What I Built
I created ClassMate AI, a multi-agent assistant designed to help students manage their academic life more efficiently. It solves common back-to-school challenges like organizing assignments, summarizing lecture notes, generating study quizzes, and even suggesting personalized learning resources based on performance trends.
ClassMate AI acts as a virtual study buddy that understands your schedule, adapts to your learning style, and keeps you on track throughout the semester.
Category
Student Success
Educator Empowerment
Crazy Creative
Demo
Live App on Heroku
Source Code on GitHub
Screenshots & Demo Video:
How I Used Heroku AI
ClassMate AI uses:
Heroku Managed Inference to run a custom GPT-based summarization agent for lecture notes.
pgvector for Heroku Postgres to store and retrieve semantic embeddings of notes and resources.
Model Context Protocol (MCP) to coordinate agents: one for note summarization, one for quiz generation, and one for resource recommendation.
Agents communicate via shared context and trigger each other based on user actions (e.g., uploading notes triggers summarization, which then triggers quiz generation).
Technical Implementation
Multi-agent architecture using MCP:
SummarizerAgent: Extracts key points from uploaded notes.
QuizAgent: Generates multiple-choice questions based on summaries.
TutorAgent: Suggests resources using vector similarity from pgvector.
Tech Stack:
Frontend: React + Tailwind
Backend: Node.js + Express
AI: OpenAI GPT + Heroku Managed Inference
Database: Heroku Postgres with pgvector
Challenges Solved:
Coordinating agent workflows using MCP
Embedding and querying semantic data efficiently
Ensuring fast inference with Heroku's AI infrastructure
Top comments (0)