Saus

Let's be honest up front: the Saus node is a placeholder. It takes zero inputs, does zero computation, and its single output is a value that no other node on earth accepts. The name is a lie - there's no sausage here. But that's exactly why you're seeing it in your node list, and understanding the trick tells you everything about the pack it ships in.

What's actually going on

SAUS ("Simple App U See") isn't a node so much as a whole second UI bolted onto your ComfyUI install. The pitch, straight from the author: ComfyUI is great for understanding what's happening inside a workflow, but the spaghetti gets old fast. SAUS takes popular workflows, hides the tangle, and presents them as clean little "apps" - a web interface where you only touch the parameters that matter, with ComfyUI still doing all the heavy lifting in the background.

So the Saus node is just the registry entry that makes the pack load. When ComfyUI starts up, the pack doesn't care about the node - it runs setup_server() instead, which is where the real work happens. That's what downloads and refreshes the "apps" (bundled workflows) from the author's GitHub repos on every launch, and what registers the routes that serve the whole interface at http://127.0.0.1:8188/saus. It's also what drops that little SAUS button into ComfyUI's action bar so you don't have to type the URL.

One genuine bit of mechanism worth knowing: on Windows the pack patches ComfyUI's folder_paths so the UI and the backend agree on forward versus backslashes. That's a real, if fiddly, compatibility fix - not marketing.

The inputs and outputs

From the schema, there's exactly one thing to know:

• Inputs: none. There's no "required" section at all.
• Output: one Saus output, which returns the string "Saus".

That output doesn't wire into anything. It's a marker so the node exists in the graph and the pack registers; dragging it into a workflow and trying to route it somewhere will just confuse you. Use it to confirm the pack loaded, then go open the actual app.

Installing it

The pack has no Python dependencies - its requirements.txt is empty, which is a small miracle for something this ambitious. The apps themselves are fetched over HTTP on startup. To install:

• ComfyUI Manager: search for "ComfyUI-SAUS" and install.
• Manual: cd ComfyUI/custom_nodes && git clone https://github.com/dsigmabcn/ComfyUI-SAUS then restart ComfyUI.
• One-click scripts: the repo ships install-saus-windows.bat and install-saus-linux.sh, which clone SAUS and the recommended companion nodes for you.

Restart, click the SAUS button (or go to /saus), and you're in.

Common issues

• Some apps need extra nodes. The README recommends installing comfyui_controlnet_aux, KJNodes, ComfyUI_essentials, ComfyUI-RMBG, AnimateDiff-Evolved, and ComfyUI-GGUF. An app that won't run usually just wants one of these.
• First launch needs internet. Apps are downloaded from GitHub on startup; no network, no apps.
• Runpod gotchas. The Runpod template assumes a fresh setup: it needs CUDA 12.8+ (it pins PyTorch 2.8), and SAUS is exposed on port 7771. If you already have ComfyUI on a network drive, the template won't install the node - install it via Manager first.

Is it worth it? If ComfyUI's canvas makes you want to throw your keyboard, maybe. If you live in the graph, the node is a no-op - ignore it and keep wiring.