Model List Reader
Your model folder, dumped into text (yes, that's the whole node)
- model_list
- model_count
Let's not pretend this is a heavyweight. Model List Reader is the quiet half of the ComfyUI-ModelMetadataReader pack, and it does exactly one thing: it lists every model ComfyUI can see in a given folder type and hands you the list as text plus a count. No API, no key, no model downloads, no hidden magic. The pack's other node - Model Metadata Reader - is the one that digs into a safetensors file and pulls out training params and download URLs. This one is the dumb companion that just tells you what files are sitting in your models folder.
You reach for it when you want to know what you actually have without leaving the graph. It's genuinely useful for debugging (did that LoRA land in the right folder?), for scripting workflows that branch on model count, or for feeding a list of filenames into some downstream node that wants text. That's the niche. It's a small niche, but it's real.
How it works
One input, model_type - a dropdown with nine choices: checkpoints, loras, vae, controlnet, embeddings, upscale_models, clip_vision, diffusion_models, and text_encoders. Default is loras. Under the hood it calls folder_paths.get_filename_list(model_type), which is the exact same function ComfyUI's core loader nodes use to populate their file dropdowns. So the list you get back is identical to what a CheckpointLoaderSimple would show for that category - nothing more, nothing less.
Two outputs:
model_list- a STRING with a numbered list of every file ComfyUI sees in that category.model_count- an INT with how many files that was.
The gotchas you'll actually hit
The outputs are text and a number, not model objects. You can't wire model_list into a loader to pick a model - it's information about your library, not a choice. Beginners sometimes expect a dropdown that outputs a model, and this just isn't that.
Two more, both straight from the source: the list text is labeled in Chinese (the pack's author writes in Chinese - the output reads roughly "model type", "model total", "model list"), so don't panic when your filenames show up wrapped in Chinese headers. That's the node, not a bug. And if the category you picked is empty or the lookup throws, you get an error string and a model_count of 0 rather than a crash.
Installing it
Through ComfyUI Manager: search for ComfyUI-ModelMetadataReader and hit install, then restart ComfyUI. Or by hand:
cd ComfyUI/custom_nodes
git clone https://github.com/kana112233/ComfyUI-ModelMetadataReader
Then restart. That's the whole install - the pack has no requirements.txt and only imports folder_paths (ComfyUI core) and safetensors (for the sibling node), both of which ship with ComfyUI. No heavy dependencies, no models to fetch. After the restart you'll find it under the utils/model category in the node menu.
When it won't help you
The one real trap: this node only sees files inside ComfyUI's configured model folders. Drop a model somewhere outside models/loras or wherever and it won't appear - same as the core loader dropdowns. If you just downloaded something and it's missing, restart ComfyUI so it rescans. And if you were hoping for a node that maps your whole library's architecture compatibility, that's not this. It's a file list. For a list, it does the job.
Inputs (1)
| Name | Type | Default | Description |
|---|---|---|---|
| model_type | COMBO | loras | 9 options: checkpoints, loras, vae, controlnet, embeddings, upscale_models, +3 |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| model_list | STRING | — |
| model_count | INT | — |