Laravel Black Forest Labs (BFL) Package

A professional Laravel wrapper for Black Forest Labs' FLUX API, providing seamless integration for AI-powered image generation and fine-tuning services.

Features

• Multiple Processing Modes: Manual, synchronous, and queue-based job processing
• Complete Model Support: All FLUX models (Pro, Dev, Ultra, Fill, Expand, Canny, Depth, Kontext)
• Fine-tuning Support: Create and manage fine-tuned models with dedicated service
• File Storage Integration: Automatic file downloading and storage with configurable drivers
• Job Tracking: Complete job lifecycle management with database persistence
• Queue Integration: Built-in Laravel queue job for background processing
• Facade Support: Convenient facades for BFL and fine-tuning operations
• Comprehensive Testing: Full test coverage with PHPUnit
• Type-Safe: Full PHP 8.1+ type hints and strict types

Installation

Install the package via Composer:

composer require lanos/laravel-bfl

Publish and Run Migrations

Publish the migration and config files:

php artisan vendor:publish --provider="Lanos\LaravelBFL\BFLServiceProvider"

Or publish them separately:

# Publish only the config
php artisan vendor:publish --tag="bfl-config"

# Publish only migrations  
php artisan vendor:publish --tag="bfl-migrations"

Run the migrations:

php artisan migrate

Environment Configuration

Add your BFL API key to your .env file:

BFL_API_KEY=your-black-forest-labs-api-key BFL_STORE_FILES=true BFL_STORAGE_DISK=local BFL_STORAGE_PATH_PREFIX=generated-bfl-images

Configuration

The package publishes a configuration file to config/bfl.php. Key configuration options:

return [
    'api_key' => env('BFL_API_KEY'),
    
    'storage' => [
        'store_files' => env('BFL_STORE_FILES', true),
        'disk' => env('BFL_STORAGE_DISK', 'local'),
        'path_prefix' => env('BFL_STORAGE_PATH_PREFIX', 'generated-bfl-images'),
        'public' => env('BFL_STORAGE_PUBLIC', false),
    ],
    
    'jobs' => [
        'queue_connection' => env('BFL_QUEUE_CONNECTION', 'default'),
        'queue_name' => env('BFL_QUEUE_NAME', 'default'),
        'polling' => [
            'max_attempts' => env('BFL_POLLING_MAX_ATTEMPTS', 120),
            'delay_seconds' => env('BFL_POLLING_DELAY_SECONDS', 3),
        ],
        'cleanup_after_days' => env('BFL_CLEANUP_AFTER_DAYS', 30),
    ],
];

Basic Usage

The package provides three different processing modes:

1. Manual Mode

Create a job and manually check its status:

use Lanos\LaravelBFL\Facades\BFL;
use Lanos\LaravelBFL\Models\BFLJob;

// Create a manual job
$job = BFL::createManualJob(
    BFLJob::MODEL_FLUX_1_PRO,
    'A majestic dragon soaring through clouds at sunset',
    [
        'width' => 1024,
        'height' => 768,
        'steps' => 50,
        'guidance' => 3.5,
    ],
    ['user_request_id' => 'req_123'], // metadata
    auth()->user() // optional user association
);

// Later, check the job status
$updatedJob = BFL::checkJobStatus($job);

if ($updatedJob->isCompleted()) {
    echo "Job completed! Files stored at: ";
    foreach ($updatedJob->file_paths ?? [] as $path) {
        echo BFL::getFileUrl($path) . "\n";
    }
} elseif ($updatedJob->isFailed()) {
    echo "Job failed: " . $updatedJob->error_message;
}

2. Synchronous Mode

Create a job and wait for completion:

// This will block until completion or timeout
$job = BFL::createSyncJob(
    BFLJob::MODEL_FLUX_1_PRO,
    'A beautiful sunset over mountains',
    [
        'aspect_ratio' => '16:9',
        'steps' => 28,
        'safety_tolerance' => 2,
    ]
);

// Job is automatically completed when this returns
if ($job->isCompleted()) {
    echo "Generated image result: " . json_encode($job->result);
}

3. Queue Mode

Create a job that processes in the background:

// Dispatch to queue for background processing
$job = BFL::createQueueJob(
    BFLJob::MODEL_FLUX_11_PRO_ULTRA,
    'An abstract artistic composition with vibrant colors',
    [
        'aspect_ratio' => '1:1',
        'raw' => false,
    ],
    [], // metadata
    auth()->user(), // user
    'redis', // queue connection (optional)
    'images'  // queue name (optional)
);

echo "Job queued with ID: {$job->id}";

// Check status later
$job->refresh();
if ($job->isCompleted()) {
    // Handle completion
}

Available Models

// Generation models
BFLJob::MODEL_FLUX_1_PRO           // 'flux-1-pro'
BFLJob::MODEL_FLUX_1_DEV           // 'flux-1-dev'
BFLJob::MODEL_FLUX_11_PRO          // 'flux-1.1-pro'
BFLJob::MODEL_FLUX_11_PRO_ULTRA    // 'flux-1.1-pro-ultra'

// Editing models  
BFLJob::MODEL_FLUX_1_FILL          // 'flux-1-fill'
BFLJob::MODEL_FLUX_1_EXPAND        // 'flux-1-expand'
BFLJob::MODEL_FLUX_1_CANNY         // 'flux-1-canny'
BFLJob::MODEL_FLUX_1_DEPTH         // 'flux-1-depth'

// Context models
BFLJob::MODEL_FLUX_KONTEXT_PRO     // 'flux-kontext-pro'
BFLJob::MODEL_FLUX_KONTEXT_MAX     // 'flux-kontext-max'

Common Parameters

$parameters = [
    // Dimensions
    'width' => 1024,
    'height' => 768,
    'aspect_ratio' => '16:9', // Alternative to width/height
    
    // Generation settings
    'steps' => 50,           // Denoising steps
    'guidance' => 3.5,       // How closely to follow prompt
    'seed' => 42,           // For reproducible results
    'safety_tolerance' => 2, // Content filtering (0=strict, 6=lenient)
    'output_format' => 'png', // 'png' or 'jpeg'
    
    // Advanced options
    'image_prompt' => base64_encode(file_get_contents('reference.jpg')),
    'prompt_upsampling' => true,
    
    // Webhooks
    'webhook_url' => 'https://your-site.com/webhooks/bfl',
    'webhook_secret' => 'your-webhook-secret',
];

Model-Specific Parameters

Fill/Inpainting models:

$parameters = [
    'image' => base64_encode(file_get_contents('input.jpg')),
    'mask' => base64_encode(file_get_contents('mask.png')), // Optional
];

Expand models:

$parameters = [
    'image' => base64_encode(file_get_contents('input.jpg')),
    'top' => 100,      // Pixels to add
    'bottom' => 100,
    'left' => 50,
    'right' => 50,
];

Control models (Canny/Depth):

$parameters = [
    'control_image' => base64_encode(file_get_contents('control.jpg')),
    'canny_low_threshold' => 50,   // For Canny model
    'canny_high_threshold' => 200,
];

Fine-tuning

The package provides comprehensive fine-tuning support through the BFLFinetune facade:

Creating Fine-tunes

use Lanos\LaravelBFL\Facades\BFLFinetune;
use Lanos\LaravelBFL\Enums\FinetuneMode;

// Create from file paths (automatically zips and encodes)
$finetune = BFLFinetune::create([
    'files' => [
        '/path/to/image1.jpg',
        '/path/to/image2.jpg',
        '/path/to/image3.jpg',
    ],
    'finetune_comment' => 'My custom style training',
    'mode' => FinetuneMode::STYLE, // or 'style' as string
    'trigger_word' => 'MYSTYLE',
    'iterations' => 500,
]);

// Or create from pre-encoded ZIP
$finetune = BFLFinetune::create([
    'file_data' => base64_encode(file_get_contents('training_data.zip')),
    'finetune_comment' => 'Character fine-tune',
    'mode' => 'character',
    'trigger_word' => 'MYCHAR',
    'iterations' => 800,
]);

Managing Fine-tunes

// Get finetune details
$details = BFLFinetune::getDetails('ft-123');

// List your finetunes
$finetunes = BFLFinetune::listMyFinetunes();

// Delete a finetune
$result = BFLFinetune::delete('ft-123');

Generating with Fine-tuned Models

// Generate with fine-tuned Pro
$job = BFLFinetune::generateWithFinetunedPro(
    'ft-123',
    'A beautiful sunset in MYSTYLE',
    ['width' => 1024, 'height' => 768]
);

// Generate with fine-tuned Ultra
$job = BFLFinetune::generateWithFinetunedUltra(
    'ft-123', 
    'Portrait in MYSTYLE',
    ['raw' => true]
);

// Generate with fine-tuned control models
$job = BFLFinetune::generateWithFinetunedCanny(
    'ft-123',
    'Landscape in MYSTYLE',
    base64_encode(file_get_contents('control.jpg'))
);

Job Status and Management

Status Checking

$job = BFLJob::find(1);

// Status checks
$job->isPending();     // Job created but not started
$job->isProcessing();  // Job is currently running
$job->isCompleted();   // Job finished successfully  
$job->isFailed();      // Job failed
$job->isFinished();    // Job is in any terminal state

// Progress tracking
echo "Progress: {$job->progress}%";
echo "Started: {$job->started_at}";
echo "Completed: {$job->completed_at}";

User Association

// Create job for specific user
$job = BFL::createManualJob(
    BFLJob::MODEL_FLUX_1_PRO,
    'User generated image',
    [],
    [],
    User::find(1)
);

// Get user's jobs
$userJobs = BFL::getUserJobs(auth()->user(), [
    'status' => BFLJob::STATUS_COMPLETED,
    'model_type' => BFLJob::MODEL_FLUX_1_PRO,
]);

Queue Configuration

Queue Settings

Configure queue processing in config/bfl.php:

'jobs' => [
    'queue_connection' => 'redis',
    'queue_name' => 'bfl-jobs',
    'polling' => [
        'max_attempts' => 120,    // Maximum poll attempts
        'delay_seconds' => 3,     // Delay between polls
    ],
],

Running Queue Workers

# Start worker for BFL jobs
php artisan queue:work --queue=bfl-jobs

# With specific connection
php artisan queue:work redis --queue=bfl-jobs

File Storage

Generated images are automatically downloaded and stored when store_files is enabled:

// Access stored files
$job = BFLJob::find(1);
$filePaths = $job->file_paths; // Array of stored file paths

// Get public URLs
foreach ($filePaths as $path) {
    echo BFL::getFileUrl($path);
}

Custom Storage Configuration

// Use S3 or other disks
'storage' => [
    'disk' => 's3',
    'public' => true,
],

Utility Methods

Cleanup

// Clean up old jobs (respects cleanup_after_days config)
$deletedCount = BFL::cleanupOldJobs();

Direct Client Access

// Access underlying FLUX client for advanced usage
$client = BFL::getClient();
$config = BFL::getConfig();

Testing

Run the package test suite:

composer test

Testing in Your Application

use Lanos\LaravelBFL\Services\BFLService;
use Lanos\LaravelBFL\Models\BFLJob;

public function test_image_generation()
{
    $mockService = $this->mock(BFLService::class);
    $mockJob = new BFLJob([
        'bfl_task_id' => 'test-task',
        'status' => BFLJob::STATUS_COMPLETED,
    ]);
    
    $mockService->allows('createManualJob')
        ->andReturn($mockJob);
    
    // Test your application logic
}

API Reference

BFL Facade Methods

// Job creation
BFL::createManualJob(string $modelType, string $prompt, array $parameters = [], array $metadata = [], ?object $user = null): BFLJob
BFL::createSyncJob(string $modelType, string $prompt, array $parameters = [], array $metadata = [], ?object $user = null): BFLJob
BFL::createQueueJob(string $modelType, string $prompt, array $parameters = [], array $metadata = [], ?object $user = null, ?string $queueConnection = null, ?string $queueName = null): BFLJob

// Job management  
BFL::checkJobStatus(BFLJob $job): BFLJob
BFL::completeJob(BFLJob $job, GetResultResponse $result): BFLJob

// Utility methods
BFL::getStorageDisk(): \Illuminate\Contracts\Filesystem\Filesystem
BFL::getFileUrl(string $filePath): string  
BFL::cleanupOldJobs(): int
BFL::getUserJobs(object $user, array $filters = []): \Illuminate\Database\Eloquent\Collection
BFL::getClient(): \Lanos\PHPBFL\FluxClient
BFL::getConfig(): array

BFLFinetune Facade Methods

// Finetune management
BFLFinetune::create(array $params): array
BFLFinetune::getDetails(string $finetuneId): array
BFLFinetune::listMyFinetunes(): array
BFLFinetune::delete(string $finetuneId): array

// Generation with finetuned models
BFLFinetune::generateWithFinetunedPro(string $finetuneId, string $prompt, array $parameters = [], array $metadata = [], ?object $user = null): BFLJob
BFLFinetune::generateWithFinetunedUltra(string $finetuneId, string $prompt, array $parameters = [], array $metadata = [], ?object $user = null): BFLJob
BFLFinetune::generateWithFinetunedDepth(string $finetuneId, string $prompt, string $controlImage, array $parameters = [], array $metadata = [], ?object $user = null): BFLJob
BFLFinetune::generateWithFinetunedCanny(string $finetuneId, string $prompt, string $controlImage, array $parameters = [], array $metadata = [], ?object $user = null): BFLJob
BFLFinetune::generateWithFinetunedFill(string $finetuneId, string $image, array $parameters = [], array $metadata = [], ?object $user = null): BFLJob

Job Model Methods

// Status checks
$job->isPending(): bool
$job->isProcessing(): bool
$job->isCompleted(): bool  
$job->isFailed(): bool
$job->isFinished(): bool

// Status updates
$job->markAsProcessing(): self
$job->markAsCompleted(?array $result = null, ?array $filePaths = null): self
$job->markAsFailed(?string $errorMessage = null): self
$job->updateProgress(float $progress): self

Contributing

We welcome contributions! Please ensure all tests pass and follow PSR-12 coding standards.

Security Vulnerabilities

Please review our security policy on how to report security vulnerabilities.

Credits

• Lanos
• All Contributors

License

The MIT License (MIT). Please see License File for more information.

Changelog

Please see CHANGELOG.md for more information on what has changed recently.