Коди - AI Programming Assistant with Learning and Memory

An intelligent Firebase Cloud Functions-based AI assistant that learns from every conversation and provides personalized programming guidance.

🚀 Features

Core Capabilities

• AI-Powered Programming Assistance: Uses Google's Gemini AI to help with HTML, CSS, JavaScript, and Python
• Bulgarian Language Support: Responds in Bulgarian for local users
• Conversational Memory: Remembers all past interactions
• Continuous Learning: Extracts and tracks programming topics from conversations
• Personalized Responses: Tailors answers based on user's learning history
• Multi-User Support: Maintains separate learning profiles for each user

Technical Features

• Firebase Cloud Functions backend
• Firestore database for conversation storage
• Google Generative AI (Gemini) integration
• RESTful API endpoints
• Callable Cloud Functions for client integration
• Automatic topic extraction and categorization
• Session management

📦 Installation

Prerequisites

• Node.js 22.x (note: currently running on Node 20.x with warnings)
• Firebase CLI
• Google Cloud account
• Gemini API key

Setup Steps

1. Clone the repository

   git clone https://github.com/papica777-eng/dpengeneering.git
   cd dpengeneering

2. Install dependencies

   cd functions
   npm install

3. Configure API Key Edit functions/index.js and replace the API key:

   const API_KEY = "YOUR_GEMINI_API_KEY_HERE";

4. Deploy to Firebase

   firebase deploy --only functions

5. Deploy Firestore indexes

   firebase deploy --only firestore:indexes

🔧 API Functions

1. callKodyAPI (Callable Function)

Main function for interacting with the AI assistant.

Parameters:

{
  userId: string,           // User identifier (optional, defaults to 'anonymous')
  sessionId: string,        // Session identifier (optional)
  chatHistory: array,       // Previous messages in the conversation
  userParts: array         // Current user message
}

Returns:

{
  text: string              // AI's response
}

Example:

const result = await firebase.functions().httpsCallable('callKodyAPI')({
    userId: 'user123',
    chatHistory: [],
    userParts: ['Как да създам функция в JavaScript?']
});
console.log(result.data.text);

2. getUserLearningStats (Callable Function)

Retrieve learning statistics for a user.

Parameters:

{
  userId: string            // User identifier
}

Returns:

{
  topics: array,            // List of topics discussed
  preferences: object,      // User preferences
  firstInteraction: timestamp,
  lastInteraction: timestamp,
  interactionCount: number
}

3. getConversationHistory (Callable Function)

Get past conversations for a user.

Parameters:

{
  userId: string,           // User identifier
  limit: number            // Maximum number of conversations (default: 10)
}

Returns:

{
  conversations: array      // Array of conversation objects
}

4. systemHealth (HTTP Function)

Check system health and database connectivity.

Endpoint: GET /systemHealth

Returns:

{
  status: string,           // 'OK' or 'ERROR'
  checks: {
    server: { status: string, message: string },
    database: { status: string, message: string }
  }
}

5. greetUserDB (HTTP Function)

Test database with a greeting function.

Endpoint: GET /greetUserDB?name=<username>

🗄️ Database Schema

Collection: conversations

Stores all user-AI conversations.

{
  userId: string,
  sessionId: string,
  timestamp: Timestamp,
  userMessage: string,
  aiResponse: string,
  chatHistory: array
}

Index: userId (ASC) + timestamp (DESC)

Collection: user_learning

Stores user learning profiles.

{
  topics: array,            // e.g., ['HTML', 'CSS', 'JavaScript']
  preferences: object,
  firstInteraction: Timestamp,
  lastInteraction: Timestamp,
  interactionCount: number
}

Collection: users

Tracks user visits (from greetUserDB function).

{
  firstVisit: Date
}

🎯 How Learning Works

1. User sends a message → System retrieves their learning profile
2. AI generates response → Enhanced with context from past interactions
3. Conversation is saved → Stored in Firestore for future reference
4. Topics are extracted → Automatically identified from the conversation
5. Profile is updated → New topics and interaction count saved

Topic Extraction

The system tracks 20+ programming keywords:

• Languages: HTML, CSS, JavaScript, Python
• Bulgarian terms: функция, клас, променлива, масив, обект, цикъл, условие
• English terms: function, class, array, loop, variable
• Technologies: Firebase, база данни, API

📚 Documentation

• LEARNING_SYSTEM.md - Comprehensive documentation of the learning system
• USAGE_EXAMPLES.js - Code examples for all API functions

🧪 Testing

Run basic tests:

cd functions
node test-learning.js

🔒 Security

• CodeQL security scanning: ✅ No vulnerabilities found
• User data is isolated by userId
• Anonymous mode available
• API key should be stored securely (use environment variables in production)

Production Recommendations

1. Enable authentication check in callKodyAPI:

   if (!context.auth) {
       throw new functions.https.HttpsError('unauthenticated', 'Please log in.');
   }

2. Store API key in environment variables
3. Implement rate limiting
4. Add data retention policies
5. Comply with data protection regulations (GDPR, etc.)

🛠️ Development

Local Development

# Start Firebase emulators
firebase emulators:start

# Test functions locally
curl http://127.0.0.1:5001/kodi-backend/us-central1/systemHealth

Project Structure

dpengeneering/
├── functions/
│   ├── index.js              # Main Cloud Functions
│   ├── package.json          # Dependencies
│   └── test-learning.js      # Basic tests
├── public/
│   ├── index.html           # Web interface
│   └── manifest/
├── firestore.indexes.json   # Database indexes
├── firestore.rules          # Security rules
├── firebase.json            # Firebase configuration
├── LEARNING_SYSTEM.md       # Learning system documentation
├── USAGE_EXAMPLES.js        # API usage examples
└── README.md               # This file

📈 Future Enhancements

• Sentiment analysis for user satisfaction tracking
• Difficulty level adaptation
• Personalized learning paths
• Quiz generation based on discussed topics
• Progress visualization dashboard
• Multi-language support expansion
• Advanced preference learning (learning style, explanation depth)
• Integration with code execution environments
• Voice interface support

👥 Contributing

Created by Камелия (Kamelia)
Maintained by the dpengineering team

📄 License

This project is private.

🙏 Acknowledgments

• Google Generative AI (Gemini) for AI capabilities
• Firebase for backend infrastructure
• The open-source community for various dependencies

📞 Support

For issues and questions, please open an issue in the GitHub repository.

---

Note: This AI assistant is designed for educational purposes to help beginners learn programming. Always verify code suggestions and follow best practices.