The WSE AI Gateway lets you use one JHU-managed gateway domain instead of each provider’s public API base URL. Before sending a request, you must create a project, allow the providers and models you need, and create a key that is permitted to use those models.
Beta Status and Limitations
The WSE AI Gateway has been tested for chat and document-processing workflows, but not every provider feature, third-party application, agent workflow, or sensitive-data workflow has been tested.
Important: Do not use the WSE AI Gateway for HIPAA-protected data or protected health information at this time.
Contact CMTS before planning a course, research project, or production workflow around any unusual or high-volume use case. This includes provider-specific multi-modal features like image or video generation, and specific third-party applications or integrations.
Before You Begin
Before you create your first request, make sure you have:
- A valid Johns Hopkins Enterprise Directory system account, often called a JHED account.
- A valid cost center for your AI Gateway project.
- A monthly budget cap for the project.
- The provider and model names you plan to use.
Creating a Project in the WSE AI Gateway
- Go to gateway.engineering.jhu.edu and sign in with your JHED credentials.
-
Click + New Project to create a project with a valid cost center and monthly budget cap.
Project Managers: To request that a cost center be attached to a project, person, or department, email CMTShelp@jh.edu.
- Add project details to the Create New Project form, including project name, cost center, and budget cap. Then select Continue.
- Select the providers you plan to use within your project. Select Continue.
-
Select the models you would like to use for each of the providers you selected in the previous step. Select Continue.
- Review your project details, providers and models. Then select Create Project.
Creating an API Key
- Go to the project page and select Add Key.
-
A Create API Key window will open. Name the key and enter your monthly spending limit. Select Continue.
- Select the model for the key being created. Limit the key to the models you actually need. Then select Create Key.
- Save the plaintext key immediately, as it will only be shown once. Check the box next to "I have copied and saved this key." Select Done.
Troubleshooting Model Access
If you receive a MODEL_NOT_ALLOWED_FOR_KEY error, the model is not allowed for either the project or the key.
To resolve the issue:
1. Open your AI Gateway project.
2. Confirm that the model is included in the project allow list.
3. Go to the Keys tab.
4. Confirm that the key allow list includes the same model.
5. Update the request to use the exact provider-prefixed model name.
6. Send the request again.
Need Help?
If you need assistance, contact CMTShelp@jhu.edu. Include the provider, model, endpoint path, and exact error message you are seeing.