Welcome to the L&S AI Inference platform, an on-premises large language model service for UCSB staff and faculty (incidental student use is supported), hosted on UCSB infrastructure. This guide shows you how to get access and start using the models.
Before you begin, please read Model Safety and Responsible Use.
No API key or additional configuration is required for the chat interface.
To call the models from code, a notebook, or a coding agent, you need an API key. The key is not required for the browser chat.
Request a key under Other Requests at https://cloud.college.ucsb.edu/contact. Keep the key secret — never commit it to source control.
See API Access for details.
All programmatic options use the same OpenAI-compatible endpoint: https://api.ai.college.ucsb.edu/v1.
| Workflow | How to use | Details |
|---|---|---|
| Browser chat only | Use the web interface. | No extra setup. |
| Python scripts / notebooks | Use the OpenAI Python client. | Python Quickstart |
| Other languages / curl | Any HTTP client. | cURL & REST Quickstart |
| Coding agents | Aider, opencode, Continue, or Pi. | See Coding agents below. |
import os
from openai import OpenAI
client = OpenAI(api_key=os.environ["LSIT_API_KEY"], base_url="https://api.ai.college.ucsb.edu/v1")
response = client.chat.completions.create(
model="gemma-4-26b-a4b-it",
messages=[{"role": "user", "content": "Hello!"}],
)
print(response.choices[0].message.content)
When in doubt, leave it out.
| Goal | Resource |
|---|---|
| Learn the chat interface | Open WebUI Guide |
| Choose the right model | Model Selection |
| Manage long conversations | Context Windows Guide |
| Use the API with Python | Python Quickstart |
| Use curl or other languages | cURL & REST Quickstart |
| Set up aider | Aider Guide |
| Set up opencode | opencode Guide |
| Set up Continue in VS Code | Continue Guide |
| Set up Pi | Pi Guide |
| Check known issues | Known Limitations |