Derp Lora Stack

The Derp Lora Stack is the pack's take on the multi-LoRA loader, and it's the node most likely to win you over if you live on LoRAs. Instead of chaining four separate LoRA loaders, you get one stack node with per-entry toggles, per-entry strength, and - the standout feature - trigger word handling that tries to fetch the trigger words for you. It's part of xcp_derp-UI (search "xcp_derp-UI" in ComfyUI Manager, by lordwedggie), under derpNodes/Loaders.

Why you'd use it

If you've ever applied a LoRA and then gone hunting for its trigger word - checking the CivitAI page, reading the model card, guessing - this is the pitch. The stack is wired to fetch LoRA triggers and ratings (there's a whole loraComponents helper doing the CivitAI lookups), syncs rating colors, and keeps your LoRAs in one visible stack you can toggle on and off per entry. It's the same idea that made rgthree's Power Lora Loader a fixture of big workflows - see the KB's comfyui-ecosystem.md for why that node travels so well - but wearing derp-UI's clothes.

How it works

The node exposes a single lora_stack output of type LORA_STACK, and it's a pure virtual shell: the Python side returns None, and the stack data rides the pack's wireless signal registry. A stack is a list of LoRA entries - each with its name, enabled state, and strength - that gets consumed downstream by the pack's LoRA loader (the xcpDerpLoraLoader, which takes the stack, loads each enabled LoRA onto model and CLIP, and even injects the associated trigger text). The stack entries themselves are managed from the node's UI: add LoRAs, toggle them, adjust strength, and drag to reorder the apply order. If a saved LoRA no longer exists on disk, the node flags it and offers to remove it rather than silently failing.

The deck also pulls per-LoRA metadata - trigger words and ratings - so you can see at a glance what a LoRA wants in your prompt. That's the feature that turns a list of files into something you can actually drive.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/lordwedggie/xcp_derp-UI-Release

Restart ComfyUI. No pip requirements, no model downloads, no npm step (JS deps are vendored). Manager users: search xcp_derp-UI.

The gotchas

First, the Router rule applies like every derp loader: no derpRouter in the workflow, no signal, no stack. Second, the LORA_STACK type is a pack-local convention - don't expect to plug it into a stock ComfyUI LoRA loader. You use this stack with the pack's own LoRA loader node; mixing ecosystems is where workflows quietly break. Third, trigger fetching depends on CivitAI data being reachable and your LoRA files being identifiable, so it works best for LoRAs that actually live on CivitAI; for local-only LoRAs you'll be typing trigger words yourself.

One workflow note: stack ordering matters. LoRA application order can shift how strongly later LoRAs interact with earlier ones, and the visual order is the apply order - top to bottom. Reorder deliberately, not just for looks.

This is a young, niche pack with thin third-party coverage; the bundled docs in derp_docs/ are the most reliable reference. But the Lora Stack is where the pack's effort is most visible, and it's genuinely one of the better-maintained parts of it.