AI Gateway
The AI Gateway is a centralized, CMU-supported service, built using LiteLLM. This tool provides developers with streamlined access to multiple AI models using a consistent, OpenAI-compatible API. It enables experimentation and production use of large language models (LLMs) through a unified interface that simplifies model selection, team management, and budget tracking.
Benefits
- Unified Access and Billing: Single API key for multiple top-tier AI models, streamlined account management, and one consolidated pay-as-you-go invoice based only on token usage.
- Financial Control: Transparently set and manage team/individual budgets, consumption limits, and rates using detailed usage dashboards and reporting.
- Data Security: Models are fully vetted and covered by CMU contract terms with Public Cloud providers to guarantee that your data is not used for training public models.
- Broad Compatibility: Supports all academic, research, and development use cases with compatibility for both on-premise and cloud-hosted models.
Eligibility
- Faculty and staff (may request and manage teams).
Best For
- Instructors who need students to access models using API keys.
- Staff writing application code that needs to call APIs programmatically.
- Researchers who want to integrate AI into their Data Analysis.
Fees
- Pay-as-you-go model, with costs depending on the model and volume used. You only pay for the input and output tokens that you generate.
- Teams must provide an Oracle string for cost recovery.
Request a Team (Team Owners)
AI Gateway access is team-based. Create multiple teams to control access, track spending, and process invoices.
To request a LiteLLM team:
- Faculty and staff should complete the AI Gateway Request and provide:
- An Oracle string
- Team owners
- Initial team members
Note: To change the membership, contact the Computing Services Help Center.
- You'll receive:
- An API key for your Andrew userID
- The base URL to use in your requests: https://ai-gateway.andrew.cmu.edu
- A list of supported models