feat(whispering): send no_verbatim for ElevenLabs scribe_v2

apps/whispering/src/lib/services/transcription/cloud/elevenlabs.ts

@@ -1,12 +1,12 @@
-import { ElevenLabsClient } from 'elevenlabs';
 import {
 	defineErrors,
 	extractErrorMessage,
 	type InferErrors,
 } from 'wellcrafted/error';
-import { type Result, tryAsync } from 'wellcrafted/result';
+import { Err, type Result, tryAsync } from 'wellcrafted/result';
 
 const MAX_FILE_SIZE_MB = 1000 as const;
+const ELEVENLABS_STT_URL = 'https://api.elevenlabs.io/v1/speech-to-text';
 
 export const ElevenLabsError = defineErrors({
 	MissingApiKey: () => ({
@@ -23,6 +23,17 @@ export const ElevenLabsError = defineErrors({
 		sizeMb,
 		maxMb,
 	}),
+	ApiError: ({
+		status,
+		body,
+	}: {
+		status: number;
+		body: string;
+	}) => ({
+		message: `ElevenLabs API error ${status}: ${body}`,
+		status,
+		body,
+	}),
 	Unexpected: ({ cause }: { cause: unknown }) => ({
 		message: extractErrorMessage(cause),
 		cause,
@@ -51,23 +62,42 @@ export const ElevenLabsTranscriptionServiceLive = {
 			});
 		}
 
-		const client = new ElevenLabsClient({ apiKey: options.apiKey });
+		const formData = new FormData();
+		formData.append('file', audioBlob);
+		formData.append('model_id', options.modelName);
+		if (options.outputLanguage !== 'auto') {
+			formData.append('language_code', options.outputLanguage);
+		}
+		formData.append('tag_audio_events', 'false');
+		formData.append('diarize', 'true');
+		// Remove filler words, false starts, and non-speech sounds. Only
+		// supported on scribe_v2 per ElevenLabs API; ignored on other models.
+		if (options.modelName === 'scribe_v2') {
+			formData.append('no_verbatim', 'true');
+		}
+
+		const { data: response, error: fetchError } = await tryAsync({
+			try: () =>
+				fetch(ELEVENLABS_STT_URL, {
+					method: 'POST',
+					headers: { 'xi-api-key': options.apiKey },
+					body: formData,
+				}),
+			catch: (cause) => ElevenLabsError.Unexpected({ cause }),
+		});
+		if (fetchError) return Err(fetchError);
+
+		if (!response.ok) {
+			const body = await response.text().catch(() => '');
+			return ElevenLabsError.ApiError({ status: response.status, body });
+		}
 
 		return tryAsync({
 			try: async () => {
-				const transcription = await client.speechToText.convert({
-					file: audioBlob,
-					model_id: options.modelName,
-					language_code:
-						options.outputLanguage !== 'auto'
-							? options.outputLanguage
-							: undefined,
-					tag_audio_events: false,
-					diarize: true,
-				});
-				return transcription.text.trim();
+				const data = (await response.json()) as { text: string };
+				return data.text.trim();
 			},
-			catch: (error) => ElevenLabsError.Unexpected({ cause: error }),
+			catch: (cause) => ElevenLabsError.Unexpected({ cause }),
 		});
 	},
 };