this error:

Traceback (most recent call last):
File "D:\Tools\ComfyUI\execution.py", line 151, in recursive_execute
output_data, output_ui = get_output_data(obj, input_data_all)
File "D:\Tools\ComfyUI\execution.py", line 81, in get_output_data
return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True)
File "D:\Tools\ComfyUI\execution.py", line 74, in map_node_over_list
results.append(getattr(obj, func)(**slice_dict(input_data_all, i)))
File "D:\Tools\ComfyUI\custom_nodes\comfyui-browser\nodes\select_inputs.py", line 39, in run
node_id, node_title, widget_name = input.split("::")
ValueError: not enough values to unpack (expected 3, got 1)

I thought that perhaps I'm defining the list of checkpoints in the wrong way.

I use Vivaldi browser (a flavor of Chrome) on macOS to connect to a local ComfyUI server on Windows.
The checkpoints are in the Windows fileserver.

Any help would be greatly appreciated. Thank you.

使用久了，难免需要手动点开输入目录，清理一些图。看到您这个插件已有的优秀功能，不禁想要一个针对输入目录的。
如能实现，万分感谢！

Win11系统，不显示底部按钮，造成无法关闭窗口。

另外，期待添加批量删除功能。想删除的图片很多的时候，一个一个删除很麻烦啊！

This seems like a very useful feature, but a bit more flexibility in terms of output path would make it even better.

It should either read the content of the --output-directory that is passed at startup time (not sure is possible) or, preferable, have an entry in the ComfyUI Manager settings menu where you can specify your output dir.

Thanks for considering the request.

I am getting the following error with the XYZ Plot node:

File "D:\Tools\ComfyUI\execution.py", line 151, in recursive_execute
output_data, output_ui = get_output_data(obj, input_data_all)
File "D:\Tools\ComfyUI\execution.py", line 81, in get_output_data
return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True)
File "D:\Tools\ComfyUI\execution.py", line 74, in map_node_over_list
results.append(getattr(obj, func)(**slice_dict(input_data_all, i)))
File "D:\Tools\ComfyUI\custom_nodes\comfyui-browser\nodes\xyz_plot.py", line 157, in run
queue_new_prompt(new_prompt)
File "D:\Tools\ComfyUI\custom_nodes\comfyui-browser\nodes\xyz_plot.py", line 110, in queue_new_prompt
return requests.post(SERVER_BASE_URL + '/prompt', data=data, proxies=proxies)
File "D:\Tools\ComfyUI\venv\lib\site-packages\requests\api.py", line 115, in post
return request("post", url, data=data, json=json, **kwargs)
File "D:\Tools\ComfyUI\venv\lib\site-packages\requests\api.py", line 59, in request
return session.request(method=method, url=url, **kwargs)
File "D:\Tools\ComfyUI\venv\lib\site-packages\requests\sessions.py", line 589, in request
resp = self.send(prep, **send_kwargs)
File "D:\Tools\ComfyUI\venv\lib\site-packages\requests\sessions.py", line 703, in send
r = adapter.send(request, **kwargs)
File "D:\Tools\ComfyUI\venv\lib\site-packages\requests\adapters.py", line 519, in send
raise ConnectionError(e, request=request)
requests.exceptions.ConnectionError: HTTPConnectionPool(host='0.0.0.0', port=8188): Max retries exceeded with url: /prompt (Caused by NewConnectionError('<urllib3.connection.HTTPConnection object at 0x000001E23270EA70>: Failed to establish a new connection: [WinError 10049] The requested address is not valid in its context'))

This is the default workflow I tested:

I'm having this problem with both Vivaldi and Safari on macOS. I remotely connect to a ComfyUI server (on Windows) in the network.
The browser plugin, part of the suite, works fine. It's just the XYZ Plot node that causes this issue.

What am I doing wrong?

Thanks for your help.

Add ability to save animated gifs or mp4 files in the xyz plot result page.

I set up this basic WF that should save AD runs as gif files that are sent to the xyz plot node.

Everything works but the xyz node only saves jpg files of the first frame rather than the gifs

AD-xyz-polt.json

vhs_filenames.py.txt

Save the vhs_filenames.py.txt in custom nodes folder and remove ".txt" (can not uplaod files with .py extension)

Thank you

Thx for this extension.

Unfortunately, this does not work with the new menue bar, as there is no button.

See screenshots:

"Browser" button visible on old side bar:

No "Browser" button on new menue bar:

Here is the screen in Comfyui Manager:

I don't see any relevant logs.

订阅github工作流，自动去gitcode.net拉取，可否取消这个设置，或者添加取消的入口，如果考虑到墙内用户，可以增加一个代理设置，用户自由选择，gitcode.net 太烂了，还不如gitee

不知道是不是我的个案问题，我浏览输出图像时，最多只能显示20个图片，且无法进行滚动浏览。

环境：
Windows11
Edge/谷歌浏览器
Comfyui：Released on '2024-01-13'
comfyui-browser也更新到了13日之后的最新版本

另，在我mac版本的comfyui下无此问题。

It's good that the output results are individual png
I also need output in one png.

https://github.com/talesofai/comfyui-browser/blob/main/utils.py#L10

SERVER_BASE_URL = f'http://{args.listen}:{args.port}'
The above code doesn't support ipv6 address.

it should be
SERVER_BASE_URL = f'http://[{args.listen}]:{args.port}'

Thanks!