Add external mic array support, UI i18n, glass settings redesign, and fixes

Backend:
- Audio: input-device selection + multichannel downmix/resample to 16k mono (devices.py, pyaudio_capture.py); facade routes default-device to native macOS AEC, external device to PyAudio (capture/__init__.py)
- API: GET /system/audio-devices, input_device_index in /system/start with persistence (audio_config.py); default system/ASR language set to Chinese (lifespan.py)
- CLI: --input-device and --list-audio-devices

Frontend:
- vue-i18n setup (en/zh), default UI + ASR language Chinese
- Unified tabbed Settings dialog (Main/Language/Prompt) replacing scattered modals; fixed-height tabs; mic + interface/recognition language icons
- Liquid-glass styling (frosted modal, glass buttons, smooth ant-fade open with backdrop-blur background)
- Wider caption area; unified bottom action-button height across Welcome and Home

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

assets/www/assets/{index-CCuJ1lip.css → index-6CUhaZ06.css}

@@ -1 +1 @@
-@charset "UTF-8";html,body{width:100%;height:100%}input::-ms-clear,input::-ms-reveal{display:none}*,*:before,*:after{box-sizing:border-box}html{font-family:sans-serif;line-height:1.15;-webkit-text-size-adjust:100%;-ms-text-size-adjust:100%;-ms-overflow-style:scrollbar;-webkit-tap-highlight-color:rgba(0,0,0,0)}body{margin:0}[tabindex="-1"]:focus{outline:none}hr{box-sizing:content-box;height:0;overflow:visible}h1,h2,h3,h4,h5,h6{margin-top:0;margin-bottom:.5em;font-weight:500}p{margin-top:0;margin-bottom:1em}abbr[title],abbr[data-original-title]{-webkit-text-decoration:underline dotted;text-decoration:underline;text-decoration:underline dotted;border-bottom:0;cursor:help}address{margin-bottom:1em;font-style:normal;line-height:inherit}input[type=text],input[type=password],input[type=number],textarea{-webkit-appearance:none}ol,ul,dl{margin-top:0;margin-bottom:1em}ol ol,ul ul,ol ul,ul ol{margin-bottom:0}dt{font-weight:500}dd{margin-bottom:.5em;margin-left:0}blockquote{margin:0 0 1em}dfn{font-style:italic}b,strong{font-weight:bolder}small{font-size:80%}sub,sup{position:relative;font-size:75%;line-height:0;vertical-align:baseline}sub{bottom:-.25em}sup{top:-.5em}pre,code,kbd,samp{font-size:1em;font-family:SFMono-Regular,Consolas,Liberation Mono,Menlo,Courier,monospace}pre{margin-top:0;margin-bottom:1em;overflow:auto}figure{margin:0 0 1em}img{vertical-align:middle;border-style:none}a,area,button,[role=button],input:not([type=range]),label,select,summary,textarea{touch-action:manipulation}table{border-collapse:collapse}caption{padding-top:.75em;padding-bottom:.3em;text-align:left;caption-side:bottom}input,button,select,optgroup,textarea{margin:0;color:inherit;font-size:inherit;font-family:inherit;line-height:inherit}button,input{overflow:visible}button,select{text-transform:none}button,html [type=button],[type=reset],[type=submit]{-webkit-appearance:button}button::-moz-focus-inner,[type=button]::-moz-focus-inner,[type=reset]::-moz-focus-inner,[type=submit]::-moz-focus-inner{padding:0;border-style:none}input[type=radio],input[type=checkbox]{box-sizing:border-box;padding:0}input[type=date],input[type=time],input[type=datetime-local],input[type=month]{-webkit-appearance:listbox}textarea{overflow:auto;resize:vertical}fieldset{min-width:0;margin:0;padding:0;border:0}legend{display:block;width:100%;max-width:100%;margin-bottom:.5em;padding:0;color:inherit;font-size:1.5em;line-height:inherit;white-space:normal}progress{vertical-align:baseline}[type=number]::-webkit-inner-spin-button,[type=number]::-webkit-outer-spin-button{height:auto}[type=search]{outline-offset:-2px;-webkit-appearance:none}[type=search]::-webkit-search-cancel-button,[type=search]::-webkit-search-decoration{-webkit-appearance:none}::-webkit-file-upload-button{font:inherit;-webkit-appearance:button}output{display:inline-block}summary{display:list-item}template{display:none}[hidden]{display:none!important}mark{padding:.2em;background-color:#feffe6}:root{font-family:Inter,system-ui,Avenir,Helvetica,Arial,sans-serif;line-height:1.5;font-weight:400;color-scheme:light dark;color:#ffffffde;background-color:#242424;font-synthesis:none;text-rendering:optimizeLegibility;-webkit-font-smoothing:antialiased;-moz-osx-font-smoothing:grayscale;-webkit-text-size-adjust:100%}a{font-weight:500;color:#646cff;text-decoration:inherit}a:hover{color:#535bf2}body{margin:0;display:flex;place-items:center;min-width:320px;height:100%;min-height:auto;color:#333;background:#fff}h1{font-size:3.2em;line-height:1.1}button{border-radius:8px;border:1px solid transparent;padding:.6em 1.2em;font-size:1em;font-weight:500;font-family:inherit;background-color:#1a1a1a;cursor:pointer;transition:border-color .25s}.card{border-bottom:solid 2px lightgray;align-items:center;justify-content:center;margin-top:40px;display:flex;max-width:1024px;width:100%}.seg-title{margin:24px 0;font-size:20px;font-weight:500}.seg-co{width:1022px;text-align:left;border-left:solid 6px midnightblue;padding-left:8px;margin-left:2px;margin-top:36px;line-height:24px}#app{margin:0 auto;padding:0;text-align:center;width:100%;height:100%}.ant-btn{padding:4px 12px}@media (prefers-color-scheme: light){:root{color:#213547;background-color:#fff}a:hover{color:#747bff}button{background-color:#f9f9f9}}.ant-card{background:#f5f6fa;height:100%}.ant-card-body{padding:24px 36px 12px!important;border-radius:0 0 8px 8px}.ant-card .ant-card-actions{background-color:#e8e8f8cc!important}.ant-popover{max-width:800px!important}.ant-form-item{background:transparent;margin-bottom:40px!important}.ant-form-item .ant-form-item-explain-error{color:#ff4d4f;text-align:left!important}.ant-form-item-label label{font-size:18px!important;color:#1a1a1a!important;font-weight:500!important}.ant-tooltip{max-width:1022px!important}.ant-page-header-heading{width:1022px!important}.highlight{background:#f8f8ff}.ant-layout-sider-collapsed{width:0!important;min-width:0!important;overflow:hidden}.ant-layout-sider-collapsed .ant-menu-item,.ant-layout-sider-collapsed .ant-menu-submenu-title{display:none}.header-nav[data-v-07594418]{display:flex;align-items:flex-start;justify-content:space-between;width:100vw;height:40px;align-items:center;position:absolute;top:0;left:0;z-index:99;-webkit-app-region:drag;cursor:move}.header-nav .window-controls[data-v-07594418],.header-nav button[data-v-07594418],.header-nav .ant-input-search[data-v-07594418],.header-nav img[data-v-07594418],.header-nav .anticon[data-v-07594418]{-webkit-app-region:no-drag;cursor:pointer}.header-nav .window-controls[data-v-07594418]{top:0;right:0;display:flex;z-index:1000;margin-left:12px}.header-nav .window-controls .window-control-btn[data-v-07594418]{width:46px;height:32px;border:none;background:transparent;color:#666;font-size:16px;cursor:pointer;display:flex;align-items:center;justify-content:center;transition:background-color .2s}.header-nav .window-controls .window-control-btn[data-v-07594418]:hover{background-color:#0000001a}.header-nav .window-controls .window-control-btn.close[data-v-07594418]:hover{background-color:#e81123;color:#fff}.header-nav .window-controls .close-icon.focus[data-v-07594418]{display:none}.header-nav .window-controls:hover .close-icon.default[data-v-07594418],.header-nav .window-controls:focus-within .close-icon.default[data-v-07594418]{display:none}.header-nav .window-controls:hover .close-icon.focus[data-v-07594418],.header-nav .window-controls:focus-within .close-icon.focus[data-v-07594418]{display:inline}.content[data-v-874ca48f]{background-color:#fff;margin:0 auto;display:flex;flex-direction:column;align-items:center;justify-content:space-between}.not-found-wrapper[data-v-aef52a59]{height:calc(100vh - 104px)}.btn-groups[data-v-839398ff]{margin-top:36px;display:flex;justify-content:flex-end;align-items:center}.prompt-title p[data-v-839398ff]{margin:0;font-size:16px;font-weight:500}.prompt-content[data-v-839398ff]{margin-top:16px}.prompt-content .prompt-title[data-v-839398ff]{margin-bottom:24px;font-size:22px;font-weight:500;text-align:center}.prompt-content .language-segment[data-v-839398ff]{display:flex;justify-content:center;margin-bottom:16px}.prompt-content .prompt-item[data-v-839398ff]{margin-top:16px}.languages[data-v-cd713caa]{margin-top:24px;margin-bottom:24px}.languages p[data-v-cd713caa]{font-size:16px;font-weight:500;margin-bottom:8px}.audio-play-btn[data-v-cd713caa]{padding:2px 8px 0;border-radius:4px;transition:all .2s;height:40px}.audio-play-btn[data-v-cd713caa]:hover{background-color:#f0f0f0}.audio-play-btn.playing[data-v-cd713caa]{background-color:#f6ffed;border-color:#1890ff}.audio-play-btn.playing .playing-icon[data-v-cd713caa]{animation:pulse-cd713caa 1.5s infinite}@keyframes pulse-cd713caa{0%{opacity:1;transform:scale(1)}50%{opacity:.7;transform:scale(1.1)}to{opacity:1;transform:scale(1)}}.btn-groups[data-v-cd713caa]{margin-top:36px;display:flex;justify-content:space-between;align-items:center}.custom-popover-list[data-v-cd713caa]{width:92px;margin:0}.custom-popover-list .custom-popover-item[data-v-cd713caa]{font-size:14px;line-height:36px;font-weight:500;color:#1e1e1e;cursor:pointer;border-radius:4px;padding:0 8px;margin:0 -8px;transition:background .2s}.custom-popover-list .custom-popover-item[data-v-cd713caa]:hover,.custom-popover-list .custom-popover-item[data-v-cd713caa]:focus{background:#e5e7eb}.welcome-wrapper[data-v-cd713caa]{width:100%;height:100%;background-image:url(./bg-BmnA8p_e.png);background-repeat:no-repeat;background-attachment:fixed;background-size:cover;background-position:center;display:flex;flex-direction:column;align-items:center;justify-content:space-between;color:#fff}.welcome-wrapper .content[data-v-cd713caa]{width:100%;height:80vh;display:flex;flex-direction:column;justify-content:space-around;margin-top:64px}.welcome-wrapper .content .inner-content[data-v-cd713caa]{display:flex;flex-direction:column;align-items:center;justify-content:center;text-align:center;padding:20px}.welcome-wrapper .content .inner-content .text-box[data-v-cd713caa]{color:#000;margin-bottom:36px}.welcome-wrapper .content .inner-content .text-box .title[data-v-cd713caa]{font-size:24px;font-weight:600;margin-bottom:24px}.welcome-wrapper .content .inner-content .text-box .sub-title[data-v-cd713caa]{font-size:15px;margin-top:10px}.welcome-wrapper .content .inner-content .btn-box[data-v-cd713caa]{width:224px;height:80px}.welcome-wrapper .actions[data-v-cd713caa]{width:100%;height:64px;display:flex;justify-content:flex-end}.ball-wrapper[data-v-34c8e583]{width:100%;height:calc(100vh - 100px);display:flex;flex-direction:column;align-items:center;justify-content:space-around}.talk-wrapper[data-v-1f502814]{width:auto;height:calc(100vh - 100px);overflow-y:scroll;padding:20px 240px 0;display:flex;flex-direction:column;align-items:flex-start;justify-content:flex-start}.talk-wrapper .cont-left[data-v-1f502814]{width:100%;margin:24px 0;display:flex;justify-content:flex-start;align-items:flex-start}.talk-wrapper .cont-left .text-left[data-v-1f502814]{color:#222;font-size:16px;font-weight:400;text-align:left;line-height:2;margin-left:12px;margin-top:6px}.talk-wrapper .cont-right[data-v-1f502814]{width:100%;margin:24px 0;display:flex;justify-content:flex-end;align-items:flex-start}.talk-wrapper .cont-right .text-right[data-v-1f502814]{color:#444;font-size:16px;font-weight:400;text-align:end;line-height:2;margin-right:12px;background:#ccc;border-radius:8px 0 8px 8px;padding:8px}.chat-wrapper[data-v-803600aa]{width:100%;height:100%;background-image:url(./bg-BmnA8p_e.png);background-repeat:no-repeat;background-attachment:fixed;background-size:cover;background-position:center;display:flex;flex-direction:column;align-items:center;justify-content:space-between;color:#fff}.chat-wrapper .content[data-v-803600aa]{width:100%;height:auto;display:flex;flex-direction:column;justify-content:space-around}.chat-wrapper .content .inner-content[data-v-803600aa]{display:flex;flex-direction:column;align-items:center;justify-content:center;text-align:center;padding:20px}.chat-wrapper .content .inner-content .text-box[data-v-803600aa]{color:#000;margin-bottom:36px}.chat-wrapper .content .inner-content .text-box .title[data-v-803600aa]{font-size:24px;font-weight:600;margin-bottom:24px}.chat-wrapper .content .inner-content .text-box .sub-title[data-v-803600aa]{font-size:15px;margin-top:10px}.chat-wrapper .content .inner-content .btn-box[data-v-803600aa]{width:224px;height:80px}.chat-wrapper .actions[data-v-803600aa]{width:100%;height:100px;display:flex;justify-content:space-between;align-items:center}.chat-wrapper .actions .holder[data-v-803600aa]{width:64px;height:48px}.chat-wrapper .actions .btns[data-v-803600aa]{width:450px;height:96px;display:flex;justify-content:space-around;align-items:flex-start}.chat-wrapper .actions .download-wrapper[data-v-803600aa]{width:64px;height:64px;display:flex;justify-content:flex-start;align-items:center;margin-right:0}.chat-wrapper .actions .download-wrapper img[data-v-803600aa]{width:24px;height:24px}.content-wrapper[data-v-d41c9ce7]{text-align:left;max-width:800px;min-width:320px;margin-bottom:64px;min-height:calc(100vh - 438px)}.content-wrapper .content-box[data-v-d41c9ce7]{padding:24px;height:240px;background-color:#e8e8e8;border-radius:16px;width:50%;margin:48px auto;min-width:300px}.content-wrapper .video-box[data-v-d41c9ce7]{max-width:800px;min-width:320px;width:90vw;height:auto}
+@charset "UTF-8";html,body{width:100%;height:100%}input::-ms-clear,input::-ms-reveal{display:none}*,*:before,*:after{box-sizing:border-box}html{font-family:sans-serif;line-height:1.15;-webkit-text-size-adjust:100%;-ms-text-size-adjust:100%;-ms-overflow-style:scrollbar;-webkit-tap-highlight-color:rgba(0,0,0,0)}body{margin:0}[tabindex="-1"]:focus{outline:none}hr{box-sizing:content-box;height:0;overflow:visible}h1,h2,h3,h4,h5,h6{margin-top:0;margin-bottom:.5em;font-weight:500}p{margin-top:0;margin-bottom:1em}abbr[title],abbr[data-original-title]{-webkit-text-decoration:underline dotted;text-decoration:underline;text-decoration:underline dotted;border-bottom:0;cursor:help}address{margin-bottom:1em;font-style:normal;line-height:inherit}input[type=text],input[type=password],input[type=number],textarea{-webkit-appearance:none}ol,ul,dl{margin-top:0;margin-bottom:1em}ol ol,ul ul,ol ul,ul ol{margin-bottom:0}dt{font-weight:500}dd{margin-bottom:.5em;margin-left:0}blockquote{margin:0 0 1em}dfn{font-style:italic}b,strong{font-weight:bolder}small{font-size:80%}sub,sup{position:relative;font-size:75%;line-height:0;vertical-align:baseline}sub{bottom:-.25em}sup{top:-.5em}pre,code,kbd,samp{font-size:1em;font-family:SFMono-Regular,Consolas,Liberation Mono,Menlo,Courier,monospace}pre{margin-top:0;margin-bottom:1em;overflow:auto}figure{margin:0 0 1em}img{vertical-align:middle;border-style:none}a,area,button,[role=button],input:not([type=range]),label,select,summary,textarea{touch-action:manipulation}table{border-collapse:collapse}caption{padding-top:.75em;padding-bottom:.3em;text-align:left;caption-side:bottom}input,button,select,optgroup,textarea{margin:0;color:inherit;font-size:inherit;font-family:inherit;line-height:inherit}button,input{overflow:visible}button,select{text-transform:none}button,html [type=button],[type=reset],[type=submit]{-webkit-appearance:button}button::-moz-focus-inner,[type=button]::-moz-focus-inner,[type=reset]::-moz-focus-inner,[type=submit]::-moz-focus-inner{padding:0;border-style:none}input[type=radio],input[type=checkbox]{box-sizing:border-box;padding:0}input[type=date],input[type=time],input[type=datetime-local],input[type=month]{-webkit-appearance:listbox}textarea{overflow:auto;resize:vertical}fieldset{min-width:0;margin:0;padding:0;border:0}legend{display:block;width:100%;max-width:100%;margin-bottom:.5em;padding:0;color:inherit;font-size:1.5em;line-height:inherit;white-space:normal}progress{vertical-align:baseline}[type=number]::-webkit-inner-spin-button,[type=number]::-webkit-outer-spin-button{height:auto}[type=search]{outline-offset:-2px;-webkit-appearance:none}[type=search]::-webkit-search-cancel-button,[type=search]::-webkit-search-decoration{-webkit-appearance:none}::-webkit-file-upload-button{font:inherit;-webkit-appearance:button}output{display:inline-block}summary{display:list-item}template{display:none}[hidden]{display:none!important}mark{padding:.2em;background-color:#feffe6}:root{font-family:Inter,system-ui,Avenir,Helvetica,Arial,sans-serif;line-height:1.5;font-weight:400;color-scheme:light dark;color:#ffffffde;background-color:#242424;font-synthesis:none;text-rendering:optimizeLegibility;-webkit-font-smoothing:antialiased;-moz-osx-font-smoothing:grayscale;-webkit-text-size-adjust:100%}a{font-weight:500;color:#646cff;text-decoration:inherit}a:hover{color:#535bf2}body{margin:0;display:flex;place-items:center;min-width:320px;height:100%;min-height:auto;color:#333;background:#fff}h1{font-size:3.2em;line-height:1.1}button{border-radius:8px;border:1px solid transparent;padding:.6em 1.2em;font-size:1em;font-weight:500;font-family:inherit;background-color:#1a1a1a;cursor:pointer;transition:border-color .25s}.card{border-bottom:solid 2px lightgray;align-items:center;justify-content:center;margin-top:40px;display:flex;max-width:1024px;width:100%}.seg-title{margin:24px 0;font-size:20px;font-weight:500}.seg-co{width:1022px;text-align:left;border-left:solid 6px midnightblue;padding-left:8px;margin-left:2px;margin-top:36px;line-height:24px}#app{margin:0 auto;padding:0;text-align:center;width:100%;height:100%}.ant-btn{padding:4px 12px}@media (prefers-color-scheme: light){:root{color:#213547;background-color:#fff}a:hover{color:#747bff}button{background-color:#f9f9f9}}.ant-card{background:#f5f6fa;height:100%}.ant-card-body{padding:24px 36px 12px!important;border-radius:0 0 8px 8px}.ant-card .ant-card-actions{background-color:#e8e8f8cc!important}.ant-popover{max-width:800px!important}.ant-form-item{background:transparent;margin-bottom:40px!important}.ant-form-item .ant-form-item-explain-error{color:#ff4d4f;text-align:left!important}.ant-form-item-label label{font-size:18px!important;color:#1a1a1a!important;font-weight:500!important}.ant-tooltip{max-width:1022px!important}.ant-page-header-heading{width:1022px!important}.highlight{background:#f8f8ff}.ant-layout-sider-collapsed{width:0!important;min-width:0!important;overflow:hidden}.ant-layout-sider-collapsed .ant-menu-item,.ant-layout-sider-collapsed .ant-menu-submenu-title{display:none}.ant-modal .ant-modal-content{background:#ffffff9e!important;backdrop-filter:blur(28px) saturate(140%);-webkit-backdrop-filter:blur(28px) saturate(140%);border:1px solid rgba(255,255,255,.6);border-radius:22px!important;box-shadow:0 16px 48px #1f26872e}.ant-modal .ant-modal-header{background:transparent!important}.ant-modal-mask{background:#14161e1f!important;backdrop-filter:blur(14px) saturate(120%);-webkit-backdrop-filter:blur(14px) saturate(120%)}.ant-select .ant-select-selector,.ant-input,textarea.ant-input,.ant-input-affix-wrapper{background:#ffffff73!important;backdrop-filter:blur(8px);-webkit-backdrop-filter:blur(8px);border:1px solid rgba(255,255,255,.7)!important}.ant-btn:not(.ant-btn-text):not(.ant-btn-link){box-shadow:0 2px 10px #1f26871a}.ant-btn-default{background:#ffffff80!important;border:1px solid rgba(255,255,255,.75)!important;backdrop-filter:blur(8px);-webkit-backdrop-filter:blur(8px)}.ant-btn-text{box-shadow:none!important;background:transparent!important}.ant-radio-group-solid .ant-radio-button-wrapper:first-child{border-top-left-radius:12px;border-bottom-left-radius:12px}.ant-radio-group-solid .ant-radio-button-wrapper:last-child{border-top-right-radius:12px;border-bottom-right-radius:12px}.header-nav[data-v-07594418]{display:flex;align-items:flex-start;justify-content:space-between;width:100vw;height:40px;align-items:center;position:absolute;top:0;left:0;z-index:99;-webkit-app-region:drag;cursor:move}.header-nav .window-controls[data-v-07594418],.header-nav button[data-v-07594418],.header-nav .ant-input-search[data-v-07594418],.header-nav img[data-v-07594418],.header-nav .anticon[data-v-07594418]{-webkit-app-region:no-drag;cursor:pointer}.header-nav .window-controls[data-v-07594418]{top:0;right:0;display:flex;z-index:1000;margin-left:12px}.header-nav .window-controls .window-control-btn[data-v-07594418]{width:46px;height:32px;border:none;background:transparent;color:#666;font-size:16px;cursor:pointer;display:flex;align-items:center;justify-content:center;transition:background-color .2s}.header-nav .window-controls .window-control-btn[data-v-07594418]:hover{background-color:#0000001a}.header-nav .window-controls .window-control-btn.close[data-v-07594418]:hover{background-color:#e81123;color:#fff}.header-nav .window-controls .close-icon.focus[data-v-07594418]{display:none}.header-nav .window-controls:hover .close-icon.default[data-v-07594418],.header-nav .window-controls:focus-within .close-icon.default[data-v-07594418]{display:none}.header-nav .window-controls:hover .close-icon.focus[data-v-07594418],.header-nav .window-controls:focus-within .close-icon.focus[data-v-07594418]{display:inline}.content[data-v-b8a456cb]{background-color:#fff;margin:0 auto;display:flex;flex-direction:column;align-items:center;justify-content:space-between}.not-found-wrapper[data-v-aef52a59]{height:calc(100vh - 104px)}.tab-body[data-v-c6781f0b]{height:360px;overflow-y:auto;padding:4px 8px 4px 2px}.setting-row[data-v-c6781f0b]{margin-bottom:20px}.setting-row>label[data-v-c6781f0b]{display:block;font-size:15px;font-weight:500;margin-bottom:8px}.setting-row>label .label-icon[data-v-c6781f0b]{margin-right:6px;color:#1890ff}.setting-row .hint[data-v-c6781f0b]{font-size:12px;color:#999;margin:8px 0 0}.setting-row .row-inline[data-v-c6781f0b]{display:flex;align-items:center;justify-content:space-between}.voice-group[data-v-c6781f0b]{display:flex;flex-direction:column;margin-top:8px}.voice-radio[data-v-c6781f0b]{display:flex;align-items:center;height:40px;line-height:40px}.voice-radio .voice-name[data-v-c6781f0b]{margin-right:8px}.audio-play-btn[data-v-c6781f0b]{padding:0 6px;border-radius:4px}.audio-play-btn.playing[data-v-c6781f0b]{background-color:#f6ffed}.settings-btn[data-v-342c7f16]{width:60px;height:60px;margin-right:24px;border-radius:50%!important;background:#ffffff80!important;border:1px solid rgba(255,255,255,.7)!important;backdrop-filter:blur(10px);-webkit-backdrop-filter:blur(10px);box-shadow:0 4px 16px #1f26871f;display:flex;align-items:center;justify-content:center}.welcome-wrapper[data-v-342c7f16]{width:100%;height:100%;background-image:url(./bg-BmnA8p_e.png);background-repeat:no-repeat;background-attachment:fixed;background-size:cover;background-position:center;display:flex;flex-direction:column;align-items:center;justify-content:space-between;color:#fff}.welcome-wrapper .content[data-v-342c7f16]{width:100%;height:80vh;display:flex;flex-direction:column;justify-content:space-around;margin-top:64px}.welcome-wrapper .content .inner-content[data-v-342c7f16]{display:flex;flex-direction:column;align-items:center;justify-content:center;text-align:center;padding:20px}.welcome-wrapper .content .inner-content .text-box[data-v-342c7f16]{color:#000;margin-bottom:36px}.welcome-wrapper .content .inner-content .text-box .title[data-v-342c7f16]{font-size:24px;font-weight:600;margin-bottom:24px}.welcome-wrapper .content .inner-content .text-box .sub-title[data-v-342c7f16]{font-size:15px;margin-top:10px}.welcome-wrapper .content .inner-content .btn-box[data-v-342c7f16]{width:224px;height:80px}.welcome-wrapper .actions[data-v-342c7f16]{width:100%;height:100px;margin-bottom:32px;display:flex;align-items:center;justify-content:flex-end}.ball-wrapper[data-v-34c8e583]{width:100%;height:calc(100vh - 100px);display:flex;flex-direction:column;align-items:center;justify-content:space-around}.talk-wrapper[data-v-05da84ae]{width:auto;width:100%;max-width:1000px;margin:0 auto;box-sizing:border-box;height:calc(100vh - 150px);overflow-y:auto;padding:20px 32px 0;display:flex;flex-direction:column;align-items:flex-start;justify-content:flex-start}.talk-wrapper .cont-left[data-v-05da84ae]{width:100%;margin:24px 0;display:flex;justify-content:flex-start;align-items:flex-start}.talk-wrapper .cont-left .text-left[data-v-05da84ae]{max-width:88%;color:#222;font-size:16px;font-weight:400;text-align:left;line-height:1.8;margin-left:12px;margin-top:6px;word-break:break-word}.talk-wrapper .cont-right[data-v-05da84ae]{width:100%;margin:24px 0;display:flex;justify-content:flex-end;align-items:flex-start}.talk-wrapper .cont-right .text-right[data-v-05da84ae]{max-width:80%;color:#444;font-size:16px;font-weight:400;text-align:start;line-height:1.8;margin-right:12px;background:#ccc;border-radius:8px 0 8px 8px;padding:8px 12px;word-break:break-word}.chat-wrapper[data-v-8b035bf4]{width:100%;height:100%;background-image:url(./bg-BmnA8p_e.png);background-repeat:no-repeat;background-attachment:fixed;background-size:cover;background-position:center;display:flex;flex-direction:column;align-items:center;justify-content:space-between;color:#fff}.chat-wrapper .content[data-v-8b035bf4]{width:100%;height:auto;display:flex;flex-direction:column;justify-content:space-around}.chat-wrapper .content .inner-content[data-v-8b035bf4]{display:flex;flex-direction:column;align-items:center;justify-content:center;text-align:center;padding:20px}.chat-wrapper .content .inner-content .text-box[data-v-8b035bf4]{color:#000;margin-bottom:36px}.chat-wrapper .content .inner-content .text-box .title[data-v-8b035bf4]{font-size:24px;font-weight:600;margin-bottom:24px}.chat-wrapper .content .inner-content .text-box .sub-title[data-v-8b035bf4]{font-size:15px;margin-top:10px}.chat-wrapper .content .inner-content .btn-box[data-v-8b035bf4]{width:224px;height:80px}.chat-wrapper .actions[data-v-8b035bf4]{width:100%;height:100px;margin-bottom:32px;display:flex;justify-content:space-between;align-items:center}.chat-wrapper .actions .holder[data-v-8b035bf4]{width:64px;height:48px}.chat-wrapper .actions .btns[data-v-8b035bf4]{width:450px;height:96px;display:flex;justify-content:space-around;align-items:center}.chat-wrapper .actions .btns[data-v-8b035bf4] .ant-btn{border-radius:50%!important;background:#ffffff80!important;border:1px solid rgba(255,255,255,.7)!important;backdrop-filter:blur(10px);-webkit-backdrop-filter:blur(10px);box-shadow:0 4px 16px #1f26871f}.chat-wrapper .actions .download-wrapper[data-v-8b035bf4]{width:64px;height:64px;display:flex;justify-content:flex-start;align-items:center;margin-right:0}.chat-wrapper .actions .download-wrapper img[data-v-8b035bf4]{width:24px;height:24px}.content-wrapper[data-v-d41c9ce7]{text-align:left;max-width:800px;min-width:320px;margin-bottom:64px;min-height:calc(100vh - 438px)}.content-wrapper .content-box[data-v-d41c9ce7]{padding:24px;height:240px;background-color:#e8e8e8;border-radius:16px;width:50%;margin:48px auto;min-width:300px}.content-wrapper .video-box[data-v-d41c9ce7]{max-width:800px;min-width:320px;width:90vw;height:auto}

assets/www/assets/{index-ByqsFGbw.js → index-STY1yYrK.js}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:215e0b4a6eee243715941860012a0d3bbee778f8880df45b0ddc8b090993405b
-size 2215701
+oid sha256:4049e7175fcba250d7bba7b6753d02e570bc5b2ff472019c32540c30b34f0e93
+size 2300342

assets/www/index.html

@@ -5,8 +5,8 @@
     <link rel="icon" type="image/svg+xml" href="./favicon.ico" />
     <meta name="viewport" content="width=device-width, initial-scale=1.0" />
     <title>VoiceDialogue</title>
-    <script type="module" crossorigin src="./assets/index-ByqsFGbw.js"></script>
-    <link rel="stylesheet" crossorigin href="./assets/index-CCuJ1lip.css">
+    <script type="module" crossorigin src="./assets/index-STY1yYrK.js"></script>
+    <link rel="stylesheet" crossorigin href="./assets/index-6CUhaZ06.css">
   </head>
   <body>
     <div id="app"></div>

frontend/src/App.vue

@@ -1,6 +1,8 @@
 <template>
-  <Header/>
-  <router-view class="content" />
+  <a-config-provider :theme="appTheme">
+    <Header/>
+    <router-view class="content" />
+  </a-config-provider>
   <!-- <Footer/> -->
 
   <!-- <a-layout>
@@ -19,6 +21,15 @@
 import Header from "@/views/Header.vue";
 import Footer from "@/views/Footer.vue";
 
+// 全局主题：统一圆角与控件高度，配合玻璃拟态（Liquid Glass）
+const appTheme = {
+  token: {
+    colorPrimary: '#1677ff',
+    borderRadius: 14,
+    controlHeight: 38,
+  },
+};
+
 // import * as api from "@/client";
 import { onBeforeMount, onMounted, watch, CSSProperties, ref} from "vue";
 import {useSettingsStore} from "@/stores/config.ts";

frontend/src/assets/ball.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:edd650ec984e26b5fde217f273e6758d0862fc856b5333e678fa0b578374e8b9
-size 23084
+oid sha256:91eaffeec742a30f475cf5e777e1605e62d3c1547b64a891c15f9a5431460b8a
+size 22455

frontend/src/config/client_config.ts

@@ -5,7 +5,7 @@ import router from "@/router";
 
 const { wsCache } = useCache();
 
-export const test_server = '127.0.0.1:8848'
+export const test_server = '127.0.0.1:8000'
 // export const test_server = '59.110.18.232:19001'
 
 axios.defaults.baseURL = import.meta.env.PROD  ?  '/api/v1' : `http://${test_server}/api/v1`;

frontend/src/i18n/index.ts

@@ -0,0 +1,35 @@
+import { createI18n } from 'vue-i18n'
+
+import en from './locales/en'
+import zh from './locales/zh'
+
+export type UiLocale = 'en' | 'zh'
+
+// 从持久化的 pinia 设置中读取界面语言，默认英文
+function getInitialLocale(): UiLocale {
+    try {
+        const raw = localStorage.getItem('settings')
+        if (raw) {
+            const parsed = JSON.parse(raw)
+            const ui = parsed?.uiLanguage
+            if (ui === 'en' || ui === 'zh') return ui
+        }
+    } catch (e) {
+        // ignore
+    }
+    return 'zh'
+}
+
+const i18n = createI18n({
+    legacy: false,
+    globalInjection: true,
+    locale: getInitialLocale(),
+    fallbackLocale: 'en',
+    messages: { en, zh },
+})
+
+export function setUiLocale(locale: UiLocale) {
+    i18n.global.locale.value = locale
+}
+
+export default i18n

frontend/src/i18n/locales/en.ts

@@ -0,0 +1,56 @@
+export default {
+    common: {
+        cancel: 'Cancel',
+        confirm: 'Confirm',
+        reset: 'Reset',
+        save: 'Save',
+        error: 'Error',
+    },
+    lang: {
+        zh: 'Chinese',
+        en: 'English',
+        auto: 'Auto',
+    },
+    welcome: {
+        title: 'Welcome',
+        subtitle: 'Click the button below to start a conversation',
+        start: 'Start Conversation',
+        startFailed: 'Failed to start the voice dialogue system',
+    },
+    settings: {
+        title: 'Settings',
+        entry: 'Settings',
+        tabs: {
+            main: 'Main',
+            language: 'Language',
+            advanced: 'Prompt',
+        },
+        general: {
+            interfaceLanguage: 'Interface Language',
+            interfaceLanguageHint: 'Language of the application interface.',
+        },
+        audio: {
+            microphone: 'Microphone (Input Device)',
+            microphoneHint: 'Choose the input device, e.g. an external microphone array.',
+            systemDefault: 'System Default',
+            channelsSuffix: 'ch',
+            defaultSuffix: 'default',
+            echoCancellation: 'Echo Cancellation',
+            echoCancellationHint: 'Uses the system AEC on the default device. For an external array, echo is handled by the array hardware.',
+        },
+        recognition: {
+            language: 'Recognition Language',
+            languageHint: 'Language used for speech recognition (ASR).',
+        },
+        voice: {
+            role: 'Voice',
+            roleHint: 'The voice used for speech synthesis (TTS).',
+            playSample: 'Play sample',
+        },
+        prompt: {
+            title: 'System Prompt',
+            hint: 'Customize the system prompt for each language.',
+        },
+        applyFailed: 'Failed to apply settings',
+    },
+}

frontend/src/i18n/locales/zh.ts

@@ -0,0 +1,56 @@
+export default {
+    common: {
+        cancel: '取消',
+        confirm: '确认',
+        reset: '重置',
+        save: '保存',
+        error: '错误',
+    },
+    lang: {
+        zh: '中文',
+        en: '英文',
+        auto: '自动',
+    },
+    welcome: {
+        title: '欢迎使用',
+        subtitle: '点击下方按钮开始对话',
+        start: '开始对话',
+        startFailed: '启动语音对话系统失败',
+    },
+    settings: {
+        title: '设置',
+        entry: '设置',
+        tabs: {
+            main: '常用',
+            language: '语言',
+            advanced: 'Prompt',
+        },
+        general: {
+            interfaceLanguage: '界面语言',
+            interfaceLanguageHint: '应用界面所使用的语言。',
+        },
+        audio: {
+            microphone: '麦克风（输入设备）',
+            microphoneHint: '选择输入设备，例如外置麦克风阵列。',
+            systemDefault: '系统默认',
+            channelsSuffix: '声道',
+            defaultSuffix: '默认',
+            echoCancellation: '回音消除',
+            echoCancellationHint: '默认设备使用系统 AEC；选择外置阵列时，回音由阵列硬件处理。',
+        },
+        recognition: {
+            language: '识别语言',
+            languageHint: '语音识别（ASR）所使用的语言。',
+        },
+        voice: {
+            role: '音色',
+            roleHint: '语音合成（TTS）所使用的音色。',
+            playSample: '试听',
+        },
+        prompt: {
+            title: '系统提示词',
+            hint: '为每种语言自定义系统提示词。',
+        },
+        applyFailed: '应用设置失败',
+    },
+}

frontend/src/main.ts

@@ -9,6 +9,7 @@ import './style.scss'
 
 import App from './App.vue'
 import router from './router'
+import i18n from './i18n'
 
 
 // import * as Sentry from "@sentry/browser";
@@ -28,4 +29,5 @@ createApp(App)
     .use(router)
     .use(Antd)
     .use(Vue3Lottie)
+    .use(i18n)
     .mount('#app')

frontend/src/stores/config.ts

@@ -8,8 +8,10 @@ export const useSettingsStore  = defineStore({
         return {
             role: '',
             language: 'zh',
+            uiLanguage: 'zh' as 'en' | 'zh',
             sider_open: true,
             echoCancel: true,
+            inputDeviceIndex: null as number | null,
         }
     },
     actions: {

frontend/src/style.scss

@@ -173,3 +173,68 @@ $FormItemWidth: 1022px;
 .ant-layout-sider-collapsed .ant-menu-submenu-title {
   display: none;
 }
+
+/* ============================================================
+   Liquid Glass —— 苹果风格玻璃拟态（全局）
+   半透明 + 背景模糊 + 柔和描边/阴影；圆角由主题 token 统一
+   ============================================================ */
+
+/* 弹窗使用 Ant 内置 fade 过渡（纯 opacity 动画、无 transform），
+   避免 transform 动画期间 backdrop-filter 失效导致的闪烁；面板与其模糊一起平滑淡入 */
+
+/* 弹窗：磨砂玻璃面板 */
+.ant-modal .ant-modal-content {
+  background: rgba(255, 255, 255, 0.62) !important;
+  backdrop-filter: blur(28px) saturate(140%);
+  -webkit-backdrop-filter: blur(28px) saturate(140%);
+  border: 1px solid rgba(255, 255, 255, 0.6);
+  border-radius: 22px !important;
+  box-shadow: 0 16px 48px rgba(31, 38, 135, 0.18);
+}
+.ant-modal .ant-modal-header {
+  background: transparent !important;
+}
+/* 遮罩：整屏磨砂——轻微变暗 + 背景模糊。
+   遮罩用 ant-fade（opacity）淡入，模糊随之平滑出现，背景文字与画面一起糊掉，不再"闪出去" */
+.ant-modal-mask {
+  background: rgba(20, 22, 30, 0.12) !important;
+  backdrop-filter: blur(14px) saturate(120%);
+  -webkit-backdrop-filter: blur(14px) saturate(120%);
+}
+
+/* 输入类控件：半透明玻璃 */
+.ant-select .ant-select-selector,
+.ant-input,
+textarea.ant-input,
+.ant-input-affix-wrapper {
+  background: rgba(255, 255, 255, 0.45) !important;
+  backdrop-filter: blur(8px);
+  -webkit-backdrop-filter: blur(8px);
+  border: 1px solid rgba(255, 255, 255, 0.7) !important;
+}
+
+/* 按钮：统一形状（圆角来自 token）+ 柔和阴影；默认按钮做玻璃质感，主按钮保持实色
+   文本/链接按钮（如音色试听的小喇叭）保持透明无阴影 */
+.ant-btn:not(.ant-btn-text):not(.ant-btn-link) {
+  box-shadow: 0 2px 10px rgba(31, 38, 135, 0.10);
+}
+.ant-btn-default {
+  background: rgba(255, 255, 255, 0.5) !important;
+  border: 1px solid rgba(255, 255, 255, 0.75) !important;
+  backdrop-filter: blur(8px);
+  -webkit-backdrop-filter: blur(8px);
+}
+.ant-btn-text {
+  box-shadow: none !important;
+  background: transparent !important;
+}
+
+/* 分段单选（中文/英文 等）两端圆角，去掉方正感 */
+.ant-radio-group-solid .ant-radio-button-wrapper:first-child {
+  border-top-left-radius: 12px;
+  border-bottom-left-radius: 12px;
+}
+.ant-radio-group-solid .ant-radio-button-wrapper:last-child {
+  border-top-right-radius: 12px;
+  border-bottom-right-radius: 12px;
+}

frontend/src/views/Home/Components/ChatText.vue

@@ -69,9 +69,13 @@ watch(() => props.chatContent, (newVal, oldVal) => {
 <style lang="scss" scoped>
 .talk-wrapper {
     width: auto;
-    height: calc(100vh - 100px);
-    overflow-y: scroll;
-    padding: 20px 240px 0 240px;
+    width: 100%;
+    max-width: 1000px;
+    margin: 0 auto;
+    box-sizing: border-box;
+    height: calc(100vh - 150px);
+    overflow-y: auto;
+    padding: 20px 32px 0;
     display: flex;
     flex-direction: column;
     align-items: flex-start;
@@ -85,13 +89,15 @@ watch(() => props.chatContent, (newVal, oldVal) => {
         justify-content: flex-start;
         align-items: flex-start;
         .text-left {
+            max-width: 88%;
             color: #222;
             font-size: 16px;
             font-weight: 400;
             text-align: left;
-            line-height: 2;
+            line-height: 1.8;
             margin-left: 12px;
             margin-top: 6px;
+            word-break: break-word;
         }
     }
 
@@ -103,16 +109,18 @@ watch(() => props.chatContent, (newVal, oldVal) => {
         align-items: flex-start;
 
         .text-right {
+            max-width: 80%;
             color: #444;
             font-size: 16px;
             font-weight: 400;
-            text-align: end;
-            line-height: 2;
+            text-align: start;
+            line-height: 1.8;
             margin-right: 12px;
             background: #ccc;
             border-radius: 8px;
             border-top-right-radius: 0;
-            padding: 8px;
+            padding: 8px 12px;
+            word-break: break-word;
         }
     }
 }

frontend/src/views/Home/index.vue

@@ -387,6 +387,7 @@ const toggleText = () => {
     .actions {
         width: 100%;
         height: 100px;
+        margin-bottom: 32px;
 
         display: flex;
         justify-content: space-between;
@@ -401,7 +402,17 @@ const toggleText = () => {
             height: 96px;
             display: flex;
             justify-content: space-around;
-            align-items: flex-start;
+            align-items: center;
+
+            // Liquid Glass 圆形按钮（与 Welcome 设置按钮统一）
+            :deep(.ant-btn) {
+                border-radius: 50% !important;
+                background: rgba(255, 255, 255, 0.5) !important;
+                border: 1px solid rgba(255, 255, 255, 0.7) !important;
+                backdrop-filter: blur(10px);
+                -webkit-backdrop-filter: blur(10px);
+                box-shadow: 0 4px 16px rgba(31, 38, 135, 0.12);
+            }
         }
         .download-wrapper {
             width: 64px;

frontend/src/views/Welcome/Components/SettingsModal.vue

@@ -0,0 +1,409 @@
+<script setup lang="ts">
+import { ref, reactive, computed, watch, onUnmounted } from "vue";
+import { Modal } from "ant-design-vue";
+import { SoundTwoTone, SoundOutlined, TranslationOutlined, AudioOutlined } from "@ant-design/icons-vue";
+import { useI18n } from "vue-i18n";
+import axios from "axios";
+import { useSettingsStore } from "@/stores/config.ts";
+import { setUiLocale, UiLocale } from "@/i18n";
+
+const props = defineProps({
+    open: { type: Boolean, default: false },
+});
+const emit = defineEmits(["update:open"]);
+
+const { t } = useI18n();
+const base_url = axios.defaults.baseURL;
+const settingsStore = useSettingsStore();
+
+const activeTab = ref<string>("main");
+const loading = ref<boolean>(false);
+
+// ---- 各项设置的本地状态（打开时从 store / 后端同步）----
+const uiLanguage = ref<UiLocale>((settingsStore.$state.uiLanguage as UiLocale) ?? "en");
+const recognitionLanguage = ref<string>(settingsStore.$state.language || "zh");
+const echoCancel = ref<boolean>(settingsStore.$state.echoCancel ?? true);
+const inputDeviceIndex = ref<number | null>(settingsStore.$state.inputDeviceIndex ?? null);
+const role = ref<string>(settingsStore.$state.role || "");
+
+const languages = reactive<string[]>([]);
+const inputDevices = reactive<any[]>([]);
+const roles = reactive<any[]>([]);
+
+// ---- Prompt ----
+const promptLang = ref<string>("zh");
+const default_prompt_en = ref<string>("");
+const default_prompt_zh = ref<string>("");
+const current_prompt_en = ref<string>("");
+const current_prompt_zh = ref<string>("");
+
+const filteredRoles = computed(() => {
+    const is_chinese = recognitionLanguage.value === "zh";
+    return roles.filter((r) => r["is_chinese_voice"] === is_chinese);
+});
+
+// 切换识别语言后，自动选中第一个匹配音色
+watch(
+    () => recognitionLanguage.value,
+    () => {
+        if (filteredRoles.value.length > 0) {
+            const exists = filteredRoles.value.find((r) => r["id"] === role.value);
+            role.value = exists ? role.value : filteredRoles.value[0]["id"];
+        } else {
+            role.value = "";
+        }
+    }
+);
+
+// 界面语言即时生效（让用户立刻看到切换效果）
+watch(uiLanguage, (v) => setUiLocale(v));
+
+// ---- 数据加载 ----
+const fetchASRLanguages = async () => {
+    try {
+        const res = await fetch(`${base_url}/asr/languages`);
+        const data = await res.json();
+        if (data?.languages) {
+            languages.splice(0, languages.length, ...data.languages);
+            // 优先沿用本地已保存/默认的识别语言（默认中文），不被后端当前值覆盖
+            const saved = settingsStore.$state.language;
+            recognitionLanguage.value = saved && data.languages.includes(saved)
+                ? saved
+                : (data.languages.includes('zh') ? 'zh' : data.languages[0]);
+        }
+    } catch (e) {
+        console.error("Error fetching ASR languages:", e);
+    }
+};
+
+const fetchTTSRoles = async () => {
+    try {
+        const res = await fetch(`${base_url}/tts/models`);
+        const data = await res.json();
+        if (data?.models) {
+            roles.splice(0, roles.length, ...data.models);
+            if (data.current_model_id) role.value = data.current_model_id;
+        }
+    } catch (e) {
+        console.error("Error fetching TTS roles:", e);
+    }
+};
+
+const fetchInputDevices = async () => {
+    try {
+        const res = await fetch(`${base_url}/system/audio-devices`);
+        const data = await res.json();
+        if (data?.devices) {
+            inputDevices.splice(0, inputDevices.length, ...data.devices);
+            const saved = settingsStore.$state.inputDeviceIndex;
+            const exists = saved != null && data.devices.some((d: any) => d.index === saved);
+            inputDeviceIndex.value = exists ? saved : (data.current_device_index ?? null);
+        }
+    } catch (e) {
+        console.error("Error fetching input devices:", e);
+    }
+};
+
+const fetchPrompts = async () => {
+    try {
+        const [cur, def] = await Promise.all([
+            fetch(`${base_url}/settings/settings/prompts`).then((r) => r.json()),
+            fetch(`${base_url}/settings/settings/prompts/default`).then((r) => r.json()),
+        ]);
+        if (cur) {
+            current_prompt_en.value = cur.english_prompt;
+            current_prompt_zh.value = cur.chinese_prompt;
+        }
+        if (def) {
+            default_prompt_en.value = def.english_prompt;
+            default_prompt_zh.value = def.chinese_prompt;
+        }
+    } catch (e) {
+        console.error("Error fetching prompts:", e);
+    }
+};
+
+const resetPrompt = (lang: string) => {
+    if (lang === "en") current_prompt_en.value = default_prompt_en.value;
+    else current_prompt_zh.value = default_prompt_zh.value;
+};
+
+// ---- 提交 / 取消 ----
+const applySettings = async () => {
+    loading.value = true;
+    try {
+        // 1. 持久化到本地 store
+        settingsStore.$state.uiLanguage = uiLanguage.value;
+        settingsStore.$state.language = recognitionLanguage.value;
+        settingsStore.$state.role = role.value || "";
+        settingsStore.$state.echoCancel = echoCancel.value;
+        settingsStore.$state.inputDeviceIndex = inputDeviceIndex.value;
+
+        // 2. 下发 TTS 音色 + ASR 语言
+        if (role.value) {
+            const r1 = await fetch(`${base_url}/tts/models/load`, {
+                method: "POST",
+                headers: { "Content-Type": "application/json" },
+                body: JSON.stringify({ model_id: role.value }),
+            });
+            if (!r1.ok) throw new Error(`TTS load failed: ${r1.status}`);
+        }
+        const r2 = await fetch(`${base_url}/asr/instance/create`, {
+            method: "POST",
+            headers: { "Content-Type": "application/json" },
+            body: JSON.stringify({ language: recognitionLanguage.value }),
+        });
+        if (!r2.ok) throw new Error(`ASR set failed: ${r2.status}`);
+
+        // 3. 保存 Prompt
+        await fetch(`${base_url}/settings/settings/prompts`, {
+            method: "POST",
+            headers: { "Content-Type": "application/json" },
+            body: JSON.stringify({
+                chinese_prompt: current_prompt_zh.value,
+                english_prompt: current_prompt_en.value,
+            }),
+        });
+
+        emit("update:open", false);
+    } catch (err) {
+        console.error("Error applying settings:", err);
+        Modal.error({ title: t("common.error"), content: t("settings.applyFailed") });
+    } finally {
+        loading.value = false;
+    }
+};
+
+const handleCancel = () => {
+    // 还原本地状态与界面语言
+    uiLanguage.value = (settingsStore.$state.uiLanguage as UiLocale) ?? "en";
+    setUiLocale(uiLanguage.value);
+    recognitionLanguage.value = settingsStore.$state.language || "zh";
+    echoCancel.value = settingsStore.$state.echoCancel ?? true;
+    inputDeviceIndex.value = settingsStore.$state.inputDeviceIndex ?? null;
+    role.value = settingsStore.$state.role || "";
+    emit("update:open", false);
+};
+
+watch(
+    () => props.open,
+    (isOpen) => {
+        if (isOpen) {
+            activeTab.value = "main";
+            uiLanguage.value = (settingsStore.$state.uiLanguage as UiLocale) ?? "en";
+            fetchASRLanguages();
+            fetchTTSRoles();
+            fetchInputDevices();
+            fetchPrompts();
+        }
+    }
+);
+
+// ---- 音色试听 ----
+const currentPlayingId = ref<string | null>(null);
+const currentAudio = ref<HTMLAudioElement | null>(null);
+const isPlaying = (id: string) => currentPlayingId.value === id;
+
+const playRefAudio = async (id: string, e: Event) => {
+    e.stopPropagation();
+    e.preventDefault();
+    try {
+        if (currentPlayingId.value === id && currentAudio.value) {
+            currentAudio.value.pause();
+            currentAudio.value = null;
+            currentPlayingId.value = null;
+            return;
+        }
+        if (currentAudio.value) {
+            currentAudio.value.pause();
+            currentAudio.value = null;
+        }
+        const audio = new Audio(`${base_url}/tts/models/${id}/reference-audio`);
+        audio.addEventListener("ended", () => {
+            currentPlayingId.value = null;
+            currentAudio.value = null;
+        });
+        await audio.play();
+        currentPlayingId.value = id;
+        currentAudio.value = audio;
+    } catch (err) {
+        currentPlayingId.value = null;
+        currentAudio.value = null;
+    }
+};
+
+onUnmounted(() => {
+    if (currentAudio.value) currentAudio.value.pause();
+});
+</script>
+
+<template>
+    <a-modal
+        :open="props.open"
+        :title="t('settings.title')"
+        :mask-closable="false"
+        :closable="true"
+        :width="600"
+        centered
+        transition-name="ant-fade"
+        @cancel="handleCancel"
+        @update:open="(v: boolean) => emit('update:open', v)"
+    >
+        <template #footer>
+            <a-button key="back" @click="handleCancel">{{ t('common.cancel') }}</a-button>
+            <a-button key="confirm" type="primary" :loading="loading" @click="applySettings">
+                {{ t('common.confirm') }}
+            </a-button>
+        </template>
+
+        <a-tabs v-model:activeKey="activeTab" class="settings-tabs">
+            <!-- 常用：输入源 + 回音消除 + 音色（大家最关心的） -->
+            <a-tab-pane key="main" :tab="t('settings.tabs.main')">
+                <div class="tab-body">
+                    <div class="setting-row">
+                        <label>{{ t('settings.audio.microphone') }}</label>
+                        <a-select v-model:value="inputDeviceIndex" style="width: 100%;">
+                            <a-select-option :value="null">{{ t('settings.audio.systemDefault') }}</a-select-option>
+                            <a-select-option v-for="dev in inputDevices" :value="dev.index" :key="dev.index">
+                                {{ dev.name }}<template v-if="dev.max_input_channels > 1"> ({{ dev.max_input_channels }}{{ t('settings.audio.channelsSuffix') }})</template><template v-if="dev.is_default"> · {{ t('settings.audio.defaultSuffix') }}</template>
+                            </a-select-option>
+                        </a-select>
+                    </div>
+                    <div class="setting-row">
+                        <div class="row-inline">
+                            <label>{{ t('settings.audio.echoCancellation') }}</label>
+                            <a-switch v-model:checked="echoCancel" />
+                        </div>
+                    </div>
+                    <div class="setting-row">
+                        <label>{{ t('settings.voice.role') }}</label>
+                        <a-radio-group v-model:value="role" class="voice-group">
+                            <a-radio v-for="r in filteredRoles" :value="r['id']" :key="r['id']" class="voice-radio">
+                                <span class="voice-name">{{ r['character_name'] }}</span>
+                                <a-button
+                                    type="text"
+                                    class="audio-play-btn"
+                                    :class="{ playing: isPlaying(r['id']) }"
+                                    @click="playRefAudio(r['id'], $event)"
+                                >
+                                    <SoundTwoTone v-if="isPlaying(r['id'])" style="font-size: 16px; color: #52c41a;" />
+                                    <SoundOutlined v-else style="font-size: 16px; color: #1890ff;" />
+                                </a-button>
+                            </a-radio>
+                        </a-radio-group>
+                    </div>
+                </div>
+            </a-tab-pane>
+
+            <!-- 语言：界面语言 + 识别语言 -->
+            <a-tab-pane key="language" :tab="t('settings.tabs.language')">
+                <div class="tab-body">
+                    <div class="setting-row">
+                        <label><TranslationOutlined class="label-icon" />{{ t('settings.general.interfaceLanguage') }}</label>
+                        <a-select v-model:value="uiLanguage" style="width: 100%;">
+                            <a-select-option value="zh">{{ t('lang.zh') }}</a-select-option>
+                            <a-select-option value="en">{{ t('lang.en') }}</a-select-option>
+                        </a-select>
+                        <p class="hint">{{ t('settings.general.interfaceLanguageHint') }}</p>
+                    </div>
+                    <div class="setting-row">
+                        <label><AudioOutlined class="label-icon" />{{ t('settings.recognition.language') }}</label>
+                        <a-select v-model:value="recognitionLanguage" style="width: 100%;">
+                            <a-select-option v-for="lan in languages" :value="lan" :key="lan">
+                                {{ t('lang.' + lan) }}
+                            </a-select-option>
+                        </a-select>
+                        <p class="hint">{{ t('settings.recognition.languageHint') }}</p>
+                    </div>
+                </div>
+            </a-tab-pane>
+
+            <!-- 高级：系统提示词 -->
+            <a-tab-pane key="advanced" :tab="t('settings.tabs.advanced')">
+                <div class="tab-body">
+                    <div class="setting-row">
+                        <label>{{ t('settings.prompt.title') }}</label>
+                        <a-radio-group button-style="solid" size="small" v-model:value="promptLang" style="margin-bottom: 12px;">
+                            <a-radio-button value="zh">{{ t('lang.zh') }}</a-radio-button>
+                            <a-radio-button value="en">{{ t('lang.en') }}</a-radio-button>
+                        </a-radio-group>
+                        <div v-show="promptLang === 'zh'">
+                            <a-textarea v-model:value="current_prompt_zh" :placeholder="default_prompt_zh"
+                                :auto-size="{ minRows: 6, maxRows: 10 }" show-count :maxlength="2000" allow-clear />
+                            <a-button size="small" @click="resetPrompt('zh')" style="margin-top: 12px;">{{ t('common.reset') }}</a-button>
+                        </div>
+                        <div v-show="promptLang === 'en'">
+                            <a-textarea v-model:value="current_prompt_en" :placeholder="default_prompt_en"
+                                :auto-size="{ minRows: 6, maxRows: 10 }" show-count :maxlength="2000" allow-clear />
+                            <a-button size="small" @click="resetPrompt('en')" style="margin-top: 12px;">{{ t('common.reset') }}</a-button>
+                        </div>
+                    </div>
+                </div>
+            </a-tab-pane>
+        </a-tabs>
+    </a-modal>
+</template>
+
+<style lang="scss" scoped>
+// 固定内容区高度，切换 Tab 时横条不再跳动
+.tab-body {
+    height: 360px;
+    overflow-y: auto;
+    padding: 4px 8px 4px 2px;
+}
+
+.setting-row {
+    margin-bottom: 20px;
+
+    // 仅作用于字段标题（直接子 label），避免影响嵌套的 radio-button 等 <label>
+    > label {
+        display: block;
+        font-size: 15px;
+        font-weight: 500;
+        margin-bottom: 8px;
+
+        .label-icon {
+            margin-right: 6px;
+            color: #1890ff;
+        }
+    }
+
+    .hint {
+        font-size: 12px;
+        color: #999;
+        margin: 8px 0 0;
+    }
+
+    .row-inline {
+        display: flex;
+        align-items: center;
+        justify-content: space-between;
+    }
+}
+
+.voice-group {
+    display: flex;
+    flex-direction: column;
+    margin-top: 8px;
+}
+
+.voice-radio {
+    display: flex;
+    align-items: center;
+    height: 40px;
+    line-height: 40px;
+
+    .voice-name {
+        margin-right: 8px;
+    }
+}
+
+.audio-play-btn {
+    padding: 0 6px;
+    border-radius: 4px;
+
+    &.playing {
+        background-color: #f6ffed;
+    }
+}
+</style>

frontend/src/views/Welcome/index.vue

@@ -2,37 +2,18 @@
 
 import router from "@/router.ts";
 import { useSettingsStore } from "@/stores/config.ts";
-import { onMounted, onUnmounted, ref, reactive, computed, watch, h } from "vue";
+import { ref } from "vue";
 import { Modal } from 'ant-design-vue';
-import { SoundTwoTone, SoundOutlined } from "@ant-design/icons-vue";
+import { useI18n } from "vue-i18n";
 import axios from "axios";
-import PromptText from "./Components/PromptText.vue";
+import SettingsModal from "./Components/SettingsModal.vue";
+import setting from "@/assets/setting.png";
 
-const base_url = axios.defaults.baseURL
+const { t } = useI18n();
+const base_url = axios.defaults.baseURL;
+const settingsStore = useSettingsStore();
 
-const settingsStore = useSettingsStore()
-
-import setting from "@/assets/setting.png"
-
-
-onMounted(async () => {
-    await fetchASRLanguages();
-    await fetchTTSRoles();
-});
-
-const chatAction = async () => {
-    const state = await startAudioChat();
-    if (!state) {
-        console.error('Failed to start audio chat system service');
-
-        Modal.error({
-            title: 'Error',
-            content: 'Failed to start audio chat system service',
-        });
-        return;
-    }
-    router.replace('/home')
-}
+const settingsOpen = ref<boolean>(false);
 const chatLoading = ref<boolean>(false);
 
 const startAudioChat = async () => {
@@ -40,265 +21,33 @@ const startAudioChat = async () => {
         chatLoading.value = true;
         const response = await fetch(`${base_url}/system/start`, {
             method: 'POST',
-            headers: {
-                'Content-Type': 'application/json',
-            },
+            headers: { 'Content-Type': 'application/json' },
             body: JSON.stringify({
-                enable_echo_cancellation: echoCancel.value
+                enable_echo_cancellation: settingsStore.$state.echoCancel ?? true,
+                input_device_index: settingsStore.$state.inputDeviceIndex ?? null
             })
         });
         if (!response.ok) {
             throw new Error(`HTTP error! status: ${response.status}`);
         }
-        const data = await response.json();
-        console.log('ASR Instance started successfully:', data);
+        await response.json();
         return true;
     } catch (error) {
-        console.error('Error starting ASR instance:', error);
+        console.error('Error starting audio chat:', error);
         return false;
     } finally {
         chatLoading.value = false;
     }
-}
-
-
-const voiceModelOpen = ref<boolean>(false);
-const modalLoading = ref<boolean>(false);
-
-const handleVoiceModalCancel = () => {
-    voiceModelOpen.value = false;
-    role.value = settingsStore.$state.role;
-    language.value = settingsStore.$state.language;
-};
-
-const handleVoiceModalSubmit = async () => {
-    console.log('Selected Language:', language.value);
-    console.log('Selected Role:', role.value);
-    console.log('Echo Cancel:', echoCancel.value);
-    settingsStore.$state.language = language.value;
-    settingsStore.$state.role = role.value || '';
-    settingsStore.$state.echoCancel = echoCancel.value;
-
-    await pushConfig(settingsStore.$state.role);
-};
-
-const pushConfig = async (model_id: string) => {
-    try {
-        modalLoading.value = true;
-        const response = await fetch(`${base_url}/tts/models/load`, {
-            method: 'POST',
-            headers: {
-                'Content-Type': 'application/json',
-            },
-            body: JSON.stringify({
-                "model_id": model_id,
-            })
-        });
-        if (!response.ok) {
-            throw new Error(`HTTP error! status: ${response.status}`);
-        }
-        const data = await response.json();
-        console.log('Config pushed successfully:', data);
-
-        const response2 = await fetch(`${base_url}/asr/instance/create`, {
-            method: 'POST',
-            headers: {
-                'Content-Type': 'application/json',
-            },
-            body: JSON.stringify({
-                "language": language.value,
-            })
-        });
-        if (!response2.ok) {
-            throw new Error(`HTTP error! status: ${response2.status}`);
-        }
-        const data2 = await response2.json();
-        console.log('ASR Language set successfully:', data2);
-
-    } catch (err) {
-        console.error('Error pushing config:', err);
-        Modal.error({
-            title: 'Error',
-            content: "Error config: " + JSON.stringify(err),
-        });
-    } finally {
-        modalLoading.value = false;
-        voiceModelOpen.value = false;
-    }
-
-    console.log('Selected Language:', language.value);
-    console.log('Selected Role:', role.value);
-}
-
-
-const language = ref<string>(settingsStore.$state.language || 'zh');
-const languages = reactive([]);
-const languageOptions = {
-    'zh': 'Chinese',
-    'en': 'English',
-    'auto': 'Auto',
-};
-const role = ref<string>(settingsStore.$state.role || '');
-const roles = reactive([])
-const echoCancel = ref<boolean>(settingsStore.$state.echoCancel ?? true);
-
-const radioStyle = reactive({
-    display: 'flex',
-    height: '40px',
-    lineHeight: '40px',
-    fontSize: '16px',
-    marginBottom: '8px',
-});
-
-const filteredRoles = computed(() => {
-    const is_chinese = language.value == 'zh';
-    return roles.filter(ro => ro['is_chinese_voice'] == is_chinese);
-});
-
-watch(
-  () => language.value,
-  (newLang) => {
-    // 语言切换后，自动选中第一个可用角色
-    if (filteredRoles.value.length > 0) {
-      const current_role_id = settingsStore.$state.role;
-      const current_role = filteredRoles.value.find(ro => ro['id'] == current_role_id);
-      if (current_role) {
-        role.value = current_role_id;
-      } else {
-        role.value = filteredRoles.value[0]['id'];
-      }
-    } else {
-      role.value = "";
-    }
-  }
-);
-
-
-const fetchTTSRoles = async () => {
-    try {
-        const response = await fetch(`${base_url}/tts/models`);
-        const data = await response.json()
-        if (data && data.models) {
-            // @ts-ignore
-            roles.splice(0, data.length, ...data.models)
-            console.log('Fetched TTS Roles:', roles);
-
-            if (data.current_model_id) {
-                role.value = data.current_model_id;
-            }
-        }
-    } catch (error) {
-        console.error('Error fetching TTS roles:', error);
-    }
-};
-
-const fetchASRLanguages = async () => {
-    try {
-        const response = await fetch(`${base_url}/asr/languages`);
-        const data = await response.json();
-        if (data && data.languages) {
-            // @ts-ignore
-            languages.splice(0, languages.length, ...data.languages);
-            console.log('Fetched ASR Languages:', data.languages);
-
-            if (data.current_asr_language) {
-                language.value = data.current_asr_language;
-            }
-        }
-    } catch (error) {
-        console.error('Error fetching ASR languages:', error);
-    }
 };
 
-const togglePopover = (item: string) => {
-    popoverVisible.value = !popoverVisible.value;
-    if (item == 'voice') {
-        voiceModelOpen.value = true;
-    } else if (item == 'prompt') {
-        promptModelOpen.value = true;
-    }
-};
-
-const popoverVisible = ref<boolean>(false);
-const promptModelOpen = ref<boolean>(false);
-
-// 音频播放状态管理
-const currentPlayingId = ref<string | null>(null);
-const currentAudio = ref<HTMLAudioElement | null>(null);
-
-// 修改音频播放逻辑
-const playRefAudio = async (id: string, e: Event) => {
-    console.log('Playing reference audio for role:', id);
-
-    e.stopPropagation();
-    e.preventDefault();
-
-    try {
-        // 如果点击的是当前正在播放的音频，则停止播放
-        if (currentPlayingId.value === id && currentAudio.value) {
-            currentAudio.value.pause();
-            currentAudio.value = null;
-            currentPlayingId.value = null;
-            console.log('Audio stopped');
-            return;
-        }
-
-        // 如果有其他音频正在播放，先停止它
-        if (currentAudio.value) {
-            currentAudio.value.pause();
-            currentAudio.value = null;
-        }
-
-        // 创建新的音频实例
-        const audio = new Audio(`${base_url}/tts/models/${id}/reference-audio`);
-
-        // 设置音频事件监听
-        audio.addEventListener('ended', () => {
-            currentPlayingId.value = null;
-            currentAudio.value = null;
-        });
-
-        audio.addEventListener('error', (error) => {
-            console.error('Audio playback error:', error);
-            currentPlayingId.value = null;
-            currentAudio.value = null;
-            Modal.error({
-                title: 'Error',
-                content: 'Failed to play reference audio',
-            });
-        });
-
-        // 开始播放
-        await audio.play();
-        currentPlayingId.value = id;
-        currentAudio.value = audio;
-        console.log('Audio played successfully');
-
-    } catch (error) {
-        console.error('Error playing audio:', error);
-        currentPlayingId.value = null;
-        currentAudio.value = null;
-        Modal.error({
-            title: 'Error',
-            content: 'Failed to play reference audio',
-        });
-    }
-};
-
-// 组件卸载时清理音频
-onUnmounted(() => {
-    if (currentAudio.value) {
-        currentAudio.value.pause();
-        currentAudio.value = null;
+const chatAction = async () => {
+    const ok = await startAudioChat();
+    if (!ok) {
+        Modal.error({ title: t('common.error'), content: t('welcome.startFailed') });
+        return;
     }
-    currentPlayingId.value = null;
-});
-
-// 计算属性：判断是否正在播放
-const isPlaying = (id: string) => {
-    return currentPlayingId.value === id;
+    router.replace('/home');
 };
-
 </script>
 
 <template>
@@ -306,178 +55,46 @@ const isPlaying = (id: string) => {
         <div class="content">
             <div class="inner-content">
                 <div class="text-box">
-                    <div class="title">
-                        欢迎使用
-                    </div>
-                    <div class="sub-title">
-                        点击下方按钮开始对话
-                    </div>
+                    <div class="title">{{ t('welcome.title') }}</div>
+                    <div class="sub-title">{{ t('welcome.subtitle') }}</div>
                 </div>
                 <div class="btn-box">
                     <a-button @click="chatAction" block :loading="chatLoading" type="primary" size="large">
-                        <span>开始对话</span>
+                        <span>{{ t('welcome.start') }}</span>
                     </a-button>
                 </div>
             </div>
         </div>
 
         <div class="actions">
-            <!-- <a-button type="text" @click="toggleSider">sider</a-button> -->
-
-             <a-button v-if="false" type="text" @click="voiceModelOpen = true"
-                style="width:44px; height: 44px; margin-right:24px;margin-bottom: 24px;">
+            <a-button type="text" @click="settingsOpen = true" class="settings-btn"
+                :title="t('settings.entry')">
                 <template #icon>
                     <img :src="setting" width="28" height="28" alt="settings" />
                 </template>
             </a-button>
-             <a-popover v-if="true" v-model:open="popoverVisible" trigger="click" ok-text="Yes" cancel-text="No" placement="bottomRight">
-                <template #content>
-                    <div class="custom-popover-list">
-                        <div class="custom-popover-item" @click="togglePopover('voice')">
-                            选择音色</div>
-                        <div class="custom-popover-item" @click="togglePopover('prompt')">Prompt调试</div>
-                    </div>
-                </template>
-                <img :src="setting" alt="item actions" style="width: 28px; height: 28px; margin-right:24px;margin-top: 16px;">
-            </a-popover>
         </div>
 
-        <a-modal v-model:open="voiceModelOpen" :title="null" :mask-closable="false" :closable="false" centered>
-            <template #footer>
-                <a-button key="back" @click="handleVoiceModalCancel">Cancel</a-button>
-                <a-button key="submit" type="primary" :loading="modalLoading" @click="handleVoiceModalSubmit">Submit</a-button>
-            </template>
-            <div class="languages">
-                <div class="echo-cancel-item">
-                    <div style="display: flex; justify-content: space-between; align-items: center;">
-                        <p style="margin: 0;">Enable Echo Cancellation:</p>
-                        <a-switch v-model:checked="echoCancel" />
-                    </div>
-                </div>
-            </div>
-            <div class="languages">
-                <div class="language-item">
-                    <p>Select Language:</p>
-                    <a-select v-model:value="language" style="width: 100%;">
-                        <a-select-option v-for="lan in languages" :value="lan" :key="lan">
-                            {{ languageOptions[lan] }}
-                        </a-select-option>
-                    </a-select>
-                </div>
-            </div>
-            <div class="languages">
-                <div class="role-item">
-                    <p>Select voice Role:</p>
-                    <a-radio-group size="large" v-model:value="role">
-                        <a-radio v-for="r in filteredRoles" :style="radioStyle" :value="r['id']" :key="r['id']">
-                            <div style="display: flex; justify-content: space-between; align-items: center; width:450px;">
-                            {{ r['character_name'] }}
-                            <a-button
-                                :key="r['id']"
-                                type="text"
-                                @click="playRefAudio(r['id'], $event)"
-                                class="audio-play-btn"
-                                :class="{ 'playing': isPlaying(r['id']) }"
-                            >
-                                <SoundTwoTone
-                                    v-if="isPlaying(r['id'])"
-                                    style="font-size: 18px; color: #52c41a;"
-                                    class="playing-icon"
-                                />
-                                <SoundOutlined
-                                    v-else
-                                    style="font-size: 18px; color: #1890ff;"
-                                />
-                            </a-button>
-                        </div>
-
-                        </a-radio>
-                    </a-radio-group>
-
-                </div>
-            </div>
-        </a-modal>
-
-        <PromptText v-model:open="promptModelOpen" />
+        <SettingsModal v-model:open="settingsOpen" />
     </div>
 </template>
 
 <style lang="scss" scoped>
-
-.languages {
-    margin-top: 24px;
-    margin-bottom: 24px;
-
-    p {
-        font-size: 16px;
-        font-weight: 500;
-        margin-bottom: 8px;
-    }
-}
-
-.audio-play-btn {
-    padding: 0px 8px;
-    padding-top:2px;
-    border-radius: 4px;
-    transition: all 0.2s;
-    height: 40px;
-
-    &:hover {
-        background-color: #f0f0f0;
-    }
-
-    &.playing {
-        background-color: #f6ffed;
-        border-color: #1890ff;
-
-        .playing-icon {
-            animation: pulse 1.5s infinite;
-        }
-    }
-}
-
-@keyframes pulse {
-    0% {
-        opacity: 1;
-        transform: scale(1);
-    }
-    50% {
-        opacity: 0.7;
-        transform: scale(1.1);
-    }
-    100% {
-        opacity: 1;
-        transform: scale(1);
-    }
-}
-
-.btn-groups {
-    margin-top: 36px;
+.settings-btn {
+    width: 60px;
+    height: 60px;
+    margin-right: 24px;
+    border-radius: 50% !important;
+    background: rgba(255, 255, 255, 0.5) !important;
+    border: 1px solid rgba(255, 255, 255, 0.7) !important;
+    backdrop-filter: blur(10px);
+    -webkit-backdrop-filter: blur(10px);
+    box-shadow: 0 4px 16px rgba(31, 38, 135, 0.12);
     display: flex;
-    justify-content: space-between;
     align-items: center;
+    justify-content: center;
 }
 
-.custom-popover-list {
-    width: 92px;
-    margin: 0;
-    .custom-popover-item {
-        font-size: 14px;
-        line-height: 36px;
-        font-weight: 500;
-        color: #1e1e1e;
-        cursor: pointer;
-        border-radius: 4px;
-        padding: 0 8px;
-        margin: 0px -8px;
-        transition: background 0.2s;
-    }
-    .custom-popover-item:hover, .custom-popover-item:focus {
-        background: #e5e7eb;
-    }
-}
-
-
 .welcome-wrapper {
     width: 100%;
     height: 100%;
@@ -523,6 +140,7 @@ const isPlaying = (id: string) => {
                     margin-top: 10px;
                 }
             }
+
             .btn-box {
                 width: 224px;
                 height: 80px;
@@ -531,10 +149,11 @@ const isPlaying = (id: string) => {
     }
 
     .actions {
-        width: 100%;;
-        height: 64px;
-
+        width: 100%;
+        height: 100px;
+        margin-bottom: 32px;
         display: flex;
+        align-items: center;
         justify-content: flex-end;
     }
 }

main.py

@@ -63,6 +63,19 @@ def main():
     parser = create_argument_parser()
     args = parser.parse_args()
 
+    # 列出音频输入设备后退出
+    if getattr(args, 'list_audio_devices', False):
+        from voice_dialogue.audio.devices import list_input_devices
+        devices = list_input_devices()
+        print(f"\n可用音频输入设备 ({len(devices)}):")
+        print(f"{'索引':>4}  {'通道':>4}  {'采样率':>7}  {'默认':>4}  名称")
+        for d in devices:
+            default_mark = '✓' if d['is_default'] else ''
+            print(f"{d['index']:>4}  {d['max_input_channels']:>4}  "
+                  f"{d['default_sample_rate']:>7}  {default_mark:>4}  {d['name']}")
+        print("\n使用 --input-device <索引> 选择设备。")
+        sys.exit(0)
+
     set_debug_mode(args.debug)
 
     print(f"""
@@ -78,8 +91,10 @@ VoiceDialogue - 语音对话系统
         if args.mode == 'cli':
             print(f"语言设置: {args.language}")
             print(f"说话人: {args.speaker}")
+            if args.input_device is not None:
+                print(f"输入设备索引: {args.input_device}")
             print("正在启动命令行语音对话系统...")
-            launch_system(args.language, args.speaker, args.disable_echo_cancellation)
+            launch_system(args.language, args.speaker, args.disable_echo_cancellation, args.input_device)
 
         elif args.mode == 'api':
             launch_api_server(

src/voice_dialogue/api/core/lifespan.py

@@ -24,8 +24,8 @@ class LifespanManager:
         startup_start_time = time.time()
 
         try:
-            # 初始化系统语言
-            system_language = get_system_language()
+            # 初始化系统语言：产品默认使用中文（不随操作系统语言变化）
+            system_language = 'zh'
             logger.info(f"系统默认语言: {system_language}")
 
             # 初始化TTS配置

src/voice_dialogue/api/core/service_factories.py

@@ -12,11 +12,15 @@ class ServiceFactories:
     """服务工厂类，封装所有服务的创建逻辑"""
 
     @staticmethod
-    def create_audio_capture(enable_echo_cancellation: bool = True) -> AudioCapture:
+    def create_audio_capture(
+            enable_echo_cancellation: bool = True,
+            input_device_index: int = None,
+    ) -> AudioCapture:
         """创建音频捕获服务"""
         return AudioCapture(
             audio_frames_queue=audio_frames_queue,
-            enable_echo_cancellation=enable_echo_cancellation
+            enable_echo_cancellation=enable_echo_cancellation,
+            input_device_index=input_device_index,
         )
 
     @staticmethod
@@ -130,11 +134,14 @@ def get_core_voice_service_definitions(system_language: str, tts_config: BaseTTS
     ]
 
 
-def get_audio_capture_service_definition(enable_echo_cancellation: bool = True) -> ServiceDefinition:
+def get_audio_capture_service_definition(
+        enable_echo_cancellation: bool = True,
+        input_device_index: int = None,
+) -> ServiceDefinition:
     """获取音频捕获服务定义"""
     return ServiceDefinition(
         name="audio_capture",
-        factory=lambda: ServiceFactories.create_audio_capture(enable_echo_cancellation),
+        factory=lambda: ServiceFactories.create_audio_capture(enable_echo_cancellation, input_device_index),
         dependencies=[],
         health_check=lambda service: hasattr(service, 'is_ready') and service.is_ready
     )

src/voice_dialogue/api/routes/system_routes.py

@@ -3,11 +3,15 @@ import time
 
 from fastapi import APIRouter, HTTPException, BackgroundTasks, Request
 
+from voice_dialogue.audio.capture import resolves_to_native_aec
+from voice_dialogue.audio.devices import list_input_devices, get_default_input_device_index, is_valid_input_device
+from voice_dialogue.config.audio_config import get_input_device_index, save_input_device_index
 from voice_dialogue.core.constants import session_manager
 from voice_dialogue.utils.logger import logger
 from ..core.service_factories import get_audio_capture_service_definition, get_speech_monitor_service_definition
 from ..schemas.system_schemas import (
-    SystemStatusResponse, SystemResponse, SystemStartRequest
+    SystemStatusResponse, SystemResponse, SystemStartRequest,
+    AudioInputDevicesResponse, AudioInputDevice
 )
 
 router = APIRouter()
@@ -60,6 +64,24 @@ async def get_system_status(request: Request):
         raise HTTPException(status_code=500, detail=f"获取系统状态失败: {str(e)}")
 
 
+@router.get("/audio-devices", response_model=AudioInputDevicesResponse, summary="获取可用音频输入设备")
+async def get_audio_devices():
+    """
+    列出系统所有可用的音频输入设备（含外置麦克风/麦克风阵列），
+    供前端选择采集设备。
+    """
+    try:
+        devices = [AudioInputDevice(**d) for d in list_input_devices()]
+        return AudioInputDevicesResponse(
+            devices=devices,
+            current_device_index=get_input_device_index(),
+            default_device_index=get_default_input_device_index(),
+        )
+    except Exception as e:
+        logger.error(f"获取音频输入设备失败: {e}", exc_info=True)
+        raise HTTPException(status_code=500, detail=f"获取音频输入设备失败: {str(e)}")
+
+
 @router.post("/start", response_model=SystemResponse, summary="启动系统")
 async def start_system(
         request: SystemStartRequest,
@@ -76,6 +98,18 @@ async def start_system(
                 message="系统已经在运行中或正在启动"
             )
 
+        # 解析输入设备：请求未指定时回退到已保存的设备
+        input_device_index = request.input_device_index
+        if input_device_index is None:
+            input_device_index = get_input_device_index()
+
+        if not is_valid_input_device(input_device_index):
+            logger.warning(f"请求的输入设备 {input_device_index} 无效，回退到系统默认设备")
+            input_device_index = None
+
+        # 持久化用户选择，供下次启动复用
+        save_input_device_index(input_device_index)
+
         # 更新状态
         _system_status["status"] = "starting"
         session_manager.reset_id()
@@ -84,7 +118,8 @@ async def start_system(
         background_tasks.add_task(
             _start_system_background,
             fastapi_request,
-            request.enable_echo_cancellation
+            request.enable_echo_cancellation,
+            input_device_index,
         )
 
         return SystemResponse(
@@ -214,7 +249,11 @@ async def restart_system(
         raise HTTPException(status_code=500, detail=f"系统重启失败: {str(e)}")
 
 
-async def _start_system_background(request: Request, enable_echo_cancellation: bool = True):
+async def _start_system_background(
+        request: Request,
+        enable_echo_cancellation: bool = True,
+        input_device_index: int = None,
+):
     """
     后台启动系统的实际逻辑 - 创建并启动audio_capture服务
     """
@@ -257,7 +296,9 @@ async def _start_system_background(request: Request, enable_echo_cancellation: b
             logger.info("语音监控服务已在运行")
         else:
             # 创建语音监控服务定义
-            enable_vad = not enable_echo_cancellation
+            # 仅当走 macOS 原生 AEC（自带 VAD）时关闭软件 VAD；
+            # 选择了外置设备走 PyAudio 时，必须启用软件 VAD。
+            enable_vad = not resolves_to_native_aec(enable_echo_cancellation, input_device_index)
             speech_monitor_def = get_speech_monitor_service_definition(enable_vad)
 
             # 启动语音监控服务
@@ -271,7 +312,7 @@ async def _start_system_background(request: Request, enable_echo_cancellation: b
             logger.info("音频捕获服务已在运行")
         else:
             # 创建audio_capture服务定义
-            audio_capture_def = get_audio_capture_service_definition(enable_echo_cancellation)
+            audio_capture_def = get_audio_capture_service_definition(enable_echo_cancellation, input_device_index)
 
             # 启动audio_capture服务
             success = service_manager.start_service(audio_capture_def)

src/voice_dialogue/api/schemas/system_schemas.py

@@ -1,4 +1,4 @@
-from typing import Optional, Literal, Dict, Any
+from typing import Optional, Literal, Dict, Any, List
 
 from pydantic import BaseModel, Field
 
@@ -17,7 +17,24 @@ class SystemStatusResponse(BaseModel):
 
 class SystemStartRequest(BaseModel):
     """系统启动请求"""
-    enable_echo_cancellation: bool = Field(default=True, description="是否启用回声消除")
+    enable_echo_cancellation: bool = Field(default=True, description="是否启用回声消除（仅在未指定输入设备时使用 macOS 原生 AEC）")
+    input_device_index: Optional[int] = Field(default=None, description="输入设备索引（如外置麦克风阵列）；为空则使用系统默认设备")
+
+
+class AudioInputDevice(BaseModel):
+    """音频输入设备信息"""
+    index: int = Field(..., description="设备索引")
+    name: str = Field(..., description="设备名称")
+    max_input_channels: int = Field(..., description="最大输入通道数")
+    default_sample_rate: int = Field(..., description="设备默认采样率")
+    is_default: bool = Field(default=False, description="是否为系统默认输入设备")
+
+
+class AudioInputDevicesResponse(BaseModel):
+    """音频输入设备列表响应"""
+    devices: List[AudioInputDevice] = Field(default_factory=list, description="可用输入设备列表")
+    current_device_index: Optional[int] = Field(default=None, description="当前已选择/保存的输入设备索引")
+    default_device_index: Optional[int] = Field(default=None, description="系统默认输入设备索引")
 
 
 class SystemResponse(BaseModel):

src/voice_dialogue/audio/capture/__init__.py

@@ -4,12 +4,43 @@
 根据配置选择并管理具体的音频捕获策略。
 """
 from multiprocessing import Queue
+from typing import Optional
 
 from voice_dialogue.utils.logger import logger
 from .aec_capture import AecCapture
 from .pyaudio_capture import PyAudioCapture
 
 
+def resolves_to_native_aec(
+        enable_echo_cancellation: bool,
+        input_device_index: Optional[int] = None,
+) -> bool:
+    """
+    判断在给定配置下是否会使用 macOS 原生 AEC 采集策略。
+
+    原生 AEC 库作用于系统默认输入设备，且自带 VAD。因此当满足以下任一情况时使用原生 AEC：
+      - 启用回声消除且未指定具体输入设备（隐式使用默认设备）；
+      - 启用回声消除且所选设备恰好就是系统默认输入设备
+        （原生 AEC 本就采集默认设备，等价覆盖）。
+
+    只有当选择了"非默认"输入设备（如外置麦克风阵列）时，才退化为 PyAudio 策略——
+    此时回声消除依赖设备自身硬件，语音活动检测改用软件 VAD。
+
+    上层据此决定 SpeechStateMonitor 是否需要启用软件 VAD
+    （enable_vad = not resolves_to_native_aec(...)）。
+    """
+    if not enable_echo_cancellation:
+        return False
+    if input_device_index is None:
+        return True
+    # 所选设备即系统默认设备时，仍可使用原生 AEC
+    try:
+        from voice_dialogue.audio.devices import get_default_input_device_index
+        return input_device_index == get_default_input_device_index()
+    except Exception:
+        return False
+
+
 class AudioCapture:
     """
     音频捕获器门面 (Facade)。
@@ -23,29 +54,44 @@ class AudioCapture:
             self,
             audio_frames_queue: Queue,
             enable_echo_cancellation: bool = True,
+            input_device_index: Optional[int] = None,
+            channels: Optional[int] = None,
     ):
         """
         初始化音频捕获器。
 
         Args:
             audio_frames_queue (Queue): 用于存放捕获的音频帧的队列。
-            enable_echo_cancellation (bool): 是否启用回声消除功能。
-                                             若为 True，则使用 AEC 原生库；
-                                             否则，使用 PyAudio。
+            enable_echo_cancellation (bool): 是否启用回声消除功能。仅在未指定
+                                             input_device_index 时生效（使用 macOS
+                                             原生 AEC 库，作用于系统默认输入设备）。
+            input_device_index (Optional[int]): 指定的输入设备索引（如外置麦克风阵列）。
+                                                一旦指定，则使用 PyAudio 策略采集该设备，
+                                                回声消除依赖设备硬件。
+            channels (Optional[int]): 采集通道数（仅 PyAudio 策略生效，多通道会降混为单声道）。
         """
+        use_native_aec = resolves_to_native_aec(enable_echo_cancellation, input_device_index)
         self._strategy = None
         try:
-            if enable_echo_cancellation:
+            if use_native_aec:
                 self._strategy = AecCapture(audio_frames_queue=audio_frames_queue)
             else:
-                self._strategy = PyAudioCapture(audio_frames_queue=audio_frames_queue)
+                self._strategy = PyAudioCapture(
+                    audio_frames_queue=audio_frames_queue,
+                    input_device_index=input_device_index,
+                    channels=channels,
+                )
             logger.info(f"音频捕获策略已选择: {self._strategy.__class__.__name__}")
         except Exception as e:
             logger.error(
-                f"初始化 {AecCapture.__name__ if enable_echo_cancellation else PyAudioCapture.__name__} 失败: {e}, 将回退到 PyAudio。")
+                f"初始化 {AecCapture.__name__ if use_native_aec else PyAudioCapture.__name__} 失败: {e}, 将回退到 PyAudio。")
             # 只有在尝试 AEC 失败时才回退
             if not isinstance(self._strategy, PyAudioCapture):
-                self._strategy = PyAudioCapture(audio_frames_queue=audio_frames_queue)
+                self._strategy = PyAudioCapture(
+                    audio_frames_queue=audio_frames_queue,
+                    input_device_index=input_device_index,
+                    channels=channels,
+                )
                 logger.info(f"已回退到音频捕获策略: {self._strategy.__class__.__name__}")
 
     def start(self):

src/voice_dialogue/audio/capture/pyaudio_capture.py

@@ -1,41 +1,130 @@
 from multiprocessing import Queue
+from typing import Optional
 
+import numpy as np
 import pyaudio
 
 from voice_dialogue.utils.logger import logger
 from .base_capture import BaseCapture
 
+# 下游 ASR / VAD 统一要求 16kHz 单声道 int16 音频
+TARGET_SAMPLE_RATE = 16000
+
 
 class PyAudioCapture(BaseCapture):
     """
     使用 PyAudio 进行标准的音频采集策略。
+
+    支持选择指定的输入设备（如外置麦克风阵列），并自动将多通道、
+    非 16kHz 的输入降混并重采样为下游所需的 16kHz 单声道 int16 数据。
     """
 
-    def __init__(self, audio_frames_queue: Queue, **kwargs):
+    def __init__(
+            self,
+            audio_frames_queue: Queue,
+            input_device_index: Optional[int] = None,
+            channels: Optional[int] = None,
+            **kwargs
+    ):
+        """
+        Args:
+            audio_frames_queue (Queue): 用于存放捕获的音频帧的队列。
+            input_device_index (Optional[int]): 输入设备索引；None 表示使用系统默认设备。
+            channels (Optional[int]): 采集通道数；None 表示自动使用设备支持的最大通道数
+                                      （麦克风阵列通常为多通道，采集后会降混为单声道）。
+        """
         super().__init__(audio_frames_queue=audio_frames_queue, **kwargs)
+        self.input_device_index = input_device_index
+        self.requested_channels = channels
+
+    def _resolve_device_params(self, p: pyaudio.PyAudio):
+        """根据所选设备解析采集通道数与采集采样率。"""
+        # 默认参数（系统默认设备、单声道、16kHz）
+        device_index = self.input_device_index
+        channels = self.requested_channels or 1
+        sample_rate = TARGET_SAMPLE_RATE
+
+        try:
+            if device_index is None:
+                device_index = int(p.get_default_input_device_info().get("index"))
+            info = p.get_device_info_by_index(device_index)
+            max_channels = int(info.get("maxInputChannels", 1)) or 1
+            # 未显式指定通道数时，采集设备的全部通道再降混（适配麦克风阵列）
+            if self.requested_channels is None:
+                channels = max_channels
+            else:
+                channels = min(self.requested_channels, max_channels)
+
+            # 优先尝试 16kHz；若设备不支持则采用设备默认采样率，后续重采样
+            device_rate = int(info.get("defaultSampleRate", TARGET_SAMPLE_RATE))
+            if not p.is_format_supported(
+                    rate=TARGET_SAMPLE_RATE,
+                    input_device=device_index,
+                    input_channels=channels,
+                    input_format=pyaudio.paInt16,
+            ):
+                sample_rate = device_rate
+        except Exception as e:
+            logger.warning(f"解析输入设备参数失败，回退到默认设备/单声道/16kHz: {e}")
+            device_index = self.input_device_index
+            channels = 1
+            sample_rate = TARGET_SAMPLE_RATE
+
+        return device_index, channels, sample_rate
 
     def _init_pyaudio(self):
         """初始化 PyAudio 并返回实例和配置。"""
         p = pyaudio.PyAudio()
-        chunk = 1024
-        sample_rate = 16000
-        return p, chunk, sample_rate
+        device_index, channels, sample_rate = self._resolve_device_params(p)
+        # 采集块大小按采集采样率取约 64ms，保证重采样后帧长足够 VAD 处理
+        chunk = max(1024, int(sample_rate * 0.064))
+        logger.info(
+            f"PyAudio 采集配置: device_index={device_index}, channels={channels}, "
+            f"sample_rate={sample_rate} -> {TARGET_SAMPLE_RATE}, chunk={chunk}"
+        )
+        return p, chunk, sample_rate, channels, device_index
 
-    def _open_stream(self, p, chunk, sample_rate):
+    def _open_stream(self, p, chunk, sample_rate, channels, device_index):
         """打开 PyAudio 音频流。"""
         return p.open(
             format=pyaudio.paInt16,
-            channels=1,
+            channels=channels,
             rate=sample_rate,
             input=True,
+            input_device_index=device_index,
             frames_per_buffer=chunk,
         )
 
-    def _capture_loop(self, stream, chunk):
+    def _to_mono_16k(self, data: bytes, channels: int, sample_rate: int) -> Optional[bytes]:
+        """将原始多通道/任意采样率的 int16 数据降混并重采样为 16kHz 单声道 int16。"""
+        samples = np.frombuffer(data, dtype=np.int16)
+        if samples.size == 0:
+            return None
+
+        # 多通道降混为单声道（按通道求平均）
+        if channels > 1:
+            frame_count = samples.size // channels
+            if frame_count == 0:
+                return None
+            samples = samples[:frame_count * channels].reshape(-1, channels)
+            mono = samples.astype(np.float32).mean(axis=1)
+        else:
+            mono = samples.astype(np.float32)
+
+        # 重采样到 16kHz
+        if sample_rate != TARGET_SAMPLE_RATE:
+            import soxr
+            mono = soxr.resample(mono, sample_rate, TARGET_SAMPLE_RATE)
+
+        return np.clip(mono, -32768, 32767).astype(np.int16).tobytes()
+
+    def _capture_loop(self, stream, chunk, channels, sample_rate):
         """PyAudio 音频捕获的主循环。"""
         logger.info("使用 PyAudio 开始音频采集...")
         self.is_ready = True
 
+        needs_processing = channels > 1 or sample_rate != TARGET_SAMPLE_RATE
+
         while not self.is_exited:
             data = stream.read(chunk, exception_on_overflow=False)
             if data is None:
@@ -44,6 +133,11 @@ class PyAudioCapture(BaseCapture):
             if self.is_paused:
                 continue
 
+            if needs_processing:
+                data = self._to_mono_16k(data, channels, sample_rate)
+                if data is None:
+                    continue
+
             self.audio_frames_queue.put(data)
 
     def _cleanup(self, stream, p):
@@ -57,11 +151,11 @@ class PyAudioCapture(BaseCapture):
         """
         线程主循环，执行 PyAudio 音频采集。
         """
-        p, chunk, sample_rate = self._init_pyaudio()
+        p, chunk, sample_rate, channels, device_index = self._init_pyaudio()
         stream = None
         try:
-            stream = self._open_stream(p, chunk, sample_rate)
-            self._capture_loop(stream, chunk)
+            stream = self._open_stream(p, chunk, sample_rate, channels, device_index)
+            self._capture_loop(stream, chunk, channels, sample_rate)
         except Exception as e:
             logger.error(f'PyAudio 音频捕获器运行时发生错误: {e}')
         finally:

src/voice_dialogue/audio/devices.py

@@ -0,0 +1,89 @@
+"""
+音频输入设备枚举工具。
+
+提供列出系统可用输入设备（包括外置麦克风/麦克风阵列）的能力，
+供 CLI、API 以及前端进行设备选择。
+"""
+from typing import List, Optional, TypedDict
+
+import pyaudio
+
+from voice_dialogue.utils.logger import logger
+
+
+class InputDeviceInfo(TypedDict):
+    """输入设备信息。"""
+    index: int
+    name: str
+    max_input_channels: int
+    default_sample_rate: int
+    is_default: bool
+
+
+def _get_default_input_index(p: pyaudio.PyAudio) -> Optional[int]:
+    """获取系统默认输入设备索引，失败时返回 None。"""
+    try:
+        return int(p.get_default_input_device_info().get("index"))
+    except Exception:
+        return None
+
+
+def list_input_devices() -> List[InputDeviceInfo]:
+    """
+    列出所有可用的音频输入设备。
+
+    Returns:
+        List[InputDeviceInfo]: 输入设备列表（仅包含 maxInputChannels > 0 的设备）。
+    """
+    devices: List[InputDeviceInfo] = []
+    p = pyaudio.PyAudio()
+    try:
+        default_index = _get_default_input_index(p)
+        for i in range(p.get_device_count()):
+            try:
+                info = p.get_device_info_by_index(i)
+            except Exception as e:
+                logger.warning(f"读取音频设备 {i} 信息失败: {e}")
+                continue
+
+            max_input_channels = int(info.get("maxInputChannels", 0))
+            if max_input_channels <= 0:
+                continue
+
+            devices.append(
+                InputDeviceInfo(
+                    index=int(info.get("index", i)),
+                    name=str(info.get("name", f"device-{i}")),
+                    max_input_channels=max_input_channels,
+                    default_sample_rate=int(info.get("defaultSampleRate", 16000)),
+                    is_default=(int(info.get("index", i)) == default_index),
+                )
+            )
+    finally:
+        p.terminate()
+
+    return devices
+
+
+def get_default_input_device_index() -> Optional[int]:
+    """获取系统默认输入设备索引。"""
+    p = pyaudio.PyAudio()
+    try:
+        return _get_default_input_index(p)
+    finally:
+        p.terminate()
+
+
+def is_valid_input_device(index: Optional[int]) -> bool:
+    """
+    校验给定索引是否为有效的输入设备。
+
+    Args:
+        index: 设备索引；None 表示使用系统默认设备，视为有效。
+
+    Returns:
+        bool: 是否有效。
+    """
+    if index is None:
+        return True
+    return any(d["index"] == index for d in list_input_devices())

src/voice_dialogue/cli/args.py

@@ -74,6 +74,20 @@ def create_argument_parser():
         default=False,
         help='禁用回声消除功能 (默认: 不禁用)'
     )
+    cli_group.add_argument(
+        '--input-device', '-i',
+        type=int,
+        default=None,
+        metavar='INDEX',
+        help='指定输入设备索引（如外置麦克风阵列）。多通道会自动降混为单声道；'
+             '指定后回声消除依赖设备硬件。用 --list-audio-devices 查看可用索引。'
+    )
+    cli_group.add_argument(
+        '--list-audio-devices',
+        action='store_true',
+        default=False,
+        help='列出可用的音频输入设备及其索引后退出'
+    )
 
     # API服务器模式参数
     api_group = parser.add_argument_group('API服务器模式参数')

src/voice_dialogue/config/audio_config.py

@@ -0,0 +1,60 @@
+"""音频设备配置管理模块。
+
+持久化用户选择的输入设备（如外置麦克风阵列），在重启后自动复用。
+"""
+import json
+from typing import Optional, TypedDict
+
+from voice_dialogue.utils.logger import logger
+from .paths import AUDIO_SETTINGS_PATH
+
+
+class AudioSettings(TypedDict, total=False):
+    """音频设置。"""
+    input_device_index: Optional[int]
+
+
+_audio_settings_cache: Optional[AudioSettings] = None
+
+
+def get_audio_settings() -> AudioSettings:
+    """加载用户音频设置（带内存缓存）。"""
+    global _audio_settings_cache
+    if _audio_settings_cache is not None:
+        return _audio_settings_cache
+
+    if not AUDIO_SETTINGS_PATH.exists():
+        _audio_settings_cache = {}
+        return _audio_settings_cache
+
+    try:
+        with open(AUDIO_SETTINGS_PATH, "r", encoding="utf-8") as f:
+            _audio_settings_cache = json.load(f)
+    except (json.JSONDecodeError, IOError) as e:
+        logger.error(f"无法加载音频设置，使用空配置: {e}")
+        _audio_settings_cache = {}
+    return _audio_settings_cache
+
+
+def get_input_device_index() -> Optional[int]:
+    """获取已保存的输入设备索引；未配置时返回 None（系统默认设备）。"""
+    value = get_audio_settings().get("input_device_index")
+    return int(value) if value is not None else None
+
+
+def save_input_device_index(input_device_index: Optional[int]) -> bool:
+    """保存用户选择的输入设备索引。"""
+    global _audio_settings_cache
+    settings = dict(get_audio_settings())
+    settings["input_device_index"] = input_device_index
+    try:
+        if not AUDIO_SETTINGS_PATH.parent.exists():
+            AUDIO_SETTINGS_PATH.parent.mkdir(parents=True, exist_ok=True)
+        with open(AUDIO_SETTINGS_PATH, "w", encoding="utf-8") as f:
+            json.dump(settings, f, ensure_ascii=False, indent=4)
+        _audio_settings_cache = settings  # type: ignore[assignment]
+        logger.info(f"音频设置已保存: input_device_index={input_device_index}")
+        return True
+    except IOError as e:
+        logger.error(f"无法保存音频设置: {e}")
+        return False

src/voice_dialogue/config/paths.py

@@ -46,6 +46,7 @@ APP_DATA_PATH = get_app_data_path()
 if not APP_DATA_PATH.exists():
     APP_DATA_PATH.mkdir(parents=True, exist_ok=True)
 USER_PROMPTS_PATH = APP_DATA_PATH / "user_prompts.json"
+AUDIO_SETTINGS_PATH = APP_DATA_PATH / "audio_settings.json"
 
 
 def load_third_party():

src/voice_dialogue/core/launcher.py

@@ -6,7 +6,7 @@
 
 import time
 
-from voice_dialogue.audio.capture import AudioCapture
+from voice_dialogue.audio.capture import AudioCapture, resolves_to_native_aec
 from voice_dialogue.config.speaker_config import get_tts_config_by_speaker_name, get_available_speaker_names
 from voice_dialogue.core.constants import (
     audio_frames_queue,
@@ -23,6 +23,7 @@ def launch_system(
         user_language: str,
         speaker: str,
         disable_echo_cancellation: bool = False,
+        input_device_index: int = None,
 ) -> None:
     """
     启动完整的语音对话系统
@@ -100,7 +101,10 @@ def launch_system(
     threads.append(audio_player)
 
     # 语音状态监测
-    enable_vad = disable_echo_cancellation
+    # 仅当走 macOS 原生 AEC（自带 VAD）时关闭软件 VAD；
+    # 指定外置设备走 PyAudio 时，必须启用软件 VAD。
+    enable_echo_cancellation = not disable_echo_cancellation
+    enable_vad = not resolves_to_native_aec(enable_echo_cancellation, input_device_index)
     speech_monitor = SpeechStateMonitor(
         audio_frame_queue=audio_frames_queue,
         user_voice_queue=user_voice_queue,
@@ -111,10 +115,10 @@ def launch_system(
     threads.append(speech_monitor)
 
     # 音频采集
-    enable_echo_cancellation = not disable_echo_cancellation
     audio_capture = AudioCapture(
         audio_frames_queue=audio_frames_queue,
-        enable_echo_cancellation=enable_echo_cancellation
+        enable_echo_cancellation=enable_echo_cancellation,
+        input_device_index=input_device_index,
     )
     audio_capture.daemon = True
     audio_capture.start()