Configuration

You configure the package with environment variables in .env. The only thing you need is the API key of at least one provider, and AI_GENERATOR_DRIVER when that provider is not OpenAI. The wizard, php artisan ai-generator:install, writes these for you; see Installation.

General settings

Config key	Environment variable	Default	What it does
default_language	AI_GENERATOR_LANGUAGE	nl	Language of a request without language, as an ISO 639-1 code.
defaults.max_words	AI_GENERATOR_MAX_WORDS	900	Word limit of a request without maxWords.
defaults.reading_level	AI_GENERATOR_LEVEL	general	Reading level of a request without readingLevel: simple, general or expert.
defaults.tone	AI_GENERATOR_TONE	informal	Tone of a request without tone: informal, neutral or formal.
driver	AI_GENERATOR_DRIVER	openai	The driver that writes the text: openai, anthropic, gemini or xai, or an alias (chatgpt, claude, google, grok). See Multiple providers.
fallbacks	AI_GENERATOR_FALLBACKS	empty	Drivers to try in order when the text call fails, comma separated, for example openai,gemini.
image_driver	AI_GENERATOR_IMAGE_DRIVER	empty	The driver that makes images: openai, gemini or xai. Empty follows the text driver, and uses OpenAI when the text driver cannot make images.
mcp.enabled	AI_GENERATOR_MCP	true	Register the local MCP server when laravel/mcp is installed.
mcp.handle	AI_GENERATOR_MCP_HANDLE	ai-generator	The name of that server, for php artisan mcp:start.

The default language is nl, Dutch. Set AI_GENERATOR_LANGUAGE=en when your site is in English.

Settings per provider

Every provider has its own block under drivers. You only fill in the providers you use.

Setting	OpenAI	Claude (Anthropic)	Gemini (Google)	Grok (xAI)
API key	OPENAI_API_KEY	ANTHROPIC_API_KEY	GEMINI_API_KEY	XAI_API_KEY
Text model	OPENAI_MODEL, default gpt-4.1-mini	ANTHROPIC_MODEL, default claude-sonnet-5	GEMINI_MODEL, default gemini-3.8-flash	XAI_MODEL, default grok-4.7
Image model	OPENAI_IMAGE_MODEL, default gpt-image-2	none, Claude makes no images	GEMINI_IMAGE_MODEL, default gemini-3.1-flash-image	XAI_IMAGE_MODEL, default grok-imagine-image-2.0
Temperature	OPENAI_TEMPERATURE, default 0.7	ANTHROPIC_TEMPERATURE, default none	GEMINI_TEMPERATURE, default none	XAI_TEMPERATURE, default none
Timeout in seconds	OPENAI_TIMEOUT, default 45	ANTHROPIC_TIMEOUT, default 45	GEMINI_TIMEOUT, default 45	XAI_TIMEOUT, default 45
Base URL	OPENAI_BASE_URL, default https://api.openai.com/v1	ANTHROPIC_BASE_URL, default https://api.anthropic.com/v1	GEMINI_BASE_URL, default https://generativelanguage.googleapis.com/v1beta	XAI_BASE_URL, default https://api.x.ai/v1

Claude has two more settings:

Config key	Environment variable	Default	What it does
drivers.anthropic.max_tokens	ANTHROPIC_MAX_TOKENS	8192	The maximum length of Claude’s answer, in tokens. Claude requires a limit.
drivers.anthropic.version	ANTHROPIC_VERSION	2023-06-01	The value of the anthropic-version header.

Temperature. An empty temperature sends none. The newest Claude models reject a temperature, so none is sent to Claude, Gemini or Grok unless you set one. OpenAI gets 0.7. OpenAI reasoning models reject a temperature too; set OPENAI_TEMPERATURE= (empty) when you use one.

Timeout. It counts per attempt, and both requests inside generate() (text and image) are tried twice. generateImage() does not use it: that method has a fixed timeout of 120 seconds and tries once.

Models age. Providers retire models. OpenAI shuts down gpt-image-1 on 2026-10-23, so since 1.3.0 the default image model is gpt-image-2, the replacement OpenAI names. Is OPENAI_IMAGE_MODEL set to gpt-image-1 in your .env? Change it to gpt-image-2, or remove the line to follow the default. The wizard lists the image models your key may use.

An example .env

# The keys of the providers you use
ANTHROPIC_API_KEY=your-claude-key
OPENAI_API_KEY=your-openai-key

# Who writes, who draws, who takes over
AI_GENERATOR_DRIVER=anthropic
AI_GENERATOR_IMAGE_DRIVER=openai
AI_GENERATOR_FALLBACKS=openai

# Optional: models
ANTHROPIC_MODEL=claude-sonnet-5
OPENAI_MODEL=gpt-4.1-mini
OPENAI_IMAGE_MODEL=gpt-image-2

# Optional: defaults for a request
AI_GENERATOR_LANGUAGE=en
AI_GENERATOR_TONE=neutral
AI_GENERATOR_LEVEL=general
AI_GENERATOR_MAX_WORDS=600

After a change in .env on a server that caches its config, run php artisan config:clear or php artisan config:cache again. Until then the application keeps the old values.

The config file

You don’t need the file in your application. Publish it when you want to set a value in PHP:

php artisan vendor:publish --tag=ai-generator-config

You get config/ai-generator.php, with the keys in alphabetical order within every group:

return [

    'default_language' => env('AI_GENERATOR_LANGUAGE', 'nl'),

    'defaults' => [
        'max_words' => env('AI_GENERATOR_MAX_WORDS', 900),
        'reading_level' => env('AI_GENERATOR_LEVEL', 'general'),
        'tone' => env('AI_GENERATOR_TONE', 'informal'),
    ],

    'driver' => env('AI_GENERATOR_DRIVER', 'openai'),

    'drivers' => [
        'anthropic' => [
            'api_key' => env('ANTHROPIC_API_KEY'),
            'base_url' => env('ANTHROPIC_BASE_URL', 'https://api.anthropic.com/v1'),
            'max_tokens' => env('ANTHROPIC_MAX_TOKENS', 8192),
            'model' => env('ANTHROPIC_MODEL', 'claude-sonnet-5'),
            'temperature' => env('ANTHROPIC_TEMPERATURE'),
            'timeout' => env('ANTHROPIC_TIMEOUT', 45),
            'version' => env('ANTHROPIC_VERSION', '2023-06-01'),
        ],
        'gemini' => [
            'api_key' => env('GEMINI_API_KEY'),
            'base_url' => env('GEMINI_BASE_URL', 'https://generativelanguage.googleapis.com/v1beta'),
            'image_model' => env('GEMINI_IMAGE_MODEL', 'gemini-3.1-flash-image'),
            'model' => env('GEMINI_MODEL', 'gemini-3.8-flash'),
            'temperature' => env('GEMINI_TEMPERATURE'),
            'timeout' => env('GEMINI_TIMEOUT', 45),
        ],
        'openai' => [
            'api_key' => env('OPENAI_API_KEY'),
            'base_url' => env('OPENAI_BASE_URL', 'https://api.openai.com/v1'),
            'image_model' => env('OPENAI_IMAGE_MODEL', 'gpt-image-2'),
            'model' => env('OPENAI_MODEL', 'gpt-4.1-mini'),
            'temperature' => env('OPENAI_TEMPERATURE', 0.7),
            'timeout' => env('OPENAI_TIMEOUT', 45),
        ],
        'xai' => [
            'api_key' => env('XAI_API_KEY'),
            'base_url' => env('XAI_BASE_URL', 'https://api.x.ai/v1'),
            'image_model' => env('XAI_IMAGE_MODEL', 'grok-imagine-image-2.0'),
            'model' => env('XAI_MODEL', 'grok-4.7'),
            'temperature' => env('XAI_TEMPERATURE'),
            'timeout' => env('XAI_TIMEOUT', 45),
        ],
    ],

    'fallbacks' => env('AI_GENERATOR_FALLBACKS', ''),

    'image_driver' => env('AI_GENERATOR_IMAGE_DRIVER'),

    'mcp' => [
        'enabled' => env('AI_GENERATOR_MCP', true),
        'handle' => env('AI_GENERATOR_MCP_HANDLE', 'ai-generator'),
    ],

];

A value you write into this file instead of env(...) wins over .env. Keep that in mind when a change in .env seems to do nothing.

A config file you published with an older version keeps working: no key was renamed or removed. A provider block that is missing from your file falls back to the defaults of the package. Publish again (with --force, after saving your changes) to see the new blocks.

A request option beats the config

The config only fills what a request leaves empty. Pass the value in the request when it differs per text:

use Darvis\LaravelAiGenerator\AiGenerator;
use Darvis\LaravelAiGenerator\ContentRequest;

$result = app(AiGenerator::class)->generate(new ContentRequest(
    topic: 'Your topic',
    tone: 'formal',
    readingLevel: 'expert',
    includeImage: false,
));

Read a setting in your own code

The package reads its config in one class, Darvis\LaravelAiGenerator\Support\AiGeneratorConfig. Use the same class when your code needs a package setting, so the default is the same everywhere:

use Darvis\LaravelAiGenerator\Support\AiGeneratorConfig;
use Darvis\LaravelAiGenerator\Support\Provider;

AiGeneratorConfig::defaultLanguage();            // 'nl'
AiGeneratorConfig::model(Provider::Anthropic);   // 'claude-sonnet-5'
AiGeneratorConfig::apiKey(Provider::OpenAi);     // null when the key is empty or not set

The API reference lists every accessor.

Another endpoint

<PREFIX>_BASE_URL sends the requests of a provider to another host, for example a proxy of your own. The endpoint has to accept exactly what the provider accepts:

• OpenAI: /responses for the text and /images/generations for the image, with the key as an Authorization: Bearer header.
• Claude: /messages, with the key in an x-api-key header.
• Gemini: /models/<model>:generateContent, with the key in an x-goog-api-key header.
• Grok: /chat/completions and /images/generations, with the key as a Bearer token.

A trailing slash in the URL is removed.