Installation

Installation

1. Local Preparation: Download Nvidia Driver

Download the Nvidia Driver for your GPU from the Nvidia official website https://www.nvidia.com/en-us/drivers/.

If deploying on a Data Center, please choose a CUDA version >12.0.

2. Download & Extract the One-Click Installation Package

You can find the download link for the ShellAgent installation package from the Github repository release. You can also download it directly from Hugging Face.

The installation package includes the ShellAgent framework engine, commonly used widgets and models.

After downloading, extract it locally to complete the installation. No need to handle environment issues.

3. Run ShellAgent

Click run.bat in the main directory, wait for the initialization to complete in the command prompt, and then access http://localhost:8099 in your browser.

4. (Optional) Import Local Models Files

For ComfyUI users, ShellAgent supports importing models from local ComfyUI projects.

Set the storage path (absolute path) of the models in settings, and we will merge the models of ShellAgent and the specified directory through symbolic links. After merging, the two directories are similar like shortcuts.

If you are using ShellAgent, after modification, you need to right-click in the canvas node and select Reload to update the selectable models.

Common Issues During Installation

1. After clicking run.bat, an error occurs: urllib.error.URLError: <urlopen error [winError 10060] A connection attempt failed because the connected party did not properly respond after a period of time, or established connection failed because connected host has failed to respond.>

   This issue is due to not finding some pre-installed models that require accessing external networks. Please check whether you have moved or deleted the models included in the compressed package.

2. After clicking run.bat, an error occurs: CUDA device assert error

   Check whether the NVIDIA driver is properly installed locally. After right-clicking on the desktop, can you see the following Nvidia options:

   
3. Failed to access http://localhost:8099

   Check whether the firewall has incorrectly blocked all local port access.

4. InstantID face analysis error:

   This error is caused by issues when Windows decompresses the node and automatically downloads the model weights:

   

Solution:

1. If you have an existing models folder from ComfyUI locally, refer to the above method to set the models in settings.

   
2. Locate the models folder under the ShellAgent (or OpenProConfig under version 0831) folder, and move the model files in insightface/models/antelopev2/antelopev2 up one level.