Base Library - AI-Powered Educational Material Generation Platform

A comprehensive microservices-based platform for generating educational materials using LangGraph workflows, featuring handwritten note recognition, material synthesis, and interactive question generation.

🏗️ Architecture Overview

This project implements a sophisticated AI-powered educational platform built on a microservices architecture with the following core components:

Core Services

1. Core AI Service - Main LangGraph workflow orchestrator
2. Article Service - File storage and export management
3. Prompt Studio Service - Dynamic prompt generation and user customization
4. PostgreSQL - Primary data storage

🔄 LangGraph Workflow Architecture

The platform is built around a sophisticated LangGraph workflow that processes educational materials through multiple AI-powered nodes:

Workflow Flow

graph TD
    A[START] --> B[Input Processing]
    B --> C[Content Generation]
    C --> D{Images Present?}
    D -->|Yes| E[Handwritten Recognition]
    D -->|No| F[Question Generation]
    E --> G[Material Synthesis]
    G --> H[Material Editing]
    H --> F[Question Generation]
    F --> I[Answer Generation]
    I --> J[END]

Loading

Node Descriptions

1. Input Processing Node (input_processing)

• Purpose: Analyzes user input and determines workflow path
• Functionality:
  • Processes text input and image uploads
  • Validates content using security guards
  • Determines next workflow step based on input type

2. Content Generation Node (generating_content)

• Purpose: Generates educational material based on exam questions
• Functionality:
  • Creates comprehensive study materials
  • Uses personalized prompts from Prompt Strudio Service
  • Determines workflow path (recognition vs. direct question generation)

3. Recognition Node (recognition_handwritten)

• Purpose: Processes handwritten notes from uploaded images
• Functionality:
  • OCR processing of handwritten content
  • Text extraction and formatting
  • Integration with image processing services

4. Synthesis Node (synthesis_material)

• Purpose: Combines generated content with recognized notes
• Functionality:
  • Merges AI-generated material with handwritten notes
  • Creates comprehensive study materials
  • Ensures content coherence and completeness

5. Edit Material Node (edit_material)

• Purpose: Interactive material editing with human-in-the-loop (HITL)
• Functionality:
  • Iterative material refinement
  • User feedback integration
  • Content improvement based on user input

6. Question Generation Node (generating_questions)

• Purpose: Creates assessment questions with HITL feedback
• Functionality:
  • Generates relevant exam questions
  • Implements feedback loop for question refinement
  • Ensures question quality and relevance

7. Answer Generation Node (answer_question)

• Purpose: Generates comprehensive answers to questions
• Functionality:
  • Creates detailed answer explanations
  • Provides educational value
  • Finalizes the learning material package

🛠️ Technical Architecture

State Management

The workflow uses a comprehensive state model (GeneralState) that tracks:

class GeneralState(BaseModel):
    # Input data
    input_content: str
    display_name: Optional[str]
    image_paths: List[str]
    
    # Processing results
    recognized_notes: str
    generated_material: str
    synthesized_material: str
    
    # Questions and answers
    questions: List[str]
    questions_and_answers: List[str]
    
    # HITL interaction
    feedback_messages: List[Any]
    edit_count: int
    needs_user_input: bool
    agent_message: Optional[str]

Security & Validation

• Security Guard: Content validation and sanitization
• Input Validation: Multi-layer content checking
• Fuzzy Matching: Advanced threat detection
• Graceful Degradation: System continues operation even with security failures

Model Configuration

The platform supports multiple LLM providers:

• OpenAI (GPT-4, GPT-3.5)
• OpenRouter (Various models)
• Fireworks AI
• DeepSeek

Each node can be configured with different models based on requirements.

🚀 Services Architecture

Core AI Service (Port 8000)

• Framework: FastAPI + LangGraph
• Functionality:
  • Workflow orchestration
  • State management
  • HITL interaction handling
  • Artifact management

Artifacts Service (Port 8001)

• Framework: FastAPI
• Functionality:
  • File storage and management
  • Export capabilities (PDF, Markdown, ZIP)
  • Thread-based organization
  • Web3 authentication support

Prompt Studio Service (Port 8002)

• Framework: FastAPI
• Functionality:
  • Dynamic prompt generation
  • User profile management
  • Template customization
  • Prompt versioning

Database (Port 5431)

• Type: PostgreSQL 16
• Functionality:
  • Workflow state persistence
  • User data storage
  • Artifact metadata
  • Authentication data

🔧 Configuration

Environment Variables

# Core Configuration
OPENAI_API_KEY=your_openai_key
DATABASE_URL=postgresql://postgres:postgres@postgres:5432/core

# Optional LLM Providers
DEEPSEEK_API_KEY=your_deepseek_key

# Security
SECURITY_ENABLED=true
SECURITY_FUZZY_THRESHOLD=0.85
SECURITY_MIN_CONTENT_LENGTH=10

# Artifacts
ARTIFACTS_BASE_PATH=/app/data/artifacts
ARTIFACTS_MAX_FILE_SIZE=10485760

Configuration Files

• configs/graph.yaml - LangGraph node configurations
• configs/prompts.yaml - Prompt templates
• configs/providers.yaml - LLM provider settings

🐳 Deployment

Docker Compose

# Start all services
docker-compose up -d

Service Health Checks

• Core: http://localhost:8000/health
• Article: http://localhost:8001/health
• Prompt Strudio: http://localhost:8002/health

📊 Monitoring & Observability

Logging

• Centralized logging in /logs directory
• Structured logging with different levels
• Service-specific log files

🔄 Workflow Execution

Human-in-the-Loop (HITL) Pattern

The platform implements sophisticated HITL patterns:

1. Generation: AI creates initial content
2. Feedback: User provides input/feedback
3. Editing: AI refines content based on feedback
4. Completion: Process continues until satisfaction

State Persistence

• PostgreSQL checkpoints for workflow state
• Resume capability for interrupted workflows
• Thread-based session management

🎯 Key Features

Educational Material Generation

• AI-powered content creation
• Handwritten note recognition
• Material synthesis and editing
• Interactive question generation

Export Capabilities

• Markdown export for documentation

Web3 Integration

• Blockchain authentication
• NFT content verification
• Decentralized ownership tracking
• Content hash validation

Security Features

• Content sanitization
• Injection attack prevention
• Fuzzy threat detection
• Graceful security degradation

🚀 Getting Started

1. Clone the repository
2. Set up environment variables
3. Run database migrations
4. Start services with Docker Compose
5. Access the API endpoints

📝 API Documentation

Core Service Endpoints

• POST /process - Start new workflow
• GET /status/{thread_id} - Check workflow status
• POST /continue/{thread_id} - Continue HITL interaction

Artifacts Service Endpoints

• GET /threads - List all threads
• POST /export/pdf/{thread_id}/{session_id} - Export to PDF
• POST /export/markdown/{thread_id}/{session_id} - Export to Markdown

Prompt Studio Endpoints

• GET /prompts/{user_id}/{node_name} - Get personalized prompts
• POST /profiles - Create user profiles
• PUT /profiles/{profile_id} - Update user profiles

🔮 Future Enhancements

• Advanced AI model integration
• Enhanced Web3 features
• Real-time collaboration
• Advanced analytics and insights
• Multi-language support
• Mobile application

---

This platform represents a cutting-edge approach to AI-powered educational content generation, combining the power of LangGraph workflows with modern microservices architecture and Web3 technologies.