Edit SD Lora Information

Every LoRA you've downloaded carries a little story that nobody wrote down: what it does, what trigger word it needs, what weight it actually likes. CivitAI shows some of that in a browser tab, but the tab doesn't follow the file into ComfyUI, and six months later you're staring at anime_style_v3.safetensors with no idea what made it special. Edit SD Lora Information is the fix: it stores that story as a .json file sitting next to the LoRA, and gives you a UI to read and write it.

The mechanism is the sidecar convention - the same "same-basename sibling file" idea that training already uses for captions (image-io-metadata.md has the full picture). For my_lora.safetensors in models/loras/, the node reads and writes my_lora.json in the same folder. The format is the community-standard shape:

{
  "description": "Anime style LoRA",
  "sd version": "SDXL",
  "activation text": "my_style",
  "preferred weight": 0.8,
  "negative text": "lowres",
  "notes": "Weight 0.6-1.0 recommended"
}

Inputs: lora_name (required, a dropdown of every LoRA in models/loras/), plus optional fields lora_prompt, description, sd_version (defaults SDXL), activation_text, preferred_weight (float, default 1, range 0–10), negative_text, and notes. When you pick a LoRA, the UI loads its sidecar and thumbnail (it scans for .preview.* and image files next to the LoRA) so you can edit and save. Outputs mirror every field - lora_name, lora_prompt, description, sd_version, activation_text, preferred_weight, negative_text, notes - so the node can sit inline and feed a LoRA loader or a note node with the saved values.

Where this earns its place:

• A trigger-word notebook that lives with the file. Activation text is the single most-forgotten bit of LoRA lore. If it's in the sidecar, you stop Googling "what's the trigger for this LoRA" forever.
• The pack-swap problem. You switch LoRAs in a workflow and forget to adjust the weight. A sidecar that says preferred weight: 0.8 - plus the node's preferred_weight output wired into your loader - makes the reminder automatic.
• It's plain JSON. Any tool, any editor, any script can read and write it. No lock-in; the node just makes it pleasant.

The honest catches:

• It doesn't edit the LoRA file itself. Safetensors can carry embedded metadata, but this node neither reads nor writes that - it manages the sidecar JSON only. If the .json is deleted, the info is gone with it. Keep the pair together when you move files.
• The UI needs the frontend extension. This is one of the pack's ⚠️ nodes - the read/thumbnail/save UI is JavaScript, and per the README it may not work under the Nodes 2.0 frontend. The Python side still functions; you'll just lose the edit buttons and preview. If you're on Nodes 2.0 and the UI looks dead, that's the known issue, not you.

The pack (2daadv/ComfyUI-GadgetNodes, MIT, one developer, brand new - no community reputation to report yet) installs via ComfyUI Manager - search "GadgetNodes" - or git clone https://github.com/2daadv/ComfyUI-GadgetNodes into custom_nodes/ + pip install -r requirements.txt + restart. For one-person LoRA hoards this node is a small quality-of-life win; for anyone who trains their own LoRAs and keeps forgetting the trigger word, it's the kind of thing you wonder how you lived without.