LoRA Quick Feedback v03

The gap between "this LoRA worked great" and "which LoRA was that again" is exactly where good testing habits die. LoRA Quick Feedback v03 is built to close that gap: right after you've generated test images, you rate the LoRA - quality, ease of use, versatility - and jot the practical notes (which checkpoint, what CFG, what resolution) while it's all still in your head. It appends that to the LoRA's library card in the pack's database, so weeks later the Stats node can tell you which of your LoRAs are actually worth keeping.

The inputs are a rating form: lora_path (which can be piped straight from the LoRA Tester node, per its tooltip), show_current_feedback (displays what's already recorded before you update - the anti-duplicate measure), and three rating dropdowns - quality_rating, ease_of_use, and versatility - each offering Keep Current / Not Rated / 1–5 Stars. update_mode decides whether your new_notes append to the existing notes, replace them, or leave them alone. tested_with_checkpoint names what you tested against, and the optional extras capture the useful specifics: style_tags, cfg_range (e.g. "7-12"), and resolution_pref (e.g. "1024x1024"). This is exactly the kind of info you'll wish you'd written down the day you can't remember why you kept a LoRA.

Outputs are lora_path (pass-through, so you can chain the node), current_feedback (what was on the card before this run), feedback_status (did the update land), and summary - the accumulated feedback, handy for a quick sanity check.

Mechanically it writes to the same lora_tester_db.json the whole LoRA Tester family uses. The smart touch is show_current_feedback defaulting to on: it makes the node safe to run repeatedly without accidentally clobbering ratings you gave last month - you see what's there before you change it.

Two honest notes. First, the ratings are only as useful as your discipline in using them - this node makes rating easy, it can't make you rate. Second, it's a single-author pack with zero community chatter, so the "Keep Current" defaults are the author's safe choice, and you should treat them as the guardrails they are. If you're testing a big batch of LoRAs, this is the lightest-weight way to leave a trail; the gallery and stats nodes turn that trail into decisions later.

Install the pack and restart, then find it under loaders/lora tester:

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

Generate your test grid, drop the LoRA's path in, rate it, and move on to the next one. Future-you will be grateful.