EduBot

A serverless AI-powered Q&A system for educational institutions

EduBot provides curriculum-specific answers using retrieval-augmented generation (RAG) with Amazon Bedrock. Students ask questions and receive answers based exclusively on their school's textbooks and course materials.

Features

• Curriculum-Based Answers: Responses drawn only from uploaded course materials
• Dual AI Backend: Gemini (free tier) with Bedrock fallback
• Serverless Architecture: AWS Lambda with container deployment
• Modern Frontend: Clean UI with book selection and collapsible explainer
• Secure & Private: All data stays within your AWS account
• RAG Pipeline: Keyword-scored retrieval + AI generation
• Production Ready: CI/CD, monitoring, error handling, and tests

EduBot only answers questions from the curriculum - no off-topic content

Architecture

• API: Python Lambda function with REST endpoints
• Storage: S3 for curriculum documents and processed indexes
• AI: Amazon Bedrock (Claude Haiku) for answer generation
• Security: IAM roles, VPC isolation, KMS encryption
• Monitoring: CloudWatch metrics and custom dashboards
• Deployment: GitHub Actions CI/CD pipeline

Quick Start

Prerequisites

• AWS CLI configured
• Docker installed
• Python 3.12+

1. Deploy Infrastructure

git clone https://github.com/YOUR_USERNAME/edubot_mvp.git
cd edubot_mvp

# Set up AWS resources
source infra/00-variables.sh
./infra/10-kms-and-s3.sh
./infra/10-iam.sh
./infra/60-lambda-api.sh

2. Upload Curriculum

# Process and upload textbooks
python tools/indexer.py \
  --pdf /path/to/textbook.pdf \
  --book-id philosophy \
  --subject philosophy \
  --s3-bucket $CURRICULUM_BUCKET

3. Test the API

# Health check
curl https://your-lambda-url/health

# Ask a question
curl -X POST https://your-lambda-url/ask \
  -H "Content-Type: application/json" \
  -d '{"question": "What is philosophy?", "book_id": "philosophy"}'

API Endpoints

• GET /health - System health and dependency status
• GET /indexes - List available curriculum indexes
• POST /ask - Ask questions based on curriculum

Development

Local Testing

# Set environment
export CURRICULUM_BUCKET=your-bucket-name

# Run tests
python run_tests.py

# Test locally with mock mode
MOCK_BEDROCK=true python -c "
import sys; sys.path.append('src/api')
import handler, json
event = {'httpMethod': 'POST', 'path': '/ask', 'body': json.dumps({'question': 'test'})}
print(handler.lambda_handler(event, None))
"

CI/CD Pipeline

Push to main branch triggers:

1. Automated testing
2. Code quality checks
3. Docker build and ECR push
4. Lambda function update

Configuration

Environment Variables

• CURRICULUM_BUCKET - S3 bucket for curriculum data
• INDEX_PREFIX - S3 prefix for processed indexes
• TOP_K - Number of sections to retrieve (default: 20)
• BEDROCK_MODEL - AI model ID (default: claude-3-haiku)
• GEMINI_API_KEY - Google Gemini API key (optional, free tier)
• MOCK_BEDROCK - Enable mock mode for development

Security

• All curriculum data encrypted with KMS
• IAM roles with minimal required permissions
• VPC isolation for Lambda function
• No data leaves your AWS account

Monitoring

• CloudWatch metrics for invocations, errors, and latency
• Custom dashboard for system health
• Structured logging with request tracking
• Automated alerts for error rates and performance

Production Considerations

• Scaling: Lambda auto-scales based on demand
• Cost: Pay-per-request pricing model
• Reliability: Multi-AZ deployment with AWS managed services
• Backup: S3 versioning and cross-region replication
• Updates: Blue/green deployments via CI/CD pipeline

Documentation

• Cost Analysis - Detailed cost breakdown and ROI
• API Documentation
• Deployment Guide
• Security Considerations
• Development Journal

Current Books

• Introduction to Philosophy - OpenStax
• World History Volume 1 - Prehistory to 1500 CE
• World History Volume 2 - 1400 CE to Present

TODO / Future Improvements

High Priority

• Semantic Search: Replace keyword scoring with proper embeddings (OpenAI/Bedrock)
• Vector Database: Add Pinecone/FAISS for true similarity search
• Docker Build Fix: Resolve Lambda image deployment issues for easier updates

Medium Priority

• User Authentication: Add Cognito for student/teacher accounts
• Usage Analytics: Track popular questions and book usage
• Answer Quality: Add citation links to specific page numbers
• Multi-Book Search: Allow searching across multiple books simultaneously

Nice to Have

• Chat History: Store conversation context for follow-up questions
• Admin Dashboard: Upload books and manage content via UI
• Mobile App: React Native wrapper for iOS/Android
• Study Tools: Flashcards, quizzes generated from curriculum

Known Issues

• Retrieval is keyword-based (not semantic) - may miss relevant sections
• Gemini free tier rate limits (15 req/min) - auto-falls back to Bedrock
• Docker image deployment requires manual steps - CI/CD needs fixing
• No caching - every question hits S3 and AI (could add Redis/ElastiCache)

License

MIT License - see LICENSE file for details.