Sign in with your CUNY account. If you don’t have access yet, let us know and we’ll get you set up.
2. Open Workspace
Look for Workspace in the top left, three rows below New Chat. Click it. This is where you’ll build your model card and knowledge collection.
Next: Once you’re signed in and can see the Workspace menu, you’re ready to set up your model card on the next slide.
Workspace → Models Navigate to Workspace in the sidebar, then open the Models tab.
Models List Click + New Model to create one, or open an existing model card.
Before We Start
Set Up Your Model Card
Knowledge collections attach to a model card in Workspace → Models. After you build your collection, complete these steps to bring everything together.
1. Name the model card
Give it a descriptive name tied to your course, for example ENGL 101 Writing Scaffold or History 202 Source Analysis Tool.
2. Select a base model
Choose a base model from the Sandbox (e.g., GLM-5, GPT-OSS-120b, or Kimi-k2.5). This is the underlying model your custom configuration will run on.
3. Add your system prompt
Find the System Prompt field under Model Params and paste the prompt you wrote last week.
4. Save your work
Click the Save button. Your model card is now ready for a knowledge collection.
Inside the Model Card Name your model, select a base model, paste your system prompt, and save. The Knowledge section below is where today’s collection will attach.
Keep this tab open. You will return to this model card at the end of today’s exercise to attach the knowledge collection you build.
Dry Run
Test Your Model in Chat
Your model is saved. Now test it to make sure it responds as expected before attaching a knowledge collection.
1. Click New Chat
Go back to the chat interface. Click New Chat in the top left to start a fresh conversation.
2. Toggle the model selector
At the top of the chat window, you’ll see the model name. Click the dropdown arrow next to it.
3. Select your model
Find your newly created model (e.g., ENGL 101 Writing Scaffold) in the dropdown list and select it.
4. Send a test message
Type a simple message related to your course. Send it and observe the response.
Test Your Model Click New Chat (1), select your model from the dropdown (2-3), and type a test message (4).
Look for: Does your custom model respond as expected? How has your system prompt shaped its output?
Part I
Knowledge Collections
What they are, where they live, and how they work
The Basics
What Is a Knowledge Collection?
A knowledge collection is a set of documents you upload to ground an AI model in your course materials. The model retrieves relevant passages from these documents when responding.
Think of it as a reference shelf for your custom AI model: the sources it draws on, the assignments it references, and the disciplinary context that shapes its responses.
Next we’ll show you where to find knowledge collections in the Sandbox and how to curate and use them with pedagogical intent.
What you build
Hidden
System Prompt
“You are a writing scaffold…”
+
New
Knowledge Collection
syllabus.pdf, prompt.pdf…
student never sees above
👤 Student“What does the assignment say about evidence?”
🤖
AI Model
Retrieves passages from uploaded course materials
🤖 Response“The assignment asks for at least three sources cited with specific page numbers. Let’s look at what you have so far.”
What the student sees
Workspace
Where to Find Knowledge Collections
Knowledge collections and model cards both live in the Workspace menu. Here’s how they connect:
Workspace → Knowledge
Where collections are created, named, and populated with course documents.
Workspace → Models
Where collections are attached to a model card via the Knowledge field, giving the model access to your materials.
Key distinction: Your system prompt defines how the model responds. The knowledge collection provides the source material it draws on. Both are configured in the same model card.
Workspace → Knowledge Click + New Knowledge to create a collection.
Name and Describe Give your collection a name, describe its contents, and set access. Click Create Knowledge when ready.
Workspace → Models Open your model card to attach the collection.
Knowledge Field Click Select Knowledge to attach your collection to the model card.
Why Bother?
Why Knowledge Collections Matter
If you create a custom model that acts as a course assistant, a student might ask: “What should I focus on for the midterm essay?”
Without Knowledge Collection
“For a midterm essay, you should generally focus on your thesis statement, use evidence from your readings, and structure your argument clearly. Make sure to address counterarguments.”
Generic advice. No connection to the assignment or the course readings.
With Knowledge Collection
“Based on the assignment prompt, your essay should analyze one primary source from the Reconstruction unit using the SOAPS framework we practiced. The prompt emphasizes evidence and sourcing. Which document are you considering?”
Grounded in the actual assignment and course methodology.
Building Materials
What Can You Upload?
Uploaded files ground AI models in context and help shape their responses. These documents are stored on CUNY’s self-hosted servers and made private by default.
Syllabi
Course schedule, learning objectives, policies, and expectations
Assignment Prompts
Instructions, requirements, and criteria for each assignment
Rubrics
Evaluation criteria so the model can reference specific expectations
Course Readings
Primary sources, articles, chapters, and excerpts students are working with
Lecture Notes
Key concepts, frameworks, and terminology from your lectures
Spreadsheets, CSV files, or structured data students analyze in labs or projects
Glossaries
Discipline-specific terminology, definitions, and key concepts for the course
Problem Sets
Exercises, practice questions, or worked examples with solutions
Lab Protocols
Step-by-step procedures, safety guidelines, and equipment instructions
Case Studies
Real-world scenarios, historical cases, or clinical examples used in coursework
Under the HoodRetrieval-Augmented Generation
How Your Documents Reach the Model
When a student asks a question, the system doesn’t feed the entire collection to the model. It searches for the most relevant passages and uses them as the basis for its response.
Chunking: Your files are split into smaller passages when uploaded
Matching: Questions are matched against those chunks
Injection: Closest matches appended to model’s context window
Response: Model generates output grounded in retrieved passages
Implication: Short, focused documents with clear headings retrieve better than long, unstructured files. In other words, the way you organize your materials matters.
👤
Student Question
“What does the assignment say about citations?”
🔍
Search Collection
Finds relevant passages from your files
🤖
AI Model + Retrieved Context
System prompt + relevant passages + student message
✅
Grounded Response
Answer references your actual course materials
Part II: Example 1
What Makes an Effective Knowledge Collection?
Starting with Composition & Writing
Composition & Writing
The Bare Minimum
WeakCollection contents:
• syllabus.pdf (14 pages, full course syllabus)
What goes wrong?
One large document retrieves poorly: retrieved passages are often irrelevant
No assignment context for the revision task
No readings or reference materials for the model to draw on
Separate documents let the model find what it needs
Assignment prompt gives the model context for the revision task
Style guide helps with formatting questions
What's still missing?
No course readings for the model to reference during analysis
No common feedback patterns to guide revision
No instructor notes on what substantive revision looks like in this course
Composition & Writing
A Collection That Grounds Revision
StrongCollection contents:
Course Context
• syllabus.pdf: schedule, learning objectives, policies
• revision-philosophy.txt: instructor notes on what revision means in this course
Assignment Materials (Essay 1: Rhetoric in Popular Media)
• essay-1-prompt.pdf: assignment instructions and requirements
• common-feedback.txt: patterns from past semesters (e.g., thesis too broad, evidence not analyzed)
Reference Materials
• mla-style-guide.pdf: citation and formatting conventions
• strong-intro-examples.txt: examples of effective introductions
• revision-checklist.pdf: the same checklist students use in peer review
Individual literary text rather than an omnibus reader
Assignment prompt provides task-specific context
Critical framework document gives the model methodological grounding
What's still missing?
No annotated examples showing how to move from observation to interpretation
No key terms for the current unit (e.g., tension, irony, ambiguity)
No instructor notes on what close reading looks like in this course
Literature & Cultural Studies
A Collection That Fosters Close Reading
StrongCollection contents:
Course Context
• syllabus.pdf: schedule, texts, learning objectives
• new-criticism-framework.txt: key concepts and terms for this unit (tension, irony, paradox, ambiguity, diction, imagery)
Assignment Materials (Close Reading Essay)
• close-reading-assignment.pdf: instructions and requirements
• annotated-passage-example.txt: model annotation showing how to move from observation to interpretation
Literary Texts (Current Unit)
• sonny-blues-baldwin.pdf: the primary text for this assignment
• passage-selections.txt: key passages the instructor has flagged for class discussion
Part III
Curating for Retrieval Quality
Tips for curating collections, and common pitfalls to avoid
Curation Best Practices
One Document, One Purpose
Upload separate files; focused documents retrieve better than omnibus ones
Add Metadata and Headings
Titles, authors, dates, and section headings serve as retrieval anchors
Supply What’s Not in the Documents
Include meta documents like “common-feedback.txt” that signpost how to use sources in the collection
Update Per Unit
Swap course materials as the semester progresses; up-to-date collections outperform semester-wide ones
Watch Out
Common Pitfalls
Dumping Everything In
Uploading every reading dilutes retrieval; start small and add materials as you test
One Giant PDF
A 200-page course reader retrieves unpredictably; short, well-labeled documents work far better
Forgetting the System Prompt
Without explicit instructions for drawing on the collection, it is just a pile of documents
Assuming Full Coverage
Only retrieved passages appear in each response; if something is critical, give it its own file
Part IV
Building Your Knowledge Collection
Three types of references to consider, then steps for how to create, curate, and use your first collection.
Types of Reference Material
Think about which type of course document you would add first
These documents give the model a picture of your course: its goals, structure, and the methods students are expected to use.
What are the course’s learning objectives?
What analytical framework or methodology is central to the course?
What course-level context would help the model support those goals?
Recommended uploads:
1. syllabus.pdf
- Course schedule, objectives, and policies
2. [framework-name].txt
- The analytical method students use
- Write it out in plain language with definitions
Consider: Is there a framework or methodology central to your course? If so, a short document (1-2 pages) explaining it in the terms you use with students could be a strong addition.
Type 2
Assignment Materials
These documents define the current task and help the model align its responses with your specific learning objectives.
What does the assignment ask students to do?
What does strong work on this assignment look like?
What patterns come up most often in your feedback?
Recommended uploads:
1. [assignment]-prompt.pdf
- The assignment instructions
2. common-feedback.txt
- 5-10 patterns you see every semester
3. strong-examples.txt (optional)
- Excerpts showing what strong work looks like
Consider: Which assignment stands to benefit? Try curating assignment instructions alongside a shortlist of common feedback patterns for starters.
Type 3
Source Materials
Upload the readings and reference materials students are working with in the current unit. This grounds the model in the actual texts.
What texts are students reading for this assignment?
Are there reference documents (timelines, glossaries, citation guides)?
Can you add brief metadata or context for each source?
Recommended uploads:
1. [reading-title].pdf
- Individual files per text (not one big reader)
- Add a header with: title, author, date, source
2. context-notes.txt (optional)
- 2-3 sentences of context per source
3. [reference-guide].pdf
- Citation style guide, glossary, or timeline
Consider: Which readings or sources are students working with right now? Individual files retrieve better than a single combined PDF.
Putting It Together
Create Your Collection
Now create the collection you’ve been thinking about.
1. Go to Workspace → Knowledge
Click + New Knowledge to start a new collection.
2. Name and describe it
Give it a clear name tied to your course and a short description of what it contains.
3. Upload your first file
Start with the document type that caught your eye - be it course context, assignment materials, or source materials.
Workspace → Knowledge Click + New Knowledge to create a collection.
Name and Describe Give your collection a name and set access to private.
Final Step
Attach Your Collection
Connect your collection to the model card you set up earlier.
1. Return to your model card
Go to Workspace → Models and open the model card you created at the start of today’s session.
2. Attach the collection
Find the Knowledge field and click Select Knowledge. Choose the collection you just created.
3. Save and test
Save the model card, open a new chat with your model, and ask a question only answerable from your collection.
Workspace → Models Open the model card you created earlier.
Attach Your Collection Click Select Knowledge to connect your collection to the model.
Coming Up
The Road Ahead
March 16
System Prompts ✓
Configured how the model responds and scaffolds learning
March 23 (Today)
Knowledge Collections ✓
Grounded the model in your course materials so it can reference real documents
March 30 (Next Week)
Skills & Tools
Build specialized skills, tools, and workflows tailored to your courses
Each workshop builds on the last. The system prompt you wrote last week now drives a model grounded in the knowledge collection you built today. Next week, you’ll extend it with custom skills and tools.
