Evan Coleman 1157ebfb01 Add model_ref output to Download LoRA node
Second output passes through the original combo value (modelId@versionId:name)
for plugging into the Upload Image node's model_name input.
- lora_name: filename for standard LoRA loader
- model_ref: Model Manager reference for Upload Image node
2026-02-09 15:12:38 -05:00
2026-02-08 20:31:57 -05:00

ComfyUI Model Manager

Custom nodes for ComfyUI that load models from a Model Manager REST API. Models are downloaded on demand and cached locally.

Nodes

  • Load Checkpoint (Model Manager) - Load checkpoint models
  • Load LoRA (Model Manager) - Load a single LoRA with strength controls
  • Load LoRA Multi (Model Manager) - Load multiple LoRAs with per-slot toggles and strength controls
  • Load VAE (Model Manager) - Load VAE models
  • Clear Cache (Model Manager) - View cache size and clear downloaded models

Features

  • Connect to any Model Manager API with URL + API key
  • Automatic local caching (models are only downloaded once)
  • In-node connect/disconnect for API authentication
  • LoRA multi-loader: add/remove slots dynamically, per-slot toggle switches, inline strength adjustment with arrow controls, baseModel-grouped dropdown, right-click context menu for move/remove, Toggle All switch

Installation

Clone this repository into your ComfyUI custom_nodes directory:

cd ComfyUI/custom_nodes
git clone https://github.com/evandcoleman/comfyui-model-manager.git
cd comfyui-model-manager
pip install -r requirements.txt

Configuration

You can configure the connection in three ways:

Click the Connect to Model Manager button on any loader node, enter your API URL and API key. Connection info is saved to config.yaml for future sessions.

2. Config file

Copy config.yaml.example to config.yaml and fill in your values:

api_url: "http://localhost:3000"
api_key: "your-api-key"

3. Environment variables

export MODEL_MANAGER_API_URL="http://localhost:3000"
export MODEL_MANAGER_API_KEY="your-api-key"

Environment variables take priority over config.yaml values.

License

MIT

S
Description
No description provided
Readme MIT
74 KiB
Languages
JavaScript 51.9%
Python 48.1%