Add ocr-translation task type and provider

lib/AppInfo/Application.php

@@ -31,6 +31,8 @@
 use OCA\Assistant\Reference\Text2StickerProvider;
 use OCA\Assistant\TaskProcessing\AudioToAudioChatProvider;
 use OCA\Assistant\TaskProcessing\ContextAgentAudioInteractionProvider;
+use OCA\Assistant\TaskProcessing\ImageToTextTranslateProvider;
+use OCA\Assistant\TaskProcessing\ImageToTextTranslateTaskType;
 use OCA\Assistant\TaskProcessing\TextToStickerProvider;
 use OCA\Assistant\TaskProcessing\TextToStickerTaskType;
 use OCA\Files\Event\LoadAdditionalScriptsEvent;
@@ -111,6 +113,9 @@ public function register(IRegistrationContext $context): void {
 		$context->registerTaskProcessingTaskType(TextToStickerTaskType::class);
 		$context->registerTaskProcessingProvider(TextToStickerProvider::class);
 		$context->registerReferenceProvider(Text2StickerProvider::class);
+
+		$context->registerTaskProcessingTaskType(ImageToTextTranslateTaskType::class);
+		$context->registerTaskProcessingProvider(ImageToTextTranslateProvider::class);
 	}
 
 	public function boot(IBootContext $context): void {

lib/Service/TaskProcessingService.php

@@ -19,6 +19,7 @@
 use OCP\TaskProcessing\Exception\UnauthorizedException;
 use OCP\TaskProcessing\Exception\ValidationException;
 use OCP\TaskProcessing\IManager;
+use OCP\TaskProcessing\IProvider;
 use OCP\TaskProcessing\Task;
 use OCP\TaskProcessing\TaskTypes\AudioToText;
 use OCP\TaskProcessing\TaskTypes\TextToTextSummary;
@@ -35,6 +36,10 @@ public function __construct(
 	) {
 	}
 
+	public function getPreferredProvider(string $taskTypeId): IProvider {
+		return $this->taskProcessingManager->getPreferredProvider($taskTypeId);
+	}
+
 	/**
 	 * @param Task $task
 	 * @return array

lib/TaskProcessing/ImageToTextTranslateProvider.php

@@ -0,0 +1,151 @@
+<?php
+
+declare(strict_types=1);
+
+/**
+ * SPDX-FileCopyrightText: 2025 Nextcloud GmbH and Nextcloud contributors
+ * SPDX-License-Identifier: AGPL-3.0-or-later
+ */
+
+namespace OCA\Assistant\TaskProcessing;
+
+use Exception;
+use OCA\Assistant\AppInfo\Application;
+use OCA\Assistant\Service\TaskProcessingService;
+use OCP\Files\File;
+use OCP\IL10N;
+use OCP\TaskProcessing\ISynchronousProvider;
+use OCP\TaskProcessing\Task;
+use OCP\TaskProcessing\TaskTypes\ImageToTextOpticalCharacterRecognition;
+use OCP\TaskProcessing\TaskTypes\TextToTextTranslate;
+use Psr\Log\LoggerInterface;
+use RuntimeException;
+
+class ImageToTextTranslateProvider implements ISynchronousProvider {
+
+	public function __construct(
+		private IL10N $l,
+		private TaskProcessingService $taskProcessingService,
+		private LoggerInterface $logger,
+	) {
+	}
+
+	public function getId(): string {
+		return Application::APP_ID . '-image2text:translate';
+	}
+
+	public function getName(): string {
+		return $this->l->t('Assistant');
+	}
+
+	public function getTaskTypeId(): string {
+		return ImageToTextTranslateTaskType::ID;
+	}
+
+	public function getExpectedRuntime(): int {
+		return 60;
+	}
+
+	public function getInputShapeEnumValues(): array {
+		$translateProvider = $this->taskProcessingService->getPreferredProvider(TextToTextTranslate::ID);
+
+		return [
+			'origin_language' => $translateProvider->getInputShapeEnumValues()['origin_language'],
+			'target_language' => $translateProvider->getInputShapeEnumValues()['target_language'],
+		];
+	}
+
+	public function getInputShapeDefaults(): array {
+		$translateProvider = $this->taskProcessingService->getPreferredProvider(TextToTextTranslate::ID);
+		return [
+			'origin_language' => $translateProvider->getInputShapeDefaults()['origin_language'],
+		];
+	}
+
+
+	public function getOptionalInputShape(): array {
+		return [];
+	}
+
+	public function getOptionalInputShapeEnumValues(): array {
+		return [];
+	}
+
+	public function getOptionalInputShapeDefaults(): array {
+		return [];
+	}
+
+	public function getOutputShapeEnumValues(): array {
+		return [];
+	}
+
+	public function getOptionalOutputShape(): array {
+		return [];
+	}
+
+	public function getOptionalOutputShapeEnumValues(): array {
+		return [];
+	}
+
+	public function process(?string $userId, array $input, callable $reportProgress): array {
+		if (!isset($input['input']) || !is_array($input['input'])) {
+			throw new RuntimeException('Invalid input');
+		}
+		foreach ($input['input'] as $i => $inputImage) {
+			if (!($inputImage instanceof File) || !$inputImage->isReadable()) {
+				throw new RuntimeException('Invalid input images');
+			}
+		}
+
+		if (!isset($input['origin_language']) || !is_string($input['origin_language'])) {
+			throw new RuntimeException('Invalid origin_language input');
+		}
+		if (!isset($input['target_language']) || !is_string($input['target_language'])) {
+			throw new RuntimeException('Invalid target_language input');
+		}
+
+		// OCR
+		$ocrInputs = array_map(static function (File $file) {
+			return $file->getId();
+		}, $input['input']);
+		try {
+			$task = new Task(
+				ImageToTextOpticalCharacterRecognition::ID,
+				['input' => $ocrInputs],
+				Application::APP_ID . ':internal',
+				$userId,
+			);
+			$taskOutput = $this->taskProcessingService->runTaskProcessingTask($task);
+			$ocrOutputs = $taskOutput['output'];
+		} catch (Exception $e) {
+			$this->logger->warning('OCR sub task failed with: ' . $e->getMessage(), ['exception' => $e]);
+			throw new RuntimeException('OCR sub task failed with: ' . $e->getMessage());
+		}
+
+		$translatedOutputs = [];
+		foreach ($ocrOutputs as $ocrOutput) {
+			try {
+				$task = new Task(
+					TextToTextTranslate::ID,
+					[
+						'input' => $ocrOutput,
+						'origin_language' => $input['origin_language'],
+						'target_language' => $input['target_language'],
+					],
+					Application::APP_ID . ':internal',
+					$userId,
+				);
+				$taskOutput = $this->taskProcessingService->runTaskProcessingTask($task);
+				$translatedOutputs[] = $taskOutput['output'];
+			} catch (Exception $e) {
+				$this->logger->warning('Translation sub task failed with: ' . $e->getMessage(), ['exception' => $e]);
+				throw new RuntimeException('Translation sub task failed with: ' . $e->getMessage());
+			}
+		}
+
+		// Translation
+		return [
+			'output' => $translatedOutputs,
+		];
+	}
+}

lib/TaskProcessing/ImageToTextTranslateTaskType.php

@@ -0,0 +1,82 @@
+<?php
+
+declare(strict_types=1);
+
+/**
+ * SPDX-FileCopyrightText: 2025 Nextcloud GmbH and Nextcloud contributors
+ * SPDX-License-Identifier: AGPL-3.0-or-later
+ */
+
+namespace OCA\Assistant\TaskProcessing;
+
+use OCA\Assistant\AppInfo\Application;
+use OCP\IL10N;
+use OCP\TaskProcessing\EShapeType;
+use OCP\TaskProcessing\ITaskType;
+use OCP\TaskProcessing\ShapeDescriptor;
+
+class ImageToTextTranslateTaskType implements ITaskType {
+	public const ID = Application::APP_ID . ':image2text:translate';
+
+	public function __construct(
+		private IL10N $l,
+	) {
+	}
+
+	/**
+	 * @inheritDoc
+	 */
+	public function getName(): string {
+		return $this->l->t('Translate image');
+	}
+
+	/**
+	 * @inheritDoc
+	 */
+	public function getDescription(): string {
+		return $this->l->t('Translate the text content of an image');
+	}
+
+	/**
+	 * @return string
+	 */
+	public function getId(): string {
+		return self::ID;
+	}
+
+	/**
+	 * @return ShapeDescriptor[]
+	 */
+	public function getInputShape(): array {
+		return [
+			'input' => new ShapeDescriptor(
+				$this->l->t('Input files'),
+				$this->l->t('The files to extract text from'),
+				EShapeType::ListOfFiles
+			),
+			'origin_language' => new ShapeDescriptor(
+				$this->l->t('Origin language'),
+				$this->l->t('The language of the origin text'),
+				EShapeType::Enum
+			),
+			'target_language' => new ShapeDescriptor(
+				$this->l->t('Target language'),
+				$this->l->t('The desired language to translate the origin text in'),
+				EShapeType::Enum
+			),
+		];
+	}
+
+	/**
+	 * @return ShapeDescriptor[]
+	 */
+	public function getOutputShape(): array {
+		return [
+			'output' => new ShapeDescriptor(
+				$this->l->t('Output texts'),
+				$this->l->t('The texts that were extracted from the files'),
+				EShapeType::ListOfTexts
+			),
+		];
+	}
+}