🚀 Milvus PHP SDK - Modern Vector Database Client

A modern, type-safe PHP SDK for Milvus vector database API v2. This library provides a clean, intuitive interface for managing collections, users, roles, and privileges in Milvus through its REST API.

Built with Saloon HTTP for robust API communication, this SDK focuses on developer experience with proper type hints, comprehensive error handling, and a fluent API design.

Perfect for applications requiring vector similarity search, AI/ML workflows, and large-scale data processing with semantic search capabilities.

Requirements

• PHP 8.1 or higher
• Milvus 2.6.x (or compatible versions)
• Laravel 10.x, 11.x, 12.x (for Laravel integration)

Versions

| Milvus Version | SDK Version | |----------------|-------------| | v2.6.x | v1.0.x |

Documentation

• Milvus REST API Reference - Official Milvus RESTful API documentation

Installation

You can install the package via Composer:

composer require voyanara/milvus-php-sdk

Docker Development Environment

For development and testing, you can quickly spin up a Milvus instance using the included Docker Compose configuration:

docker-compose up -d

This will start Milvus with all necessary dependencies (etcd, MinIO) and expose it on the default port 19530.

Laravel Integration

This package includes Laravel service provider for seamless integration:

# After installation, publish the configuration file
php artisan milvus-php-sdk:install

This command will publish the configuration file to config/milvus-php-sdk.php where you can set your Milvus connection parameters.

Configuration

Add the following environment variables to your .env file:

Roadmap

✅ Implemented Features

• User Management - Complete user operations (create, describe, drop, list, update password)
• Role Management - Full role-based access control (create, drop, describe, list, grant/revoke privileges)
• Collection Management - Complete collection operations and schema management (create, drop, describe, list, load, release, rename, etc.)
• Vector Operations - Complete vector data operations (insert, search, query, upsert, get, delete, hybrid search)

🚧 In Development

• Index Management - Vector index creation and optimization

📋 Planned Features

• Alias Management - Collection alias operations
• Database Management - Multi-database support
• Import Operations - Bulk data import functionality
• Partition Management - Data partitioning for better performance
• Resource Group - Resource allocation and management

Quick Start

Basic Usage

<?php

use Voyanara\MilvusSdk\Milvus;

// Initialize the client
$milvus = new Milvus(
    token: 'your_token_here',
    host: 'http://localhost',
    port: '19530'
);

// Example: List all users
$users = $milvus->user()->list();

Using with Zilliz Cloud

For Zilliz Cloud (hosted Milvus), configure the client with your cloud credentials:

<?php

use Voyanara\MilvusSdk\Milvus;

$milvus = new Milvus(
    token: "db_randomstring:your_password",
    host: 'https://in03.serverless.gcp-us-west1.cloud.zilliz.com',
    port: '443'
);

// Now you can use all SDK features with Zilliz Cloud
$collections = $milvus->collection()->list();

Using with Laravel

Once configured, you can use the Milvus facade throughout your Laravel application:

<?php

use Voyanara\MilvusSdk\Facades\Milvus;

// User management
$users = Milvus::user()->list();
$user = Milvus::user()->describe('username');

// Role management  
$roles = Milvus::role()->list();
$role = Milvus::role()->describe('role_name');

// Collection management
$collections = Milvus::collection()->list();
$collection = Milvus::collection()->describe('collection_name');

// Create a new collection with schema and index
$schema = [
    'fields' => [
        [
            'fieldName' => 'id',
            'dataType' => 'Int64', 
            'isPrimary' => true
        ],
        [
            'fieldName' => 'vector',
            'dataType' => 'FloatVector',
            'elementTypeParams' => ['dim' => '128']
        ]
    ]
];

$indexParams = [
    [
        'fieldName' => 'vector',
        'indexName' => 'vector_index', 
        'metricType' => 'L2'
    ]
];

Milvus::collection()->createCollection('my_collection', $schema, $indexParams);

// Vector operations
// Insert vector data
$vectorData = [
    [
        'id' => 1,
        'vector' => [0.1, 0.2, 0.3, 0.4, 0.5],
        'metadata' => 'document1'
    ],
    [
        'id' => 2, 
        'vector' => [0.6, 0.7, 0.8, 0.9, 1.0],
        'metadata' => 'document2'
    ]
];

Milvus::vector()->insert('my_collection', $vectorData);

// Search for similar vectors
$queryVector = [[0.1, 0.2, 0.3, 0.4, 0.5]];
$searchResults = Milvus::vector()->search(
    collectionName: 'my_collection',
    data: $queryVector,
    annsField: 'vector',
    limit: 10,
    outputFields: ['id', 'metadata']
);

// Upsert (insert or update) vector data
$upsertData = [
    [
        'id' => 1,
        'vector' => [0.2, 0.3, 0.4, 0.5, 0.6], // Updated vector
        'metadata' => 'document1_updated'
    ]
];

Milvus::vector()->upsert('my_collection', $upsertData);

Vector Operations Examples

Vector Search

Perform semantic search using vector embeddings:

<?php

use Voyanara\MilvusSdk\Milvus;

// Initialize Milvus client
$milvus = new Milvus(
    token: 'root:Milvus',
    host: 'http://localhost',
    port: '19530'
);

// Prepare query vectors (can be multiple vectors)
$queryVectors = [
    [0.3580376395471989, -0.6023495712049978, 0.18414012509913835],
    [0.19886812562848388, 0.06023560599112088, 0.6976963061752597]
];

// Basic vector search
$response = $milvus->vector()->search(
    collectionName: 'documents_collection',
    data: $queryVectors,
    annsField: 'content_vector',
    limit: 5,
    outputFields: ['id', 'title', 'category']
);

// Search with filtering
$response = $milvus->vector()->search(
    collectionName: 'documents_collection', 
    data: $queryVectors,
    annsField: 'content_vector',
    filter: "category == 'technology' and publish_date >= '2024-01-01'",
    limit: 10,
    outputFields: ['id', 'title', 'content']
);

// Advanced search with custom parameters
$searchParams = [
    'metricType' => 'L2',
    'params' => [
        'radius' => 0.1,
        'range_filter' => 0.9
    ]
];

$response = $milvus->vector()->search(
    collectionName: 'documents_collection',
    data: $queryVectors,
    annsField: 'content_vector', 
    searchParams: $searchParams,
    limit: 20,
    offset: 10  // Pagination support
);

// Process results
foreach ($response->json('data') as $result) {
    echo "Document ID: {$result['id']}, Distance: {$result['distance']}\n";
    echo "Title: {$result['title']}\n";
}

Vector Upsert

Insert new vectors or update existing ones based on primary key:

<?php

use Voyanara\MilvusSdk\Milvus;

// Initialize Milvus client
$milvus = new Milvus(
    token: 'root:Milvus',
    host: 'http://localhost', 
    port: '19530'
);

// Prepare document vectors for upsert
$documents = [
    [
        'id' => 1,
        'content_vector' => [0.1, 0.2, 0.3, 0.4, 0.5],
        'title' => 'Introduction to AI',
        'category' => 'technology',
        'publish_date' => '2024-01-15'
    ],
    [
        'id' => 2,
        'content_vector' => [0.6, 0.7, 0.8, 0.9, 1.0], 
        'title' => 'Machine Learning Basics',
        'category' => 'technology',
        'publish_date' => '2024-02-01'
    ],
    [
        'id' => 3,
        'content_vector' => [0.2, 0.4, 0.6, 0.8, 0.1],
        'title' => 'Deep Learning Guide', 
        'category' => 'technology',
        'publish_date' => '2024-03-10'
    ]
];

// Upsert documents (will insert new or update existing based on ID)
$response = $milvus->vector()->upsert(
    collectionName: 'documents_collection',
    data: $documents
);

echo "Upserted {$response->json('data.upsertCount')} documents\n";
print_r($response->json('data.upsertIds'));

// Upsert to specific partition
$response = $milvus->vector()->upsert(
    collectionName: 'documents_collection',
    data: $documents,
    partitionName: 'tech_partition'
);

// Single document upsert
$singleDocument = [
    [
        'id' => 100,
        'content_vector' => [0.3, 0.1, 0.4, 0.7, 0.2],
        'title' => 'Updated Document Title',
        'category' => 'science'
    ]
];

$milvus->vector()->upsert('documents_collection', $singleDocument);

Creating a Milvus Collection

Quick example of creating a collection with vector field:

<?php

use Voyanara\MilvusSdk\Milvus;

// Initialize client
$milvus = new Milvus(
    token: 'root:Milvus',
    host: 'http://localhost',
    port: '19530'
);

// Define collection schema
$schema = [
    'fields' => [
        [
            'fieldName' => 'id',
            'dataType' => 'Int64',
            'isPrimary' => true
        ],
        [
            'fieldName' => 'title',
            'dataType' => 'VarChar',
            'elementTypeParams' => ['max_length' => 200]
        ],
        [
            'fieldName' => 'content_vector',
            'dataType' => 'FloatVector',
            'elementTypeParams' => ['dim' => 768] // 768-dimensional vectors
        ]
    ]
];

// Define vector index
$indexParams = [
    [
        'fieldName' => 'content_vector',
        'indexName' => 'content_vector_index',
        'metricType' => 'L2'
    ]
];

// Create collection
$response = $milvus->collection()->createCollection(
    collectionName: 'my_documents',
    schema: $schema,
    indexParams: $indexParams
);

// Load collection into memory for operations
$milvus->collection()->loadCollection('my_documents');

echo "Collection 'my_documents' created successfully!\n";

License

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