Skip to content

Manage ComfyUI using ComfyUI Launcher

ComfyUI Launcher is the core management tool for ComfyUI administrators. Use it to control the ComfyUI service across the cluster, and manage models, plugins, the runtime environment, and network settings.

This guide walks you through using ComfyUI Launcher for service management, routine administration, and troubleshooting.

Learning objectives

In this guide, you will learn how to:

  • Start and stop the ComfyUI service for all users in the cluster.
  • Verify and adjust network settings so GitHub, PyPI, and Hugging Face are accessible.
  • Add models in different ways and understand where models are stored in Olares Files.
  • Manage plugins from the ComfyUI Launcher library or install them from GitHub.
  • Manage Python dependencies by installing, updating, removing packages, and analyzing missing dependencies for plugins.
  • Troubleshoot common issues by checking dependency conflicts, resetting ComfyUI, reinstalling it, and collecting diagnostic information for support.

Start and stop the service

As the administrator, you must start the ComfyUI service before you or other members can access it via the client interface.

  • Start ComfyUI service

    Go to Home and click the START button in the upper-right corner.

    Start ComfyUI service

    Notes on first run

    • Initial startup of ComfyUI Launcher typically takes 10-20 seconds for environment initialization.
    • If the system prompts that essential models are missing, you can click START ANYWAY. However, workflows may fail without these base models. Download the required models before starting the service.
  • Stop ComfyUI service

    Go to Home and click the STOP button when ComfyUI is not in use. This releases VRAM and memory resources for other applications.

    Stop ComfyUI service

Configure network

ComfyUI needs access to GitHub for plugins, PyPI for Python packages, and Hugging Face for models. Before installing any of these, check the connection status in Network Manager.

  1. Go to Network.

  2. Check the connectivity to GitHub, PyPI, and Hugging Face.

  3. If any service shows Inaccessible, fix your network or proxy settings, and then click SAVE & CHECK to test again.

    Configure network

  4. Repeat until the status of each service becomes Accessible.

Network recheck

Manage files and directories

ComfyUI Launcher provides quick access to key directories and uses a shared model folder so models can be reused across ComfyUI and SD Web UI.

Access file locations

On the Home tab, the File Type section provides direct access to the main directories used by ComfyUI:

EntryLocation in FilesDescription
Root/Files/External/olares/ai/
comfyui/ComfyUI
Root directory of the ComfyUI installation.
Plugin/Files/External/olares/ai/
comfyui/ComfyUI/custom_nodes
Stores installed plugins (custom nodes and extensions).
Model/Files/External/olares/ai/modelCentralized storage for AI models shared by ComfyUI and other AI apps.
Output/Files/External/olares/ai/
output/comfyui
Default destination for images and other assets generated by ComfyUI.
Input/Files/External/olares/ai/
comfyui/ComfyUI/input
Folder for source images used in image-to-image or in-painting workflows.

Clicking any of these entries opens the Files app directly in the corresponding directory.

Understand the model directory structure

ComfyUI in Olares uses a file structure that differs from the standard installation. This change allows models to be shared between ComfyUI and SD Web UI.

When uploading external models manually or selecting a destination in Custom Download, place files into the correct subfolder.

TIP

Make sure the model type matches the target folder so ComfyUI can detect and use the files correctly.

Standard ComfyUI directoryDirectory in FilesModel type
checkpoints/main/Base checkpoints (SD 1.5, SDXL, Flux)
loras/lora/LoRA weights
vae/vae/VAE models
embeddings/embeddings/Textual Inversion embeddings
hypernetworks/hypernetworks/Hypernetwork weights
controlnet/controlnet/ControlNet models
clip_vision/clip_vision/CLIP Vision models (used for IP-Adapter)
style_models/style_models/Style or effect models
upscale_models/upscale_models/Upscalers (ESRGAN, SwinIR)
ipadapter/ipadapter/IP-Adapter models
facerestore_models/facerestore_models/Face restoration models (GFPGAN, CodeFormer)
inpaint/inpaint/Inpainting-specific models
text_encoders/clip/, text_encoders/Text encoder models (for example, CLIP and T5 encoders)
diffusion_models/unet/, diffusion_models/Diffusion model weights, including UNet-based models

Manage models

ComfyUI supports multiple ways to add models. Choose the method that best matches the model source and your workflow.

MethodUse this whenNotes
ComfyUI LauncherThe model is a public Hugging Face
model that does not require login or
a token.
In most cases, this is the easiest way to download the model directly to Olares.
Upload local modelsThe model requires login, a token,
approval, or manual download, or it
comes from a source not supported
by ComfyUI Launcher.
Download the model file to your local device first, then upload it through Files or LarePass.
Downloader nodesA custom node provides its own
model downloader.
Follow the custom node documentation for setup, storage location, and model requirements.

Download with ComfyUI Launcher

Use this method for public Hugging Face models that do not require login or a token.

ComfyUI Launcher provides three common ways to download models directly to Olares.

Download a resource package

Use this method when you want to install a ready-to-use package provided by Olares, such as a starter bundle or a workflow-specific package.

  1. Go to Home > Resource Package.

  2. Find the package you want, then click VIEW.

  3. On the Package Details page, click GET ALL to download all required files. You can track the progress in the status bar.

    Download progress

Download from the model library

Use this method when you want to download an individual public model directly in ComfyUI Launcher.

  1. Go to Models > Model library, then scroll to Available models.

  2. Search for the model by name, or browse by category.

  3. Find the model you want, then click the download button to download it.

    Library download

Download by direct URL

Use this method when you already have a direct download URL for a model file that ComfyUI Launcher can access directly.

  1. Go to Models > Custom Download.

  2. Paste the model URL.

  3. Select the destination path based on the model type. If you're not sure which folder to choose, see Understand the model directory structure.

  4. Click DOWNLOAD MODEL.

    Custom download

Upload local models

Use this method when the model requires login, a token, approval, or manual download, or when the model source is not supported by ComfyUI Launcher.

  1. Download the model file to your local device. If needed, see Models cannot be downloaded in ComfyUI Launcher.
  2. Open Files from the Launchpad.
  3. Navigate to External/olares/ai/model.
  4. Open the folder that matches the model type. If you're not sure which folder to use, see Understand the model directory structure.
  5. Upload the model file to the target folder.

Use downloader nodes

Some custom nodes can download models automatically.

Follow the documentation provided by the node author for setup, storage location, and model requirements.

If the workflow still reports a missing model after download, check whether the file was saved to the expected location and whether it matches the model type expected by the workflow.

Delete models

To delete a model:

  1. Go to Models > Model library.

  2. Under the Installed models section, find the model you want to delete, and click the delete icon.

    Delete a model

Manage plugins

ComfyUI Launcher provides flexible ways to manage plugins in Plugins.

Plugin status

Manage available plugins

To manage available plugins registered in ComfyUI Launcher:

  1. Go to Plugins > Plugin library.
  2. Under Available Plugins, find the plugin you want.

For any plugin, you can:

  • Click the visibility button to view plugin details.
  • Visit the GitHub repository if it is available.

Depending on the plugin status, different actions are available:

  • Not installed: Click Install to get the latest version, or Switch version to choose a specific one.
  • Installed: Disable, Uninstall, or Switch version.
  • Disabled: Click Enable to reactivate it, or Uninstall to remove it.

At the top of the section, you can also:

  • Click UPDATE ALL PLUGINS to update all installed plugins.
  • Click REFRESH to refresh the plugin list.

Install plugins from GitHub

To install plugins directly from GitHub repositories:

  1. Go to Plugins > Custom Install.

  2. Enter the GitHub repository URL of the plugin.

  3. (Optional) Specify the branch. If unsure, keep the default.

  4. Click INSTALL PLUGIN.

    Download plugin

Manage environment

ComfyUI runs on a set of Python libraries. Manage them on the Environment page.

Install Python packages

  1. Go to Environment > INSTALL NEW PACKAGE.

  2. In the pop-up window, enter the package name and version number (optional), and then click INSTALL.

    Install new package

Manage installed Python packages

  1. Go to Environment.

  2. Under the Installed Python packages tab, find the Python library you want to manage.

  3. Click the arrow_upward button on the right to update the library, or the delete button to remove it.

    Manage installed packages

Analyze dependency installation status

  1. Go to Environment > Dependency analysis.

  2. Click ANALYZE NOW to start analyzing.

  3. From the plugins list on the left, find the problematic plugin highlighted in red, and click on it.

  4. From Dependency list, find the missing library for the plugin, and click the Install button on the right. You can also click FIX ALL to automatically install all missing libraries.

    Analyze dependencies

Troubleshooting flow

Use the following flow when you encounter an issue and need a general recovery path.

TIP

For symptom-specific solutions such as ComfyUI cannot start, models cannot be downloaded, or high CPU temperature, see Common issues first.

Check dependency conflicts

If problems start after installing new plugins, the issue may be caused by dependency conflicts.

Run a dependency analysis to identify and fix the problem. For detailed steps, see Analyze dependency installation status.

Reset ComfyUI configuration

If the issue is still not resolved after the checks above, reset ComfyUI to its initial state.

Perform with caution

Resetting ComfyUI is irreversible. All plugins, custom configurations, and Python dependencies will be removed. Models stored in the shared model folder are not affected.

Get diagnostic details

If you plan to contact support, export your ComfyUI logs before resetting. See Collect information for support.

To reset ComfyUI:

  1. In ComfyUI Launcher, go to Home and click more_vert in the upper-right corner, then click Wipe and restore.

  2. In the prompt window, click WIPE AND RESTORE.

    Wipe and restore

  3. Enter CONFIRM, then click CONFIRM.

    Second confirmation

After the reset completes, restart ComfyUI for the changes to take effect.

Reinstall ComfyUI completely

If the issue persists after wipe and restore, uninstall and reinstall ComfyUI completely.

  1. Go to Market > My Olares.
  2. Click the dropdown arrow next to ComfyUI's operation button and select Uninstall.
  3. In the Uninstall window, select Also remove all local data, then click Confirm.
  4. Open Files from the Launchpad and go to External/olares/ai.
  5. Delete the comfyui folder.
  6. Reinstall ComfyUI from Market.
  7. Once installation is complete, open ComfyUI Launcher and start the service.

Collect information for support

If you cannot resolve the issue and need to escalate it to the support team, prepare the following diagnostic information.

Export ComfyUI logs

Logs contain the backend running status and error traces.

  1. In ComfyUI Launcher, go to Home and click more_vert in the upper-right corner, then click View logs.

    View Logs

  2. Click the refresh button to ensure you have the latest output.

  3. Click the download button to save the log file.

    Export Logs

Optional: Get the workflow error report

If a specific workflow failed, include a screenshot of the workflow error report to help the support team identify the problem.

  1. In the ComfyUI client, click Active to open the Job Queue.

  2. Select the failed task from the list.

  3. Click Report error, then click Show Report to expand the details.

    Workflow error report