RunningHub File Previewer

First thing to know: this node is not for your local ComfyUI. It lives in the Addoor/RHAPI corner of the Addoor pack, and "RHAPI" means RunningHub API - the cloud ComfyUI host at runninghub.ai. If you're running Comfy on your own machine, you don't need this. If you're building a workflow that lives on RunningHub and you want a particular file to show up in the results, that's the whole job.

Here's the problem it solves. On a cloud host, the graph runs on someone else's box, not yours. Your local disk doesn't exist over there - a fact people rediscover the hard way. As one RunningHub user put it when their local prompt-loader node stopped working: "it doesn't work on RunningHub because it can't connect to my PC." So the platform grows its own little ecosystem of nodes for shoving files in and pulling files out of the cloud environment. RunningHubFilePreviewer is the "pull out" side: point it at a path that exists inside the run, and it makes that file visible instead of leaving it stranded in some temp directory the platform never shows you.

How it works

It's an output node - a terminal, the end of a branch, same structural role as SaveImage. Output nodes are the things ComfyUI actually executes the graph to satisfy; everything upstream of them runs because they demand it. The difference from SaveImage is that this one doesn't care what kind of file it is. A zip, a text log, a CSV, a video, a model checkpoint you just downloaded mid-workflow - if it's a real file at the path you give, this node registers it as something the run produced so RunningHub's results panel can pick it up. Think of it as "hey platform, this file is an output, don't throw it away."

It also passes the path back out as a string, so you can chain it or log it rather than dead-ending completely.

The inputs and outputs that matter

There's exactly one input and one output, which tells you how focused this thing is:

• file_path (STRING, required) - the path to the file you want surfaced. This is a path inside the cloud run, not a path on your laptop. It's usually something an upstream node produced and reported back - a save node's output path, a downloaded file, a zip the Addoor AD_ZipSave node just wrote.
• Output: STRING - the node emits the path back out. Handy if you want to feed it into another node or just keep it in view; mostly you leave it unwired and let the node do its output-node thing.

That's the entire surface. No format toggles, no options. It previews whatever you point it at.

How to install it

It ships inside the whole ComfyUI-Addoor pack (the author calls it 葵花宝典, the "Sunflower Manual" - a grab-bag of CSV, batch-image, file-op, and Hugging Face utility nodes, with the RHAPI nodes bolted on more recently). Two ways in:

• ComfyUI Manager - open the Manager, search ComfyUI-Addoor, install, restart.
• Git - cd ComfyUI/custom_nodes && git clone https://github.com/Eagle-CN/ComfyUI-Addoor, then cd ComfyUI-Addoor && pip install -r requirements.txt, and restart ComfyUI.

No model downloads, no giant dependency tree - the pack is pure-Python utility glue. On RunningHub itself you typically don't install anything; you just search for the node in their editor, since the platform maintains the custom nodes for you. The install steps above are for the rare case you're running Addoor locally to build and test before pushing a workflow up.

Common issues & troubleshooting

The one that gets everybody is the local-vs-cloud path confusion. If you feed it a path that made sense on your own machine - C:\Users\you\... or /home/you/... - it'll find nothing, because that filesystem isn't there. Wire the path from an upstream node that generated the file in the run, not one you typed from memory.

A blank file_path gives it nothing to preview, so leaving the default empty string in place is a no-op. And because it's an output node, if it's the only output in your graph and its path is wrong, you can end up with a run that "succeeds" but shows you nothing useful - check that the file actually exists at that path before blaming the node.

Honestly, if RunningHub isn't part of your stack, skip this one entirely. It's a good, small, single-purpose tool - for exactly one platform. Outside that context it has nothing to do.