Overpass

A robust Laravel package for integrating Python AI capabilities through a secure subprocess bridge. Instead of reimplementing complex AI libraries in PHP, this package provides a clean interface to delegate AI operations to Python's rich ecosystem (OpenAI, LangChain, scikit-learn, etc.).

🎬 Demo Application

Check out Ask-My-Doc - a sophisticated document Q&A system built with Overpass that showcases real-world AI integration:

Ask-My-Doc demonstrates Overpass in action with:

• 📄 Document ingestion and intelligent chunking
• 🔍 Semantic search using OpenAI embeddings
• 💬 AI-powered Q&A with source citations
• ⚡ Real-time Python bridge status monitoring
• 🎨 Beautiful dark-themed UI inspired by Linear

Note: This package is not affiliated with the existing decodelabs/overpass package, which provides a PHP bridge to Node.js. Our "Overpass" package serves a different purpose - bridging between PHP and Python for AI operations.

🚀 Features

• 🐍 Python Integration: Seamlessly execute Python AI scripts from Laravel
• 🔒 Secure: API keys passed via environment variables, never in command lines
• ⚡ Async Ready: Built for Laravel queues and background processing
• 🛡️ Robust Error Handling: Graceful degradation and comprehensive logging
• 🧪 Production Tested: Battle-tested patterns for reliability
• 🎯 Laravel Native: Follows Laravel conventions with facades, service providers, and Artisan commands

📦 Installation

You can install the package via composer:

composer require bmadigan/overpass

Install the package with example Python scripts:

php artisan overpass:install --with-examples

Publish the config file:

php artisan vendor:publish --tag="overpass-config"

⚙️ Configuration

Add your OpenAI API key to your .env file:

OPENAI_API_KEY=your-openai-api-key
OVERPASS_SCRIPT_PATH=/path/to/your/python/script.py

Configure the package in config/overpass.php:

return [
    'script_path' => env('OVERPASS_SCRIPT_PATH', base_path('overpass-ai/main.py')),
    'timeout' => env('OVERPASS_TIMEOUT', 90),
    'max_output_length' => env('OVERPASS_MAX_OUTPUT', 10000),
    // ... more options
];

🎯 Usage

Basic Usage

use Bmadigan\Overpass\Facades\Overpass;

// Test the connection
$healthCheck = Overpass::testConnection();

// Generate embeddings
$embedding = Overpass::generateEmbedding('Hello, world!');

// Execute custom operations
$result = Overpass::execute('custom_operation', ['data' => 'value']);

Chat Operations

$response = Overpass::chat([
    'message' => 'What is machine learning?',
    'session_id' => 'user-123',
    'context' => ['previous' => 'conversation']
]);

echo $response['response']; // AI-generated response

Vector Search

$results = Overpass::vectorSearch('machine learning concepts', [
    'limit' => 10,
    'threshold' => 0.8
]);

Queue Integration

use Bmadigan\Overpass\Services\PythonAiBridge;

// In a Laravel job
class ProcessAiTask implements ShouldQueue
{
    public function handle(PythonAiBridge $bridge)
    {
        $result = $bridge->analyzeData($this->data);
        // Process result...
    }
}

🐍 Python Script Structure

Your Python script should follow this pattern:

import sys
import json

def health_check(data):
    return {
        'success': True,
        'data': {'status': 'healthy'}
    }

def analyze_data(data):
    # Your AI logic here
    return {
        'success': True,
        'data': {'result': 'analysis complete'}
    }

def main():
    operation = sys.argv[1]
    data = json.loads(sys.argv[2]) if len(sys.argv) > 2 else {}

    operations = {
        'health_check': health_check,
        'analyze_data': analyze_data,
    }

    if operation in operations:
        result = operations[operation](data)
        print(json.dumps(result))
    else:
        print(json.dumps({'success': False, 'error': 'Unknown operation'}))

if __name__ == '__main__':
    main()

🧪 Testing

# Test the bridge connection
php artisan overpass:test

# Run package tests
composer test

# Run tests with coverage
composer test-coverage

📚 Advanced Usage

Custom Operations

// Define custom operations in your Python script
$result = Overpass::execute('sentiment_analysis', [
    'text' => 'This product is amazing!',
    'options' => ['return_confidence' => true]
]);

Error Handling

try {
    $result = Overpass::generateEmbedding($text);
} catch (\Exception $e) {
    Log::error('Embedding generation failed', ['error' => $e->getMessage()]);
    // Implement fallback logic
}

Dependency Injection

class MyService
{
    public function __construct(
        private PythonAiBridge $bridge
    ) {}

    public function processData(array $data): array
    {
        return $this->bridge->execute('process', $data);
    }
}

🔧 Artisan Commands

# Install and setup the package
php artisan overpass:install --with-examples

# Test the bridge connection
php artisan overpass:test --verbose

# Check configuration
php artisan overpass:test --timeout=60

🚨 Troubleshooting

Connection Failures:

• Ensure Python 3 is installed and accessible
• Check that your Python script path is correct
• Verify Python dependencies are installed
• Test your OpenAI API key

Performance Issues:

• Increase timeout for complex operations
• Use Laravel queues for long-running tasks
• Consider output length limits

Memory Issues:

• Set appropriate max_output_length in config
• Use streaming for large datasets
• Monitor Python process memory usage

📊 Performance Considerations

• Timeouts: Set appropriate timeouts based on operation complexity
• Memory: Monitor both PHP and Python memory usage
• Queues: Use Laravel queues for expensive AI operations
• Caching: Cache embedding results and AI responses when appropriate

🔒 Security

• API keys are passed via environment variables, never command line arguments
• All input data is JSON-encoded to prevent injection attacks
• Process isolation ensures Python failures don't crash Laravel
• Comprehensive logging for audit trails

🤝 Contributing

Contributions are welcome! Please see CONTRIBUTING for details.

🐛 Security Vulnerabilities

Please review our security policy on how to report security vulnerabilities.

📄 Credits

• Your Name
• All Contributors

📜 License

The MIT License (MIT). Please see License File for more information.