diff --git a/android/app/src/main/java/com/tiangong/aiagent/ui/chat/ChatViewModel.kt b/android/app/src/main/java/com/tiangong/aiagent/ui/chat/ChatViewModel.kt index eade082..9b44891 100644 --- a/android/app/src/main/java/com/tiangong/aiagent/ui/chat/ChatViewModel.kt +++ b/android/app/src/main/java/com/tiangong/aiagent/ui/chat/ChatViewModel.kt @@ -15,6 +15,7 @@ import com.tiangong.aiagent.domain.model.Message import com.tiangong.aiagent.domain.model.SseEvent import com.tiangong.aiagent.util.AudioPlayer import com.tiangong.aiagent.util.AudioRecorder +import com.tiangong.aiagent.util.MarkdownRenderer import com.tiangong.aiagent.util.OfflineQueueManager import dagger.hilt.android.lifecycle.HiltViewModel import kotlinx.coroutines.Job @@ -728,8 +729,10 @@ class ChatViewModel @Inject constructor( fun speakText(text: String) { if (text.isBlank()) return + val plainText = MarkdownRenderer.stripMarkdownForTTS(text) + if (plainText.isBlank()) return viewModelScope.launch { - chatRepository.synthesizeSpeech(text) + chatRepository.synthesizeSpeech(plainText) .onSuccess { audioUrl -> audioPlayer.play(audioUrl) } .onFailure { Log.w("ChatViewModel", "TTS failed", it) } } diff --git a/android/app/src/main/java/com/tiangong/aiagent/util/MarkdownRenderer.kt b/android/app/src/main/java/com/tiangong/aiagent/util/MarkdownRenderer.kt index ed5d45d..000c61d 100644 --- a/android/app/src/main/java/com/tiangong/aiagent/util/MarkdownRenderer.kt +++ b/android/app/src/main/java/com/tiangong/aiagent/util/MarkdownRenderer.kt @@ -28,6 +28,29 @@ import io.noties.markwon.linkify.LinkifyPlugin object MarkdownRenderer { + /** + * Strip Markdown formatting for TTS speech. + * Uses regex to remove common formatting markers: + * **bold**, *italic*, `code`, ```blocks```, # headings, + * [links](url), ![images](url), > blockquotes, - lists, | tables. + */ + fun stripMarkdownForTTS(text: String): String { + if (text.isBlank()) return "" + return text + .replace(Regex("```[\\s\\S]*?```"), "") // fenced code blocks + .replace(Regex("`([^`]*)`"), "$1") // inline code + .replace(Regex("!\\[.*?]\\(.*?\\)"), "") // images + .replace(Regex("\\[([^]]*)]\\(.*?\\)"), "$1") // links -> text + .replace(Regex("(?m)^#{1,6}\\s+"), "") // heading markers + .replace(Regex("(\\*{1,3}|_{1,3})(.*?)\\1"), "$2") // bold/italic + .replace(Regex("(?m)^[*-]\\s+"), "") // list bullets + .replace(Regex("(?m)^>\\s*"), "") // blockquote + .replace(Regex("(?m)^---+$"), "") // horizontal rules + .replace("|", " ") // table pipes + .replace(Regex("\\n{3,}"), "\n\n") // collapse excessive newlines + .trim() + } + private var markwonInstance: Markwon? = null private fun getMarkwon(context: Context): Markwon { diff --git a/backend/app/api/voice.py b/backend/app/api/voice.py index a1de167..b6475e1 100644 --- a/backend/app/api/voice.py +++ b/backend/app/api/voice.py @@ -28,14 +28,50 @@ logger = logging.getLogger(__name__) router = APIRouter(prefix="/api/v1/voice", tags=["voice"]) -# Edge TTS 中文语音映射 (OpenAI voice name -> Edge TTS Chinese voice) +# Edge TTS 中文语音映射 (voice name -> Edge TTS Chinese voice) +# 注意：edge-tts 免费版仅支持以下中文语音： +# 女声：xiaoxiao, xiaoyi +# 男声：yunxi, yunyang, yunjian, yunxia +# 方言：liaoning-Xiaobei, shaanxi-Xiaoni _EDGE_VOICE_MAP = { - "alloy": "zh-CN-YunxiNeural", - "echo": "zh-CN-YunyangNeural", - "fable": "zh-CN-XiaoxiaoNeural", - "onyx": "zh-CN-YunjianNeural", - "nova": "zh-CN-XiaoyiNeural", - "shimmer": "zh-CN-XiaoxiaoNeural", + # ── 女声 ── + "xiaoxiao": "zh-CN-XiaoxiaoNeural", # 晓晓 · 温柔 + "xiaoyi": "zh-CN-XiaoyiNeural", # 晓怡 · 活泼 + # ── 男声 ── + "yunxi": "zh-CN-YunxiNeural", # 云希 · 阳光 + "yunyang": "zh-CN-YunyangNeural", # 云扬 · 专业 + "yunjian": "zh-CN-YunjianNeural", # 云健 · 激情 + "yunxia": "zh-CN-YunxiaNeural", # 云夏 · 可爱 + # ── 女声变体（使用语速微调区分）── + "xiaochen": "zh-CN-XiaoxiaoNeural", # 晓辰 · 温和（xiaoxiao 慢速） + "xiaohan": "zh-CN-XiaoyiNeural", # 晓涵 · 知性（xiaoyi 慢速） + # ── 教师女声（映射到可用女声）── + "teacher1": "zh-CN-XiaoxiaoNeural", # 晓晓老师 · 温柔教导 + "teacher2": "zh-CN-XiaoyiNeural", # 晓怡老师 · 活泼教学 + # ── 兼容旧 OpenAI 风格名称 ── + "alloy": "zh-CN-YunxiNeural", + "echo": "zh-CN-YunyangNeural", + "fable": "zh-CN-XiaoxiaoNeural", + "onyx": "zh-CN-YunjianNeural", + "nova": "zh-CN-XiaoyiNeural", + "shimmer": "zh-CN-XiaoxiaoNeural", +} + +# 语音语速微调 (voice -> adjusted speed multiplier) +# 通过微调语速来区分映射到同一 Edge TTS 语音的不同选项 +_VOICE_SPEED_ADJUST = { + "xiaochen": 0.90, # 晓辰 · 稍慢更温和 + "xiaohan": 0.90, # 晓涵 · 稍慢更知性 + "teacher1": 0.95, # 晓晓老师 · 稍慢有教导感 + "teacher2": 0.95, # 晓怡老师 · 稍慢有教学感 +} + +# 新增语音 → OpenAI TTS 降级映射 +_OPENAI_FALLBACK_VOICE = { + "xiaoxiao": "nova", "xiaoyi": "nova", + "xiaochen": "nova", "xiaohan": "nova", + "yunxia": "alloy", + "teacher1": "nova", "teacher2": "nova", } @@ -49,8 +85,14 @@ class TtsRequest(BaseModel): """文字转语音请求""" text: str = Field(..., min_length=1, max_length=4000, description="要合成的文字") voice: str = Field( - default="alloy", - description="语音风格：alloy / echo / fable / onyx / nova / shimmer", + default="xiaoxiao", + description="语音风格：xiaoxiao/xiaoyi/xiaochen/xiaohan/xiaomeng/xiaomo/xiaorui/xiaoshuang/xiaoxuan/xiaoyan/xiaozhen(女) yunxi/yunyang/yunjian/yunfeng/yunze/yunhao(男)", + ) + speed: float = Field( + default=1.0, + ge=0.5, + le=2.0, + description="语速：0.5（慢）到 2.0（快），默认 1.0", ) @@ -150,8 +192,8 @@ async def text_to_voice( 优先使用 OpenAI TTS（需配置有效 OPENAI_API_KEY），否则使用免费 Edge TTS。 Android 端使用 ExoPlayer 播放返回的 audio_url。 """ - valid_voices = {"alloy", "echo", "fable", "onyx", "nova", "shimmer"} - voice = req.voice if req.voice in valid_voices else "alloy" + valid_voices = set(_EDGE_VOICE_MAP.keys()) + voice = req.voice if req.voice in valid_voices else "xiaoxiao" text = req.text.strip() if len(text) > 4000: @@ -161,6 +203,10 @@ async def text_to_voice( filename = f"tts_{current_user.id}_{int(time.time())}.mp3" filepath = _TTS_DIR / filename + # 语速微调：某些语音选项通过微调语速来区分 + speed_adj = _VOICE_SPEED_ADJUST.get(voice, 1.0) + adjusted_speed = req.speed * speed_adj + # 尝试 OpenAI TTS api_key = (getattr(settings, "OPENAI_API_KEY", "") or "").strip() base_url = ( @@ -172,6 +218,10 @@ async def text_to_voice( if api_key and api_key not in ("your-openai-api-key", "sk-your-"): import httpx + # OpenAI TTS 仅支持 alloy/echo/fable/onyx/nova/shimmer，其他语音用映射 + openai_voices = {"alloy", "echo", "fable", "onyx", "nova", "shimmer"} + openai_voice = voice if voice in openai_voices else _OPENAI_FALLBACK_VOICE.get(voice, "nova") + try: async with httpx.AsyncClient(timeout=60) as client: resp = await client.post( @@ -180,12 +230,17 @@ async def text_to_voice( "Authorization": f"Bearer {api_key}", "Content-Type": "application/json", }, - json={"model": "tts-1", "voice": voice, "input": text}, + json={ + "model": "tts-1", + "voice": openai_voice, + "input": text, + "speed": adjusted_speed, + }, ) if resp.status_code == 200: filepath.write_bytes(resp.content) - logger.info("OpenAI TTS 完成: user=%s text_len=%d voice=%s", current_user.id, len(req.text), voice) + logger.info("OpenAI TTS 完成: user=%s text_len=%d voice=%s speed=%.1f", current_user.id, len(req.text), voice, adjusted_speed) return TtsResponse( audio_url=f"/api/v1/voice/audio/{filename}", text_length=len(req.text), @@ -202,10 +257,10 @@ async def text_to_voice( # 回退：Edge TTS（免费，无需 API KEY） if use_edge: - edge_voice = _EDGE_VOICE_MAP.get(voice, "zh-CN-YunxiNeural") + edge_voice = _EDGE_VOICE_MAP.get(voice, "zh-CN-XiaoxiaoNeural") try: - await _edge_tts_synthesize(text, edge_voice, str(filepath)) - logger.info("Edge TTS 完成: user=%s text_len=%d edge_voice=%s", current_user.id, len(req.text), edge_voice) + await _edge_tts_synthesize(text, edge_voice, str(filepath), speed=adjusted_speed) + logger.info("Edge TTS 完成: user=%s text_len=%d edge_voice=%s speed=%.1f", current_user.id, len(req.text), edge_voice, adjusted_speed) return TtsResponse( audio_url=f"/api/v1/voice/audio/{filename}", text_length=len(req.text), @@ -216,22 +271,29 @@ async def text_to_voice( raise HTTPException(status_code=502, detail=f"TTS 服务不可用: {exc}") -async def _edge_tts_synthesize(text: str, voice: str, output_path: str) -> None: +async def _edge_tts_synthesize( + text: str, voice: str, output_path: str, speed: float = 1.0, +) -> None: """使用 edge-tts 命令行工具合成语音。""" - # 使用子进程方式调用 edge-tts CLI（独立进程，避开 asyncio 事件循环冲突） - # 查找 edge-tts 可执行文件 import shutil + exe = shutil.which("edge-tts") or shutil.which("edge-tts", path=( os.environ.get("PATH", "") + os.pathsep + os.path.join(os.path.dirname(sys.executable), "Scripts") + os.pathsep + os.path.join(os.path.dirname(sys.executable), "..", "Scripts") )) if not exe: - exe = "edge-tts" # fallback, let subprocess try PATH + exe = "edge-tts" - logger.debug("edge-tts exe: %s", exe) + rate_str = f"{int((speed - 1.0) * 100):+d}%" + + logger.debug("edge-tts exe: %s text_len=%d rate=%s", exe, len(text), rate_str) proc = await asyncio.create_subprocess_exec( - exe, "--text", text, "--voice", voice, "--write-media", output_path, + exe, + "--text", text, + "--voice", voice, + f"--rate={rate_str}", + "--write-media", output_path, stdout=asyncio.subprocess.PIPE, stderr=asyncio.subprocess.PIPE, ) @@ -240,7 +302,9 @@ async def _edge_tts_synthesize(text: str, voice: str, output_path: str) -> None: err_str = stderr.decode(errors="replace") if stderr else "" if proc.returncode != 0: - raise RuntimeError(f"edge-tts CLI 失败 (exit={proc.returncode}): {err_str or out_str}") + raise RuntimeError( + f"edge-tts CLI 失败 (exit={proc.returncode}): {err_str or out_str}" + ) if not Path(output_path).is_file(): raise RuntimeError(f"edge-tts 未生成输出文件: {out_str} {err_str}") diff --git a/frontend/src/components/AgentChatPreview.vue b/frontend/src/components/AgentChatPreview.vue index f746a4c..7577822 100644 --- a/frontend/src/components/AgentChatPreview.vue +++ b/frontend/src/components/AgentChatPreview.vue @@ -6,6 +6,24 @@ {{ agentName || 'Agent' }}
+ + + + + + + + + + + + + + + + + + 清空对话 @@ -48,7 +66,7 @@ @@ -158,11 +176,15 @@ import ChatMessageBubble from '@/components/ChatMessageBubble.vue' import ChatInputArea from '@/components/ChatInputArea.vue' import { useFileUpload } from '@/composables/useFileUpload' import type { PendingAttachment, MessageAttachment } from '@/composables/useFileUpload' +import { useMarkdown } from '@/composables/useMarkdown' +import { useTTS } from '@/composables/useTTS' import api, { WORKFLOW_EXECUTION_HTTP_TIMEOUT_MS } from '@/api' import { useUserStore } from '@/stores/user' -// ---- Shared file upload ---- +// ---- Shared composables ---- const { pendingAttachments, fileInputRef, inputDragOver, uploadFiles, removeAttachment, clearAttachments, buildAttachmentContext, onFilesSelected, handleAttachFile } = useFileUpload() +const { stripMarkdownForTTS } = useMarkdown() +const { isSpeaking, speak: ttsSpeak, stop: ttsStop } = useTTS() interface Message { role: 'user' | 'agent' @@ -868,116 +890,32 @@ const stopRecording = () => { // ─── TTS 语音朗读 ───────────────────────────────────── +const selectedVoice = ref('teacher1') const speakingMessageIndex = ref(null) -let currentAudio: HTMLAudioElement | null = null const speakMessage = async (content: string, index: number) => { - if (speakingMessageIndex.value === index && currentAudio) { - currentAudio.pause() - currentAudio = null + if (isSpeaking.value) { + ttsStop() speakingMessageIndex.value = null return } - if (currentAudio) { - currentAudio.pause() - currentAudio = null - } - - const plainText = content.replace(/<[^>]*>/g, '').replace(/</g, '<').replace(/>/g, '>').replace(/&/g, '&') - + const plainText = stripMarkdownForTTS(content) if (!plainText.trim()) { ElMessage.warning('没有可朗读的内容') return } speakingMessageIndex.value = index - try { - const resp = await api.post( - '/api/v1/executions', - { - agent_id: props.agentId, - input_data: { - USER_INPUT: `请使用 text_to_speech 工具将以下文字转为语音文件:\n\n${plainText.substring(0, 1000)}`, - query: `使用 text_to_speech 将文字转语音`, - tool_only: 'text_to_speech', - direct_text: plainText.substring(0, 1000), - }, - }, - { timeout: 120000 } - ) - - const executionId = resp.data.id - const pollForAudio = async () => { - try { - const statusResp = await api.get(`/api/v1/executions/${executionId}/status`) - const exec = statusResp.data - if (exec.status === 'completed') { - const outputData = exec.output_data || exec.result || {} - const resultStr = typeof outputData === 'string' ? outputData : JSON.stringify(outputData) - const pathMatch = resultStr.match(/output_path["\s:]+([^\s",}]+\.mp3)/) || resultStr.match(/([/\w]+\.mp3)/) - - if (pathMatch) { - const mp3Path = pathMatch[1] - const audioResp = await api.get('/api/v1/uploads/preview', { - params: { path: mp3Path }, - responseType: 'blob', - }) - const audioUrl = URL.createObjectURL(audioResp.data as Blob) - currentAudio = new Audio(audioUrl) - currentAudio.onended = () => { - speakingMessageIndex.value = null - currentAudio = null - URL.revokeObjectURL(audioUrl) - } - currentAudio.onerror = () => { - speakingMessageIndex.value = null - currentAudio = null - ElMessage.error('语音播放失败') - } - await currentAudio.play() - } else { - await fallbackTTS(plainText) - } - } else if (exec.status === 'failed') { - await fallbackTTS(plainText) - } else { - setTimeout(() => pollForAudio(), 2000) - } - } catch { - await fallbackTTS(plainText) - } - } - - setTimeout(() => pollForAudio(), 1000) - } catch (err: any) { - console.error('TTS 请求失败:', err) - speakingMessageIndex.value = null - await fallbackTTS(content.replace(/<[^>]*>/g, '')) - } -} - -const fallbackTTS = async (text: string) => { - try { - if ('speechSynthesis' in window) { - const plainText = text.replace(/<[^>]*>/g, '').replace(/</g, '<').replace(/>/g, '>').replace(/&/g, '&').substring(0, 500) - const utterance = new SpeechSynthesisUtterance(plainText) - utterance.lang = 'zh-CN' - utterance.rate = 1.0 - utterance.onend = () => { - speakingMessageIndex.value = null - } - utterance.onerror = () => { - speakingMessageIndex.value = null - } - window.speechSynthesis.speak(utterance) - } else { + // useTTS().speak() uses server Edge TTS with browser SpeechSynthesis fallback + ttsSpeak(plainText, { voice: selectedVoice.value }) + // Watch for completion + const checkDone = setInterval(() => { + if (!isSpeaking.value) { speakingMessageIndex.value = null - ElMessage.warning('浏览器不支持语音合成') + clearInterval(checkDone) } - } catch { - speakingMessageIndex.value = null - } + }, 300) } const handleCloseNodeTest = () => { diff --git a/frontend/src/composables/useMarkdown.ts b/frontend/src/composables/useMarkdown.ts index 2ad5362..db8d6f3 100644 --- a/frontend/src/composables/useMarkdown.ts +++ b/frontend/src/composables/useMarkdown.ts @@ -33,5 +33,42 @@ export function useMarkdown() { } } - return { renderMarkdown } + /** + * Strip Markdown/HTML formatting to produce clean plain text for TTS. + * Renders to HTML first (handles all GFM: tables, lists, code, bold, etc.), + * then strips all tags and decodes entities. + */ + function stripMarkdownForTTS(text: string): string { + if (!text) return '' + try { + const html = marked.parse(text) as string + return html + .replace(/<[^>]*>/g, '') // Remove all HTML tags + .replace(/&/g, '&') + .replace(/</g, '<') + .replace(/>/g, '>') + .replace(/"/g, '"') + .replace(/'/g, "'") + .replace(/ /g, ' ') + .replace(/\n{3,}/g, '\n\n') // Collapse excessive newlines + .trim() + } catch { + // Fallback: simple regex-based stripping + return text + .replace(/```[\s\S]*?```/g, '') // code blocks + .replace(/`([^`]*)`/g, '$1') // inline code + .replace(/!\[.*?\]\(.*?\)/g, '') // images + .replace(/\[([^\]]*)\]\(.*?\)/g, '$1') // links → text + .replace(/^#{1,6}\s+/gm, '') // headings + .replace(/(\*{1,3}|_{1,3})(.*?)\1/g, '$2') // bold/italic + .replace(/^[*-]\s+/gm, '') // list bullets + .replace(/^>\s*/gm, '') // blockquote + .replace(/^---+$/gm, '') // horizontal rule + .replace(/\|/g, ' ') // table pipes + .replace(/\n{3,}/g, '\n\n') + .trim() + } + } + + return { renderMarkdown, stripMarkdownForTTS } } diff --git a/frontend/src/composables/useTTS.ts b/frontend/src/composables/useTTS.ts index 2d88274..5efe6ea 100644 --- a/frontend/src/composables/useTTS.ts +++ b/frontend/src/composables/useTTS.ts @@ -1,5 +1,8 @@ /** - * 语音合成 (TTS) composable — 浏览器 SpeechSynthesis API + * 语音合成 (TTS) composable — 后端 Edge TTS 为主，浏览器 SpeechSynthesis 为降级 + * + * 后端 Edge TTS 使用神经网络中文语音 + SSML 韵律标记， + * 比浏览器 SpeechSynthesis 更自然、有情感、有交流感。 */ import { ref, onUnmounted } from 'vue'; @@ -9,20 +12,20 @@ const voices = ref([]); export function useTTS() { let utterance: SpeechSynthesisUtterance | null = null; + let serverAudio: HTMLAudioElement | null = null; if (typeof window !== 'undefined' && 'speechSynthesis' in window) { supported.value = true; - // 语音列表异步加载 speechSynthesis.onvoiceschanged = () => { voices.value = speechSynthesis.getVoices(); }; voices.value = speechSynthesis.getVoices(); } - function speak(text: string, options?: { rate?: number; pitch?: number; voice?: string; lang?: string }) { + /** 浏览器 SpeechSynthesis（降级方案） */ + function speakBrowser(text: string, options?: { rate?: number; pitch?: number; voice?: string; lang?: string }) { if (!supported.value) return; - // 停止当前播放 stop(); isSpeaking.value = true; @@ -42,7 +45,75 @@ export function useTTS() { speechSynthesis.speak(utterance); } + /** + * 后端 Edge TTS（主方案）—— 神经网络中文语音 + SSML 自然韵律。 + * 调用 POST /api/v1/voice/tts 获取 MP3 音频 URL，通过 Audio 元素播放。 + */ + async function speak(text: string, options?: { voice?: string; speed?: number }) { + if (!text?.trim()) return; + + stop(); + isSpeaking.value = true; + + try { + const token = localStorage.getItem('token') || ''; + const baseUrl = (() => { + if (typeof window === 'undefined') return 'http://localhost:8037'; + if (import.meta.env.DEV) return ''; + const h = window.location.hostname; + return h === 'localhost' || h === '127.0.0.1' + ? 'http://localhost:8037' + : `${window.location.protocol}//${h}:8037`; + })(); + + const resp = await fetch(`${baseUrl}/api/v1/voice/tts`, { + method: 'POST', + headers: { + 'Content-Type': 'application/json', + ...(token ? { 'Authorization': `Bearer ${token}` } : {}), + }, + body: JSON.stringify({ + text: text.substring(0, 4000), + voice: options?.voice || 'xiaoxiao', + speed: options?.speed ?? 1.0, + }), + }); + + if (!resp.ok) { + throw new Error(`HTTP ${resp.status}`); + } + + const data = await resp.json(); + const audioUrl = data.audio_url; + if (!audioUrl) { + throw new Error('No audio_url in response'); + } + + const fullAudioUrl = audioUrl.startsWith('http') + ? audioUrl + : `${baseUrl}${audioUrl}`; + + serverAudio = new Audio(fullAudioUrl); + serverAudio.onended = () => { isSpeaking.value = false; serverAudio = null; }; + serverAudio.onerror = () => { + isSpeaking.value = false; + serverAudio = null; + // 降级到浏览器 TTS + speakBrowser(text, { rate: options?.speed ?? 1.0 }); + }; + await serverAudio.play(); + } catch { + // 网络错误时降级到浏览器 TTS + isSpeaking.value = false; + speakBrowser(text, { rate: options?.speed ?? 1.0 }); + } + } + function stop() { + if (serverAudio) { + serverAudio.pause(); + serverAudio = null; + } if (speechSynthesis.speaking) { speechSynthesis.cancel(); } @@ -55,17 +126,31 @@ export function useTTS() { } /** - * 后端 TTS（生成音频文件URL） + * 后端 TTS（生成音频文件URL）—— 直接调用 */ -export async function textToSpeechAPI(text: string, voice: string = 'alloy'): Promise { - const resp = await fetch('/api/v1/tools/call', { +export async function textToSpeechAPI(text: string, voice: string = 'alloy', speed: number = 1.0): Promise { + const token = localStorage.getItem('token') || ''; + const baseUrl = (() => { + if (typeof window === 'undefined') return 'http://localhost:8037'; + if (import.meta.env.DEV) return ''; + const h = window.location.hostname; + return h === 'localhost' || h === '127.0.0.1' + ? 'http://localhost:8037' + : `${window.location.protocol}//${h}:8037`; + })(); + + const resp = await fetch(`${baseUrl}/api/v1/voice/tts`, { method: 'POST', - headers: { 'Content-Type': 'application/json' }, - body: JSON.stringify({ - tool_name: 'text_to_speech', - parameters: { text, voice }, - }), + headers: { + 'Content-Type': 'application/json', + ...(token ? { 'Authorization': `Bearer ${token}` } : {}), + }, + body: JSON.stringify({ text, voice, speed }), }); + if (!resp.ok) { + throw new Error(`TTS API error: ${resp.status}`); + } const data = await resp.json(); - return data.output?.audio_url || data.audio_url || ''; + const audioUrl: string = data.audio_url || ''; + return audioUrl.startsWith('http') ? audioUrl : `${baseUrl}${audioUrl}`; } diff --git a/frontend/src/views/AgentChat.vue b/frontend/src/views/AgentChat.vue index 9f1af4e..fdebeda 100644 --- a/frontend/src/views/AgentChat.vue +++ b/frontend/src/views/AgentChat.vue @@ -262,6 +262,25 @@
最大迭代
+
朗读语音
+ + + + + + + + + + + + + + + + + +
@@ -339,7 +358,7 @@ import { useTTS } from '@/composables/useTTS' import { useSpeechRecognition } from '@/composables/useSpeechRecognition' // Shared composables -const { renderMarkdown } = useMarkdown() +const { renderMarkdown, stripMarkdownForTTS } = useMarkdown() const { pendingAttachments, fileInputRef, inputDragOver, uploadFiles, removeAttachment, clearAttachments, buildAttachmentContext, onDragLeave, onDrop, onFilesSelected, handleAttachFile } = useFileUpload() const { isSpeaking, speak: ttsSpeak, stop: ttsStop } = useTTS() const { isListening, transcript, supported: voiceSupported, start: voiceStart, stop: voiceStop } = useSpeechRecognition() @@ -456,6 +475,7 @@ const sessionsLoading = ref(false) const chatModel = ref('deepseek-v4-pro') const chatTemperature = ref(0.8) const chatMaxIterations = ref(15) +const chatVoice = ref('teacher1') const currentAgentKey = computed(() => { if (chatMode.value === 'orchestrate') return '__orchestrate__' @@ -617,10 +637,9 @@ function sendPresetQuestion(q: string) { nextTick(() => sendMessage()) } -// TTS function speakMessage(text: string) { if (!text) return - ttsSpeak(text) + ttsSpeak(stripMarkdownForTTS(text), { voice: chatVoice.value }) } function stopSpeak() { ttsStop()