Thanks to visit codestin.com
Credit goes to github.com

Skip to content

tetraksys/edu-ai-adk

 
 

Repository files navigation

📘 Edu.AI — Multi-Agent Educational System for ENEM

🇧🇷 Quality education for all, powered by autonomous agents.

Edu.AI is an intelligent multi-agent platform designed to democratize access to high-quality ENEM preparation in Brazil. Built using Google’s Agent Development Kit (ADK), Edu.AI offers personalized essay correction, simulated exams, study plans, and content generation — simulating the experience of an elite preparatory course, but powered entirely by AI.


🚀 Features

  • ✍️ AI-powered Essay Correction Corrects essays according to ENEM's 5 official competencies, with scores and detailed feedback.

  • 📚 ENEM-style Prompt Generation Creates contextualized essay prompts with motivational texts.

  • 🧪 Simulated Exams by Area & Theme Builds customized tests with answer keys and explanations.

  • 🧩 Interdisciplinary Question Generation Combines areas like History + Physics, or Chemistry + Sociology.

  • 🎓 Personalized Study Paths Analyzes performance and suggests targeted learning strategies.

  • 🎥 Didactic Content Generation Produces summaries, flashcards, slides, and visuals.

  • 🗣️ AI Rewriting Assistant Helps students rewrite and improve their essays.

  • 📈 Progress Tracking Dashboard Tracks scores over time and gives actionable recommendations.

  • 📤 Seamless Frontend Delivery Clean UI powered by Next.js and deployed via Vercel.


🧠 Multi-Agent System Overview

Agent Role
EssayEvaluatorAgent Evaluates essays with official rubric and feedback
PromptBuilderAgent Generates contextual ENEM-style essay prompts
SimulatedExamAgent Produces realistic exams with grading and explanations
InterdisciplinaryAgent Creates cross-disciplinary questions
ContentGeneratorAgent Builds content: summaries, flashcards, slides
RephraserAgent Suggests improved phrasing for essays or short answers
ProgressTrackerAgent Tracks performance and calculates averages
PersonalTutorAgent Recommends study paths based on user history

🗺️ Architecture

🧭 Agents work collaboratively via an orchestrator and shared state.

Architecture Diagram

  • Frontend: Next.js + Tailwind (Vercel)
  • Backend: FastAPI with ADK agents
  • Storage: Google Cloud Storage, SQLite
  • OCR: Cloud Vision for extracting essays from images
  • LLMs: Gemini 2.5-flash

🏃‍♀️ Installation

To run the agents, clone the repository and install the dependencies:

pip install -r requirements.txt

Then, start the FastAPI server:

uvicorn main:app --reload --port 8080

The frontend is available at https://edu-ai-adk.vercel.app/.


🏆 What We’re Proud Of

  • Built a truly modular and scalable multi-agent architecture with 8+ intelligent agents.
  • Delivered a full-stack AI product with a clean and functional UI.
  • Created a solution with real potential to impact education at scale in Brazil.

Built by Giovanna Moeller.

About

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • TypeScript 75.9%
  • Python 21.3%
  • CSS 2.7%
  • JavaScript 0.1%