README

A Laravel package providing a unified, chainable interface for multiple LLM (Large Language Model) providers: OpenAI, Anthropic (Claude), Gemini, and Ollama.

Requirements

• Laravel 10.x, 11.x, or 12.x (This package only supports Laravel 10 and above)
• PHP 8.1 or higher

Tested on Laravel 12.x. Other versions may work, but are not officially tested.

Installation

composer require lzhx00/laravel-llm-client

Laravel will auto-discover and register the package.
If you have disabled auto-discovery, add the following to config/app.php:

'providers' => [
    // ...
    Lzhx00\LLMClient\LLMClientServiceProvider::class,
],

'aliases' => [
    // ...
    'LLMClient' => Lzhx00\LLMClient\Facades\LLMClient::class,
],

Configuration

Publish the config file (optional, for customization):

php artisan vendor:publish --tag=llm-client-config

Set your API keys and provider settings in .env or config/llm.php:

LLM_DEFAULT_PROVIDER=openai
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GOOGLE_API_KEY=AIza...
OLLAMA_BASE_URL=http://localhost:11434

config/llm.php Example

Each provider has its own default_model, embedding_model, and options.

return [
    'default' => env('LLM_DEFAULT_PROVIDER', 'openai'),
    'providers' => [
        'openai' => [
            'api_key' => env('OPENAI_API_KEY'),
            'default_model' => 'gpt-3.5-turbo',
            'embedding_model' => 'text-embedding-3-small',
            'options' => [
                'temperature' => 0.7,
                // ...other OpenAI-specific options
            ],
        ],
        'ollama' => [
            'base_url' => env('OLLAMA_BASE_URL', 'http://localhost:11434'),
            'default_model' => 'llama3',
            'embedding_model' => 'nomic-embed-text',
            'options' => [
                'temperature' => 0.5,
                // ...other Ollama-specific options
            ],
        ],
        // ...other providers
    ],
];

Usage

Basic Text Generation

$response = LLMClient::generate('Say hello in English.');

Specify Provider

$response = LLMClient::use('ollama')->generate('Say hello in English.');

Chainable Model/Options (Recommended)

$response = LLMClient::model('llama3')->with(['temperature' => 0.5])->generate('Say hello.');

• model() only affects chat/completion.
• embedModel() only affects embed.
• with() sets provider-specific options (except model/embeddingModel).

Embeddings

$vector = LLMClient::use('ollama')->embed('hello world');

// Specify embedding model 
$vector = LLMClient::embedModel('nomic-embed-text')->embed('hello world');

Streaming Response

LLMClient::generateStream('Tell me a joke.', [], function($chunk) {
    echo $chunk;
});

List Models

$models = LLMClient::use('gemini')->models();

Supported Providers

• OpenAI (ChatGPT)
• Anthropic (Claude)
• Gemini (Google)
• Ollama

⚠️ Note: Only the Ollama provider has been fully tested.
Other providers are implemented based on official docs, but not tested with real API keys.

📄 License

MIT License