feat(speech-to-text): SpeechToTextManager::transcribeFile calls TaskProcessingManager::runTask

Signed-off-by: Julien Veyssier <julien-nc@posteo.net>

5 files changed, 61 insertions, 9 deletions

diff --git a/lib/private/SpeechToText/SpeechToTextManager.php b/lib/private/SpeechToText/SpeechToTextManager.php
index d6cda473875..c4397112a18 100644
--- a/lib/private/SpeechToText/SpeechToTextManager.php
+++ b/lib/private/SpeechToText/SpeechToTextManager.php
@@ -24,6 +24,9 @@ use OCP\SpeechToText\ISpeechToTextManager;
 use OCP\SpeechToText\ISpeechToTextProvider;
 use OCP\SpeechToText\ISpeechToTextProviderWithId;
 use OCP\SpeechToText\ISpeechToTextProviderWithUserId;
+use OCP\TaskProcessing\IManager as ITaskProcessingManager;
+use OCP\TaskProcessing\Task;
+use OCP\TaskProcessing\TaskTypes\AudioToText;
 use Psr\Container\ContainerExceptionInterface;
 use Psr\Container\NotFoundExceptionInterface;
 use Psr\Log\LoggerInterface;
@@ -41,6 +44,7 @@ class SpeechToTextManager implements ISpeechToTextManager {
 		private IJobList $jobList,
 		private IConfig $config,
 		private IUserSession $userSession,
+		private ITaskProcessingManager $taskProcessingManager,
 	) {
 	}
 
@@ -112,7 +116,28 @@ class SpeechToTextManager implements ISpeechToTextManager {
 		}
 	}
 
-	public function transcribeFile(File $file): string {
+	public function transcribeFile(File $file, ?string $userId = null, string $appId = 'core'): string {
+		// try to run a TaskProcessing core:audio2text task
+		// this covers scheduling as well because OC\SpeechToText\TranscriptionJob calls this method
+		try {
+			$taskProcessingTask = new Task(
+				AudioToText::ID,
+				['input' => $file->getId()],
+				$appId,
+				$userId,
+				'from-SpeechToTextManager||' . $file->getId() . '||' . ($userId ?? '') . '||' . $appId,
+			);
+			$resultTask = $this->taskProcessingManager->runTask($taskProcessingTask);
+			if ($resultTask->getStatus() === Task::STATUS_SUCCESSFUL) {
+				$output = $resultTask->getOutput();
+				if (isset($output['output']) && is_string($output['output'])) {
+					return $output['output'];
+				}
+			}
+		} catch (Throwable $e) {
+			$this->logger->debug('Failed to run a Speech-to-text job from STTManager with TaskProcessing for file ' . $file->getId(), ['exception' => $e]);
+		}
+
 		if (!$this->hasProviders()) {
 			throw new PreConditionNotMetException('No SpeechToText providers have been registered');
 		}
diff --git a/lib/private/SpeechToText/TranscriptionJob.php b/lib/private/SpeechToText/TranscriptionJob.php
index a46fd737865..6e899ef6e96 100644
--- a/lib/private/SpeechToText/TranscriptionJob.php
+++ b/lib/private/SpeechToText/TranscriptionJob.php
@@ -63,7 +63,7 @@ class TranscriptionJob extends QueuedJob {
 				);
 				return;
 			}
-			$result = $this->speechToTextManager->transcribeFile($file);
+			$result = $this->speechToTextManager->transcribeFile($file, $userId, $appId);
 			$this->eventDispatcher->dispatchTyped(
 				new TranscriptionSuccessfulEvent(
 					$fileId,
diff --git a/lib/private/TaskProcessing/Manager.php b/lib/private/TaskProcessing/Manager.php
index 873e0720f08..fb0a4da4c4e 100644
--- a/lib/private/TaskProcessing/Manager.php
+++ b/lib/private/TaskProcessing/Manager.php
@@ -88,7 +88,6 @@ class Manager implements IManager {
 		IAppDataFactory $appDataFactory,
 		private IRootFolder $rootFolder,
 		private \OCP\TextToImage\IManager $textToImageManager,
-		private \OCP\SpeechToText\ISpeechToTextManager $speechToTextManager,
 		private IUserMountCache $userMountCache,
 		private IClientService $clientService,
 		private IAppManager $appManager,
@@ -369,12 +368,35 @@ class Manager implements IManager {
 		return $newProviders;
 	}
 
+	/**
+	 * This is almost a copy of SpeechToTextManager->getProviders
+	 * to avoid a dependency cycle between SpeechToTextManager and TaskProcessingManager
+	 */
+	private function _getRawSpeechToTextProviders(): array {
+		$context = $this->coordinator->getRegistrationContext();
+		if ($context === null) {
+			return [];
+		}
+		$providers = [];
+		foreach ($context->getSpeechToTextProviders() as $providerServiceRegistration) {
+			$class = $providerServiceRegistration->getService();
+			try {
+				$providers[$class] = $this->serverContainer->get($class);
+			} catch (NotFoundExceptionInterface|ContainerExceptionInterface|\Throwable $e) {
+				$this->logger->error('Failed to load SpeechToText provider ' . $class, [
+					'exception' => $e,
+				]);
+			}
+		}
+
+		return $providers;
+	}
 
 	/**
 	 * @return IProvider[]
 	 */
 	private function _getSpeechToTextProviders(): array {
-		$oldProviders = $this->speechToTextManager->getProviders();
+		$oldProviders = $this->_getRawSpeechToTextProviders();
 		$newProviders = [];
 		foreach ($oldProviders as $oldProvider) {
 			$newProvider = new class($oldProvider, $this->rootFolder, $this->appData) implements IProvider, ISynchronousProvider {
diff --git a/lib/private/TextProcessing/Manager.php b/lib/private/TextProcessing/Manager.php
index c73a6a9ee37..52ca62eaede 100644
--- a/lib/private/TextProcessing/Manager.php
+++ b/lib/private/TextProcessing/Manager.php
@@ -137,10 +137,13 @@ class Manager implements IManager {
 				$this->logger->debug('Running a TextProcessing (' . $taskTypeClass . ') task with TaskProcessing');
 				$taskProcessingResultTask = $this->taskProcessingManager->runTask($taskProcessingTask);
 				if ($taskProcessingResultTask->getStatus() === \OCP\TaskProcessing\Task::STATUS_SUCCESSFUL) {
-					$task->setOutput($taskProcessingResultTask->getOutput()['output'] ?? '');
-					$task->setStatus(OCPTask::STATUS_SUCCESSFUL);
-					$this->taskMapper->update(DbTask::fromPublicTask($task));
-					return $task->getOutput();
+					$output = $taskProcessingResultTask->getOutput();
+					if (isset($output['output']) && is_string($output['output'])) {
+						$task->setOutput($output['output']);
+						$task->setStatus(OCPTask::STATUS_SUCCESSFUL);
+						$this->taskMapper->update(DbTask::fromPublicTask($task));
+						return $output['output'];
+					}
 				}
 			} catch (\Throwable $e) {
 				$this->logger->error('TextProcessing to TaskProcessing failed', ['exception' => $e]);
diff --git a/lib/public/SpeechToText/ISpeechToTextManager.php b/lib/public/SpeechToText/ISpeechToTextManager.php
index 043dac0ba14..6bd95197695 100644
--- a/lib/public/SpeechToText/ISpeechToTextManager.php
+++ b/lib/public/SpeechToText/ISpeechToTextManager.php
@@ -59,11 +59,13 @@ interface ISpeechToTextManager {
 
 	/**
 	 * @param File $file The media file to transcribe
+	 * @param ?string $userId The user that triggered this request
+	 * @param string $appId The app that triggered this request
 	 * @returns string The transcription of the passed media file
 	 * @throws PreConditionNotMetException If no provider was registered but this method was still called
 	 * @throws InvalidArgumentException If the file could not be found or is not of a supported type
 	 * @throws RuntimeException If the transcription failed for other reasons
 	 * @since 27.0.0
 	 */
-	public function transcribeFile(File $file): string;
+	public function transcribeFile(File $file, ?string $userId, string $appId): string;
 }