feat: voice mode bottom sheet with hold/tap/auto talk modes

dashboard/frontend/src/lib/voice.js

@@ -1,6 +1,12 @@
 const LANGS = ["en-US", "zh-CN", "zh-HK"];
 const LANG_LABELS = ["EN", "普", "粤"];
+const TALK_MODES = ["auto", "tap", "hold"];
+const MODE_LABELS = { auto: "Auto", tap: "Tap", hold: "Hold" };
 const NOISE = /^[\s│┃┆┇┊┋─━┄┅┈┉═╌╍╔╗╚╝╠╣╦╩╬├┤┬┴┼╭╮╯╰░▒▓█▌▐▀▄⠋⠙⠹⠸⠼⠴⠦⠧⠇⠏◐◑◒◓⣾⣽⣻⢿⡿⣟⣯⣷⠁⠂⠄⡀⢀⠠⠐⠈\-–—_=+*#~<>/.\\|,;:!?(){}\[\]]+$/;
+const MODE_COMMANDS = {
+  "hold mode": "hold", "tap mode": "tap", "auto mode": "auto",
+  "按住模式": "hold", "点击模式": "tap", "自动模式": "auto",
+};
 
 export class VoiceSession {
   constructor(onChange) {
@@ -10,6 +16,9 @@ export class VoiceSession {
     this.listening = false;
     this.sttAvailable = !!(window.SpeechRecognition || window.webkitSpeechRecognition);
     this.langIdx = 0;
+    this.voiceMode = false;
+    this.transcript = "";
+    this.talkMode = "auto";
     this.term = null;
     this.ws = null;
     this._lastLine = 0;
@@ -17,10 +26,12 @@ export class VoiceSession {
     this._recog = null;
     this._onWriteDispose = null;
     this._ttsUnlocked = false;
+    this._silenceTimer = null;
   }
 
   get lang() { return LANGS[this.langIdx]; }
   get langLabel() { return LANG_LABELS[this.langIdx]; }
+  get modeLabel() { return MODE_LABELS[this.talkMode]; }
 
   attach(term, ws) {
     this.term = term;
@@ -31,6 +42,131 @@ export class VoiceSession {
 
   _notify() { this.onChange?.(); }
 
+  // --- Voice Mode ---
+  enterVoiceMode() {
+    this.voiceMode = true;
+    this.ttsEnabled = true;
+    if (!this._ttsUnlocked) {
+      speechSynthesis.speak(new SpeechSynthesisUtterance(""));
+      this._ttsUnlocked = true;
+    }
+    if (this.term) {
+      this._lastLine = this.term.buffer.active.baseY + this.term.buffer.active.cursorY;
+    }
+    if (this.talkMode === "auto") this._startSTT(true);
+    this._notify();
+  }
+
+  exitVoiceMode() {
+    this.voiceMode = false;
+    this.ttsEnabled = false;
+    this.transcript = "";
+    speechSynthesis.cancel();
+    this.speaking = false;
+    this._stopSTT();
+    this._notify();
+  }
+
+  setTalkMode(mode) {
+    this.talkMode = mode;
+    this._stopSTT();
+    this.transcript = "";
+    if (mode === "auto" && this.voiceMode) this._startSTT(true);
+    this._notify();
+  }
+
+  cycleTalkMode() {
+    const i = (TALK_MODES.indexOf(this.talkMode) + 1) % TALK_MODES.length;
+    this.setTalkMode(TALK_MODES[i]);
+  }
+
+  // Hold mode
+  startHold() {
+    if (this.talkMode !== "hold") return;
+    this._startSTT(false);
+  }
+
+  releaseHold() {
+    if (this.talkMode !== "hold") return;
+    this._sendAndStop();
+  }
+
+  // Tap mode
+  tapMic() {
+    if (this.talkMode !== "tap") return;
+    if (this.listening) { this._sendAndStop(); }
+    else { this._startSTT(false); }
+  }
+
+  _sendAndStop() {
+    const text = this.transcript.trim();
+    this._stopSTT();
+    if (text && this.ws?.readyState === 1) this._typeText(text + "\r");
+    this.transcript = "";
+    this._notify();
+  }
+
+  // --- STT ---
+  _startSTT(continuous) {
+    const SR = window.SpeechRecognition || window.webkitSpeechRecognition;
+    if (!SR) return;
+    speechSynthesis.cancel();
+    this.speaking = false;
+    this._recog = new SR();
+    this._recog.continuous = continuous;
+    this._recog.interimResults = true;
+    this._recog.lang = this.lang;
+    this._finalText = "";
+    this._recog.onresult = (e) => {
+      let interim = "";
+      for (let i = e.resultIndex; i < e.results.length; i++) {
+        const t = e.results[i][0].transcript;
+        if (e.results[i].isFinal) {
+          // Check voice commands
+          const cmd = MODE_COMMANDS[t.trim().toLowerCase()];
+          if (cmd) { this.setTalkMode(cmd); return; }
+          this._finalText += t;
+        } else { interim = t; }
+      }
+      this.transcript = (this._finalText + interim).trim();
+      this._notify();
+      // Auto mode: reset silence timer on new results
+      if (this.talkMode === "auto" && this._finalText) {
+        clearTimeout(this._silenceTimer);
+        this._silenceTimer = setTimeout(() => this._autoSend(), 2000);
+      }
+    };
+    this._recog.onerror = () => { this.listening = false; this._notify(); };
+    this._recog.onend = () => {
+      if (this.listening && continuous) {
+        // Auto mode: send accumulated text, restart
+        if (this.talkMode === "auto" && this._finalText.trim()) this._autoSend();
+        try { this._recog.start(); } catch { this.listening = false; this._notify(); }
+      } else { this.listening = false; this._notify(); }
+    };
+    this._recog.start();
+    this.listening = true;
+    this._notify();
+  }
+
+  _stopSTT() {
+    this.listening = false;
+    clearTimeout(this._silenceTimer);
+    this._recog?.stop();
+    this._recog = null;
+    this._finalText = "";
+  }
+
+  _autoSend() {
+    clearTimeout(this._silenceTimer);
+    const text = this._finalText.trim();
+    this._finalText = "";
+    this.transcript = "";
+    if (text && this.ws?.readyState === 1) this._typeText(text + "\r");
+    this._notify();
+  }
+
+  // --- TTS ---
   _onOutput() {
     if (!this.ttsEnabled || this.listening) return;
     clearTimeout(this._debounce);
@@ -66,24 +202,6 @@ export class VoiceSession {
     speechSynthesis.speak(u);
   }
 
-  toggleTts() {
-    this.ttsEnabled = !this.ttsEnabled;
-    if (!this.ttsEnabled) { speechSynthesis.cancel(); this.speaking = false; }
-    if (this.ttsEnabled) {
-      // iOS Safari requires first speak() in a user gesture to unlock audio
-      if (!this._ttsUnlocked) {
-        const u = new SpeechSynthesisUtterance("");
-        speechSynthesis.speak(u);
-        this._ttsUnlocked = true;
-      }
-      if (this.term) {
-        const buf = this.term.buffer.active;
-        this._lastLine = buf.baseY + buf.cursorY;
-      }
-    }
-    this._notify();
-  }
-
   _typeText(text) {
     let i = 0;
     const next = () => {
@@ -95,58 +213,21 @@ export class VoiceSession {
     next();
   }
 
-  toggleMic() {
-    this.listening ? this.stopListening() : this.startListening();
-  }
-
-  startListening() {
-    const SR = window.SpeechRecognition || window.webkitSpeechRecognition;
-    if (!SR) return;
-    speechSynthesis.cancel();
-    this.speaking = false;
-    this._recog = new SR();
-    this._recog.continuous = true;
-    this._recog.interimResults = false;
-    this._recog.lang = this.lang;
-    this._recog.onresult = (e) => {
-      for (let i = e.resultIndex; i < e.results.length; i++) {
-        if (e.results[i].isFinal) {
-          const text = e.results[i][0].transcript.trim();
-          if (text && this.ws?.readyState === 1) this._typeText(text + "\r");
-        }
-      }
-    };
-    this._recog.onerror = () => { this.listening = false; this._notify(); };
-    this._recog.onend = () => {
-      if (this.listening) {
-        try { this._recog.start(); } catch { this.listening = false; this._notify(); }
-      }
-    };
-    this._recog.start();
-    this.listening = true;
-    this._notify();
-  }
-
-  stopListening() {
-    this.listening = false;
-    this._recog?.stop();
-    this._recog = null;
-    this._notify();
-  }
-
   cycleLang() {
     this.langIdx = (this.langIdx + 1) % LANGS.length;
-    if (this._recog && this.listening) {
+    if (this.listening) {
-      this.stopListening();
+      const wasContinuous = this._recog?.continuous;
-      this.startListening();
+      this._stopSTT();
+      this._startSTT(!!wasContinuous);
     }
     this._notify();
   }
 
   destroy() {
     clearTimeout(this._debounce);
+    clearTimeout(this._silenceTimer);
     this._onWriteDispose?.dispose();
-    this.stopListening();
+    this._stopSTT();
     speechSynthesis.cancel();
     this.term = null;
     this.ws = null;

dashboard/frontend/src/routes/Terminal.svelte

@@ -144,33 +144,15 @@
     </div>
     {#if activeTab && tabs.find(t => t.id === activeTab)?.connected && activeVoice()}
       {@const v = activeVoice()}
-      <div class="ml-auto flex items-center gap-1">
+      <div class="ml-auto">
         <button
-          class="px-2 py-1 rounded text-sm transition-colors {v.ttsEnabled ? (v.speaking ? 'text-indigo-400' : 'text-emerald-400') : 'text-surface-500 hover:text-surface-300'}"
+          class="px-2 py-1 rounded text-sm transition-colors {v.voiceMode ? 'text-rose-400' : v.sttAvailable ? 'text-surface-500 hover:text-surface-300' : 'text-surface-700 cursor-not-allowed'}"
-          title={v.ttsEnabled ? 'Disable TTS' : 'Enable TTS'}
+          title={!v.sttAvailable ? 'STT not supported' : v.voiceMode ? 'Voice mode on' : 'Enter voice mode'}
-          onclick={() => { v.toggleTts(); voiceTick++; }}
-        >
-          {#if !v.ttsEnabled}
-            <svg class="w-4 h-4" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2"><path d="M11 5L6 9H2v6h4l5 4V5z"/><line x1="23" y1="9" x2="17" y2="15"/><line x1="17" y1="9" x2="23" y2="15"/></svg>
-          {:else if v.speaking}
-            <svg class="w-4 h-4 animate-pulse" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2"><path d="M11 5L6 9H2v6h4l5 4V5z"/><path d="M15.54 8.46a5 5 0 010 7.07"/><path d="M19.07 4.93a10 10 0 010 14.14"/></svg>
-          {:else}
-            <svg class="w-4 h-4" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2"><path d="M11 5L6 9H2v6h4l5 4V5z"/><path d="M15.54 8.46a5 5 0 010 7.07"/></svg>
-          {/if}
-        </button>
-        <button
-          class="px-2 py-1 rounded text-sm transition-colors {v.listening ? 'text-rose-400 animate-pulse' : v.sttAvailable ? 'text-surface-500 hover:text-surface-300' : 'text-surface-700 cursor-not-allowed'}"
-          title={!v.sttAvailable ? 'STT not supported in this browser' : v.listening ? 'Stop mic' : 'Start mic'}
           disabled={!v.sttAvailable}
-          onclick={() => { v.toggleMic(); voiceTick++; }}
+          onclick={() => { if (!v.voiceMode) v.enterVoiceMode(); else v.exitVoiceMode(); voiceTick++; }}
         >
           <svg class="w-4 h-4" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2"><path d="M12 1a3 3 0 00-3 3v8a3 3 0 006 0V4a3 3 0 00-3-3z"/><path d="M19 10v2a7 7 0 01-14 0v-2"/><line x1="12" y1="19" x2="12" y2="23"/><line x1="8" y1="23" x2="16" y2="23"/></svg>
         </button>
-        <button
-          class="px-1.5 py-0.5 rounded text-xs font-medium transition-colors text-surface-400 hover:text-surface-200 hover:bg-surface-800"
-          title="Switch language"
-          onclick={() => { v.cycleLang(); voiceTick++; }}
-        >{v.langLabel}</button>
       </div>
     {/if}
   </div>
@@ -185,9 +167,78 @@
     <div class={activeTab === tab.id ? '' : 'hidden'}>
       <div
         use:initTerminal={tab}
-        class="h-[calc(100vh-10rem)] rounded-xl overflow-hidden shadow-lg border border-surface-700"
+        class="rounded-xl overflow-hidden shadow-lg border border-surface-700 transition-all duration-300"
-        style="background: #0f172a;"
+        style="background: #0f172a; height: calc(100vh - {activeVoice()?.voiceMode ? '25rem' : '10rem'});"
       ></div>
     </div>
   {/each}
+
+  {#if activeVoice()?.voiceMode}
+    {@const v = activeVoice()}
+    <div class="mt-2 rounded-xl bg-surface-800/95 border border-surface-700 shadow-xl overflow-hidden animate-slide-up" style="height: 180px;">
+      <!-- Header -->
+      <div class="flex items-center justify-between px-4 py-2 border-b border-surface-700/50">
+        <span class="text-sm font-medium text-surface-300">Voice Mode</span>
+        <div class="flex items-center gap-2">
+          <button
+            class="px-1.5 py-0.5 rounded text-xs font-medium text-surface-400 hover:text-surface-200 hover:bg-surface-700"
+            onclick={() => { v.cycleLang(); voiceTick++; }}
+          >{v.langLabel}</button>
+          <button
+            class="px-1.5 py-0.5 rounded text-xs font-medium text-surface-400 hover:text-surface-200 hover:bg-surface-700"
+            onclick={() => { v.cycleTalkMode(); voiceTick++; }}
+          >{v.modeLabel}</button>
+          <button
+            class="px-1.5 py-0.5 rounded text-xs text-surface-500 hover:text-rose-400"
+            onclick={() => { v.exitVoiceMode(); voiceTick++; }}
+          >&times;</button>
+        </div>
+      </div>
+      <!-- Transcript -->
+      <div class="px-4 py-2 h-[60px] overflow-y-auto">
+        {#if v.transcript}
+          <p class="text-sm text-surface-400 italic">"{v.transcript}"</p>
+        {:else}
+          <p class="text-sm text-surface-600 italic">{v.speaking ? 'Speaking...' : 'Waiting for speech...'}</p>
+        {/if}
+      </div>
+      <!-- Mic area -->
+      <div class="flex flex-col items-center gap-1 py-2">
+        {#if v.talkMode === "hold"}
+          <button
+            aria-label="Hold to talk"
+            class="w-14 h-14 rounded-full flex items-center justify-center transition-colors {v.listening ? 'bg-rose-500/30 text-rose-400' : 'bg-surface-700 text-surface-300 hover:bg-surface-600'}"
+            onpointerdown={() => { v.startHold(); voiceTick++; }}
+            onpointerup={() => { v.releaseHold(); voiceTick++; }}
+            onpointerleave={() => { if (v.listening) { v.releaseHold(); voiceTick++; } }}
+          >
+            <svg class="w-6 h-6" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2"><path d="M12 1a3 3 0 00-3 3v8a3 3 0 006 0V4a3 3 0 00-3-3z"/><path d="M19 10v2a7 7 0 01-14 0v-2"/><line x1="12" y1="19" x2="12" y2="23"/><line x1="8" y1="23" x2="16" y2="23"/></svg>
+          </button>
+          <span class="text-xs text-surface-500">{v.listening ? 'Release to send' : 'Hold to talk'}</span>
+        {:else if v.talkMode === "tap"}
+          <button
+            aria-label="Tap to talk"
+            class="w-14 h-14 rounded-full flex items-center justify-center transition-colors {v.listening ? 'bg-rose-500/30 text-rose-400' : 'bg-surface-700 text-surface-300 hover:bg-surface-600'}"
+            onclick={() => { v.tapMic(); voiceTick++; }}
+          >
+            <svg class="w-6 h-6" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2"><path d="M12 1a3 3 0 00-3 3v8a3 3 0 006 0V4a3 3 0 00-3-3z"/><path d="M19 10v2a7 7 0 01-14 0v-2"/><line x1="12" y1="19" x2="12" y2="23"/><line x1="8" y1="23" x2="16" y2="23"/></svg>
+          </button>
+          <span class="text-xs text-surface-500">{v.listening ? 'Tap to send' : 'Tap to talk'}</span>
+        {:else}
+          <div class="w-14 h-14 rounded-full flex items-center justify-center {v.listening ? 'bg-emerald-500/20 text-emerald-400 animate-pulse' : 'bg-surface-700 text-surface-500'}">
+            <svg class="w-6 h-6" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2"><path d="M12 1a3 3 0 00-3 3v8a3 3 0 006 0V4a3 3 0 00-3-3z"/><path d="M19 10v2a7 7 0 01-14 0v-2"/><line x1="12" y1="19" x2="12" y2="23"/><line x1="8" y1="23" x2="16" y2="23"/></svg>
+          </div>
+          <span class="text-xs text-surface-500">{v.listening ? 'Listening...' : v.speaking ? 'Speaking...' : 'Ready'}</span>
+        {/if}
+      </div>
+    </div>
+  {/if}
 </div>
+
+<style>
+  @keyframes slide-up {
+    from { transform: translateY(100%); opacity: 0; }
+    to { transform: translateY(0); opacity: 1; }
+  }
+  .animate-slide-up { animation: slide-up 0.25s ease-out; }
+</style>