AI Chatbot for Glasshouse
HyperCloud and VM Squared offer the ability to install an AI Chatbot into the Glasshouse web GUI for users to ask questions about the interface and how to accomplish task within the infrastructure.
By default no model is loaded on the dashboard and clicking on the "Help" icon will show the toggle grayed out.
To enable the chatbot, a model will need to be installed on the dashboard via the CLI.
Requirements
Client-side Hardware Requirements:
-
Minimum 8 GB RAM (16 GB recommended)
-
WebGPU-compatible GPU
- Intel Iris Xe or newer
- NVIDIA GTX
10xx
+ - AMD RX 5000+
- Apple M1+
Browser Requirements:
- Chromium-based browsers only:
- Google Chrome
- Microsoft Edge
- Brave
- Opera
- Browser version: Chrome version 113+ (tested with version 135+)
WebGPU support must be enabled (enabled by default in most modern Chrome-like browsers)
Upload the files
Before installation, navigate to the Storage tab of the web GUI and click on the Files submenu; this is where the model's .zip
file will be uploaded for installation.
A tester model can be found at: https://cdn.softiron.com/ai/si_model_preview_v0.1.zip
Click on the + button at the top of the page to create a new file. Here you can provide a logical name for the model's file package and select the datastore.
Install the model
With the model's files have been uploaded into the dashboard's datastore and is labeled as READY for use, login to the cluster's dashboard CLI to install the AI chatbot. It is a single command that will detect all available models and allow the user to select which one to install. Enter the desired number and input yes to proceed with installation.
[test] root@si-dashboard:~# install-ai-model
Available Images:
1. SoftIron AI Model Demo
Enter the number of an image to use (Ctrl-C to cancel): 1
Proceeding with image: SoftIron AI Model Demo
Continue? (type yes to continue): yes
Installing model...
Success.
Enable the model
After successfully installing the AI Chatbot via the dashboard's CLI, open a supported browser to the cluster's dashboard and the toggle switch for enabling AI Help will now be interactive. Toggle the switch on and wait for the model to load in the browser. The first time the model is enabled and loaded into a browser can take a while (up to 45+ min).
Once the model is loaded, the user can prompt the AI Chatbot from the Help window pane and the responses will appear within the frame.