AI Gateway

Important

The AI Gateway is an experimental feature for Coder Workspaces. An API key is required to use the gateway. Access is currently restricted to library staff.

Coder Workspaces can use an AI Gateway (LiteLLM) for API access to large language models. The gateway is only accessible from your Coder Workspace.

The gateway URL is http://litellm.dreamlab.ucsb.edu:4000

AI-assisted Coding Tools

Most AI-assisted coding tools can be configured to use the AI gateway.

OpenCode

OpenCode is open-source coding agent that works with a wide range of model providers. These instructions are from LiteLLM’s OpenCodeQuickStart.

1. Install OpenCode

# download and install opencode
curl -fsSL https://opencode.ai/install | bash

# apply updated $PATH
source ~/.bashrc

2. Configure OpenCode

Configure OpenCode:

# create opencode config folder
mkdir -p ~/.config/opencode

# edit the config file (opencode.json)
nano ~/.config/opencode/opencode.json

Paste the following into nano as the contents of ~/.config/opencode/opencode.json.

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "litellm": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "DREAM Lab LiteLLM",
      "options": {
        "baseURL": "http://litellm.dreamlab.ucsb.edu:4000/v1"
      },
      "models": {
        "gemini-3-flash-preview": {
          "name": "Gemini 3 Flash"
        }
      }
    }
  }
}

Save the file (Ctrl + O, then Ctrl + X in nano); then start OpenCode:

opencode

In OpenCode, enter /connect to start the Connect a provider configuration:

Enter DREAM Lab in the search box, to filter the list to just the provider we set in the configuration file. Press Enter to select the provider.
In the API key input, enter your personal API key and hit Enter.
Hit enter again to select the model, Gemini 3 Flash
You should now be back in OpenCode’s main screen.Confirm that the connection is working with a simple question (“Are we connected?”)

Use /exit to quit OpenCode. The configured provider and model should persist with your next session.

Claude Code

Claude Code is Anthropic’s coding agent. It works primarily with Anthropic models (Opus, Sonnet, and Haiku), which some users may have access to through the AI gateway. These instructions are based on Claude Code’s documentation for connecting to a LiteLLM gateway.

1. Install Claude Code

Start the terminal and install Claude Code if necessary:

curl -fsSL https://claude.ai/install.sh | bash

2. Configure Claude Code

You need to set three environment variables so Claude Code can connect to authenticate with the AI gateway. You can add these to the bottom of your ~/.bashrc file so they take effect automatically when you start the terminal.

Open ~/.bashrc in a text editor:

nano ~/.bashrc

Scroll to the bottom of the file and add the following lines. Replace "YOUR_API_KEY..." with the key you received from DREAM Lab staff.

export ANTHROPIC_AUTH_TOKEN="YOUR_API_KEY..."
export ANTHROPIC_BASE_URL=http://litellm.dreamlab.ucsb.edu:4000
export CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS=1

Save the file (Ctrl + O, then Ctrl + X in nano) and reload it:

source ~/.bashrc

By adding these variables to ~/.bashrc, you won’t need to enter the API key or gateway address every time you log in to your Coder workspace.

3. Start Claude Code

You should start Claude Code from inside a project directory:

claude

The workspace will ask if it’s a safe environment. Select “Yes, I trust this folder” and Claude will start.

Once Claude starts, try asking it a question. You do not need to run /init like it says at the top.

Tip

If you see a Missing API Token error in red to the right of the terminal, you are not connected to Anthropic and will need to either set up the environment variables in ~/.bashrc (as described above) or exit Claude and manually export the variables in your terminal session.

LiteLLM API with curl

You can use curl to call the LiteLLM API directly. The LiteLLM API is documented here: https://docs.litellm.ai/docs/supported_endpoints

# get spend ($) for your API key
export LITELLM_API_KEY="YOUR API KEY..."
curl -s -H "Authorization: Bearer ${LITELLM_API_KEY}" "http://litellm.dreamlab.ucsb.edu:4000/key/info?key=${LITELLM_API_KEY}" | jq .info.spend

# list available models
curl -s -H "Authorization: Bearer ${LITELLM_API_KEY}" http://litellm.dreamlab.ucsb.edu:4000/v1/models | jq .data[].id

LiteLLM API with Python

This code is taken from the LiteLLM documentation.

First, create a .env file in your project directory with your API key:

LITELLM_PROXY_API_KEY="YOUR_API_KEY..."

Then use the following Python code to connect to the gateway:

import litellm
from litellm import completion
from dotenv import load_dotenv

# load environment variables from .env file
load_dotenv()

# set custom api base to your proxy
litellm.api_base = "http://litellm.dreamlab.ucsb.edu:4000"

messages = [{"content": "Hello, how are you?", "role": "user"}]

# litellm proxy call
response = completion(
    model="litellm_proxy/gemini-3-flash-preview",
    messages=messages,
)