LoRA Info Viewer v03

LoRA filenames are the worst documentation ever written. style_0713_v2_flux.safetensors tells you almost nothing, and if you've been downloading for a year you have no idea which file has the trigger words you need or which one is the Pony variant. LoRA Info Viewer v03 is the lookup side of that problem: give it a path to a LoRA file and it spits out everything the AAA Metadata System has catalogued about it.

It lives in the "loaders/lora tester" category and belongs to a family of nodes that treats your LoRA folder as a searchable library. LoRATester_v03 picks and loads LoRAs for you, the info setters record what each one is, and this node is the read-only window into that catalog. If the tester feels like the "load" button, this is the "inspect" button.

How it works

Under the hood there's no magic - no API calls, no model loading. The node reads from a local JSON database (lora_tester_db.json) that ships in the pack's nodes folder and gets populated by the LoRA tester and batch-info nodes. When you enter a path, it looks up that file's entry and returns the recorded fields. It also hunts for "associated images" the old-fashioned way: same basename as the LoRA with a .png, .jpg, .webp, etc. extension, plus numbered variants like lora-1.png or lora_2.jpg. Those get loaded into an IMAGE batch you can actually look at.

So the honest framing: it's only as smart as your database. Out of the box the DB is empty, and this node shows you nothing until you've run a tester or setter over your files.

Inputs and outputs that matter

One required input: lora_path, a plain string with the full path to the .safetensors (or whatever) file. No dropdown, no folder browser - paste the absolute path.

The outputs are the interesting part:

• lora_name and architecture - what the file is called and whether it's SDXL, Flux, Pony, etc.
• category - style, character, concept, and so on.
• trigger_words - the strings you'd drop into your prompt. This is usually the one you actually want.
• compatible_items and notes - whatever whoever tagged it wrote down.
• associated_images - the IMAGE batch mentioned above, handy for wiring into a preview node.

A common pattern is feeding trigger_words into a text concat before your CLIP encode, or using architecture to gate which workflow branch runs.

Installing it

The node ships inside the AAA Metadata System pack by Eric Hiss (GitHub: EricRollei). Easiest route is ComfyUI Manager - search "AAA Metadata System" - or clone it by hand:

cd ComfyUI/custom_nodes
git clone https://github.com/EricRollei/AAA_Metadata_System.git
cd AAA_Metadata_System
pip install -r requirements.txt

Restart ComfyUI and you're done. Core deps (Pillow, opencv-python, numpy, imagehash, requests, and friends) install automatically; this particular node needs nothing exotic.

Common issues

The big one is empty output: you pointed it at a valid file and got blank strings. Nine times out of ten that's because nothing has written an entry for that file yet. Run LoRATester_v03 (or the batch LoRA info setter) once so it scans and tags your folder, then re-run this. Also double-check the path is a real absolute path the node can stat - a typo just silently returns blanks rather than erroring. And remember the node is a viewer, not a loader: it won't apply the LoRA to a model, so don't expect a MODEL output to come out of it.