@hanzo/rooms 0.1.3 → 0.1.5

package/dist/speech.js

@@ -1,11 +1,15 @@
 'use client';
 import { Fragment as _Fragment, jsx as _jsx, jsxs as _jsxs } from "react/jsx-runtime";
-// Agent Speech & Voice Synthesis
+// Agent Speech
 //
 // Every agent in the Hanzo ecosystem has an assigned voice (`voiceOf(name)` from
-// `cast.tsx`). This module provides speech synthesis for agents across Chat,
-// Talk channels, and Meet video calls:
-// - Direct browser SpeechSynthesis with agent-tailored pitch, rate, and voice matching
+// `cast.tsx`), and the platform reads it: `/v1/audio/speech` through
+// @hanzo/voice's `speech()` transport and `mouth()`, in that voice. The
+// browser's own speechSynthesis reads only when the platform refuses, in a
+// voice of the same register and accent, and the refusal is reported rather
+// than passed off as the platform's voice.
+//
+// - The Listen button on a message, the Meet speaker test and its briefing
 // - Audio wave visualization during playback
 // - Clean cancellation and state management
 // - Text normalization (strips code fences, markdown syntax so speech sounds natural)
@@ -13,7 +17,10 @@ import { useEffect, useRef, useState, useCallback } from 'react';
 import { Volume2, Square } from 'lucide-react';
 import { View, XStack } from '@hanzo/gui';
 import { Button } from '@hanzo/ui';
+import { useIamToken } from '@hanzo/iam/react';
+import { mouth, speech, REFUSED } from '@hanzo/voice';
 import { GOOD } from './lib/mix.js';
+import { base } from './lib/ai.js';
 import { voiceOf, voiceProfileOf } from './cast.js';
 /** Strips code blocks, links, headers, and markdown markup for natural audio speech. */
 export function cleanForSpeech(raw) {
@@ -41,24 +48,41 @@ export function cleanForSpeech(raw) {
         .replace(/\s+/g, ' ')
         .trim();
 }
-/** Currently active speech synthesis utterance tracker. */
-let activeUtterance = null;
+/** Browser voice names by register, for the stand-in. Word-bounded: "Female" holds "male". */
+const REGISTER = {
+    female: /\b(female|samantha|victoria|karen|zira|serena|moira|tessa)\b/i,
+    male: /\b(male|daniel|alex|fred|david|arthur|oliver)\b/i,
+};
+/**
+ * The browser voice that reads when the platform refuses: English, the cast
+ * voice's accent where this browser has it, its register where a name says so.
+ * Read per sentence, because the browser loads its voice list asynchronously.
+ */
+export function standIn({ gender, accent }, voices) {
+    const english = voices.filter((v) => v.lang.toLowerCase().startsWith('en'));
+    const local = english.filter((v) => v.lang.toLowerCase().replace('_', '-') === (accent === 'british' ? 'en-gb' : 'en-us'));
+    const pick = (list) => list.find((v) => REGISTER[gender].test(v.name));
+    return (pick(local) ?? pick(english) ?? local[0] ?? english[0] ?? voices[0])?.name;
+}
+/** The reply being read aloud on this page. One at a time: a new one stops the last. */
+let active = null;
 const activeListeners = new Set();
+const notify = () => {
+    for (const listener of activeListeners)
+        listener();
+};
 export function stopAgentSpeech() {
-    if (typeof window !== 'undefined' && 'speechSynthesis' in window) {
-        try {
-            window.speechSynthesis.cancel();
-        }
-        catch { }
-    }
-    activeUtterance = null;
-    for (const notify of activeListeners) {
-        notify();
-    }
+    const lips = active;
+    active = null;
+    lips?.hush();
+    notify();
 }
 /**
- * Speak text in the assigned voice of an agent.
+ * Speak text in the assigned voice of an agent, through the platform.
  * Returns a cancel function.
+ *
+ * Call it from the click that asks for it: the player is made inside that
+ * gesture, which is what lets Safari play the reply when it arrives.
  */
 export function speakAgent(text, options = {}) {
     if (typeof window === 'undefined')
@@ -67,77 +91,38 @@ export function speakAgent(text, options = {}) {
     const clean = cleanForSpeech(text);
     if (!clean)
         return () => { };
-    if (!('speechSynthesis' in window)) {
-        options.onError?.(new Error('SpeechSynthesis not supported in this browser'));
-        return () => { };
-    }
     const profile = voiceProfileOf(options.agent);
-    const utterance = new SpeechSynthesisUtterance(clean);
-    utterance.pitch = profile.pitch;
-    utterance.rate = profile.rate;
-    // Attempt to select best matching system voice based on gender / voice tone
-    try {
-        const voices = window.speechSynthesis.getVoices();
-        if (voices.length > 0) {
-            const isFemale = profile.gender === 'female';
-            const candidate = voices.find((v) => {
-                const name = v.name.toLowerCase();
-                const lang = v.lang.toLowerCase();
-                if (!lang.startsWith('en'))
-                    return false;
-                if (isFemale) {
-                    return name.includes('female') || name.includes('samantha') || name.includes('victoria') || name.includes('karen') || name.includes('zira');
-                }
-                else {
-                    return name.includes('male') || name.includes('daniel') || name.includes('alex') || name.includes('fred') || name.includes('david');
-                }
-            }) || voices.find((v) => v.lang.toLowerCase().startsWith('en')) || voices[0];
-            if (candidate) {
-                utterance.voice = candidate;
-            }
-        }
-    }
-    catch { }
-    utterance.onstart = () => {
-        activeUtterance = utterance;
-        options.onStart?.();
-    };
-    utterance.onend = () => {
-        if (activeUtterance === utterance) {
-            activeUtterance = null;
-        }
+    const lips = mouth({
+        speech: speech({ ...base(), ...(options.token ? { token: options.token } : {}) }),
+        voice: () => standIn(profile, window.speechSynthesis?.getVoices() ?? []),
+        refused: options.onRefusal,
+    });
+    active = lips;
+    options.onStart?.();
+    lips.say(clean, profile.voice).then(() => {
+        if (active === lips)
+            active = null;
         options.onEnd?.();
-        for (const notify of activeListeners) {
-            notify();
-        }
-    };
-    utterance.onerror = (e) => {
-        if (activeUtterance === utterance) {
-            activeUtterance = null;
-        }
-        options.onError?.(e);
-        for (const notify of activeListeners) {
-            notify();
-        }
-    };
-    try {
-        window.speechSynthesis.speak(utterance);
-    }
-    catch (err) {
+        notify();
+    }, (err) => {
+        if (active === lips)
+            active = null;
         options.onError?.(err);
-    }
+        notify();
+    });
     return () => {
-        if (activeUtterance === utterance) {
+        if (active === lips)
             stopAgentSpeech();
-        }
     };
 }
 /**
  * Hook for managing agent speech playback in UI components.
  */
 export function useAgentSpeech(defaultAgent) {
+    const { token } = useIamToken();
     const [speaking, setSpeaking] = useState(false);
     const [currentText, setCurrentText] = useState(null);
+    const [refusal, setRefusal] = useState(null);
     const cancelRef = useRef(null);
     useEffect(() => {
         const listener = () => {
@@ -160,8 +145,10 @@ export function useAgentSpeech(defaultAgent) {
         stop();
         setCurrentText(text);
         setSpeaking(true);
+        setRefusal(null);
         cancelRef.current = speakAgent(text, {
             agent: agentName || defaultAgent,
+            token,
             onStart: () => setSpeaking(true),
             onEnd: () => {
                 setSpeaking(false);
@@ -171,8 +158,9 @@ export function useAgentSpeech(defaultAgent) {
                 setSpeaking(false);
                 setCurrentText(null);
             },
+            onRefusal: setRefusal,
         });
-    }, [defaultAgent, stop]);
+    }, [defaultAgent, stop, token]);
     const toggle = useCallback((text, agentName) => {
         if (speaking && currentText === text) {
             stop();
@@ -181,7 +169,7 @@ export function useAgentSpeech(defaultAgent) {
             speak(text, agentName);
         }
     }, [speaking, currentText, speak, stop]);
-    return { speaking, currentText, speak, stop, toggle };
+    return { speaking, currentText, refusal, speak, stop, toggle };
 }
 /** The five bars' resting heights, as a share of the 12px the wave stands in. */
 const BARS = [0.4, 0.9, 0.6, 1, 0.5];
@@ -219,11 +207,14 @@ export function AudioWave({ active = true, color = 'var(--state-success)' }) {
  * Allows one-click audio playback in the agent's voice.
  */
 export function SpeakButton({ text, agentName, compact = false, }) {
-    const { speaking, toggle } = useAgentSpeech(agentName);
+    const { speaking, refusal, toggle } = useAgentSpeech(agentName);
     const voice = voiceOf(agentName);
+    // A browser voice standing in for a refused platform sounds like success;
+    // the label is the one place that says otherwise.
+    const stood = refusal ? ` ${REFUSED[refusal.covered ? 'covered' : 'lost']}` : '';
     // An @hanzo/ui Button: ghost at rest, the quiet pressed ground while it
     // speaks. The word for speaking is the one status here, so it alone takes
     // the success ink (GOOD holds 4.5:1 on either theme); the icons follow the
     // label through currentColor.
-    return (_jsx(Button, { variant: speaking ? 'secondary' : 'ghost', size: compact ? 'icon-sm' : 'sm', color: speaking ? GOOD : '$soft', onClick: () => toggle(text, agentName), "aria-label": speaking ? 'Stop speaking' : `Listen to ${agentName || 'agent'} (${voice})`, "aria-pressed": speaking, title: speaking ? 'Stop speaking' : `Listen in ${agentName || 'agent'}'s voice (${voice})`, children: speaking ? (_jsxs(_Fragment, { children: [_jsx(AudioWave, { active: true, color: GOOD }), !compact && 'Speaking…', _jsx(Square, { size: 10, fill: "currentColor" })] })) : (_jsxs(_Fragment, { children: [_jsx(Volume2, { size: 13 }), !compact && 'Listen'] })) }));
+    return (_jsx(Button, { variant: speaking ? 'secondary' : 'ghost', size: compact ? 'icon-sm' : 'sm', color: speaking ? GOOD : '$soft', onClick: () => toggle(text, agentName), "aria-label": (speaking ? 'Stop speaking' : `Listen to ${agentName || 'agent'} (${voice})`) + stood, "aria-pressed": speaking, "data-refusal": refusal ? refusal.service : undefined, title: (speaking ? 'Stop speaking' : `Listen in ${agentName || 'agent'}'s voice (${voice})`) + stood, children: speaking ? (_jsxs(_Fragment, { children: [_jsx(AudioWave, { active: true, color: GOOD }), !compact && 'Speaking…', _jsx(Square, { size: 10, fill: "currentColor" })] })) : (_jsxs(_Fragment, { children: [_jsx(Volume2, { size: 13 }), !compact && 'Listen'] })) }));
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@hanzo/rooms",
-  "version": "0.1.3",
+  "version": "0.1.5",
   "description": "The workspace rooms — Home, Chat, Inbox, Contacts, Meet, Cal, Drive, Board, Work, Bots — and the frame that holds them, once, for every Hanzo surface.",
   "type": "module",
   "main": "./dist/index.js",
@@ -56,7 +56,7 @@
     "@hanzo/plans": ">=1.8.9",
     "@hanzo/ui": ">=8.27.51",
     "@hanzo/usage": ">=0.1.17",
-    "@hanzo/voice": ">=0.1.8",
+    "@hanzo/voice": ">=0.1.10",
     "@hanzogui/shell": "^8.3.20",
     "lucide-react": ">=1.42.0",
     "react": ">=19.2.8",
@@ -67,6 +67,7 @@
     "@hanzo/appearance": "0.2.6",
     "@hanzo/build": "0.2.34",
     "@hanzo/composer": "0.2.0",
+    "@hanzo/design": "0.5.33",
     "@hanzo/event": "0.3.56",
     "@hanzo/gui": "8.3.5",
     "@hanzo/iam": "0.21.28",
@@ -74,9 +75,10 @@
     "@hanzo/plans": "1.8.9",
     "@hanzo/ui": "8.27.51",
     "@hanzo/usage": "0.1.17",
-    "@hanzo/voice": "0.1.8",
+    "@hanzo/voice": "0.1.10",
     "@hanzogui/config": "8.3.5",
     "@hanzogui/core": "8.3.5",
+    "@hanzogui/lucide-icons-2": "8.3.5",
     "@hanzogui/shell": "8.3.20",
     "@hanzogui/web": "8.3.5",
     "@types/node": "26.5.0",
@@ -87,9 +89,7 @@
     "react-dom": "19.2.8",
     "react-native-web": "0.21.2",
     "typescript": "7.0.2",
-    "vitest": "4.1.11",
-    "@hanzo/design": "0.5.33",
-    "@hanzogui/lucide-icons-2": "8.3.5"
+    "vitest": "4.1.11"
   },
   "repository": {
     "type": "git",