comfyanonymous / ComfyUI

The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface.
https://www.comfy.org/
GNU General Public License v3.0
56.03k stars 5.92k forks source link

[FEATURE] Adding TABS in the bottom of the UI (similar to EXCEL tabs) in order to be able to switch between multiple workflows #1239

Open StableInfo opened 1 year ago

StableInfo commented 1 year ago

Same as EXCEL tabs, where you can select from one page to another one, I believe ComfyUI heavily needs one, it is especially good if you dont want to lose a workflow from your sight and be able to switch between workflows, it is useful for beginners when testing new workflows, comparing, looking out for differences and new nodes. It is generally useful for people who alternate between multiple workflows.

Can you implement it please? Thanks

Ferniclestix commented 1 year ago

This would be super cool, you could add them to the same queue and it shouldn't be incompatible

fuami commented 1 year ago

Pretty sure this is basically the same as #1173

StableInfo commented 1 year ago

Indeed, it shows how much needed it is indeed needed! Hopefully it will make it. btw, does the repo master update comfy often?

ltdrdata commented 1 year ago

There are a few tricks for this.

You can use --listen 0.0.0.0 and access it in the form of 127.0.0.1:8188, 127.0.0.2:8188, and so on.

The workflow itself will likely be separated. However, since the execution cache is shared, there's a limitation where running the workflow in one tab will reset the cache in other tabs. This is a non-negotiable aspect due to memory constraints, as avoiding simultaneous retention is necessary.

Alternatively, you can register them in the hosts file like this and access them as comfy1:8188, comfy2:8188, and so on.

127.0.0.1 comfy1 
127.0.0.1 comfy2
127.0.0.1 comfy3

In my case, I use the latter method.

Lex-DRL commented 4 months ago

@ltdrdata Thanks for the workaround suggestion.

... but still, it's just what it is: a workaround. And quite a hacky one, especially for "non-technical" users who struggle even with cloning a github repo. We could use your approach for now, but I believe the issue is still relevant. As much as before.