# 🔥 ComfyUI-LevelX-Nodes
**The Ultimate "Fire & Forget" LoRA System for ComfyUI.** Automatically manages LoRA loading, trigger word injection, and database synchronization across multiple generation engines (SDXL, FLUX, Qwen, Z-Image).
---
## ✨ Key Features
### 1. 🧠 Intelligent Auto-Loader
Stop copy-pasting trigger words manually. These nodes automatically:
* **Load the LoRA Model:** Handles model & clip strength.
* **Inject Trigger Words:**
* **Prefix Mode:** The 1st LoRA's trigger is forced to the *front* of your prompt (critical for style/character).
* **Suffix Mode:** 2nd and 3rd LoRA triggers are appended to the *end*.
* **Smart Deduplication:** If you already typed the trigger, it won't add it again.
* **Engine Filtering:** Dedicated nodes for **SDXL**, **FLUX**, **FLUX 2**, **Qwen**, and **Z-Image** so you never load the wrong format.
### 2. 🕵️♂️ Advanced Trigger Manager
A powerful new node (`Level X Trigger Manager`) that builds your database for you.
* **Scan Local:** Instantaneously extracts triggers from:
* `.civitai.info` files.
* `.txt` sidecar files.
* Internal `.safetensors` metadata (ModelSpec & Kohya).
* **Scan Online:** Queries the Civitai API live to find triggers for unknown LoRAs.
* **Force Re-Scan:** One-click option to rebuild your entire database from scratch.
### 3. 📂 Self-Healing Database
* **Location:** Stores data in `lora_trigger.json` right next to the nodes.
* **Auto-Update:** If a loader finds a new trigger during generation (via metadata), it silently updates the JSON file so the next run is instant.
* **Smart Caching:** Only reloads the JSON from disk if the file has actually changed, ensuring maximum performance.
---
## 🛠️ Installation
1. **Navigate** to your custom nodes directory:
```bash
cd ComfyUI/custom_nodes/
```
2. **Clone** or **Copy** the `ComfyUI-LevelX-Nodes` folder here.
3. **Install Dependencies** (if missing):
```bash
pip install safetensors
```
4. **Restart ComfyUI**.
---
## 🚀 How to Use
### A. The Auto-Loaders (Generation)
1. **Add Node:** Right-click > `Level X` > `Loaders`.
2. **Select Engine:**
* `🔥 Universal`: Shows ALL LoRAs (good for testing).
* `⚡ FLUX`: Shows only `/FLUX` folder LoRAs.
* `🎨 SDXL`: Shows only `/SDXL` folder LoRAs.
* (Also available: `FLUX 2`, `Qwen`, `Z-Image`).
3. **Connect:**
* **Input:** Model & CLIP from your checkpoint.
* **Prompt:** Connect your text prompt (Primitive/String).
* **Output:** Connect Model/CLIP to KSampler, and `final_prompt` to your CLIP Text Encode.
**Visual Flow:**
[Checkpoint] ==(Model/CLIP)==> [🔥 Level X Auto-LoRA] ==(Model/CLIP)==> [KSampler] [Prompt String] =============> [ Node ] ==(final_prompt)==> [CLIP Text Encode]
### B. The Manager (Database Building)
1. **Add Node:** Right-click > `Level X` > `Utils` > `Trigger Manager`.
2. **Select Operation:**
* `SCAN LOCAL (Fast)`: Quick update from local files.
* `SCAN ONLINE (Slow)`: Deep search Civitai for missing tags.
* `FORCE RE-SCAN`: Wipe and rebuild everything.
3. **Run:** Queue the prompt (no inputs needed). Check the output string or console for the report:
> "Scan Complete: Added 45 new triggers [Online: True]"
---
## ⚙️ Technical Details
### Database Logic
The system uses `lora_trigger.json` as the source of truth.
* **Format:** `{"SDXL/style.safetensors": "trigger_word, style"}`.
* **Matching:** Uses fuzzy matching (ignores case, spaces vs underscores).
* **Sync:** The manager node writes to this file. The loaders read from it.
### Folder Structure
To use the filtered nodes effectively, organize your `ComfyUI/models/loras/` directory like this:
models/loras/ ├── SDXL/ │ └── character_a.safetensors ├── FLUX/ │ └── style_b.safetensors ├── FLUX2/ │ └── ... └── Qwen/ └── ...
*(If you dump everything in root, just use the `🔥 Universal` node).*LIKE AND BUZZ ME FOR SUPPORT !