ComfyUI本体のAnimaLLLiteApplyとのノードID衝突を回避するためリネーム

ComfyUI本体がbeta版のAnima LLLiteノード（comfy_extras/nodes_model_patch.py の
AnimaLLLiteApply）を追加したことで、本ノードのIDが衝突し登録がスキップされる
ようになった（ComfyUIはビルトインとIDが重複するカスタムノードを黙って無視する）。

- ノードID / Pythonクラス名: AnimaLLLiteApply → AnimaLLLiteApply_sdscripts
- 表示名: "Apply Anima ControlNet-LLLite (sd-scripts)"
- READMEに破壊的変更の注意書きを追加（旧ワークフローはノードの置き直しが必要）
- 本体ノードは同じv2重みフォーマットのネイティブ再実装である旨もREADMEに記載

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

README.md

@@ -12,6 +12,34 @@ This is intended as a **minimal reference implementation**. Community nodes
 with extra features (per-step scheduling, multi-cond, region masks, per-block
 selection, …) are welcome and can build on this codebase.
 
+## ⚠️ Breaking change: node renamed to `AnimaLLLiteApply_sdscripts`
+
+ComfyUI core has added its own (beta) Anima LLLite node named
+`AnimaLLLiteApply` (in `comfy_extras/nodes_model_patch.py`). ComfyUI silently
+skips custom-node registrations whose node ID collides with a built-in one,
+so this node's original ID `AnimaLLLiteApply` stopped loading. The node has
+therefore been renamed:
+
+* Node ID: `AnimaLLLiteApply` → **`AnimaLLLiteApply_sdscripts`**
+* Display name: **"Apply Anima ControlNet-LLLite (sd-scripts)"**
+
+**Existing workflows saved with the old node need to be fixed manually:**
+the old node type now resolves to the built-in core node, which has a
+different input signature (it takes a `MODEL_PATCH` from `ModelPatchLoader`
+instead of loading the weights file directly, and has no `mask` /
+`preserve_wrapper` inputs), so old workflows will fail validation. Delete
+the old node, add **Apply Anima ControlNet-LLLite (sd-scripts)**, and
+rewire it. There is unfortunately no way to keep the old ID as an alias
+because the built-in node now owns it.
+
+Note that the core node is a native reimplementation of the **same v2
+weight format** this node loads (sd-scripts-trained checkpoints work with
+both). The main practical differences: the core node loads weights from
+`models/model_patches/` via `ModelPatchLoader` (this node reads from
+`models/controlnet/`), integrates through ComfyUI's attention/MLP patch
+hooks instead of a `model_function_wrapper`, and currently supports
+single-frame (T=1) latents only.
+
 ## Install
 
 Clone into `ComfyUI/custom_nodes/`:
@@ -25,7 +53,8 @@ Place LLLite weights (`.safetensors`) under `ComfyUI/models/controlnet/`.
 
 ## Node
 
-**Apply Anima ControlNet-LLLite** (`loaders` category)
+**Apply Anima ControlNet-LLLite (sd-scripts)** (`loaders` category, node ID
+`AnimaLLLiteApply_sdscripts`)
 
 | Input | Type | Notes |
 |---|---|---|
@@ -70,7 +99,7 @@ silently substituting an empty mask.
 
 ### Cascading multiple LLLite nodes
 
-Multiple **Apply Anima ControlNet-LLLite** nodes can be chained on the same
+Multiple **Apply Anima ControlNet-LLLite (sd-scripts)** nodes can be chained on the same
 `MODEL` to apply more than one LLLite at the same time (e.g. a pose LLLite
 and a depth LLLite, or several inpaint LLLites with different conditions).
 Just feed the output `MODEL` of one node into the `model` input of the next.

nodes.py

@@ -119,7 +119,7 @@ def _build_inpaint_cond_image(rgb_pm1: torch.Tensor, mask01: torch.Tensor,
     return torch.cat([rgb_pm1, mask_pm1], dim=1)
 
 
-class AnimaLLLiteApply:
+class AnimaLLLiteApply_sdscripts:
     @classmethod
     def INPUT_TYPES(cls):
         return {
@@ -279,9 +279,9 @@ class AnimaLLLiteApply:
 
 
 NODE_CLASS_MAPPINGS = {
-    "AnimaLLLiteApply": AnimaLLLiteApply,
+    "AnimaLLLiteApply_sdscripts": AnimaLLLiteApply_sdscripts,
 }
 
 NODE_DISPLAY_NAME_MAPPINGS = {
-    "AnimaLLLiteApply": "Apply Anima ControlNet-LLLite",
+    "AnimaLLLiteApply_sdscripts": "Apply Anima ControlNet-LLLite (sd-scripts)",
 }