LF Nodes

<!-- markdownlint-disable MD033 -->

LF Nodes

<div align="center">

</div> <div align="center">

</div> <div align="center">

</div>

Overview

A suite of custom nodes for ComfyUI aimed at enhancing user experience with more interactive and visually engaging widgets.

Most UI elements used by the frontend belong to the LF Widgets webcomponents library, a modern collection of modular and customizable webcomponents built on Stencil.js specifically to integrate with LF Nodes.

The nodes span quite a few categories:

• Analytics nodes: Visualize and track data, like checkpoint/LoRA usage or image histograms.
• Configuration nodes: Manage CivitAI metadata, and control the suite via the Control Panel.
• Filter nodes: Apply various filters to images, including blur, sharpen, and edge detection.
• Image manipulation nodes: Tools to manipulate images, such as filter and resize nodes.
• IO Operations nodes: Load and save files to/from the file system.
• JSON nodes: Tools to manipulate and display JSON data.
• Latent manipulation nodes: Latent decoding and encoding tools.
• LLM nodes: Interface with locally running LLMs, like the Messenger node, which also manages characters.
• Logic nodes: Control flow using simple switches.
• Primitive nodes: Work with primitive data types, offering features like history.
• Region nodes: Image region selection and editing tools.
• Seed generation nodes: Generate seeds for complex workflows.
• Selector nodes: Resource selection widgets with metadata display for models.
• Visual novel nodes: Author, validate, and preview deterministic target-neutral narrative graphs.

To see example workflows, check the example_workflows folder. For generic texture serialization, see DDS Output. For the current visual-novel node inputs and identity contract, see Visual Novel Nodes.

Table of Contents

• LF Nodes
  • Overview
  • Table of Contents
  • Installation
    • ComfyUI Manager
    • Manual
      • Method 1: Download ZIP
      • Method 2: Git Clone
  • Workflow Runner
    • Adding New Workflows
  • Image editor
  • Notes
  • Contributing
  • License
  • Documentation

Installation

ComfyUI Manager

• Open ComfyUI Manager.
• Search LF Nodes.
• Install the node suite and restart ComfyUI.

Manual

Method 1: Download ZIP

• Go to the releases page.
• Download the latest release ZIP file.
• Extract the contents of the ZIP file into the ComfyUI/custom_nodes folder.
• Restart ComfyUI.

Method 2: Git Clone

• Go to the ComfyUI/custom_nodes folder.
• Open a terminal.
• Copy and paste this command git clone https://github.com/lucafoscili/lf-nodes.git.

Workflow Runner

The Workflow Runner miniapp is included in this package but is disabled by default. To enable it you must set the environment variable WORKFLOW_RUNNER_ENABLED=true before starting ComfyUI.

By default the runner will not register its HTTP routes or static frontend when WORKFLOW_RUNNER_ENABLED is not set or set to a false value. This makes the runner opt-in and prevents accidental exposure of the runner endpoints.

Configuration is loaded from the repository-level .env file (at the project root). Set WORKFLOW_RUNNER_ENABLED=true to enable the runner.

If you do enable it, ensure authentication/allowed-users are configured (see docs/WORKFLOW_RUNNER.md) to avoid unauthorised access.

Adding New Workflows

Reusable, project-agnostic workflows can ship with LF Nodes. To add one to the Workflow Runner, follow these steps:

1. Create the workflow JSON file
   Export your ComfyUI workflow and save it as modules/workflow_runner/workflows/<workflow_name>.json

2. Create the workflow Python module
   Create a corresponding Python file modules/workflow_runner/workflows/<workflow_name>.py with:

   • Workflow configuration function: _configure(prompt, inputs) that maps user inputs to workflow node inputs
   • Input cells: Define WorkflowCell objects for each user input (uploads, text fields, checkboxes, etc.)
   • Output cells: Define WorkflowCell objects for each output (images, strings, JSON, etc.)
   • Workflow definition: Create a WorkflowNode object with metadata and export it as WORKFLOW

3. Register the workflow
   Add your workflow module name to the _WORKFLOW_MODULES tuple in modules/workflow_runner/workflows/__init__.py

4. Update frontend types (if adding new output types)
   If your workflow produces new output types:

   • Add the output interface to web/workflow-runner/src/types/api.ts
   • Update WorkflowNodeOutputs interface to include your new type
   • Update output rendering in web/workflow-runner/src/elements/components.ts and main.outputs.ts

5. Update node outputs (if needed)
   If using custom nodes, ensure they return data in the expected format:

   • Set OUTPUT_IS_LIST appropriately for batch/list outputs
   • Return structured data via ui.lf_output for frontend consumption

Example commit: See commit 2fbb49e which adds the caption_image_vision workflow, demonstrating all these steps including updating LF_DisplayString to support string outputs and frontend components to render them.

Project-owned workflows do not need to be copied into this published package. Store each workflow's .py definition and .json graph together in an external directory, then set WORKFLOW_RUNNER_EXTRA_WORKFLOW_ROOTS in the repository-level .env. The setting accepts comma- or semicolon-separated absolute directories. Set WORKFLOW_RUNNER_EXTRA_WORKFLOW_GROUPS to a matching ordered list of safe display labels when the drawer should group each registered root separately. Packaged LF Nodes workflows remain grouped by their domain; external definitions are shown only under their custom registration group. External definitions use the same workflows.custom import contract as local custom workflows, so their existing relative imports and workflow IDs can remain stable.

External roots are trusted Python source: every discovered .py module can run code during Workflow Runner startup. Point the setting only at directories you control, use unique module filenames and workflow IDs, and restart ComfyUI after changing the configured roots or their contents.

Preflighting Workflow JSON

Before queueing a large workflow, run the queue-free structural preflight from the LF Nodes repository root:

python -m modules.workflow_runner.scripts.workflow_preflight <workflow.json>

The command validates node/link integrity, verifies that every referenced LF node type still exists, and exercises the same workflow-to-prompt conversion as the runner. Add --require-all-lf-nodes when the file is intended to cover every currently shipped LF node; missing coverage then becomes a failing exit status.

Image editor

The image editor node (LF_LoadAndEditImages) allows users to load images from disk and perform editing operations such as inpainting, adjusting saturations, brightness, contrast, and more, all through an interactive interface. It's possible to select the images from the file system tree and then send them downstream to other nodes for further processing.

Notes

The LLM nodes were tested with Koboldcpp, but any Open AI-compatible endpoint that does not require authentication/an API key should work. The model used in the workflows samples is UCLA-AGI/Llama-3-Instruct-8B-SPPO-Iter3 with ChaoticNeutrals/LLaVA-Llama-3-8B-mmproj-Updated.

Contributing

Contributions to this repository are welcome, feel free to submit pull requests or open issues for discussion! To setup the environment clone this repository, then from the root open a terminal and run the command

pip install -r requirements.txt

This will install all the required dependencies for the Python backend.

To build the frontend, install Node.js with Corepack available, then run the command

corepack yarn install --immutable

This command will install all the dependencies. Note that the repository includes the compiled frontend sources, so you can skip this step if you don't plan to modify the frontend.

corepack yarn build

This command will compile all the frontend sources and generate/refresh the actual web directory.

License

MIT License

Documentation

For an overview of how this nodes suite is structured and how it integrates with the UI, see:

• Architecture
• DDS Output
• Frontend Proxy
• Image Editor
• Node Template
• Proxy
• Visual Novel Nodes
• Workflow Runner