Dartmouth Research Computing & Data (RCD) logo
  • Home 
  • HPC 
  • AI 
  • Services 
  • Status 
  • Blog 
  • Events 
  • About Us 

  •   Search this site
  •  
  1.   Using Dartmouth Chat
  1. Home
  2. GenAI@Research Computing
  3. Using Dartmouth Chat
  4. How do I build my own custom model?

How do I build my own custom model?

On this page
Overview   Adding a Custom Model   Creating a Custom Model   Using your Model   Sharing your Model  

Overview  

You can create your own model in Dartmouth Chat using Workspace. Custom models are layered on top of existing models present in Dartmouth Chat. Once created, your custom model will be visible and shareable from the drop-down of available models. You can set a number of parameters, including visibility/security, system prompts, and related knowledge bases.

Adding a Custom Model  

To create a custom Model:

  • Navigate to the Workspace Models area (Choose “Workspace” from the left sidebar and select “Models”)
  • Click the “+” sign to add a new model
Adding a New Model

Creating a Custom Model  

From the resulting screen, enter information about your model. There are a number of configuration options:

Creating a New Model
  • Name. The name that will appear in the model drop-down.
  • Base Model (From). The underlying Dartmouth Chat model that will be used to construct responses. As each model has different strengths, you will likely need to experiment to find the best underlying model to meet your needs.
  • Description. A brief description that will appear as a tooltip next to the model’s name and in the header when your model is selected.
  • Image/Avatar. An image/icon providing context.
  • Tags. Custom tags that can be used to identify and group similar models.
  • Visibility. By default, the visibility of your model is set as Private. If you select a Group or Groups, you can subsequently set up Read/Write permissions by group to share and/or manage your model collaboratively.
  • System Prompt. These are guidelines and rules for the model’s responses. Your system prompt should address things like role, scope, response format, and rules. Consider using Dartmouth Chat to develop an initial draft of your system prompt.
  • Prompt Suggestions. Suggested prompts to get user conversations started.
  • Knowledge. If you have existing, relevant knowledge bases to provide context for your model, add them here. These user-curated collections will provide context for the model’s response.
  • Capabilities. Based on the capabilities of the underlying Dartmouth Chat model (e.g., Vision, Image Generation, etc.), choose what capabilities your model will expose to the user.

Using your Model  

Once you’ve saved your model, you can access it from the drop-down list of available models.

Responses will be identified by the image/avatar, name, and description provided; shaped by your system prompt; and informed by any knowledge bases you attached to the model.

Accessing your New Model Reviewing Results from your New Model

Sharing your Model  

Custom models can be shared with others in a variety of ways. It starts with setting the Visibility of your model (i.e., setting visibility at Public or Private level, and providing relevant groups with Read or Write access). From there you can choose to Copy a Link to share or Exporting to a JSON file for later import. To do this, go back to the Workspace > Models tab, choose a model, and choose from the choices under the “…” link.

Sharing your New Model

Didn’t find what you needed? Please reach out to research.computing@dartmouth.edu.

 What is Workspace, and how do I access it?
How do I create and manage knowledge bases? 
On this page:
Overview   Adding a Custom Model   Creating a Custom Model   Using your Model   Sharing your Model  
Copyright © 2025 Dartmouth Research Computing & Data | Powered by Hinode.
Dartmouth Research Computing & Data (RCD)
Code copied to clipboard