add new admin setting to choose transcription model, enforce whisper …

…when OpenAI is used Signed-off-by: Julien Veyssier <[email protected]>
nextcloud · Nov 18, 2024 · 68a51de · 68a51de
1 parent 0292b2f
commit 68a51de
Show file tree

Hide file tree

Showing 4 changed files with 85 additions and 2 deletions.
diff --git a/lib/Service/OpenAiAPIService.php b/lib/Service/OpenAiAPIService.php
@@ -520,9 +520,13 @@ public function transcribe(
 		if ($this->isQuotaExceeded($userId, Application::QUOTA_TYPE_TRANSCRIPTION)) {
 			throw new Exception($this->l10n->t('Audio transcription quota exceeded'), Http::STATUS_TOO_MANY_REQUESTS);
 		}
+		// enforce whisper for OpenAI
+		if ($this->isUsingOpenAi()) {
+			$model = Application::DEFAULT_TRANSCRIPTION_MODEL_ID;
+		}
 
 		$params = [
-			'model' => $model,
+			'model' => $model === Application::DEFAULT_MODEL_ID ? Application::DEFAULT_TRANSCRIPTION_MODEL_ID : $model,
 			'file' => $audioFileContent,
 			'response_format' => 'verbose_json',
 			// Verbose needed for extraction of audio duration

diff --git a/lib/Service/OpenAiSettingsService.php b/lib/Service/OpenAiSettingsService.php
@@ -38,6 +38,7 @@ class OpenAiSettingsService {
 		'service_name' => 'string',
 		'api_key' => 'string',
 		'default_completion_model_id' => 'string',
+		'default_stt_model_id' => 'string',
 		'default_image_model_id' => 'string',
 		'default_image_size' => 'string',
 		'max_tokens' => 'integer',
@@ -109,6 +110,13 @@ public function getAdminDefaultCompletionModelId(): string {
 		return $this->appConfig->getValueString(Application::APP_ID, 'default_completion_model_id', Application::DEFAULT_COMPLETION_MODEL_ID) ?: Application::DEFAULT_COMPLETION_MODEL_ID;
 	}
 
+	/**
+	 * @return string
+	 */
+	public function getAdminDefaultSttModelId(): string {
+		return $this->appConfig->getValueString(Application::APP_ID, 'default_stt_model_id') ?: Application::DEFAULT_MODEL_ID;
+	}
+
 	/**
 	 * @return string
 	 */
@@ -263,6 +271,7 @@ public function getAdminConfig(): array {
 			'service_name' => $this->getServiceName(),
 			'api_key' => $this->getAdminApiKey(),
 			'default_completion_model_id' => $this->getAdminDefaultCompletionModelId(),
+			'default_stt_model_id' => $this->getAdminDefaultSttModelId(),
 			'default_image_model_id' => $this->getAdminDefaultImageModelId(),
 			'default_image_size' => $this->getAdminDefaultImageSize(),
 			'max_tokens' => $this->getMaxTokens(),
@@ -389,6 +398,15 @@ public function setAdminDefaultCompletionModelId(string $defaultCompletionModelI
 		$this->appConfig->setValueString(Application::APP_ID, 'default_completion_model_id', $defaultCompletionModelId);
 	}
 
+	/**
+	 * @param string $defaultSttModelId
+	 * @return void
+	 */
+	public function setAdminDefaultSttModelId(string $defaultSttModelId): void {
+		// No need to validate. As long as it's a string, we're happy campers
+		$this->appConfig->setValueString(Application::APP_ID, 'default_stt_model_id', $defaultSttModelId);
+	}
+
 	/**
 	 * @param string $defaultImageModelId
 	 * @return void
@@ -561,6 +579,9 @@ public function setAdminConfig(array $adminConfig): void {
 		if (isset($adminConfig['default_completion_model_id'])) {
 			$this->setAdminDefaultCompletionModelId($adminConfig['default_completion_model_id']);
 		}
+		if (isset($adminConfig['default_stt_model_id'])) {
+			$this->setAdminDefaultSttModelId($adminConfig['default_stt_model_id']);
+		}
 		if (isset($adminConfig['default_image_model_id'])) {
 			$this->setAdminDefaultImageModelId($adminConfig['default_image_model_id']);
 		}

diff --git a/lib/TaskProcessing/AudioToTextProvider.php b/lib/TaskProcessing/AudioToTextProvider.php
@@ -8,6 +8,7 @@
 use OCA\OpenAi\AppInfo\Application;
 use OCA\OpenAi\Service\OpenAiAPIService;
 use OCP\Files\File;
+use OCP\IAppConfig;
 use OCP\TaskProcessing\ISynchronousProvider;
 use OCP\TaskProcessing\TaskTypes\AudioToText;
 use Psr\Log\LoggerInterface;
@@ -18,6 +19,7 @@ class AudioToTextProvider implements ISynchronousProvider {
 	public function __construct(
 		private OpenAiAPIService $openAiAPIService,
 		private LoggerInterface $logger,
+		private IAppConfig $appConfig,
 	) {
 	}
 
@@ -75,8 +77,10 @@ public function process(?string $userId, array $input, callable $reportProgress)
 		}
 		$inputFile = $input['input'];
 
+		$model = $this->appConfig->getValueString(Application::APP_ID, 'default_stt_model_id', Application::DEFAULT_MODEL_ID) ?: Application::DEFAULT_MODEL_ID;
+
 		try {
-			$transcription = $this->openAiAPIService->transcribeFile($userId, $inputFile);
+			$transcription = $this->openAiAPIService->transcribeFile($userId, $inputFile, false, $model);
 			return ['output' => $transcription];
 		} catch (Exception $e) {
 			$this->logger->warning('OpenAI\'s Whisper transcription failed with: ' . $e->getMessage(), ['exception' => $e]);

diff --git a/src/components/AdminSettings.vue b/src/components/AdminSettings.vue
@@ -298,6 +298,46 @@
 					</NcButton>
 				</div>
 			</div>
+			<div>
+				<h2>
+					{{ t('integration_openai', 'Audio transcription') }}
+				</h2>
+				<div v-if="models"
+					class="line line-select">
+					<NcSelect
+						v-model="selectedModel.stt"
+						class="model-select"
+						:clearable="state.default_image_model_id !== DEFAULT_MODEL_ITEM.id"
+						:options="formattedModels"
+						:input-label="t('integration_openai', 'Default transcription model to use')"
+						:no-wrap="true"
+						input-id="openai-stt-model-select"
+						@input="onModelSelected('stt', $event)" />
+					<a v-if="state.url === ''"
+						:title="t('integration_openai', 'More information about OpenAI models')"
+						href="https://beta.openai.com/docs/models"
+						target="_blank">
+						<NcButton type="tertiary" aria-label="openai-info">
+							<template #icon>
+								<HelpCircleIcon />
+							</template>
+						</NcButton>
+					</a>
+					<a v-else
+						:title="t('integration_openai', 'More information about LocalAI models')"
+						href="https://localai.io/model-compatibility/index.html"
+						target="_blank">
+						<NcButton type="tertiary" aria-label="localai-info">
+							<template #icon>
+								<HelpCircleIcon />
+							</template>
+						</NcButton>
+					</a>
+				</div>
+				<NcNoteCard v-else type="info">
+					{{ t('integration_openai', 'No models to list') }}
+				</NcNoteCard>
+			</div>
 			<div>
 				<h2>
 					{{ t('integration_openai', 'Usage limits') }}
@@ -463,6 +503,7 @@ export default {
 			selectedModel: {
 				text: null,
 				image: null,
+				stt: null,
 			},
 			apiKeyUrl: 'https://platform.openai.com/account/api-keys',
 			quotaInfo: null,
@@ -540,8 +581,15 @@ export default {
 						|| this.models[1]
 						|| this.models[0]
 
+					const defaultSttModelId = this.state.default_stt_model_id || response.data?.default_stt_model_id
+					const sttModelToSelect = this.models.find(m => m.id === defaultSttModelId)
+						|| this.models.find(m => m.id.match(/whisper/i))
+						|| this.models[1]
+						|| this.models[0]
+
 					this.selectedModel.text = this.modelToNcSelectObject(completionModelToSelect)
 					this.selectedModel.image = this.modelToNcSelectObject(imageModelToSelect)
+					this.selectedModel.stt = this.modelToNcSelectObject(sttModelToSelect)
 
 					// save if url/credentials were changed OR if the values are not up-to-date in the stored settings
 					if (shouldSave
@@ -570,17 +618,23 @@ export default {
 				} else if (type === 'text') {
 					this.selectedModel.text = this.modelToNcSelectObject(DEFAULT_MODEL_ITEM)
 					this.state.default_completion_model_id = DEFAULT_MODEL_ITEM.id
+				} else if (type === 'stt') {
+					this.selectedModel.stt = this.modelToNcSelectObject(DEFAULT_MODEL_ITEM)
+					this.state.default_stt_model_id = DEFAULT_MODEL_ITEM.id
 				}
 			} else {
 				if (type === 'image') {
 					this.state.default_image_model_id = selected.id
 				} else if (type === 'text') {
 					this.state.default_completion_model_id = selected.id
+				} else if (type === 'stt') {
+					this.state.default_stt_model_id = selected.id
 				}
 			}
 			this.saveOptions({
 				default_completion_model_id: this.state.default_completion_model_id,
 				default_image_model_id: this.state.default_image_model_id,
+				default_stt_model_id: this.state.default_stt_model_id,
 			})
 		},
 		loadQuotaInfo() {