laravel-openai-vector-search: A PHP repository from MongoDB Developer Relations

About Laravel

Laravel is a web application framework with expressive, elegant syntax. We believe development must be an enjoyable and creative experience to be truly fulfilling. Laravel takes the pain out of development by easing common tasks used in many web projects, such as:

Learning Laravel

Laravel has the most extensive and thorough documentation and video tutorial library of all modern web application frameworks, making it a breeze to get started with the framework.

To leverage MongoDB with Laravel see the following resources

Tour Planner with MongoDB Vector Search

This project demonstrates the use of MongoDB Vector Search in a Laravel backend with a Vue.js frontend to create an intelligent tour planning application. It showcases how to leverage vector search capabilities for enhanced search and recommendation features in travel applications. See the article Introducing the Tour Planner With MongoDB Vector Search

Project Structure

/tour-planner: Laravel backend
/tour-planner/frontend/tour-planner-frontend: Vue.js frontend

Features

City search with vector-based similarity
Points of interest recommendation
Intelligent trip planning using OpenAI integration

Vector Search and Laravel sdk

The points of interest controller search cities by embeddings generated on city - attraction - desc concat, and gets the searched term as $search

$embedding = $this->generateEmbedding($search);
    
    // Use Atlas Search to search for points of interest within a city
    $points = DB::collection('points_of_interest')
    ->raw(
        function ( $collection) use ($embedding) {
            
            return $collection->aggregate([
                    [
                        '$vectorSearch' => [
                            'index' => 'vector_index',
                            'path' => 'embedding',
                            'queryVector' => $embedding,
                            'numCandidates' => 20,
                            'limit' => 5
                            ],
                        ]
                    ,
                    ['$project' => [
                        'name' => 1,
                        'description' => 1,
                        'rating' => 1,
                        'location' => 1,
                        'score' => ['$meta' => 'vectorSearchScore']
                    ]]
                ]
                );
        },
    )->toArray();

Retrieval Augmented Generation and Laravel with OpenAI

Using the relevant context from MongoDB to augment the trip planning JSON response

protected function generatePossibleTrip($cities,$context, $days){
        $result = OpenAI::chat()->create([
            'model' => 'gpt-4o-mini',
            'temperature' => 0,
            'response_format' => ['type' => 'json_object'],
            'messages' => [
                [
                    'role' => 'system',
                    'content' => 'You are a travel agent helping a customer plan a trip to a city. If it will be hard to visit that in the number of days Have a one day plan stating the problem. The customer will provide you with points of interest to visit in json.'
                ],
                [
                    'role' => 'system',
                    'content' => 'take this schema, for flights add orig_airport_code and dest_airport_code: 
    "tripPlan": {
      "destination": [{
        "city": "string",
        "country": "string"
                    }],
      "pointsOfInterest": [
        {
          "name": "string",
          "description": "string",
          "location": {
            "coordinates": [number, number]
          },
          "rating": number
        }
      ],
      // only in relevant direction
      "flights" : [ "src_airport_code": "string",
                "dest_airport_code": "string" 
      ],
      "itinerary": [
        {
          "day": number,
          "destination": "string",
          "activities": [
            {
              "time": "string",
              "activity": "string",
              "duration": "string",
              // if flight
               "src_airport_code": "string",
                dest_airport_code: "string" 
            }
              ...
          ]

        }
          ...
      ]
    }
 '
                ],
                [
                    'role' => 'user',
                    'content' => 'For cities: ' . json_encode($cities) . '| Take this POIs: ' . $context . ' and build a plan for the a trip of ' . $days . 'days. '
                ]
            ]
        ]);

        return $result->choices[0]->message->content;
    
    }

Prerequisites

PHP 8.1+
- MongoDB Extension and prereqisites.
Composer
Node.js and npm
MongoDB Atlas cluster
OpenAI API key

Backend Setup (Laravel)

Navigate to the cloned directory:

cd laravel-openai-vector-search

Install PHP dependencies:

composer install

Copy the .env.example file to .env and configure your environment variables:

OPENAI_API_KEY=your_openai_api_key
DB_URI=your_atlas_uri

Generate an application key:

php artisan key:generate

Run database migrations and seeders (if any):

php artisan db:seed

Create Atlas vector search index on database: trip_planner collection: points_of_interest:

** Index name : vector_index **

{
  "fields": [
    {
      "type": "vector",
      "path": "embedding",
      "numDimensions": 1536,
      "similarity": "cosine"
    }
  ]
}

Start the Laravel development server:

php artisan serve

Frontend Setup (Vue.js)

Navigate to the frontend directory:

cd frontend/trip-planner-frontend
npm install
npm run serve

The application should be visible at http://localhost:8080 .

Disclaimer

Not a MongoDB official product. Use at your own risk

mongodb-developer/laravel-openai-vector-search