Inspiration
We are inspired by the feeling of confusion that many students have felt during a given semester where they can't keep up with the work load and start falling behind on deadlines and assignments. Our team seeks to remedy this.
Abstract
Canvas Companion is an integrated Canvas LMS AI Dialogue application that allows students to access and review their assignments, courses, grades, quizzes, and more with natural language in a conversational context with a specialized LLM assistant.
Stack
On the frontend, the application is powered by React.js/Next.js, with a Node.js backend. REST API calls are made to Canvas to fetch updated student data, while LLM API calls are made through OpenAI using gpt-3.5-turbo and the new function calling feature, which is how the agent can access the Canvas data on demand.
Frontend React: Maya, Ryan
Backend Canvas: Nana, Hemanth
Backend OpenAI: Sebastian
Challenges
We ran into challenges with parsing the copious amounts of unstructured data from Canvas API. There were also many scopes that we considered scaling to, including user authentication and database solutions, but ultimately did not finish the components in the interest of time and focus on the minimum viable product.
Accomplishments
We are proud of the fact that most of our team came in with zero knowledge of React or Next.js or JavaScript, and came out as the lead frontend developers by the end. Those who knew Next.js already were the lead backend developers.
Learning
We learned a lot about full-stack development, JavaScript, React, Next.js, REST APIs, and so much more, including interpersonal integrity when on such a time crunch.
Conclusion + Future
This is just a small demonstration of what is possible when LLM agents are given access to data beyond their pre-trained scope. With this, we hope to empower students to get a broader, more intuitive view of their academic career, so that they may spend more of their time learning and improving instead of stressing, especially for those who may not be familiar with the Canvas platform. With elevated access to the API, there are so many opportunities to dig deeper into the details of a student's potential to uncover insights and analysis they may benefit from knowing.
Built With
- css
- html
- instructure-canvas
- javascript
- jsx
- next.js
- node.js
- openai
- react
Log in or sign up for Devpost to join the conversation.