Contextual Agent SDK

A powerful SDK for building AI agents with seamless voice-text context switching capabilities. Enable your agents to maintain conversation context when switching between voice and text modalities.

🚀 Key Features

Core Capabilities

🔄 Seamless Voice-Text Context Switching - The main innovation
🤖 Multi-LLM Support - OpenAI, Anthropic, Ollama, Google, Azure, Custom APIs
🧠 Intelligent Context Management - Pluggable context providers with auto-discovery
💾 Persistent Session Management - Redis, MongoDB, or in-memory storage
🎙️ Configurable Speech Integration - STT/TTS with multiple provider support
📝 Auto-Documentation Discovery - README, docs, changelogs automatically indexed
⚡ Event-Driven Architecture - Real-time monitoring and analytics
🔧 TypeScript First - Full type safety and IntelliSense support

Production Ready

🏗️ Modular Architecture - Use only what you need
🔐 Secure by Default - Environment-based configuration
📊 Performance Monitoring - Built-in metrics and analytics
🛡️ Error Handling - Graceful fallbacks and retry mechanisms
🧪 Comprehensive Testing - Unit and integration tests included
📚 Rich Documentation - Examples, guides, and API reference

🎯 The Innovation: Context Bridging

The Problem: Traditional AI agents lose context when switching between voice and text interactions.

Our Solution: Intelligent context bridging that preserves conversation state and adapts responses for different modalities.

// Start conversation via text
await agent.processMessage('I need help with order #12345', 'text', 'session-123');

// 🔄 THE MAGIC: Switch to voice mid-conversation
await agent.switchModality('voice', 'session-123');
// Agent responds: "I see you're asking about order 12345. Let me help you with that..."

// Continue seamlessly with voice
const audioFile = new File([audioBuffer], 'voice.wav');
await agent.processMessage(audioFile, 'voice', 'session-123');

📦 Installation

npm install contextual-agent-sdk

🚀 Quick Start

Basic Setup

import { ContextualAgent } from 'contextual-agent-sdk';

const agent = new ContextualAgent({
  name: 'Support Agent',
  systemPrompt: 'You are a helpful customer support agent.',
  llm: {
    providers: {
      'openai': {
        type: 'openai',
        apiKey: process.env.OPENAI_API_KEY
      }
    },
    defaultProvider: 'openai'
  }
});

// Text interaction
const response = await agent.processMessage(
  'Hello, I need help',
  'text',
  'session-123'
);

console.log(response.data.message.content);

Advanced Setup with All Features

import { 
  ContextualAgent, 
  KnowledgeBaseProvider, 
  DatabaseContextProvider 
} from 'contextual-agent-sdk';

// 1. Setup context providers
const knowledgeBase = new KnowledgeBaseProvider({
  id: 'docs',
  name: 'Documentation',
  priority: 90,
  options: {
    autoDiscoverDocs: {
      enabled: true,
      rootPath: process.cwd(),
      patterns: ['README.md', 'docs/**/*.md']
    }
  }
});

const database = new DatabaseContextProvider({
  id: 'user_data',
  name: 'User Database',
  priority: 80,
  connection: {
    type: 'custom',
    customQuery: async (query) => {
      // Your database logic
      return { preferences: {}, history: [] };
    }
  }
});

// 2. Setup speech providers
const speechToText = {
  async transcribe(audioInput, options) {
    // OpenAI Whisper implementation
    const formData = new FormData();
    formData.append('file', audioInput);
    formData.append('model', options?.model || 'whisper-1');
    
    const response = await fetch('https://api.openai.com/v1/audio/transcriptions', {
      method: 'POST',
      headers: { 'Authorization': `Bearer ${process.env.OPENAI_API_KEY}` },
      body: formData
    });
    
    return await response.json();
  }
};

// 3. Create agent with full configuration
const agent = new ContextualAgent({
  name: 'Advanced Support Agent',
  systemPrompt: 'You are an intelligent support agent with access to documentation and user data.',
  
  // Multi-LLM setup with fallback
  llm: {
    providers: {
      'openai': { type: 'openai', apiKey: process.env.OPENAI_API_KEY },
      'anthropic': { type: 'anthropic', apiKey: process.env.ANTHROPIC_API_KEY },
      'ollama': { type: 'ollama', baseURL: 'http://localhost:11434' }
    },
    defaultProvider: 'openai',
    fallbackProvider: 'ollama'
  },
  
  // Context management
  contextManager: {
    providers: [knowledgeBase, database]
  },
  
  // Speech capabilities
  modalityRouter: {
    speechToText,
    defaultSTTOptions: {
      language: 'en-US',
      model: 'whisper-1'
    }
  },
  
  // Persistent storage
  storage: {
    type: 'redis',
    config: { url: process.env.REDIS_URL }
  }
});

// 4. Use the agent
const textResponse = await agent.processMessage(
  'What are the installation steps?',
  'text',
  'session-123'
);

// 5. Switch to voice seamlessly
const voiceResponse = await agent.switchModality('voice', 'session-123');

// 6. Process voice input
const audioFile = new File([audioBuffer], 'question.wav');
const voiceReply = await agent.processMessage(audioFile, 'voice', 'session-123');

🧠 Context Management

Auto-Discovery of Documentation

const docsProvider = new KnowledgeBaseProvider({
  id: 'project_docs',
  options: {
    autoDiscoverDocs: {
      enabled: true,
      patterns: [
        'README.md', 'CHANGELOG.md', 'LICENSE',
        'docs/**/*.md', 'api/**/*.json'
      ]
    }
  }
});

// Automatically indexes:
// ✅ README files
// ✅ Documentation folders  
// ✅ API specifications
// ✅ Changelog and contributing guides
// ✅ License information

Custom Context Providers

class CRMProvider implements ContextProvider {
  async getContext(params) {
    const customer = await crm.getCustomer(params.userId);
    return {
      content: {
        tier: customer.tier,
        tickets: customer.openTickets,
        satisfaction: customer.satisfactionScore
      },
      metadata: {
        source: 'crm',
        timestamp: new Date()
      }
    };
  }
}

🎙️ Speech Integration

Multiple STT/TTS Providers

// OpenAI Whisper + TTS
const openaiSpeech = {
  speechToText: whisperProvider,
  textToSpeech: openaiTTSProvider
};

// Google Cloud Speech
const googleSpeech = {
  speechToText: googleSTTProvider,
  textToSpeech: googleTTSProvider
};

// AWS Transcribe + Polly
const awsSpeech = {
  speechToText: transcribeProvider,
  textToSpeech: pollyProvider
};

// Mix and match
const agent = new ContextualAgent({
  modalityRouter: {
    speechToText: googleSTTProvider,  // Google for STT
    textToSpeech: openaiTTSProvider,  // OpenAI for TTS
    defaultSTTOptions: {
      language: 'en-US',
      model: 'latest_long'
    }
  }
});

🤖 Multi-LLM Support

Provider Management

const agent = new ContextualAgent({
  llm: {
    providers: {
      'primary': { type: 'openai', apiKey: '...' },
      'backup': { type: 'anthropic', apiKey: '...' },
      'local': { type: 'ollama', baseURL: 'http://localhost:11434' },
      'custom': { type: 'custom', baseURL: 'https://my-llm-api.com' }
    },
    defaultProvider: 'primary',
    fallbackProvider: 'backup'
  }
});

// Runtime provider management
agent.addLLMProvider('new-provider', { type: 'openai', apiKey: '...' });
agent.setDefaultLLMProvider('new-provider');

// Test providers
const status = await agent.testLLMProvider('primary');
console.log(status.success); // true/false

💾 Session Storage

Multiple Storage Options

// Redis (Production)
const agent = new ContextualAgent({
  storage: {
    type: 'redis',
    config: {
      url: process.env.REDIS_URL,
      ssl: true,
      maxConnections: 10
    }
  }
});

// MongoDB (Document Storage)
const agent = new ContextualAgent({
  storage: {
    type: 'mongodb',
    config: {
      url: process.env.MONGODB_URL,
      ssl: true
    }
  }
});

// Memory (Development)
const agent = new ContextualAgent({
  storage: { type: 'memory' }
});

📊 Event System & Monitoring

Real-time Event Monitoring

// Session events
agent.on('session_started', (event) => {
  console.log(`New session: ${event.sessionId}`);
});

// Context switching events
agent.on('modality_switched', (event) => {
  console.log(`Switched from ${event.data.from} to ${event.data.to}`);
});

// Context bridging events
agent.on('context_bridged', (event) => {
  console.log(`Context bridged: ${event.data.bridgeType}`);
});

// Performance monitoring
agent.on('performance_metric', (event) => {
  console.log(`Response time: ${event.data.responseTime}ms`);
});

// Error handling
agent.on('error_occurred', (event) => {
  console.error(`Error in session ${event.sessionId}:`, event.data.error);
});

🔧 Configuration

Environment Variables

Create a .env file:

# LLM Providers
OPENAI_API_KEY=sk-your-openai-key
ANTHROPIC_API_KEY=sk-ant-your-anthropic-key
GOOGLE_AI_API_KEY=your-google-key
OLLAMA_HOST=http://localhost:11434

# Speech Services
AZURE_SPEECH_KEY=your-azure-key
AZURE_SPEECH_REGION=your-region
GOOGLE_APPLICATION_CREDENTIALS=path/to/service-account.json
ASSEMBLYAI_API_KEY=your-assemblyai-key
DEEPGRAM_API_KEY=your-deepgram-key

# Storage
REDIS_URL=redis://localhost:6379
MONGODB_URL=mongodb://localhost:27017/agents

# Defaults
DEFAULT_LLM_PROVIDER=openai
DEFAULT_STT_LANGUAGE=en-US
DEFAULT_TTS_VOICE=alloy

TypeScript Configuration

interface AgentConfig {
  name: string;
  systemPrompt: string;
  llm?: LLMConfig;
  contextManager?: ContextManagerConfig;
  modalityRouter?: ModalityRouterConfig;
  storage?: StorageConfig;
}

🏗️ Architecture

┌─────────────────┐    ┌─────────────────┐    ┌─────────────────┐
│  ContextualAgent │────│ SessionManager  │────│  StorageProvider│
│                 │    │                 │    │  (Redis/Mongo)  │
│ Main SDK Class  │    │ Session & State │    │                 │
└─────────────────┘    └─────────────────┘    └─────────────────┘
         │                        │                        │
         │                        │                        │
┌─────────────────┐    ┌─────────────────┐    ┌─────────────────┐
│ ModalityRouter  │    │ ContextManager  │    │   LLMManager    │
│                 │    │                 │    │                 │
│ Voice/Text      │    │ Context Bridge  │    │ Multi-Provider  │
│ Processing      │    │ & Providers     │    │ Management      │
└─────────────────┘    └─────────────────┘    └─────────────────┘

🧪 Testing

# Run all tests
npm test

# Run specific test suites
npm run test:core
npm run test:integration

# Run with coverage
npm run test:coverage

# Validate setup
npm run validate-setup

📚 Examples & Demos

# Basic multi-LLM demo
npm run demo:multi-llm

# Investor presentation demo
npm run demo:investor

# Interactive HTTP server
npm run demo

🔗 Use Cases

Customer Support

const supportAgent = new ContextualAgent({
  systemPrompt: 'You are a customer support agent with access to user data and documentation.',
  contextManager: {
    providers: [
      new DatabaseContextProvider({ /* CRM integration */ }),
      new KnowledgeBaseProvider({ /* FAQ & docs */ })
    ]
  }
});

// Handle support tickets across voice and text

Educational Assistant

const tutorAgent = new ContextualAgent({
  systemPrompt: 'You are an educational assistant helping students learn.',
  contextManager: {
    providers: [
      new KnowledgeBaseProvider({ /* Curriculum content */ }),
      new DatabaseContextProvider({ /* Student progress */ })
    ]
  }
});

// Seamless transition from reading materials to voice Q&A

Code Assistant

const codeAgent = new ContextualAgent({
  contextManager: {
    providers: [
      new KnowledgeBaseProvider({
        options: {
          autoDiscoverDocs: {
            enabled: true,
            patterns: ['**/*.md', 'src/**/*.ts', 'docs/**/*']
          }
        }
      })
    ]
  }
});

// Code review via text, explanations via voice

🚀 Production Deployment

Docker Setup

FROM node:18-alpine
WORKDIR /app
COPY package*.json ./
RUN npm ci --production
COPY . .
RUN npm run build
EXPOSE 3000
CMD ["npm", "start"]

Environment Configuration

# docker-compose.yml
version: '3.8'
services:
  agent:
    build: .
    environment:
      - OPENAI_API_KEY=${OPENAI_API_KEY}
      - REDIS_URL=redis://redis:6379
    depends_on:
      - redis
  
  redis:
    image: redis:7-alpine

🤝 Contributing

Fork the repository
Create feature branch: git checkout -b feature/amazing-feature
Commit changes: git commit -m 'Add amazing feature'
Push to branch: git push origin feature/amazing-feature
Open a Pull Request

📄 License & Attribution

This project is licensed under the MIT License - see the LICENSE file for details.

🏷️ Attribution Requirements

If you use this SDK in your project, please provide attribution to simcoder technologies a subsidiary of SCS Group as the creator. You can do this in one of the following ways:

In Your Application:

<!-- In your app footer or about section -->
Powered by <a href="https://github.com/simcoder/contextual-agent-sdk">Contextual Agent SDK</a> by simcoder technologies

In Your README:

## Credits
- Uses [Contextual Agent SDK](https://github.com/simcoder/contextual-agent-sdk) by simcoder technologies for seamless voice-text AI interactions

In Package.json:

{
  "credits": {
    "contextual-agent-sdk": "simcoder technologies - https://github.com/simcoder/contextual-agent-sdk"
  }
}

Programmatic Attribution (Easy Way):

import { getAttribution, getHTMLAttribution, logAttribution } from 'contextual-agent-sdk';

// Get attribution info
const attribution = getAttribution();
console.log(attribution.author); // "simcoder technologies a subsidiary of SCS Group"

// Generate HTML for your app
const htmlCredit = getHTMLAttribution();
// Returns: "Powered by <a href="...">Contextual Agent SDK</a> by simcoder technologies"

// Log attribution during development
logAttribution(); // Shows attribution info in console

// For React/Vue components
function AboutPage() {
  const attribution = getAttribution();
  return (
    <div>
      <p>This app uses {attribution.library} by simcoder technologies</p>
      <a href={attribution.url}>Learn more about the technology</a>
    </div>
  );
}

🙏 Why Attribution Matters

This SDK represents significant innovation in AI agent technology. Attribution helps:

✅ Give credit where credit is due
✅ Help others discover this technology
✅ Support continued development and improvement
✅ Build a community around the innovation

Thank you for using Contextual Agent SDK responsibly!

🌟 Why Choose Contextual Agent SDK?

✅ First SDK to solve seamless voice-text context switching
✅ Production Ready with enterprise-grade features
✅ Provider Agnostic - use any LLM or speech service
✅ TypeScript First - full type safety and great DX
✅ Modular Design - use only what you need
✅ Comprehensive Documentation - examples and guides
✅ Active Development - regular updates and improvements

Ready to build the future of conversational AI? Get started with Contextual Agent SDK today!

📊 Monitoring & Analytics

For SDK Publishers (simcoder technologies)

Track the usage and adoption of your SDK with these tools:

📈 NPM Download Analytics

NPM Download Counts API:

# Get daily downloads
curl https://api.npmjs.org/downloads/point/last-day/contextual-agent-sdk

# Get weekly downloads  
curl https://api.npmjs.org/downloads/point/last-week/contextual-agent-sdk

# Get monthly downloads
curl https://api.npmjs.org/downloads/point/last-month/contextual-agent-sdk

# Get historical data
curl https://api.npmjs.org/downloads/range/2024-01-01:2024-12-31/contextual-agent-sdk

Third-Party Analytics Tools:
- npm-stat.com - Historical download statistics
- npmtrends.com - Compare download trends
- pkgstats.com - Package discovery and stats
- npm-pack-count - Automated download tracking

🔍 GitHub Repository Analytics

Built-in GitHub Insights:
- Repository > Insights > Traffic (14-day data retention)
- Stars history and Fork tracking
- Clone statistics and Visitor analytics
Extended Analytics Tools:
- YHYPE - Extended GitHub analytics (>14 days)
- OSS Insight - Deep repository insights
- GitHub Traffic Dashboard - Custom traffic analytics
- u8views.com - Profile view tracking

Profile View Tracking:

<!-- Add to your GitHub profile README -->
![Profile Views](https://u8views.com/api/v1/github/profiles/YOUR-USER-ID/views/day-week-month-total-count.svg)

⚖️ Attribution Compliance Monitoring

Automated License Scanning:
- SCANOSS - 100% open source SCA tool
- FOSSology - License compliance toolkit
- ScanCode - Industry-leading code scanner
Code Attribution Detection:
- Vendetect - Detect copied/vendored code
- Custom GitHub search for attribution mentions
- Google search alerts for your SDK name

📡 Built-in Telemetry (Privacy-Compliant)

The SDK includes optional, privacy-first telemetry:

import { reportUsage } from 'contextual-agent-sdk/utils/attribution';

// Optional usage reporting (production only, user-consented)
await reportUsage({
  projectName: 'my-chatbot',
  environment: 'production',
  features: ['voice-switching', 'context-management'],
  disableTelemetry: false // Set to true to disable
});

Privacy Features:

✅ Opt-in only - Disabled by default
✅ Production only - Never runs in development
✅ Anonymous - No personal data collected
✅ Non-blocking - Silently fails without affecting app
✅ Transparent - Full source code available

For SDK Users (Ensuring Compliance)

🏷️ Attribution Helpers

Use built-in utilities to ensure proper attribution:

import { 
  getAttribution, 
  getHTMLAttribution, 
  getMarkdownAttribution,
  getAttributionBadge,
  validateAttribution 
} from 'contextual-agent-sdk/utils/attribution';

// Get attribution info
const attribution = getAttribution();
console.log(attribution.author); // "simcoder technologies a subsidiary of SCS Group"

// Generate HTML for your app
const htmlCredit = getHTMLAttribution();
// Returns: "Powered by <a href="...">Contextual Agent SDK</a> by simcoder technologies"

// Add badge to README
const badge = getAttributionBadge('markdown');
// Returns: [![Powered by Contextual Agent SDK](https://img.shields.io/badge/...)](https://github.com/simcoder/contextual-agent-sdk)

// Validate your package.json includes proper attribution
const validation = validateAttribution(require('./package.json'));
if (!validation.isValid) {
  console.log('Attribution suggestions:', validation.suggestions);
}

// Log attribution during development
logAttribution();

🎯 Easy Attribution Examples

React Component:

function AboutPage() {
  const attribution = getAttribution();
  return (
    <div>
      <p>This app uses {attribution.library} by simcoder technologies</p>
      <a href={attribution.url}>Learn more about the technology</a>
    </div>
  );
}

Package.json Credits:

{
  "credits": {
    "contextual-agent-sdk": "simcoder technologies - https://github.com/simcoder/contextual-agent-sdk"
  }
}

🔐 License Enforcement Strategy

Legal Protection

✅ MIT License with Attribution Clause - Legally enforceable
✅ Clear attribution requirements in LICENSE file
✅ Multiple attribution options (flexible compliance)

Technical Monitoring

✅ Automated scanning with tools like SCANOSS and ScanCode
✅ GitHub search alerts for usage without attribution
✅ Download analytics to track adoption
✅ Code fingerprinting to detect unauthorized copies

Community Enforcement

✅ Clear documentation makes compliance easy
✅ Helpful tools reduce friction for proper attribution
✅ Community reporting via GitHub issues
✅ Professional follow-up for violations

🎯 Recommended Monitoring Setup

For Comprehensive Tracking:

Set up NPM download monitoring:

npm install -g npm-download-stats
npm-download-stats contextual-agent-sdk

Enable GitHub repository insights:
- Go to Settings > Features > Insights
- Enable Traffic and Community analytics
Use YHYPE for extended analytics:
- Sign up at yhype.me
- Connect your GitHub account
- Track repository metrics beyond 14 days

Set up Google Alerts:

"contextual-agent-sdk" -site:github.com -site:npmjs.com
"simcoder technologies" "AI agent"

Deploy license scanning:

# Install SCANOSS
pip install scanoss

# Scan for your code usage
scanoss scan /path/to/suspect/repository

🏷️ Attribution Requirements

If you use this SDK in your project, please provide attribution to simcoder technologies a subsidiary of SCS Group as the creator. You can do this in one of the following ways:

In Your Application:

<!-- In your app footer or about section -->
Powered by <a href="https://github.com/simcoder/contextual-agent-sdk">Contextual Agent SDK</a> by simcoder technologies

In Your README:

## Credits
- Uses [Contextual Agent SDK](https://github.com/simcoder/contextual-agent-sdk) by simcoder technologies for seamless voice-text AI interactions

In Package.json:

{
  "credits": {
    "contextual-agent-sdk": "simcoder technologies - https://github.com/simcoder/contextual-agent-sdk"
  }
}

Programmatic Attribution (Easy Way):

import { getAttribution, getHTMLAttribution, logAttribution } from 'contextual-agent-sdk';

// Get attribution info
const attribution = getAttribution();
console.log(attribution.author); // "simcoder technologies a subsidiary of SCS Group"

// Generate HTML for your app
const htmlCredit = getHTMLAttribution();
// Returns: "Powered by <a href="...">Contextual Agent SDK</a> by simcoder technologies"

// Log attribution during development
logAttribution(); // Shows attribution info in console

// For React/Vue components
function AboutPage() {
  const attribution = getAttribution();
  return (
    <div>
      <p>This app uses {attribution.library} by simcoder technologies</p>
      <a href={attribution.url}>Learn more about the technology</a>
    </div>
  );
}

🙏 Why Attribution Matters

This SDK represents significant innovation in AI agent technology. Attribution helps:

✅ Give credit where credit is due
✅ Help others discover this technology
✅ Support continued development and improvement
✅ Build a community around the innovation

Thank you for using Contextual Agent SDK responsibly!

🌟 Why Choose Contextual Agent SDK?

✅ First SDK to solve seamless voice-text context switching
✅ Production Ready with enterprise-grade features
✅ Provider Agnostic - use any LLM or speech service
✅ TypeScript First - full type safety and great DX
✅ Modular Design - use only what you need
✅ Comprehensive Documentation - examples and guides
✅ Active Development - regular updates and improvements

Ready to build the future of conversational AI? Get started with Contextual Agent SDK today!

This comprehensive monitoring approach ensures you can track usage, enforce attribution, and protect your intellectual property while maintaining the open-source nature of your SDK! 🚀

Name		Name	Last commit message	Last commit date
Latest commit History 6 Commits
examples		examples
scripts		scripts
src		src
tests		tests
.env.example		.env.example
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md
package-lock.json		package-lock.json
package.json		package.json
test-runner.js		test-runner.js
tsconfig.json		tsconfig.json

License

simcoder/contextual-agent-sdk

Folders and files

Latest commit

History

Repository files navigation

Contextual Agent SDK

🚀 Key Features

Core Capabilities

Production Ready

🎯 The Innovation: Context Bridging

📦 Installation

🚀 Quick Start

Basic Setup

Advanced Setup with All Features

🧠 Context Management

Auto-Discovery of Documentation

Custom Context Providers

🎙️ Speech Integration

Multiple STT/TTS Providers

🤖 Multi-LLM Support

Provider Management

💾 Session Storage

Multiple Storage Options

📊 Event System & Monitoring

Real-time Event Monitoring

🔧 Configuration

Environment Variables

TypeScript Configuration

🏗️ Architecture

🧪 Testing

📚 Examples & Demos

🔗 Use Cases

Customer Support

Educational Assistant

Code Assistant

🚀 Production Deployment

Docker Setup

Environment Configuration

🤝 Contributing

📄 License & Attribution

🏷️ Attribution Requirements

In Your Application:

In Your README:

In Package.json:

Programmatic Attribution (Easy Way):

🙏 Why Attribution Matters

🌟 Why Choose Contextual Agent SDK?

📊 Monitoring & Analytics

For SDK Publishers (simcoder technologies)

📈 NPM Download Analytics

🔍 GitHub Repository Analytics

⚖️ Attribution Compliance Monitoring

📡 Built-in Telemetry (Privacy-Compliant)

For SDK Users (Ensuring Compliance)

🏷️ Attribution Helpers

🎯 Easy Attribution Examples

🔐 License Enforcement Strategy

Legal Protection

Technical Monitoring

Community Enforcement

🎯 Recommended Monitoring Setup

For Comprehensive Tracking:

🏷️ Attribution Requirements

In Your Application:

In Your README:

In Package.json:

Programmatic Attribution (Easy Way):

🙏 Why Attribution Matters

🌟 Why Choose Contextual Agent SDK?

About

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Languages

Packages