Debug Metadata (LoraManager)

This one's a developer's flashlight, not a node you'll wire into a finished workflow. Its whole purpose, per its own description, is to verify that the pack's metadata processor is doing its job - you feed it images, and it shows you the generation metadata the pack would extract from your graph. If you've ever wondered what exactly Save Image (LoraManager) is going to embed, this is how you look before you commit.

Most people will never need it. But when you're building the metadata half of this pack - say the model name isn't showing up in your saved files, or a prompt field lands in the wrong place - this node lets you inspect the extracted values directly instead of saving a file, opening it in another tool, and reverse-engineering what went wrong.

The metadata this pack writes is the CivitAI/A1111-style generation info - model, prompt, seed, sampler settings - that lets an image carry its own recipe. Getting that extraction right is fiddlier than it sounds, because the values live scattered across different nodes in your graph, and the processor has to find them. This node is how you confirm it did before you rely on it for a whole batch.

How it works

It's an output node: connect your images into it and run the graph. Rather than saving anything, it surfaces the metadata the processor gathered - the same data Save Image (LoraManager) would write into a file. Think of it as a dry run for the metadata pipeline. It produces no output wire; the result is what it displays.

Inputs and outputs that matter

• images (IMAGE, required) - the images whose metadata you want to inspect. Usually straight off your VAE decode.

No outputs. It's a terminal debug/display node - the value is what it shows you, not a wire you route onward.

How to install it

Via ComfyUI Manager: search lora-manager, Install, restart. Or by hand:

cd ComfyUI/custom_nodes
git clone https://github.com/willmiao/ComfyUI-Lora-Manager.git
cd ComfyUI-Lora-Manager
pip install -r requirements.txt

then restart. Light Python install, nothing to download.

Common issues & troubleshooting

It shows less metadata than I expected. The processor extracts from what's actually in your graph. If a field is empty here, it'll be empty in your saved files too - which is the point of checking. Trace it back to the node that should be supplying that value (the checkpoint loader for the model name, the sampler for the seed, and so on).

Should I leave it in my workflow? No - it's a diagnostic. Once you've confirmed the metadata looks right, wire up Save Image (LoraManager) for real and remove the debug node. It doesn't save anything, so leaving it in just adds a dead end.

Nothing happens when I run it. Make sure images are actually connected and reaching it. As an output node it only fires when the graph executes a path that leads to it.