Enhance audio and video device selection in MobileCallPage

- Refactor `CallDeviceSelect` to `CallAudioDeviceSelect`, allowing simultaneous selection of microphone and output devices.
- Introduce dropdown menus for selecting audio input and output devices, improving user experience.
- Update `useVoicePreview` hook to manage audio output devices, including support for selecting and applying output devices.
- Modify `MobileCallPage` to integrate new device selection components, ensuring seamless audio and video call functionality.

frontend/src/components/pages/MobileCallPage.tsx

@@ -1,8 +1,17 @@
 "use client";
 
 import { useCallback, useEffect, useRef } from "react";
-import { Camera, Loader2, Mic, Phone, PhoneOff, Video } from "lucide-react";
+import { Camera, Headphones, Loader2, Mic, Phone, PhoneOff, Video } from "lucide-react";
 import { Button } from "@/components/ui/button";
+import {
+  DropdownMenu,
+  DropdownMenuContent,
+  DropdownMenuLabel,
+  DropdownMenuRadioGroup,
+  DropdownMenuRadioItem,
+  DropdownMenuSeparator,
+  DropdownMenuTrigger,
+} from "@/components/ui/dropdown-menu";
 import {
   Select,
   SelectContent,
@@ -13,22 +22,99 @@ import {
 import { useCameraPreview } from "@/hooks/use-camera-preview";
 import { useVoicePreview } from "@/hooks/use-voice-preview";
 
-function CallDeviceSelect({
+function CallAudioDeviceSelect({
-  kind,
+  micValue,
+  micDevices,
+  onMicSelect,
+  outputValue,
+  outputDevices,
+  onOutputSelect,
+  outputSupported,
+  disabled,
+}: {
+  micValue: string;
+  micDevices: MediaDeviceInfo[];
+  onMicSelect: (deviceId: string) => void | Promise<void>;
+  outputValue: string;
+  outputDevices: MediaDeviceInfo[];
+  onOutputSelect: (deviceId: string) => void;
+  outputSupported: boolean;
+  disabled?: boolean;
+}) {
+  return (
+    <DropdownMenu>
+      <DropdownMenuTrigger asChild disabled={disabled}>
+        <button
+          type="button"
+          aria-label="选择麦克风和耳机"
+          title="选择麦克风和耳机"
+          className="flex size-12 items-center justify-center rounded-full border border-white/15 bg-black/45 text-white shadow-xl backdrop-blur-md transition-colors hover:bg-black/60 focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-white/40 disabled:pointer-events-none disabled:opacity-50 min-[380px]:size-14"
+        >
+          <Mic className="size-5" />
+        </button>
+      </DropdownMenuTrigger>
+      <DropdownMenuContent
+        side="top"
+        align="start"
+        className="max-w-[min(20rem,calc(100vw-2rem))] min-w-56"
+      >
+        <DropdownMenuLabel>麦克风</DropdownMenuLabel>
+        <DropdownMenuRadioGroup
+          value={micValue || "default"}
+          onValueChange={(next) =>
+            void onMicSelect(next === "default" ? "" : next)
+          }
+        >
+          <DropdownMenuRadioItem value="default">默认麦克风</DropdownMenuRadioItem>
+          {micDevices.map((device, index) => (
+            <DropdownMenuRadioItem key={device.deviceId} value={device.deviceId}>
+              {device.label || `麦克风 ${index + 1}`}
+            </DropdownMenuRadioItem>
+          ))}
+        </DropdownMenuRadioGroup>
+        {outputSupported && (
+          <>
+            <DropdownMenuSeparator />
+            <DropdownMenuLabel className="flex items-center gap-2">
+              <Headphones className="size-3.5" />
+              耳机 / 扬声器
+            </DropdownMenuLabel>
+            <DropdownMenuRadioGroup
+              value={outputValue || "default"}
+              onValueChange={(next) =>
+                onOutputSelect(next === "default" ? "" : next)
+              }
+            >
+              <DropdownMenuRadioItem value="default">
+                默认输出设备
+              </DropdownMenuRadioItem>
+              {outputDevices.map((device, index) => (
+                <DropdownMenuRadioItem
+                  key={device.deviceId}
+                  value={device.deviceId}
+                >
+                  {device.label || `输出设备 ${index + 1}`}
+                </DropdownMenuRadioItem>
+              ))}
+            </DropdownMenuRadioGroup>
+          </>
+        )}
+      </DropdownMenuContent>
+    </DropdownMenu>
+  );
+}
+
+function CallCameraDeviceSelect({
   value,
   devices,
   onSelect,
   disabled,
 }: {
-  kind: "microphone" | "camera";
   value: string;
   devices: MediaDeviceInfo[];
   onSelect: (deviceId: string) => void | Promise<void>;
   disabled?: boolean;
 }) {
-  const isMic = kind === "microphone";
-  const label = isMic ? "选择麦克风" : "选择摄像头";
-
   return (
     <Select
       value={value || "default"}
@@ -39,30 +125,24 @@ function CallDeviceSelect({
     >
       <SelectTrigger
         size="default"
-        aria-label={label}
+        aria-label="选择摄像头"
-        title={label}
+        title="选择摄像头"
         className="w-12 justify-center overflow-hidden rounded-full border-white/15 bg-black/45 p-0 text-white shadow-xl backdrop-blur-md hover:bg-black/60 focus-visible:ring-white/40 data-[size=default]:h-12 min-[380px]:w-14 min-[380px]:data-[size=default]:h-14 [&>svg:last-child]:hidden"
       >
         <span className="sr-only">
           <SelectValue />
         </span>
-        {isMic ? (
+        <Camera className="size-5" />
-          <Mic className="size-5" />
-        ) : (
-          <Camera className="size-5" />
-        )}
       </SelectTrigger>
       <SelectContent
         side="top"
-        align={isMic ? "start" : "end"}
+        align="end"
         className="max-w-[min(20rem,calc(100vw-2rem))]"
       >
-        <SelectItem value="default">
+        <SelectItem value="default">默认摄像头</SelectItem>
-          {isMic ? "默认麦克风" : "默认摄像头"}
-        </SelectItem>
         {devices.map((device, index) => (
           <SelectItem key={device.deviceId} value={device.deviceId}>
-            {device.label || `${isMic ? "麦克风" : "摄像头"} ${index + 1}`}
+            {device.label || `摄像头 ${index + 1}`}
           </SelectItem>
         ))}
       </SelectContent>
@@ -77,8 +157,12 @@ export function MobileCallPage({ assistantId }: { assistantId: string }) {
     status,
     error: previewError,
     audioInputs,
+    audioOutputs,
     selectedDeviceId,
+    selectedOutputDeviceId,
     selectDevice,
+    selectOutputDevice,
+    supportsOutputSelection,
     connect,
     replaceVideoStream,
     disconnect,
@@ -203,11 +287,14 @@ export function MobileCallPage({ assistantId }: { assistantId: string }) {
 
       {inCall && (
         <div className="absolute bottom-[max(1.25rem,env(safe-area-inset-bottom))] left-1/2 flex -translate-x-1/2 items-center gap-3 min-[380px]:gap-5 landscape:bottom-[max(1rem,env(safe-area-inset-bottom))]">
-          <CallDeviceSelect
+          <CallAudioDeviceSelect
-            kind="microphone"
+            micValue={selectedDeviceId}
-            value={selectedDeviceId}
+            micDevices={audioInputs}
-            devices={audioInputs}
+            onMicSelect={selectDevice}
-            onSelect={selectDevice}
+            outputValue={selectedOutputDeviceId}
+            outputDevices={audioOutputs}
+            onOutputSelect={selectOutputDevice}
+            outputSupported={supportsOutputSelection}
             disabled={connecting}
           />
           <button
@@ -219,8 +306,7 @@ export function MobileCallPage({ assistantId }: { assistantId: string }) {
           >
             <PhoneOff className="size-6 min-[380px]:size-7" />
           </button>
-          <CallDeviceSelect
+          <CallCameraDeviceSelect
-            kind="camera"
             value={cameraDeviceId}
             devices={cameraDevices}
             onSelect={selectCamera}

frontend/src/hooks/use-voice-preview.ts

@@ -99,10 +99,14 @@ export function useVoicePreview(
   // 远端(助手 TTS)媒体流:除挂到 <audio> 播放外,也暴露给波形可视化
   const [remoteStream, setRemoteStream] = useState<MediaStream | null>(null);
   const [messages, setMessages] = useState<ChatMessage[]>([]);
-  // 可选麦克风列表与当前选择(空串表示交给浏览器选默认设备)
+  // 可选麦克风/扬声器列表与当前选择(空串表示交给浏览器选默认设备)
   const [audioInputs, setAudioInputs] = useState<MediaDeviceInfo[]>([]);
+  const [audioOutputs, setAudioOutputs] = useState<MediaDeviceInfo[]>([]);
   const [selectedDeviceId, setSelectedDeviceId] = useState<string>("");
+  const [selectedOutputDeviceId, setSelectedOutputDeviceId] =
+    useState<string>("");
   const selectedDeviceIdRef = useRef("");
+  const selectedOutputDeviceIdRef = useRef("");
   const audioRef = useRef<HTMLAudioElement | null>(null);
   const pcRef = useRef<RTCPeerConnection | null>(null);
   const wsRef = useRef<WebSocket | null>(null);
@@ -118,14 +122,32 @@ export function useVoicePreview(
     onNodeActiveRef.current = onNodeActive;
   }, [onNodeActive]);
 
-  // 枚举可用麦克风。未授权前 label 为空,授权(连接)后再刷新即可拿到名称。
+  const supportsOutputSelection =
+    typeof HTMLAudioElement !== "undefined" &&
+    "setSinkId" in HTMLAudioElement.prototype;
+
+  const applyOutputDevice = useCallback(async (deviceId: string) => {
+    const audio = audioRef.current;
+    if (!audio || !supportsOutputSelection) return;
+    try {
+      await audio.setSinkId(deviceId);
+    } catch {
+      /* 忽略切换失败(设备不可用或不支持) */
+    }
+  }, [supportsOutputSelection]);
+
+  // 枚举可用麦克风与扬声器。未授权前 label 为空,授权(连接)后再刷新即可拿到名称。
   const refreshDevices = useCallback(async () => {
     if (!navigator.mediaDevices?.enumerateDevices) return;
     try {
       const devices = await navigator.mediaDevices.enumerateDevices();
       // 未授权时设备可能没有 deviceId(空串),无法选择,直接过滤掉
       const inputs = devices.filter((d) => d.kind === "audioinput" && d.deviceId);
+      const outputs = devices.filter(
+        (d) => d.kind === "audiooutput" && d.deviceId,
+      );
       setAudioInputs(inputs);
+      setAudioOutputs(outputs);
       // 选中的设备已被拔出时,回退到浏览器默认设备
       if (
         selectedDeviceIdRef.current &&
@@ -133,16 +155,31 @@ export function useVoicePreview(
       ) {
         setSelectedDeviceId("");
       }
+      if (
+        selectedOutputDeviceIdRef.current &&
+        !outputs.some((d) => d.deviceId === selectedOutputDeviceIdRef.current)
+      ) {
+        setSelectedOutputDeviceId("");
+        void applyOutputDevice("");
+      }
     } catch {
       /* 忽略枚举失败 */
     }
-  }, []);
+  }, [applyOutputDevice]);
 
   // connect/refreshDevices 在回调里读最新选择,避免把它们挂进依赖反复重建
   useEffect(() => {
     selectedDeviceIdRef.current = selectedDeviceId;
   }, [selectedDeviceId]);
 
+  useEffect(() => {
+    selectedOutputDeviceIdRef.current = selectedOutputDeviceId;
+  }, [selectedOutputDeviceId]);
+
+  useEffect(() => {
+    void applyOutputDevice(selectedOutputDeviceId);
+  }, [applyOutputDevice, selectedOutputDeviceId, remoteStream]);
+
   useEffect(() => {
     if (!navigator.mediaDevices?.enumerateDevices) return;
     // eslint-disable-next-line react-hooks/set-state-in-effect
@@ -336,7 +373,6 @@ export function useVoicePreview(
       };
 
       // 应用消息通道:收后端转写(聊天记录),发文字输入。
-      // 由浏览器侧主动创建,后端 SmallWebRTCConnection 的 on("datachannel") 会接住。
       const channel = pc.createDataChannel("chat");
       dataChannelRef.current = channel;
       channel.onopen = () => {
@@ -433,7 +469,9 @@ export function useVoicePreview(
         setRemoteStream(remote);
         if (audioRef.current) {
           audioRef.current.srcObject = remote;
-          void audioRef.current.play().catch(() => {});
+          void applyOutputDevice(selectedOutputDeviceIdRef.current).then(() =>
+            audioRef.current?.play().catch(() => {}),
+          );
         }
       };
 
@@ -491,7 +529,7 @@ export function useVoicePreview(
     } finally {
       startingRef.current = false;
     }
-  }, [assistantId, fail, closeOnRemoteEnd, refreshDevices]);
+  }, [assistantId, applyOutputDevice, fail, closeOnRemoteEnd, refreshDevices]);
 
   const replaceVideoStream = useCallback(
     async (videoStream: MediaStream | null) => {
@@ -554,6 +592,15 @@ export function useVoicePreview(
     [],
   );
 
+  const selectOutputDevice = useCallback(
+    (deviceId: string) => {
+      setSelectedOutputDeviceId(deviceId);
+      selectedOutputDeviceIdRef.current = deviceId;
+      void applyOutputDevice(deviceId);
+    },
+    [applyOutputDevice],
+  );
+
   // 发送文字消息:后端先打断当前播报,再按用户输入触发新回复。
   // 成功返回 true;通道未就绪(未开始对话/连接中)返回 false。
   const sendText = useCallback((text: string): boolean => {
@@ -575,9 +622,13 @@ export function useVoicePreview(
     remoteStream,
     messages,
     audioInputs,
+    audioOutputs,
     selectedDeviceId,
+    selectedOutputDeviceId,
     setSelectedDeviceId,
     selectDevice,
+    selectOutputDevice,
+    supportsOutputSelection,
     sendText,
     connect,
     replaceVideoStream,