Conditioning Passer

You will not get a cool image out of this node. It doesn't sample, encode, or touch a pixel. What it does is carry a CONDITIONING value from A to B without letting the graph forget what that value is - and in a big workflow, that small guarantee is worth more than it looks.

What it is

A passer is the plumbing version of a reroute: it takes an input and hands it straight back out, no transformation. Where it differs from a raw wire is that it gives you a place to break a connection and rejoin it, which matters when you're routing one conditioning output to several places, or when you want a bypass switch you can flip without rewiring. Eclipse ships a whole family of these - model, clip, vae, latent, image, float, and so on - and the Conditioning Passer is the one for positive/negative conditioning.

The important design detail: it's typed, not a wildcard. A lot of "pass anything" nodes accept * and silently erase the type, which breaks downstream type-checking and makes graphs harder to debug. This one declares a CONDITIONING template on both the input and the output, so whatever you connect on the left comes out the right still knowing it's conditioning.

How it works

Under the hood it's about as simple as a node gets. The input conditioning is marked "Conditioning input to be passed through," and the execute method is literally one line: return whatever came in. The output socket is also named conditioning. Nothing is cloned, nothing is validated against your model - it's a straight pipe.

Where people actually use these: in workflows that toggle branches. Wire the conditioning from a CLIP Text Encode into the passer, then run the passer's output into a switch or sampler. If you want to mute that branch, you mute or bypass the passer instead of hunting through the graph for the original wire. It also keeps your graph readable when one conditioning feeds three samplers - the single source fans out through the passer without you having to redraw five diagonal wires.

The one thing to know

There's no config here. One input, one output, both conditioning. If you see the node doing nothing in a workflow, that's correct behavior - verify what's connected on both ends, not the node itself.

It's part of the Eclipse pack (formerly RvTools, rewritten and renamed in v4.0.0), so it installs with the whole suite.

Install

The easiest path is ComfyUI Manager - search ComfyUI_Eclipse and install, then restart. Or clone it yourself:

cd ComfyUI/custom_nodes
git clone https://github.com/r-vage/ComfyUI_Eclipse

Then restart ComfyUI. You'll find it under Eclipse → Router → Typed. No model downloads, no extra Python packages beyond what the pack's requirements.txt lists (torch, numpy, Pillow, opencv-python and friends), and ComfyUI usually already has those.

Troubleshooting

• Node doesn't appear: restart ComfyUI and check the server console for import errors. A missing package in your environment is the usual suspect - install it with pip install -r ComfyUI_Eclipse/requirements.txt into the same Python that runs ComfyUI.
• Conditioning "doesn't do anything": that's the job. If your sampler is getting the wrong prompt, check what's upstream of the passer, not the passer itself.
• Loading an old workflow with red nodes: if it predates v4.0.0, run the built-in Workflow Migration Tool or python tools/migrate_workflow.py <your_workflow.json> - RvTools-era node IDs were cleaned up.