Quickstart¶

Get from zero to a working, document-grounded assistant in about five minutes.

TL;DR

Illinois Chat is the easiest way to train your own LLM and share it like a Google Doc.

Go to chat.illinois.edu and sign in. Create a new project — a project is an assistant scoped to a topic, course, or team, with its own documents, settings, and access controls. See Concepts → Projects.

2. Upload documents¶

Open the Materials page of your project and add content:

Drag and drop files — PDF, Word, PowerPoint, Excel, CSV, text, HTML, code files, and even videos (transcribed automatically).
Crawl a website — enter a starting URL and the built-in crawler ingests linked pages. See Web Crawling.
Connect Canvas — import course files, pages, and modules directly. See Canvas Integration.

Documents are automatically chunked, embedded, and indexed. See Uploading Materials for supported formats and details.

3. Configure an LLM provider¶

For the best experience, bring your own API key (OpenAI, Anthropic, Azure OpenAI, and others are supported) in your project's settings. Free NCSA-hosted open models are also available with no key required. See Concepts → LLM Providers.

Your keys are yours

Provider keys are used only to serve your project's requests. Your data is never used to train models.

4. Chat and review citations¶

Ask a question in the Chat tab. Your query is embedded, matched against your project's documents, and the most relevant passages are streamed to the LLM, which answers with citations. Expand the Sources section under an answer to see each cited passage with its filename, page number, and a link to the original document.

5. Customize the system prompt (optional)¶

Tailor your assistant's behavior on the Prompting page — for example, enable tutor mode so the assistant guides students toward answers instead of giving them away.

Every project has a permanent URL you can share. Control who gets access — private with an approved email list, or public to anyone with the link. See Sharing & Access Control.

7. Integrate via API (optional)¶

Generate an API key on your project's API page and call the chat endpoint from your own applications:

curl -X POST https://chat.illinois.edu/api/chat-api/chat \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o-mini",
    "messages": [{"role": "user", "content": "What is in these documents?"}],
    "course_name": "your-project-name",
    "api_key": "uc_YOUR_API_KEY",
    "openai_key": "sk-YOUR_PROVIDER_KEY"
  }'

See the API Reference for full details.

Next steps¶

FAQs — common questions about cost, support, and security.
Video walkthroughs — short screencasts of common workflows.
Self-hosting — run the entire platform on your own infrastructure.