URI ITS LLM Gateway

A self-hosted, OpenAI-compatible API gateway to large language models on AWS Bedrock — with per-user keys, budget controls, and full cost visibility for the University of Rhode Island community.

Enter Portal

What is this?

The LLM Gateway is URI ITS’s centralized entry point for AI model access. It proxies requests to AWS Bedrock through LiteLLM, exposing a standard OpenAI Chat Completions API so existing tools and SDKs work unchanged.

  • Per-user API keys tied to your URI identity, with tiered model access.
  • Budget management — personal credit balances plus delegated budget pools and groups for teams.
  • Cost & usage tracking — detailed per-model spend and token reporting.

Example uses

Any OpenAI-compatible tool works — point it at https://llmgw.its.uri.edu/v1 and use your personal key.

# Chat completion from the command line
curl https://llmgw.its.uri.edu/v1/chat/completions \
  -H "Authorization: Bearer $LLMGW_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "its_direct/pt3-claude-sonnet-4.6-1m-us",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'
# pip install openai
from openai import OpenAI

client = OpenAI(
    base_url="https://llmgw.its.uri.edu/v1",
    api_key="sk-...",   # your gateway key
)
resp = client.chat.completions.create(
    model="its_direct/pt3-claude-sonnet-4.6-1m-us",
    messages=[{"role": "user", "content": "Summarize this..."}],
)
print(resp.choices[0].message.content)
// ~/.config/opencode/opencode.json
{
  "provider": {
    "llmgw": {
      "npm": "@ai-sdk/openai-compatible",
      "options": {
        "baseURL": "https://llmgw.its.uri.edu/v1",
        "apiKey": "sk-..."
      },
      "models": {
        "its_direct/pt3-claude-sonnet-4.6-1m-us": {}
      }
    }
  }
}

Then select llmgw/its_direct/pt3-claude-sonnet-4.6-1m-us as your model in OpenCode.

Compatible clients

The gateway speaks the OpenAI Chat Completions API. Any tool that lets you set a custom OpenAI base URL + API key works directly — point it at https://llmgw.its.uri.edu/v1.

Works directly (set base URL + key):

Coding / IDEsOpenCode, Cline, Continue, Cursor, Aider, Zed
Chat UIsLibreChat, Open WebUI, AnythingLLM, Chatbox, LobeChat
CLIllm, aichat, chatblade, curl
SDKsOpenAI SDK, LangChain, LlamaIndex, Vercel AI SDK
Automationn8n, Flowise

Via a router/proxy — for vendor-locked tools:

Tools tied to a specific vendor’s API (e.g. Google Antigravity — Gemini-only, Claude Code, Gemini CLI, GitHub Copilot) can connect through an OpenAI-compatible router such as 9router / n9router or CLIProxyAPI. Set the router’s upstream to https://llmgw.its.uri.edu/v1 with your gateway key; it translates the tool’s native calls to the gateway.

In every client, the model name must match what the gateway serves (e.g. its_direct/pt3-claude-sonnet-4.6-1m-us) — not generic IDs like gpt-4o. Full list in the portal.

Top URI units

URI departments accounting for the most usage over the last 30 days (by request volume).

Loading…

Usage — last 30 days

Requests / day

Tokens / day

Top models in use

Most-used models over the last 30 days (by request volume). Full list in the portal.

Loading…