README

A high-performance language detection service written in PHP using Unix sockets and Efficient Language Detector.

Installation

composer require keevitaja/language-detector-socket

Important

Enable OPcache for CLI to avoid excessive CPU usage. The LanguageDetector library loads and compiles large language model data files on every request. Without OPcache, PHP will recompile these files repeatedly, causing significant CPU overhead.

opcache.enable_cli=1

Usage

Start the Server

use Keevitaja\LanguageDetectorSocket\Server;

$server = new Server([
    'socket' => '/tmp/language-detector.sock',
]);

$server->run();

Use the Client

use Keevitaja\LanguageDetectorSocket\Client;

$client = new Client([
    'socket' => '/tmp/language-detector.sock',
])->make();

$scores = $client->detect('Hello world');
// ['en' => 0.95, 'nl' => 0.03, ...]

$client->close();

Configuration defaults

[
    // Unix socket path
    'socket' => '/tmp/language.detector.sock',

    // Number of worker processes
    'processes' => 1,

    // Connection timeout in seconds
    'timeout' => 10,

    // Limit detection to specific languages (null = all or ['en', 'nl'])
    'locales' => null,

    // Language model size (SMALL, MEDIUM, LARGE)
    'eldDataFile' => EldDataFile::SMALL,

    // Language code format (ISO639_1, ISO639_3)
    'eldFormat' => EldFormat::ISO639_1,

    // Path to store the worker process ID file
    'workerPidFile' => '/tmp/language.detector.worker.pid',

    // Path to store the worker log file
    'workerLogFile' => '/tmp/language.detector.worker.log',

    // Path to store the worker an stdout log file
    'workerStdoutFile' => '/tmp/language.detector.worker.stdout.log',

    // Whether to run the worker as a daemon process
    'workerDemonize' => false,
]