check if audio chat is available to show the record button, filter out ContextAgentAudioInteraction in the assistant task type list, change the audio html tag border radius

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

Author: julien-nc

lib/Listener/BeforeTemplateRenderedListener.php

@@ -66,6 +66,7 @@ public function handle(Event $event): void {
 			$indexingComplete = $this->appConfig->getValueInt('context_chat', 'last_indexed_time', 0) !== 0;
 			$this->initialStateService->provideInitialState('contextChatIndexingComplete', $indexingComplete);
 			$this->initialStateService->provideInitialState('contextAgentToolSources', $this->assistantService->informationSources);
+			$this->initialStateService->provideInitialState('isAudioChatAvailable', $this->assistantService->isAudioChatAvailable());
 			$autoplayAudioChat = $this->config->getUserValue($this->userId, Application::APP_ID, 'autoplay_audio_chat', '1') === '1';
 			$this->initialStateService->provideInitialState('autoplayAudioChat', $autoplayAudioChat);
 		}

lib/Listener/ChattyLLMTaskListener.php

@@ -18,7 +18,6 @@
 use OCP\EventDispatcher\IEventListener;
 use OCP\TaskProcessing\Events\TaskSuccessfulEvent;
 use OCP\TaskProcessing\Task;
-use OCP\TaskProcessing\TaskTypes\TextToSpeech;
 use Psr\Log\LoggerInterface;
 
 /**
@@ -96,6 +95,7 @@ public function handle(Event $event): void {
 					$queryMessage = $this->messageMapper->getMessageById($queryMessageId);
 					$queryMessage->setContent(trim($taskOutput['input_transcript'] ?? ''));
 					$this->messageMapper->update($queryMessage);
+					// TODO update session title if it's the first message
 				}
 			} else {
 				$content = trim($taskOutput['output'] ?? '');
@@ -138,7 +138,7 @@ private function runTtsIfNeeded(int $sessionId, Message $message, string $taskTy
 		$attachments = $lastNonEmptyMessage->jsonSerialize()['attachments'] ?? [];
 		foreach ($attachments as $attachment) {
 			if ($attachment['type'] === 'Audio') {
-				// we found an audio attachment
+				// we found an audio attachment, response should be audio
 				$this->runTtsTask($message, $userId);
 				return;
 			}
@@ -152,7 +152,7 @@ private function runTtsIfNeeded(int $sessionId, Message $message, string $taskTy
 	 */
 	private function runTtsTask(Message $message, ?string $userId): void {
 		$task = new Task(
-			TextToSpeech::ID,
+			\OCP\TaskProcessing\TaskTypes\TextToSpeech::ID,
 			['input' => $message->getContent()],
 			Application::APP_ID . ':internal',
 			$userId,

lib/Service/AssistantService.php

@@ -212,6 +212,17 @@ public function cancelNotifyWhenReady(int $taskId, string $userId): void {
 		$this->taskNotificationMapper->deleteByTaskId($taskId);
 	}
 
+	public function isAudioChatAvailable(): bool {
+		$availableTaskTypes = $this->taskProcessingManager->getAvailableTaskTypes();
+		// we have at least the simple audio chat task type and the 3 sub task types available
+		return class_exists('OCP\\TaskProcessing\\TaskTypes\\AudioToAudioChat')
+			&& array_key_exists(\OCP\TaskProcessing\TaskTypes\AudioToAudioChat::ID, $availableTaskTypes)
+			&& array_key_exists(AudioToText::ID, $availableTaskTypes)
+			&& class_exists('OCP\\TaskProcessing\\TaskTypes\\TextToSpeech')
+			&& array_key_exists(\OCP\TaskProcessing\TaskTypes\TextToSpeech::ID, $availableTaskTypes)
+			&& array_key_exists(TextToTextChat::ID, $availableTaskTypes);
+	}
+
 	/**
 	 * @return array<AssistantTaskProcessingTaskType>
 	 */
@@ -281,6 +292,10 @@ public function getAvailableTaskTypes(): array {
 					&& $typeId === \OCP\TaskProcessing\TaskTypes\ContextAgentInteraction::ID) {
 					continue;
 				}
+				if (class_exists('OCP\\TaskProcessing\\TaskTypes\\ContextAgentAudioInteraction')
+					&& $typeId === \OCP\TaskProcessing\TaskTypes\ContextAgentAudioInteraction::ID) {
+					continue;
+				}
 				if (class_exists('OCP\\TaskProcessing\\TaskTypes\\AudioToAudioChat')
 					&& $typeId === \OCP\TaskProcessing\TaskTypes\AudioToAudioChat::ID) {
 					continue;

src/components/ChattyLLM/ChattyLLMInputForm.vue

@@ -775,6 +775,14 @@ export default {
 					getChatURL(`/sessions/${lastHumanMessage.session_id}/messages/${lastHumanMessage.id}`),
 				)
 				lastHumanMessage.content = updatedMessage.data.content
+				// TODO update session title (just i the frontend data, the db session is updated in the backend listener)
+				/*
+				const firstHumanMessage = this.messages.length === 1 && this.messages[0].role === Roles.HUMAN
+				if (firstHumanMessage) {
+					const session = this.sessions.find((session) => session.id === sessionId)
+					session.title = content
+				}
+				*/
 			}
 		},
 

src/components/ChattyLLM/InputArea.vue

@@ -18,7 +18,7 @@
 			@update:model-value="$emit('update:chatContent', $event)"
 			@submit="$emit('submit', $event)" />
 		<div class="input-area__button-box">
-			<NcButton v-if="chatContent"
+			<NcButton v-if="!isAudioChatAvailable || chatContent"
 				class="input-area__button-box__button"
 				:aria-label="submitBtnAriaText"
 				:disabled="disabled || !chatContent.trim()"
@@ -49,6 +49,7 @@ import isMobile from '../../mixins/isMobile.js'
 import { generateOcsUrl } from '@nextcloud/router'
 import axios from '@nextcloud/axios'
 import { showError } from '@nextcloud/dialogs'
+import { loadState } from '@nextcloud/initial-state'
 
 /*
 maxlength calculation (just a rough estimate):
@@ -103,6 +104,7 @@ export default {
 			thinkingText: t('assistant', 'Processing…'),
 			submitBtnAriaText: t('assistant', 'Submit'),
 			isRecording: false,
+			isAudioChatAvailable: loadState('assistant', 'isAudioChatAvailable', false),
 		}
 	},
 

src/components/fields/AudioDisplay.vue

@@ -78,7 +78,7 @@ export default {
 
 <style scoped lang="scss">
 audio {
-	border-radius: 16px;
+	border-radius: 100px;
 	&.shadowed {
 		border: 2px solid var(--color-primary-element);
 	}