1

Author: steve02081504

src/public/pages/scripts/features/speechRecognition.mjs

@@ -143,6 +143,7 @@ export async function openSpeechRecognitionSession({
 	ws.binaryType = 'arraybuffer'
 
 	let finalSettled = false
+	let closedIntentionally = false
 	/** @type {(value: any) => void} */
 	let resolveFinal
 	/** @type {(reason?: any) => void} */
@@ -181,7 +182,9 @@ export async function openSpeechRecognitionSession({
 	})
 
 	ws.addEventListener('close', () => {
-		if (!finalSettled) rejectFinal(new Error('SpeechRecognition websocket closed unexpectedly'))
+		if (finalSettled) return
+		if (closedIntentionally) resolveFinal({ text: '' })
+		else rejectFinal(new Error('SpeechRecognition websocket closed unexpectedly'))
 	})
 
 	ws.send(JSON.stringify({ type: 'start', language, hotwords }))
@@ -210,6 +213,8 @@ export async function openSpeechRecognitionSession({
 		 * @returns {void}
 		 */
 		close: () => {
+			closedIntentionally = true
+			if (!finalSettled) resolveFinal({ text: '' })
 			try { ws.close() } catch { /* ignore */ }
 		},
 	}

src/public/parts/serviceGenerators/SpeechRecognition/fallback/main.mjs

@@ -42,15 +42,15 @@ const configTemplate = {
  */
 async function GetSource(config, { username, SaveConfig }) {
 	const unnamedSources = []
-	let sources
-	try {
-		sources = await Promise.all(config.sources.map(source => loadSpeechRecognitionSourceFromNameOrConfigData(username, source, unnamedSources, {
-			SaveConfig
-		})))
-	} catch (e) {
+	const settled = await Promise.allSettled(config.sources.map(source => loadSpeechRecognitionSourceFromNameOrConfigData(username, source, unnamedSources, {
+		SaveConfig
+	})))
+	const failed = settled.find(entry => entry.status === 'rejected')
+	if (failed) {
 		await Promise.all(unnamedSources.map(source => source?.Unload?.()))
-		throw e
+		throw failed.reason
 	}
+	const sources = settled.map(entry => /** @type {PromiseFulfilledResult<SpeechRecognitionSource_t>} */entry.value)
 	return {
 		type: 'speech-recognition',
 		info: buildSourceInfo(product_info, { name: config.name, provider: config.provider || 'unknown' }),

src/public/parts/serviceGenerators/SpeechRecognition/openai-whisper/main.mjs

@@ -54,7 +54,11 @@ async function GetSource(config) {
 			const wav = pcmToWav(pcm)
 			const form = new FormData()
 			form.append('model', model)
-			form.append('language', normalizeLang(options.language, 'auto'))
+			const rawLang = String(options.language || '').trim().toLowerCase()
+			if (rawLang && rawLang !== 'auto') {
+				const language = normalizeLang(options.language, 'short')
+				if (/^[a-z]{2}$/.test(language)) form.append('language', language)
+			}
 			if (options.hotwords?.length)
 				form.append('prompt', options.hotwords.join(', '))
 			form.append('file', new Blob([wav], { type: 'audio/wav' }), 'audio.wav')

src/public/parts/serviceGenerators/SpeechRecognition/shared/recognizeHelpers.mjs

@@ -127,10 +127,10 @@ export function openWs(url, opts = {}) {
  * @returns {string} 文本
  */
 export function extractRtasrText(resultData) {
-	let out = ''
-	for (const rt of resultData?.cn?.st?.rt || [])
-		for (const ws of rt.ws || [])
-			for (const cw of ws.cw || [])
-				out += cw.w || ''
-	return out
+	let text = ''
+	for (const recognitionResult of resultData?.cn?.st?.rt || [])
+		for (const wordSegment of recognitionResult.ws || [])
+			for (const candidateWord of wordSegment.cw || [])
+				text += candidateWord.w || ''
+	return text
 }

src/public/parts/serviceGenerators/SpeechRecognition/xfyun-lfasr-fast/main.mjs

@@ -58,16 +58,16 @@ function parseJsonField(value) {
 function parseFastResult(result) {
 	const parsed = parseJsonField(result)
 	if (!parsed) return typeof result === 'string' ? result : ''
-	let out = ''
+	let text = ''
 	for (const item of parsed.lattice || []) {
 		const best = parseJsonField(item.json_1best)
 		if (!best) continue
-		for (const rt of best.st?.rt || [])
-			for (const ws of rt.ws || [])
-				for (const cw of ws.cw || [])
-					out += cw.w || ''
+		for (const recognitionResult of best.st?.rt || [])
+			for (const wordSegment of recognitionResult.ws || [])
+				for (const candidateWord of wordSegment.cw || [])
+					text += candidateWord.w || ''
 	}
-	return out
+	return text
 }
 
 /**

src/public/parts/serviceGenerators/SpeechRecognition/xfyun-rtasr-std/main.mjs

@@ -42,11 +42,18 @@ const configTemplate = {
  */
 function sleep(ms, signal) {
 	return new Promise((resolve, reject) => {
-		const timer = setTimeout(resolve, ms)
-		signal?.addEventListener('abort', () => {
+		/**
+		 * @returns {void}
+		 */
+		const onAbort = () => {
 			clearTimeout(timer)
 			reject(signal.reason instanceof Error ? signal.reason : new Error('aborted'))
-		}, { once: true })
+		}
+		const timer = setTimeout(() => {
+			signal?.removeEventListener('abort', onAbort)
+			resolve()
+		}, ms)
+		signal?.addEventListener('abort', onAbort)
 	})
 }
 

src/public/parts/serviceGenerators/SpeechRecognition/xfyun-rtasr/main.mjs

@@ -43,11 +43,18 @@ const configTemplate = {
  */
 function sleep(ms, signal) {
 	return new Promise((resolve, reject) => {
-		const timer = setTimeout(resolve, ms)
-		signal?.addEventListener('abort', () => {
+		/**
+		 * @returns {void}
+		 */
+		const onAbort = () => {
 			clearTimeout(timer)
 			reject(signal.reason instanceof Error ? signal.reason : new Error('aborted'))
-		}, { once: true })
+		}
+		const timer = setTimeout(() => {
+			signal?.removeEventListener('abort', onAbort)
+			resolve()
+		}, ms)
+		signal?.addEventListener('abort', onAbort)
 	})
 }
 

src/public/parts/serviceGenerators/SpeechRecognition/xfyun-spark/main.mjs

@@ -77,7 +77,7 @@ function decodePayload(msg) {
 function applySegment(segments, decoded, lastText) {
 	const { piece, pgs, rg, sn } = decoded
 	if (pgs === 'rpl' && Array.isArray(rg) && rg.length >= 2) {
-		for (let i = rg[0]; i <= rg[1]; i++) segments.delete(i)
+		for (let segmentIndex = rg[0]; segmentIndex <= rg[1]; segmentIndex++) segments.delete(segmentIndex)
 		segments.set(sn, piece)
 		return buildSegmentText(segments)
 	}

src/public/parts/shells/chat/public/hub/composerFiles.mjs

@@ -32,8 +32,8 @@ async function setVoiceBtnIcon(recording) {
 }
 
 let isRecording = false
-/** @type {Awaited<ReturnType<typeof startVoiceRecording>> | null} */
-let voiceSession = null
+/** @type {Promise<Awaited<ReturnType<typeof startVoiceRecording>>> | null} */
+let voiceSessionPromise = null
 
 /**
  * @returns {HTMLElement|null} 附件预览容器
@@ -151,24 +151,34 @@ export async function toggleVoiceRecording() {
 	if (!voiceButton) return
 
 	if (isRecording) {
-		const session = voiceSession
-		voiceSession = null
-		await setVoiceBtnIcon(false)
+		const sessionPromise = voiceSessionPromise
+		voiceSessionPromise = null
 		isRecording = false
-		const file = await session?.stop()
-		if (!file) return
-		const [added] = await addFilesFromEvent({ target: { files: [file] } })
-		if (added) await maybeRecognizeVoiceAttachment(added.file, added.element, file)
+		await setVoiceBtnIcon(false)
+		try {
+			const session = await sessionPromise
+			const file = await session?.stop()
+			if (!file) return
+			const [added] = await addFilesFromEvent({ target: { files: [file] } })
+			if (added) await maybeRecognizeVoiceAttachment(added.file, added.element, file)
+		}
+		catch { /* 启动失败或 stop 失败时忽略 */ }
 		return
 	}
 
+	isRecording = true
+	const starting = startVoiceRecording()
+	voiceSessionPromise = starting
 	try {
-		isRecording = true
-		voiceSession = await startVoiceRecording()
+		await starting
+		if (voiceSessionPromise !== starting) return
 		await setVoiceBtnIcon(true)
 	}
 	catch {
+		if (voiceSessionPromise !== starting) return
+		voiceSessionPromise = null
 		isRecording = false
+		await setVoiceBtnIcon(false)
 		showToastI18n('error', 'chat.voiceRecording.errorAccessingMicrophone')
 	}
 }
@@ -178,11 +188,14 @@ export async function toggleVoiceRecording() {
  * @returns {Promise<void>}
  */
 export async function stopVoiceIfRecording() {
-	if (isRecording && voiceSession) {
-		const session = voiceSession
-		voiceSession = null
-		isRecording = false
-		await setVoiceBtnIcon(false)
-		await session.stop().catch(() => null)
+	if (!isRecording) return
+	const sessionPromise = voiceSessionPromise
+	voiceSessionPromise = null
+	isRecording = false
+	await setVoiceBtnIcon(false)
+	try {
+		const session = await sessionPromise
+		await session?.stop()
 	}
+	catch { /* ignore */ }
 }

src/public/parts/shells/chat/public/hub/messages/render/file.mjs

@@ -132,7 +132,7 @@ let revealScheduled = false
  * 合并调度 `revealMessageAudioSpeechRecognitionItems(#messages)`：同一微任务队列内只执行一次。
  * @returns {void}
  */
-function scheduleRevealMessageAudioSpeechRecognitionItems() {
+export function scheduleRevealMessageAudioSpeechRecognitionItems() {
 	if (revealScheduled) return
 	revealScheduled = true
 	queueMicrotask(() => {
@@ -220,9 +220,7 @@ export async function renderMessageFileIdsHtml(message) {
 		rows.push(await renderSingleFileAttachmentHtml(groupId, id, meta, mime, alt))
 	}
 	if (!rows.length) return ''
-	const html = `<div class="message-files flex flex-col gap-1 mt-1">${rows.join('')}</div>`
-	scheduleRevealMessageAudioSpeechRecognitionItems()
-	return html
+	return `<div class="message-files flex flex-col gap-1 mt-1">${rows.join('')}</div>`
 }
 
 /**