Blasp Profanity Filter for Laravel

If 'events' => true is set in your configuration, Blasp fires the following events:

Note: ModelProfanityDetected always fires and is not gated by the events config setting.

Blasp uses different drivers to detect various types of profanity:

• regex driver (Default): Detects obfuscated profanity including substitutions (f4ck), separators (f-u-c-k), doubled letters (ffuucckkiinngg), and combinations. It allows up to 3 separator characters between letters.
• pattern driver: Performs fast, exact word-boundary matching only.
• phonetic driver: Uses metaphone() and Levenshtein distance to catch sound-alike evasions (e.g., phuck, sheit). Sensitivity is configurable in config/blasp.php under drivers.phonetic.
• pipeline driver: Chains multiple drivers together using union merge semantics. Text is flagged if any driver finds a match. If multiple drivers detect the same word at the same position, only the longest match is kept.

Register the blasp middleware alias to filter incoming request fields. You can specify the action (reject or sanitize) and the minimum severity level.

```php
// Using the short alias (rejects by default)
Route::post('/comment', CommentController::class)
    ->middleware('blasp');

// With parameters: action, severity
Route::post('/comment', CommentController::class)
    ->middleware('blasp:sanitize,mild');

// Using the class directly
use Blaspsoft\Blasp\Middleware\CheckProfanity;
Route::post('/comment', CommentController::class)
    ->middleware(CheckProfanity::class);

Middleware Actions:

• reject (default): Returns a 422 JSON response with field errors.
• sanitize: Replaces profane fields in the request with masked versions and continues.

Blasp caches check() results by default to improve performance. If the same text is checked with the same configuration (language, driver, severity, allow/block lists), the cached result is returned immediately.

Note: Result caching is automatically bypassed when using a CallbackMask because closures cannot be serialized.

To clear caches:

• Via Artisan: php artisan blasp:clear (clears both dictionary and result caches).
• Programmatically: Dictionary::clearCache();

After installation, publish the configuration and language files using Artisan. You can choose to publish everything, only the configuration, or only the language files.

# Everything (config + language files)
php artisan vendor:publish --tag="blasp"

# Config only
php artisan vendor:publish --tag="blasp-config"

# Language files only
php artisan vendor:publish --tag="blasp-languages"

Use the Blasp facade to perform a basic profanity check. The check() method returns a Result object containing details about the detection and methods to clean the text.

use Blaspsoft\
Blasp\Facades\Blasp;

$result = Blasp::check('This is a fucking sentence');

$result->isOffensive();  // true
$result->clean();         // "This is a ******* sentence"
$result->original();      // "This is a fucking sentence"
$result->score();         // 30
$result->count();         // 1
$result->uniqueWords();   // ['fucking']
$result->severity();      // Severity::High

Install the Blasp profanity filter for Laravel using Composer. Requires PHP 8.2+ and Laravel 8.0+.

composer require blaspsoft/blasp

The PipelineDriver is a composite driver that allows you to run multiple profanity detection drivers in a single pass. It executes each configured driver, collects all detected MatchedWord instances, and resolves overlaps using a greedy, longest-first approach.

When using a PipelineDriver, the resulting Result object will contain the union of all matches from the sub-drivers, deduplicated so that overlapping words are not double-masked. The final score is recalculated based on the unique set of matches found across all drivers.

use Blaspsoft\Blasp\Drivers\PipelineDriver;
// Assuming $otherDrivers is an array of DriverInterface implementations
$pipeline = new PipelineDriver($otherDrivers);

$result = $pipeline->detect($text, $dictionary, $mask, $options);

In your test suites, you can use Blasp::fake() to prevent actual detection logic from running. This allows you to simulate clean or offensive results.

Common Assertions:

• Blasp::assertChecked(): Asserts that a check was performed.
• Blasp::assertCheckedTimes(int $times): Asserts the number of checks performed.
• Blasp::assertCheckedWith(string $text): Asserts that a specific text was checked.

use Blaspsoft\Blasp\Facades\Blasp;
use Blaspsoft\Blasp\Core\Result;

// Replace with a fake — all checks return clean by default
Blasp::fake();

// Pre-configure specific responses
Blasp::fake([
    'bad text'   => Result::withMatches(['fuck']),
    'clean text' => Result::none('clean text'),
]);

$result = Blasp::check('bad text');
$result->isOffensive(); // true

// Assertions
Blasp::assertChecked();
Blasp::assertCheckedTimes(1);
Blasp::assertCheckedWith('bad text');