Lora Name List
Remember what each of your 40 LoRAs was trained for
- Lora Name
LoRAs are the worst offenders when it comes to "what does this file actually do?" You download them by the dozen, they all have cryptic names, and half of them silently depend on a specific base model or trigger phrase you've long since forgotten. Lora Name List is the comfyui-mdnotes node that gives you a fighting chance: a dropdown of everything in models/loras, and a right-click menu item - "✒️ Show note of lora" - that opens a markdown note for the selected one.
The note habit is the point, and it pays off fastest with LoRAs because the metadata that matters is the stuff you can't read off the filename: what it was trained on, what trigger it needs, what strength it tolerates, whether it's a LoCon or a plain LoRA. Write that down once in the note, and you never have to re-research a model you already own. The KB's lora-training notes hammer the same theme - a LoRA's usefulness lives in its training provenance, not its download page title - so this node is really a tiny organizational habit wrapped in a dropdown.
Mechanically it's the same story as its checkpoint and diffusion-model siblings. The one input, lora_name, is a dropdown auto-populated from models/loras. The widget name contains lora, which is the trigger comfyui-mdnotes' extension scans for, so the note menu attaches here just like it does on the native LoraLoader. Nice touch: if a loader has several lora widgets, the extension adds a separate menu entry per LoRA - "Show note of lora1", "Show note of lora2" - so multi-LoRA workflows stay legible. The output, Lora Name, is typed * (ComfyUI's wildcard), meaning it'll slot into any input that takes a model-name string; convert a loader's name widget to an input and you can drive it from this node's selection.
The backend matching is the same across the pack: your selection goes to /mdnotes/current_model, the server scores every .md file in the LoRA's folder against the model filename using Python's difflib.SequenceMatcher, and opens the top match if it clears the similarity threshold (default 0.5, adjustable in settings). No match means a fresh blank note, named after the LoRA, saved into the folder. That last part is worth repeating: your notes are just .md files sitting next to the .safetensors in models/loras. Editable in any text editor, greppable, and easy to back up - which is exactly what you want for a folder that grows by the week.
Install is the whole-pack routine. ComfyUI Manager: search "comfyui-mdnotes," install, restart. Or by hand:
cd ComfyUI/custom_nodes
git clone https://github.com/Endericedragon/comfyui-mdnotes
# restart ComfyUI
No pip dependencies, no model downloads, nothing to compile - the shipped web/ build is what actually runs. The usual mdnotes gotchas apply: keep ComfyUI current (the pack uses the newer backend API), and if the editor window renders blank you're hitting the CDN that Vditor loads its assets from - turn on "cache vditor resources locally" in settings. If you pull a blank note when you know one exists, the filename match missed the threshold, so name the note close to the LoRA's filename. For a habit that'll save you hours of squinting at filenames, that's a small price.
Inputs (1)
| Name | Type | Default | Description |
|---|---|---|---|
| lora_name | COMBO | 0 | 0 options: |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| Lora Name | * | — |