Model List

Model List is a dropdown for LoRA models that's backed by a CSV file instead of a folder scan. You maintain a list where every row is a LoRA with a human-readable name, a path, and prefix/suffix text for prompt engineering. The node turns that into a pick-list, and when you select a row it spits out the path (M), prefix (P) and suffix (S) as strings. Critically, it doesn't load any model - ComfyUI's own LoRA loader still does that. This node just hands the loader a name from your list.

How it works

That division of labor is the whole thing to understand, because it's the difference between "works" and "empty dropdown." The full chain, per the README:

1. The CSV's model path column holds a filename the LoraLoader would normally show in its own dropdown.
2. Pick a row here; the node outputs that path as M.
3. M feeds a CR String To Combo node (which turns a string into a one-item list).
4. That list plugs into the LoraLoader's lora_name input.

The actual LoRA application happens downstream in the LoraLoader. This node's entire job is the pick-list plus the free prefix/suffix, which it prepends and appends to your prompt so the right trigger words ride along automatically.

Setup

Setup is file-based, and this is where people stumble. The pack ships a modellocation.txt that points at /mnt/private/models/lora/ - that's the author's own server path. This pack comes from Finetuners.AI, a company that trains models for hire, and this node reads like internal tooling that got published: a curated, human-readable list for handing a client "load the LoRA" as a two-click task. On your machine that path won't exist. You edit modellocation.txt to point at a folder containing modelist.csv - same name, modelist.csv, inside that folder - with exactly these columns:

Name on list,model path,prefix,sufix
cartoon 1,style/cartoon1.safetensors,a cartoon image of,sbsg style

Yes, sufix is misspelled in the original. Keep the typo, because the code reads row['sufix']. "Fix" it to suffix and the lookup throws - the exception is swallowed, and you get an empty dropdown with no error message at all. A sample CSV lives in the pack's workflow folder if you want to copy it.

Gotchas

• The dropdown only contains what's in the CSV, and the CSV is re-read when ComfyUI refreshes node definitions. Edit the file, then restart ComfyUI to see changes.
• The CSV model path values must match what the LoraLoader dropdown would list, relative to your models/loras folder. A path that doesn't exist there means a silently dead LoRA - it'll select, output, and do nothing.
• Missing or malformed CSV → one empty dropdown entry and empty M/P/S outputs, with no in-UI hint why.
• The trailing comma the node appends to M is a hack so a string-to-combo node can split it. Skip that hop and wire M straight into lora_name and it won't match.

When to reach for it

If you already use rgthree's Power Lora Loader or just the stock LoraLoader, you don't need this - folder scans cover most people. Model List earns its place when you want a curated list with friendly names and attached trigger text, which is a genuinely nice thing for teams and for anyone tired of squinting at lora_foo_v2_final_really.safetensors.

Install

Install is the same as the rest of the Finetuners suite - no pip dependencies, no models to download:

cd ComfyUI/custom_nodes
git clone https://github.com/FinetunersAI/ComfyUI_Finetuners_Suite

Restart ComfyUI and it appears under "finetuners". Ignore the README's install command, which points at a stale finetunersTest repo; the Suite repo above is the right one (and what ComfyUI Manager installs when you search "ComfyUI_Finetuners_Suite").