The knowledge base is the primary source of truth your agent uses to answer questions, explain policies, and guide students accurately. Strong knowledge foundations are essential for consistent agent behavior and a high-quality student experience.
This article explains what to include in your knowledge base and how to approach population thoughtfully.
⚠️ Important: every asset must have a tag applied. Every single asset you upload to the knowledge base must have at least one tag applied to it. Tags are required for your agent to correctly scope and retrieve content. If you have questions about how tags work, refer to the Knowledge base tags article. If you need further assistance with tagging, reach out to the CollegeVine Success Team at [email protected].
How the knowledge base works in CollegeVine
Agents rely on the knowledge base to ground their responses in institution-approved information. When an agent needs to explain a policy, deadline, or process, it retrieves relevant content from the knowledge you've provided.
The agent's effectiveness depends directly on the clarity, coverage, and quality of your knowledge assets.
Going about populating your agent's knowledge base
Your Knowledge section is the foundation of your AI agent's contextual knowledge. In short, it's what enables your agents(s) to make informed decisions.
It provides a central place to upload assets, schedule staff interviews, share emails, and store key information about your institution. This training data helps your agent generate accurate answers, reflect your brand voice, and engage constituents in a way that aligns with your team's expertise.
Training assets can include:
Websites
Files
Staff interviews
Previously sent email communications
File upload process tips:
You can upload multiple files at once.
You can see all uploaded files, remove them, and add more before saving.
You will see a spinner while your files are uploading. Once the upload is complete, a confirmation message will appear showing the number of files successfully uploaded.
Asset essentials
Once added, your assets will have a status of "Uploaded." Once processed, your assets will be flagged as "Ready."
Only "Ready" assets can be tested.
Only when you view and publish your changes will that asset be committed to production. The extra publish step allows you to test knowledge assets in a risk-free way!
Organizing your knowledge base with tags
Tags give you a way to organize your knowledge base assets and control which information your agent draws from in different contexts.
What tags do
Tags act as labels you can apply to assets in your knowledge base. Once tagged, assets can be used to scope a specific Workflow — meaning you can configure an agent to only pull from content that carries a particular tag. For example, a financial aid Workflow could be bound to a "Financial Aid" tag, ensuring it only references relevant content rather than your entire knowledge base.
Managing tags
To create or edit a tag, go to the Knowledge Base section of your platform. You can:
Create new tags with a name, description, and color
Apply tags to assets directly from the asset table
Use bulk actions to select multiple assets at once, making it easier to tag, delete, or refresh a batch of content
Website assets
Websites allow your agent to learn directly from public-facing content such as admissions or student affairs pages, academic program descriptions, tuition information, and policies.
Why you will upload unique URLs
Why you will upload unique URLs
Each website will be added with its own unique URL. For example, one for application requirements, one for campus visits, one for academic programs, and so forth. Adding unique URLs ensures your agent is trained on the most up-to-date information available, and not mis-trained on orphan pages or legacy content that might surface if the agent crawled the entire website. Adding these unique URLs puts you in control of the right types of data you'd like the agent to ingest.
How the agent reads website content
How the agent reads website content
The agent reads from the HTML source, not the visible page layout. This means the agent will read:
All text present in the page's raw HTML at the time of scraping
Content hidden by CSS, such as collapsed accordion sections, if the text is already in the HTML
The agent may not read:
Content loaded only after a user interacts with the page
Text inserted dynamically via JavaScript that is not part of the initial HTML
If the information you need isn't in the HTML, consider adding it as a document or including it in an interview.
Website upload tips
URLs must start with http:// or https://
Duplicate URLs are automatically detected
Invalid inputs show a red outline + error message, and Save is disabled until fixed
How to add a website
How to add a website
When adding URLs, always include the full address starting with https://.
File assets
Files allow you to upload static documents and structured data that the agent can learn from. These are helpful for policy documents, program sheets, course catalogs, fee charts, and any written guidance your institution relies on.
Why you can upload internal documents
Why you can upload internal documents
Because your AI agent is just learning from the content of the files, you can upload both formal/marketing-approved assets or internal documents (i.e.- internal process guide, internal training guide for new employees) that aren't intended to be outward-facing.
Accepted file types
Accepted file types
.pdf
.csv
.txt
.md
.html




