Ollama-Laravel Package

Ollama-Laravel is a Laravel package that provides a seamless integration with the Ollama API. It includes functionalities for model management, prompt generation, format setting, and more. This package is perfect for developers looking to leverage the power of the Ollama API in their Laravel applications.

If you use laravel 10.x, please use the following version V1.0.9

https://github.com/cloudstudio/ollama-laravel/releases/tag/v1.0.9

Installation

composer require cloudstudio/ollama-laravel

Configuration

php artisan vendor:publish --tag="ollama-laravel-config"

Published config file:

return [
    'model' => env('OLLAMA_MODEL', 'llama2'),
    'url' => env('OLLAMA_URL', 'http://127.0.0.1:11434'),
    'default_prompt' => env('OLLAMA_DEFAULT_PROMPT', 'Hello, how can I assist you today?'),
    'connection' => [
        'timeout' => env('OLLAMA_CONNECTION_TIMEOUT', 300),
        'verify_ssl' => env('OLLAMA_VERIFY_SSL', true),
        'auth' => [
            'type' => env('OLLAMA_AUTH_TYPE', null), // 'basic' or 'bearer'
            'username' => env('OLLAMA_AUTH_USERNAME', null),
            'password' => env('OLLAMA_AUTH_PASSWORD', null),
            'token' => env('OLLAMA_AUTH_TOKEN', null),
        ],
        'headers' => [
            // Add any custom headers here
            // 'X-Custom-Header' => 'value',
        ],
    ],
];

Authentication

Note: The Ollama server itself does not include built-in authentication. The authentication features in this package are designed to work with reverse proxy setups (like Nginx, Caddy, or Apache) that add authentication in front of your Ollama server.

The package supports both Basic Authentication and Bearer Token Authentication when accessing Ollama through a reverse proxy. This is particularly useful when:

• Deploying Ollama in a production environment
• Securing access to your Ollama instance
• Running Ollama behind a corporate firewall

Common reverse proxy setups:

• Nginx with basic auth or token validation
• Caddy with authentication middleware
• Apache with auth modules

Basic Authentication

To use Basic Authentication with your reverse proxy, set the following environment variables:

OLLAMA_AUTH_TYPE=basic
OLLAMA_AUTH_USERNAME=your_username
OLLAMA_AUTH_PASSWORD=your_password

Bearer Token Authentication

To use Bearer Token Authentication, set the following environment variables:

OLLAMA_AUTH_TYPE=bearer
OLLAMA_AUTH_TOKEN=your_bearer_token

Custom Headers

You can add custom headers in the configuration file:

'headers' => [
    'X-Custom-Header' => 'value',
    'Another-Header' => 'another-value',
],

Usage

Basic Usage

use Cloudstudio\Ollama\Facades\Ollama;

/** @var array $response */
$response = Ollama::agent('You are a weather expert...')
    ->prompt('Why is the sky blue?')
    ->model('llama2')
    ->options(['temperature' => 0.8])
    ->stream(false)
    ->ask();

Vision Support

/** @var array $response */
$response = Ollama::model('llava:13b')
    ->prompt('What is in this picture?')
    ->image(public_path('images/example.jpg')) 
    ->ask();

// "The image features a close-up of a person's hand, wearing bright pink fingernail polish and blue nail polish. In addition to the colorful nails, the hand has two tattoos – one is a cross and the other is an eye."

Chat Completion

$messages = [
    ['role' => 'user', 'content' => 'My name is Toni Soriano and I live in Spain'],
    ['role' => 'assistant', 'content' => 'Nice to meet you , Toni Soriano'],
    ['role' => 'user', 'content' => 'where I live ?'],
];

$response = Ollama::agent('You know me really well!')
    ->model('llama2')
    ->chat($messages);

// "You mentioned that you live in Spain."

### Chat Completion

Chat Completion with tools

$messages = [
    ['role' => 'user', 'content' => 'What is the weather in Toronto?'],
];

$response = Ollama::model('llama3.1')
    ->tools([
        [
            "type"     => "function",
            "function" => [
                "name"        => "get_current_weather",
                "description" => "Get the current weather for a location",
                "parameters"  => [
                    "type"       => "object",
                    "properties" => [
                        "location" => [
                            "type"        => "string",
                            "description" => "The location to get the weather for, e.g. San Francisco, CA",
                        ],
                        "format"   => [
                            "type"        => "string",
                            "description" => "The format to return the weather in, e.g. 'celsius' or 'fahrenheit'",
                            "enum"        => ["celsius", "fahrenheit"],
                        ],
                    ],
                    "required"   => ["location", "format"],
                ],
            ],
        ],
    ])
    ->chat($messages);

Streamable responses

use Cloudstudio\Ollama\Facades\Ollama;
use Illuminate\Console\BufferedConsoleOutput;

/** @var \GuzzleHttp\Psr7\Response $response */
$response = Ollama::agent('You are a snarky friend with one-line responses')
    ->prompt("I didn't sleep much last night")
    ->model('llama3')
    ->options(['temperature' => 0.1])
    ->stream(true)
    ->ask();

$output = new BufferedConsoleOutput();
$responses = Ollama::processStream($response->getBody(), function($data) use ($output) {
    $output->write($data['response']);
});

$output->write("\n");
$complete = implode('', array_column($responses, 'response'));
$output->write("<info>$complete</info>");