College project
Build an AI Chatbot for Your College Project
A working chatbot trained on real documents, embedded in a page you built, with an integration small enough to explain in a viva. Free tier is enough to demo it.
How it works
The fourth one is what decides how the demo goes, and it is the one people skip.
Pick a real document set
Your college handbook, a department FAQ, or the rules for a club. A project that answers real questions demos far better than one trained on filler.
Create and configure the chatbot
Upload the sources, set the name and welcome message, and choose the response style. This is the part you will be asked to explain.
Embed it in a page you built
One snippet in your own HTML, React, or Next.js project. The integration is small enough to walk through line by line in a viva.
Prepare the demo
Have three questions ready that it answers well and one it correctly refuses. Showing the limit is what makes an evaluator believe the rest.
The integration
This is the code you will be asked about. It is short enough to know line by line.
<link rel="stylesheet" href="https://cdn.cortexvia.in/chatbotsdk-web.css" />
<script src="https://cdn.cortexvia.in/chatbotsdk-web.iife.js"></script>
<div id="cv-chat"></div>
<script>
window.ChatbotSDK.mount("#cv-chat", {
chatbotKey: "ctx-cb-xxxxxxxxxxxxxxxx",
config: {
mode: "chatbot",
position: "bottom-right"
}
});
</script>Two CDN files, a container element, and a mount call carrying your chatbot key. Replace the placeholder key with your own.
- mount takes a selector and an options object, and returns a handle
- The container must exist in the page before the mount call runs
- React and Next.js need the mount inside an effect — see those guides
What to explain
Four things an evaluator is likely to probe. Knowing these is the difference between a pass and a good mark.
Retrieval, not memorisation
The documents are indexed and the relevant parts are retrieved for each question. The model writes the answer from that, which is why it can cite your material.
Grounding and refusal
Answers stay inside the uploaded sources and it says when a question falls outside them. This is the property that separates it from a plain chatbot.
The integration layer
A CDN script exposes a global with a mount call taking a key and a config object. It returns a handle you use to update or unmount the widget.
The limits you chose
Being able to say what it cannot do — images in a scan, anything outside the sources — reads as understanding rather than as a gap.
Be straight
This matters more than any feature on this page.
Cortexvia is a hosted platform. If you use it, your project is the page, the integration, the document set you chose, and your understanding of how retrieval and grounding work — not the model behind it. Say that plainly in your report and your viva. Presenting a platform as something you built yourself is the one thing that turns a good project into an academic integrity problem, and evaluators ask exactly the questions that expose it. If your brief requires you to implement the retrieval layer yourself, build that part and use this only as a reference for what the finished behaviour should look like.
Questions
What students ask before committing to this for a submission.
More for students
The same platform, pointed at the rest of your material.
Get started
Upload one real document, embed it in a page, and prepare four questions.
Use a document your evaluator recognises. A chatbot that answers real questions about your own college lands far better than one trained on sample text.