Merge pull request #413 from nextcloud/enh/noid/audio-adjustments

Fix audio recording and deal with audio watermarking in sub tasks

Author: julien-nc

.github/workflows/appstore-build-publish.yml

@@ -13,7 +13,7 @@ on:
     types: [published]
 
 env:
-  PHP_VERSION: 8.1
+  PHP_VERSION: 8.2
 
 jobs:
   build_and_publish:

appinfo/info.xml

@@ -62,7 +62,7 @@ Known providers:
 
 More details on how to set this up in the [admin docs](https://docs.nextcloud.com/server/latest/admin_manual/ai/index.html)
 ]]>	</description>
-	<version>2.10.0</version>
+	<version>2.11.0-dev</version>
 	<licence>agpl</licence>
 	<author>Julien Veyssier</author>
 	<namespace>Assistant</namespace>

composer.lock

@@ -30,6 +30,7 @@ public function handle(Event $event): void {
 
 		$csp = new ContentSecurityPolicy();
 		$csp->addAllowedWorkerSrcDomain('blob:');
+		$csp->addAllowedConnectDomain('blob:');
 		$event->addPolicy($csp);
 	}
 }

lib/Listener/CSPListener.php

@@ -143,6 +143,10 @@ public function process(?string $userId, array $input, callable $reportProgress)
 				Application::APP_ID . ':internal',
 				$userId,
 			);
+			// the setIncludeWatermark method was introduced in NC 33
+			if (method_exists($task, 'setIncludeWatermark')) {
+				$task->setIncludeWatermark(false);
+			}
 			$taskOutput = $this->taskProcessingService->runTaskProcessingTask($task);
 			$outputAudioFileId = $taskOutput['speech'];
 

lib/TaskProcessing/AudioToAudioChatProvider.php

@@ -143,6 +143,10 @@ public function process(?string $userId, array $input, callable $reportProgress)
 					Application::APP_ID . ':internal',
 					$userId,
 				);
+				// the setIncludeWatermark method was introduced in NC 33
+				if (method_exists($task, 'setIncludeWatermark')) {
+					$task->setIncludeWatermark(false);
+				}
 				$ttsTaskOutput = $this->taskProcessingService->runTaskProcessingTask($task);
 				$outputAudioFileId = $ttsTaskOutput['speech'];
 				$outputAudioFileContent = $this->taskProcessingService->getOutputFileContent($outputAudioFileId);

lib/TaskProcessing/ContextAgentAudioInteractionProvider.php

@@ -132,7 +132,7 @@ export default {
 			const url = generateOcsUrl('/apps/assistant/api/v1/input-file')
 			const formData = new FormData()
 			formData.append('data', blob)
-			formData.append('filename', 'chat-input.mp3')
+			formData.append('filename', 'chat-input.wav')
 			axios.post(url, formData).then(response => {
 				this.$emit('submit-audio', response.data.ocs.data.fileId)
 			}).catch(error => {

src/audioUtils.js

@@ -49,7 +49,6 @@ import NcButton from '@nextcloud/vue/components/NcButton'
 
 import { showError } from '@nextcloud/dialogs'
 
-import { convertWavToMp3 } from '../../audioUtils.js'
 import { MediaRecorder, register } from 'extendable-media-recorder'
 import { connect } from 'extendable-media-recorder-wav-encoder'
 
@@ -210,8 +209,7 @@ export default {
 			this.killStreams()
 			if (!this.aborted) {
 				const wavBlob = new Blob(this.chunks, { type: this.mediaRecorder.mimeType })
-				const mp3Blob = await convertWavToMp3(wavBlob)
-				this.$emit('new-recording', mp3Blob)
+				this.$emit('new-recording', wavBlob)
 				this.$emit('update:is-recording', false)
 			}
 			this.resetComponentData()

src/components/ChattyLLM/InputArea.vue

@@ -218,7 +218,7 @@ export default {
 			const url = generateOcsUrl('/apps/assistant/api/v1/input-file')
 			const formData = new FormData()
 			formData.append('data', blob)
-			formData.append('filename', 'recording.mp3')
+			formData.append('filename', 'recording.wav')
 			axios.post(url, formData).then(response => {
 				const fileId = response.data.ocs.data.fileId
 				if (this.value === null) {