asr-settings.tsx 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319
  1. 'use client';
  2. import { useState, useRef } from 'react';
  3. import { Label } from '@/components/ui/label';
  4. import { Input } from '@/components/ui/input';
  5. import { Button } from '@/components/ui/button';
  6. import {
  7. Select,
  8. SelectContent,
  9. SelectItem,
  10. SelectTrigger,
  11. SelectValue,
  12. } from '@/components/ui/select';
  13. import { useI18n } from '@/lib/hooks/use-i18n';
  14. import { useSettingsStore } from '@/lib/store/settings';
  15. import { ASR_PROVIDERS } from '@/lib/audio/constants';
  16. import type { ASRProviderId } from '@/lib/audio/types';
  17. import { Mic, MicOff, CheckCircle2, XCircle, Eye, EyeOff } from 'lucide-react';
  18. import { cn } from '@/lib/utils';
  19. import { createLogger } from '@/lib/logger';
  20. const log = createLogger('ASRSettings');
  21. interface ASRSettingsProps {
  22. selectedProviderId: ASRProviderId;
  23. }
  24. export function ASRSettings({ selectedProviderId }: ASRSettingsProps) {
  25. const { t } = useI18n();
  26. const asrLanguage = useSettingsStore((state) => state.asrLanguage);
  27. const asrProvidersConfig = useSettingsStore((state) => state.asrProvidersConfig);
  28. const setASRProviderConfig = useSettingsStore((state) => state.setASRProviderConfig);
  29. const asrProvider = ASR_PROVIDERS[selectedProviderId] ?? ASR_PROVIDERS['openai-whisper'];
  30. const isServerConfigured = !!asrProvidersConfig[selectedProviderId]?.isServerConfigured;
  31. const [showApiKey, setShowApiKey] = useState(false);
  32. const [isRecording, setIsRecording] = useState(false);
  33. const [asrResult, setASRResult] = useState('');
  34. const [testStatus, setTestStatus] = useState<'idle' | 'testing' | 'success' | 'error'>('idle');
  35. const [testMessage, setTestMessage] = useState('');
  36. const mediaRecorderRef = useRef<MediaRecorder | null>(null);
  37. // Reset state when provider changes (derived state pattern)
  38. const [prevProviderId, setPrevProviderId] = useState(selectedProviderId);
  39. if (selectedProviderId !== prevProviderId) {
  40. setPrevProviderId(selectedProviderId);
  41. setShowApiKey(false);
  42. setTestStatus('idle');
  43. setTestMessage('');
  44. setASRResult('');
  45. }
  46. const handleToggleASRRecording = async () => {
  47. if (isRecording) {
  48. if (mediaRecorderRef.current && mediaRecorderRef.current.state === 'recording') {
  49. mediaRecorderRef.current.stop();
  50. }
  51. setIsRecording(false);
  52. } else {
  53. setASRResult('');
  54. setTestStatus('testing');
  55. setTestMessage('');
  56. if (selectedProviderId === 'browser-native') {
  57. const SpeechRecognitionCtor =
  58. (window as unknown as Record<string, unknown>).SpeechRecognition ||
  59. (window as unknown as Record<string, unknown>).webkitSpeechRecognition;
  60. if (!SpeechRecognitionCtor) {
  61. setTestStatus('error');
  62. setTestMessage(t('settings.asrNotSupported'));
  63. return;
  64. }
  65. // eslint-disable-next-line @typescript-eslint/no-explicit-any -- Vendor-prefixed API without standard typings
  66. const recognition = new (SpeechRecognitionCtor as new () => any)();
  67. recognition.lang = asrLanguage || 'zh-CN';
  68. recognition.onresult = (event: {
  69. results: {
  70. [index: number]: { [index: number]: { transcript: string } };
  71. };
  72. }) => {
  73. const transcript = event.results[0][0].transcript;
  74. setASRResult(transcript);
  75. setTestStatus('success');
  76. setTestMessage(t('settings.asrTestSuccess'));
  77. };
  78. recognition.onerror = (event: { error: string }) => {
  79. setTestStatus('error');
  80. setTestMessage(t('settings.asrTestFailed') + ': ' + event.error);
  81. };
  82. recognition.onend = () => {
  83. setIsRecording(false);
  84. };
  85. recognition.start();
  86. setIsRecording(true);
  87. } else {
  88. try {
  89. const stream = await navigator.mediaDevices.getUserMedia({
  90. audio: true,
  91. });
  92. const mediaRecorder = new MediaRecorder(stream);
  93. mediaRecorderRef.current = mediaRecorder;
  94. const audioChunks: Blob[] = [];
  95. mediaRecorder.ondataavailable = (event) => {
  96. audioChunks.push(event.data);
  97. };
  98. mediaRecorder.onstop = async () => {
  99. stream.getTracks().forEach((track) => track.stop());
  100. const audioBlob = new Blob(audioChunks, { type: 'audio/webm' });
  101. const formData = new FormData();
  102. formData.append('audio', audioBlob, 'recording.webm');
  103. formData.append('providerId', selectedProviderId);
  104. formData.append(
  105. 'modelId',
  106. asrProvidersConfig[selectedProviderId]?.modelId || asrProvider.defaultModelId,
  107. );
  108. formData.append('language', asrLanguage);
  109. const apiKeyValue = asrProvidersConfig[selectedProviderId]?.apiKey;
  110. if (apiKeyValue?.trim()) formData.append('apiKey', apiKeyValue);
  111. const baseUrlValue = asrProvidersConfig[selectedProviderId]?.baseUrl;
  112. if (baseUrlValue?.trim()) formData.append('baseUrl', baseUrlValue);
  113. try {
  114. const response = await fetch('/api/transcription', {
  115. method: 'POST',
  116. body: formData,
  117. });
  118. if (response.ok) {
  119. const data = await response.json();
  120. setASRResult(data.text);
  121. setTestStatus('success');
  122. setTestMessage(t('settings.asrTestSuccess'));
  123. } else {
  124. setTestStatus('error');
  125. const errorData = await response
  126. .json()
  127. .catch(() => ({ error: response.statusText }));
  128. setTestMessage(errorData.details || errorData.error || t('settings.asrTestFailed'));
  129. }
  130. } catch (error) {
  131. log.error('ASR test failed:', error);
  132. setTestStatus('error');
  133. setTestMessage(t('settings.asrTestFailed'));
  134. }
  135. };
  136. mediaRecorder.start();
  137. setIsRecording(true);
  138. } catch (error) {
  139. log.error('Failed to access microphone:', error);
  140. setTestStatus('error');
  141. setTestMessage(t('settings.microphoneAccessFailed'));
  142. }
  143. }
  144. }
  145. };
  146. return (
  147. <div className="space-y-6 max-w-3xl">
  148. {/* Server-configured notice */}
  149. {isServerConfigured && (
  150. <div className="rounded-lg border border-blue-200 bg-blue-50 dark:border-blue-800 dark:bg-blue-950/30 p-3 text-sm text-blue-700 dark:text-blue-300">
  151. {t('settings.serverConfiguredNotice')}
  152. </div>
  153. )}
  154. {/* API Key & Base URL */}
  155. {(asrProvider.requiresApiKey || isServerConfigured) && (
  156. <>
  157. <div className="grid grid-cols-2 gap-4">
  158. <div className="space-y-2">
  159. <Label className="text-sm">{t('settings.asrApiKey')}</Label>
  160. <div className="relative">
  161. <Input
  162. name={`asr-api-key-${selectedProviderId}`}
  163. type={showApiKey ? 'text' : 'password'}
  164. autoComplete="new-password"
  165. autoCapitalize="none"
  166. autoCorrect="off"
  167. spellCheck={false}
  168. placeholder={
  169. isServerConfigured ? t('settings.optionalOverride') : t('settings.enterApiKey')
  170. }
  171. value={asrProvidersConfig[selectedProviderId]?.apiKey || ''}
  172. onChange={(e) =>
  173. setASRProviderConfig(selectedProviderId, {
  174. apiKey: e.target.value,
  175. })
  176. }
  177. className="font-mono text-sm pr-10"
  178. />
  179. <button
  180. type="button"
  181. onClick={() => setShowApiKey(!showApiKey)}
  182. className="absolute right-2 top-1/2 -translate-y-1/2 text-muted-foreground hover:text-foreground"
  183. >
  184. {showApiKey ? <EyeOff className="h-4 w-4" /> : <Eye className="h-4 w-4" />}
  185. </button>
  186. </div>
  187. </div>
  188. <div className="space-y-2">
  189. <Label className="text-sm">{t('settings.asrBaseUrl')}</Label>
  190. <Input
  191. name={`asr-base-url-${selectedProviderId}`}
  192. autoComplete="off"
  193. autoCapitalize="none"
  194. autoCorrect="off"
  195. spellCheck={false}
  196. placeholder={asrProvider.defaultBaseUrl || t('settings.enterCustomBaseUrl')}
  197. value={asrProvidersConfig[selectedProviderId]?.baseUrl || ''}
  198. onChange={(e) =>
  199. setASRProviderConfig(selectedProviderId, {
  200. baseUrl: e.target.value,
  201. })
  202. }
  203. className="text-sm"
  204. />
  205. </div>
  206. </div>
  207. {/* Request URL Preview */}
  208. {(() => {
  209. const effectiveBaseUrl =
  210. asrProvidersConfig[selectedProviderId]?.baseUrl || asrProvider.defaultBaseUrl || '';
  211. if (!effectiveBaseUrl) return null;
  212. let endpointPath = '';
  213. switch (selectedProviderId) {
  214. case 'openai-whisper':
  215. endpointPath = '/audio/transcriptions';
  216. break;
  217. case 'qwen-asr':
  218. endpointPath = '/services/aigc/multimodal-generation/generation';
  219. break;
  220. }
  221. if (!endpointPath) return null;
  222. return (
  223. <p className="text-xs text-muted-foreground break-all">
  224. {t('settings.requestUrl')}: {effectiveBaseUrl + endpointPath}
  225. </p>
  226. );
  227. })()}
  228. </>
  229. )}
  230. {/* Test ASR */}
  231. <div className="space-y-2">
  232. <Label className="text-sm">{t('settings.testASR')}</Label>
  233. <div className="flex gap-2">
  234. <Input
  235. value={asrResult}
  236. readOnly
  237. placeholder={t('settings.asrResultPlaceholder')}
  238. className="flex-1 bg-muted/50"
  239. />
  240. <Button
  241. onClick={handleToggleASRRecording}
  242. disabled={
  243. asrProvider.requiresApiKey &&
  244. !asrProvidersConfig[selectedProviderId]?.apiKey?.trim() &&
  245. !isServerConfigured
  246. }
  247. className="gap-2 w-[140px]"
  248. >
  249. {isRecording ? (
  250. <>
  251. <MicOff className="h-4 w-4" />
  252. {t('settings.stopRecording')}
  253. </>
  254. ) : (
  255. <>
  256. <Mic className="h-4 w-4" />
  257. {t('settings.startRecording')}
  258. </>
  259. )}
  260. </Button>
  261. </div>
  262. </div>
  263. {testMessage && (
  264. <div
  265. className={cn(
  266. 'rounded-lg p-3 text-sm overflow-hidden',
  267. testStatus === 'success' &&
  268. 'bg-green-50 text-green-700 border border-green-200 dark:bg-green-950/50 dark:text-green-400 dark:border-green-800',
  269. testStatus === 'error' &&
  270. 'bg-red-50 text-red-700 border border-red-200 dark:bg-red-950/50 dark:text-red-400 dark:border-red-800',
  271. )}
  272. >
  273. <div className="flex items-start gap-2 min-w-0">
  274. {testStatus === 'success' && <CheckCircle2 className="h-4 w-4 mt-0.5 shrink-0" />}
  275. {testStatus === 'error' && <XCircle className="h-4 w-4 mt-0.5 shrink-0" />}
  276. <p className="flex-1 min-w-0 break-all">{testMessage}</p>
  277. </div>
  278. </div>
  279. )}
  280. {/* Model Selection */}
  281. {asrProvider.models.length > 0 && (
  282. <div className="space-y-2">
  283. <Label className="text-sm">{t('settings.ttsModel')}</Label>
  284. <Select
  285. value={asrProvidersConfig[selectedProviderId]?.modelId || asrProvider.defaultModelId}
  286. onValueChange={(value) => setASRProviderConfig(selectedProviderId, { modelId: value })}
  287. >
  288. <SelectTrigger>
  289. <SelectValue />
  290. </SelectTrigger>
  291. <SelectContent>
  292. {asrProvider.models.map((model) => (
  293. <SelectItem key={model.id} value={model.id}>
  294. {model.name}
  295. </SelectItem>
  296. ))}
  297. </SelectContent>
  298. </Select>
  299. </div>
  300. )}
  301. </div>
  302. );
  303. }