README

HTTP client for Yandex Cloud Vision OCR (sync + async). Designed as a small, predictable, PSR-friendly Composer library with clear errors, DTOs, and optional concurrency via a runner interface.

Features

• Sync OCR (recognizeText)
• Async OCR (startTextRecognition + polling with wait / waitMany)
• PSR-18 transport + PSR-17 factories
• Deterministic DTOs with raw payload/meta
• Typed exceptions
• No hard dependency on event loops (optional concurrency via runner)

Requirements

• PHP 8.4+
• PSR-18 HTTP client
• PSR-17 factories

Installation

composer require php-vision/ya-ocr-vision-client

You must also install a PSR-18 client and PSR-17 factories, for example:

composer require guzzlehttp/guzzle nyholm/psr7

Authentication

Two options:

• IAM Token
  Use Authorization: Bearer <IAM_TOKEN>.
• API Key
  Use Authorization: Api-Key <API_KEY> (no folderId header).

This library only signs requests and forwards headers; it does not call IAM endpoints.

Quick Start (Sync)

<?php

declare(strict_types=1);

use GuzzleHttp\Client as GuzzleClient;
use Nyholm\Psr7\Factory\Psr17Factory;
use PhpVision\YandexVision\Auth\ApiKeyCredentialProvider;
use PhpVision\YandexVision\Ocr\Enum\LanguageCode;
use PhpVision\YandexVision\Ocr\OcrOptions;
use PhpVision\YandexVision\Ocr\OcrService;
use PhpVision\YandexVision\Transports\HttpTransport;

require __DIR__ . '/vendor/autoload.php';

$httpClient = new GuzzleClient();
$psr17Factory = new Psr17Factory();

$transport = new HttpTransport($httpClient);
$credentials = new ApiKeyCredentialProvider('YOUR_API_KEY');

$ocr = new OcrService($transport, $credentials, $psr17Factory, $psr17Factory);

$bytes = file_get_contents(__DIR__ . '/image.png');
$options = OcrOptions::create()->withLanguageCodes(LanguageCode::RU, LanguageCode::EN);
$response = $ocr->recognizeText($bytes, 'image/png', $options);

var_dump($response->getPayload());

Async OCR (Start + Poll)

<?php

declare(strict_types=1);

use GuzzleHttp\Client as GuzzleClient;
use Nyholm\Psr7\Factory\Psr17Factory;
use PhpVision\YandexVision\Auth\ApiKeyCredentialProvider;
use PhpVision\YandexVision\Ocr\OcrService;
use PhpVision\YandexVision\Transports\HttpTransport;

require __DIR__ . '/vendor/autoload.php';

$httpClient = new GuzzleClient();
$psr17Factory = new Psr17Factory();

$transport = new HttpTransport($httpClient);
$credentials = new ApiKeyCredentialProvider('YOUR_API_KEY');

$ocr = new OcrService($transport, $credentials, $psr17Factory, $psr17Factory);

$handle = $ocr->startTextRecognitionFromFile(__DIR__ . '/image.png');
$result = $ocr->wait($handle->getOperationId(), 60);

var_dump($result->getPayload());

Options

Every OCR call accepts an OcrOptions object:

• languageCodes (array of LanguageCode enums)
• model (OcrModel enum, default is page)
• requestId (string, forwarded as x-request-id)

Concurrency (waitMany)

waitMany() uses a runner to execute waits. Default is sequential:

$results = $ocr->waitMany(['op-1', 'op-2'], 60);

You can provide a custom runner that performs concurrent execution.

Multi-page Documents

For some multi-page OCR responses, the API may return NDJSON/JSONL (one JSON object per line, one line per page).
The client parses this automatically and returns:

$payload = $response->getPayload();
$pages = $payload['pages']; // array of page objects in API order

Error Handling

Exceptions are thrown for common failures:

• ApiException — non-2xx responses or operation errors
• HttpException — transport failures
• ValidationException — invalid input or JSON
• TimeoutException — async polling timed out

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.