@burdenoff/fe-libs 2026.1029.6 → 2026.1029.7

package/dist/shared/assistant/ui/AssistantSpeechProvider.js

@@ -3,27 +3,29 @@ import { useAssistantTr as t } from "../i18n.js";
 import { useAssistantStore as n } from "../store.js";
 import { isVoiceTurn as r } from "../voice/turnVisibility.js";
 import { HOSTED_VOICES as i } from "../speech/hostedVoiceCatalogue.js";
-import { VOICE_LIST_TIMEOUT_MS as a, languageDisplayName as o, selectVoiceForLocale as s, speechGate as ee, speechLangForLocale as c, voiceAvailability as te, voicesForLocale as l } from "../voice/speechVoices.js";
-import { getSpeechEngine as u, subscribeSpeechEngine as ne } from "../speech/createSpeechEngine.js";
-import { speechFeed as d } from "../speech/speechFeed.js";
-import { hasPendingReplyRequests as f, releaseReplyRequestsOutside as p, replyWantsAvatar as re, settleReplyRequest as m, subscribeReplyRequests as h } from "../speech/replySpeechRequests.js";
+import { VOICE_LIST_TIMEOUT_MS as a, languageDisplayName as o, selectVoiceForLocale as s, speechGate as c, speechLangForLocale as l, voiceAvailability as ee, voicesForLocale as u } from "../voice/speechVoices.js";
+import { getSpeechEngine as d, subscribeSpeechEngine as te } from "../speech/createSpeechEngine.js";
+import { speechFeed as f } from "../speech/speechFeed.js";
+import { hasPendingReplyRequests as ne, releaseReplyRequestsOutside as p, replyWantsAvatar as re, settleReplyRequest as m, subscribeReplyRequests as h } from "../speech/replySpeechRequests.js";
 import { isTurnSuperseded as g } from "../voice/supersededTurns.js";
 import { chunkForSpeech as ie, toSpeakableText as _ } from "../voice/speakableText.js";
 import { voicePreviewText as v } from "../voice/voicePreview.js";
-import { findAssistantLanguage as ae } from "../voice/languages.js";
-import { resolveLanguagePair as y } from "../voice/languagePair.js";
-import { speechLabelsFor as oe } from "../voice/speechLabels.js";
-import { resolveSpeechLocale as se } from "../voice/replyLanguage.js";
-import { IDLE_PLAYBACK as ce, isSpeakingMessage as le, publishedChunks as ue, speechPlaybackReducer as de } from "../voice/speechPlayback.js";
-import { isTurnInFlight as fe } from "../turnState.js";
-import { parseToolActivity as pe } from "./AssistantToolActivity.js";
+import { findAssistantLanguage as y } from "../voice/languages.js";
+import { resolveLanguagePair as b } from "../voice/languagePair.js";
+import { speechLabelsFor as x } from "../voice/speechLabels.js";
+import { resolveSpeechLocale as S } from "../voice/replyLanguage.js";
+import { IDLE_PLAYBACK as ae, isSpeakingMessage as C, publishedChunks as oe, speechPlaybackReducer as se } from "../voice/speechPlayback.js";
+import { isTurnInFlight as ce } from "../turnState.js";
+import { parseToolActivity as le } from "./AssistantToolActivity.js";
 import "../speech/index.js";
-import { beginListenTurn as me, finishListenTurn as b, markTurn as he } from "../turnTimeline.js";
-import { registerTurnTimingPageExit as ge } from "../turnTimingReporter.js";
-import { createContext as _e, useCallback as x, useContext as ve, useEffect as S, useMemo as C, useReducer as ye, useRef as w, useState as T, useSyncExternalStore as be } from "react";
-import { jsx as xe } from "react/jsx-runtime";
+import { createPhrasePlayback as ue } from "../voice/phrasePlayback.js";
+import { phraseSpeechEnabled as de } from "../voice/phraseSpeechConfig.js";
+import { beginListenTurn as fe, describeTurn as pe, finishListenTurn as w, markTurn as me } from "../turnTimeline.js";
+import { registerTurnTimingPageExit as he } from "../turnTimingReporter.js";
+import { createContext as ge, useCallback as T, useContext as _e, useEffect as E, useMemo as D, useReducer as ve, useRef as O, useState as k, useSyncExternalStore as ye } from "react";
+import { jsx as be } from "react/jsx-runtime";
 //#region src/shared/assistant/ui/AssistantSpeechProvider.tsx
-var Se = {
+var xe = {
 	codeBlock: "",
 	image: "",
 	link: "",
@@ -32,30 +34,30 @@ var Se = {
 	emailAt: "at",
 	emailDot: "dot",
 	nestedItem: ""
-}, Ce = 4e3, E = () => {}, we = {
-	subscribe: () => E,
+}, Se = 4e3, A = () => {}, Ce = {
+	subscribe: () => A,
 	getSnapshot: () => !1
-}, Te = {
+}, we = {
 	subscribe: h,
-	getSnapshot: f
+	getSnapshot: ne
 };
-function Ee(e, t) {
+function Te(e, t) {
 	return re(t) ? {
 		...e,
 		autoSpeakEnabled: !0
 	} : e;
 }
-function De(e, t) {
+function Ee(e, t) {
 	let n = e.activeConversationId;
 	if (!n || n !== t.activeConversationId) return null;
 	let r = e.processingByConversationId[n], i = t.processingByConversationId[n];
-	if (r?.status !== "complete" || !fe(i?.status)) return null;
+	if (r?.status !== "complete" || !ce(i?.status)) return null;
 	let a = r.messageId ?? i?.messageId ?? null;
 	if (!a) return null;
 	let o = e.conversations.find((e) => e.id === n)?.messages.find((e) => e.id === a);
 	return o && o.role === "assistant" ? o : null;
 }
-var Oe = _e({
+var De = ge({
 	isSupported: !1,
 	availability: "unsupported",
 	isBlocked: !0,
@@ -80,404 +82,462 @@ var Oe = _e({
 	speakFinishedReply: () => {},
 	previewVoice: () => {}
 });
-function D() {
-	return ve(Oe);
+function j() {
+	return _e(De);
 }
-var O = ({ children: f, visible: h = !0, engine: _e }) => {
-	let ve = be(ne, u, u), E = _e ?? ve, D = t(), { locale: O } = e(), [k, A] = ye(de, ce), [j, M] = T(() => E.listVoices()), [N, ke] = T(!1), [Ae, P] = T(() => /* @__PURE__ */ new Map()), [je, F] = T(() => /* @__PURE__ */ new Set()), [Me, Ne] = T(null), I = w(0), L = w(null), R = w(null), z = w(void 0), B = w(null), V = w(null);
-	S(() => () => {
-		V.current !== null && clearTimeout(V.current);
+var M = ({ children: ne, visible: h = !0, engine: ge }) => {
+	let _e = ye(te, d, d), A = ge ?? _e, j = t(), { locale: M } = e(), [N, P] = ve(se, ae), [F, Oe] = k(() => A.listVoices()), [ke, Ae] = k(!1), [je, I] = k(() => /* @__PURE__ */ new Map()), [Me, Ne] = k(() => /* @__PURE__ */ new Set()), [Pe, Fe] = k(null), L = O(0), Ie = O(null), Le = O(null), R = O(void 0), z = O(null), B = O(null), V = O(/* @__PURE__ */ new Set()), H = O(null);
+	E(() => () => {
+		H.current !== null && clearTimeout(H.current);
 	}, []);
-	let [Pe, Fe] = T(null), H = w(null), Ie = w(() => {}), U = x(() => {
-		H.current?.(), H.current = null;
-	}, []), W = n((e) => e.autoSpeakEnabled), Le = n((e) => e.setAutoSpeak), [G, Re] = T(!1), ze = w(!1), Be = x((e) => {
-		ze.current = e;
-	}, []), Ve = n((e) => e.activeConversationId), K = C(() => E.isSupported(), [E]), He = E.wouldSpeakOnAvatar === void 0 ? we : Te, Ue = be(He.subscribe, He.getSnapshot, He.getSnapshot);
-	S(() => {
-		if (!K) return;
+	let [Re, ze] = k(null), U = O(null), Be = O(() => {}), W = T(() => {
+		U.current?.(), U.current = null;
+	}, []), G = n((e) => e.autoSpeakEnabled), Ve = n((e) => e.setAutoSpeak), [K, He] = k(!1), Ue = O(!1), We = T((e) => {
+		Ue.current = e;
+	}, []), Ge = n((e) => e.activeConversationId), q = D(() => A.isSupported(), [A]), Ke = A.wouldSpeakOnAvatar === void 0 ? Ce : we, qe = ye(Ke.subscribe, Ke.getSnapshot, Ke.getSnapshot);
+	E(() => {
+		if (!q) return;
 		let e = () => {
-			M(E.listVoices()), P((e) => e.size === 0 ? e : /* @__PURE__ */ new Map());
+			Oe(A.listVoices()), I((e) => e.size === 0 ? e : /* @__PURE__ */ new Map());
 		};
-		return e(), E.subscribeVoices(e);
-	}, [E, K]);
-	let We = n((e) => e.assistantLanguage), Ge = n((e) => e.replyLanguage), q = n((e) => e.preferredVoices), Ke = n((e) => e.setPreferredVoice), J = C(() => y(We, Ge, O).reply, [
-		We,
-		Ge,
-		O
-	]), Y = J.code, qe = x((e) => ae(e)?.tag ?? c(e), []), Je = x((e) => ae(e)?.rate ?? .9, []), Ye = C(() => te(K, j, J.tag, N), [
-		K,
-		j,
-		J,
-		N
-	]), Xe = C(() => ({
-		codeBlock: D("speech.codeBlock", "Code block"),
-		image: D("speech.image", "image"),
-		link: D("speech.link", "link"),
-		largeTable: D("speech.largeTable", "Table with {{rows}} rows and {{columns}} columns"),
-		truncated: D("speech.truncated", "The rest of this message is not read aloud"),
-		nestedItem: D("speech.nestedItem", "Sub-item {{ordinal}}"),
-		emailAt: D("speech.emailAt", "at"),
-		emailDot: D("speech.emailDot", "dot")
-	}), [D]), X = x(() => {
-		U(), R.current = null, b(z.current, "stopped"), z.current = void 0, E.cancel(), m(L.current, "stopped"), A({ type: "stop" });
-	}, [U, E]), Ze = x((e) => {
-		let t = E.listVoices().find((t) => t.name === e) ?? null;
+		return e(), A.subscribeVoices(e);
+	}, [A, q]);
+	let Je = n((e) => e.assistantLanguage), Ye = n((e) => e.replyLanguage), J = n((e) => e.preferredVoices), Xe = n((e) => e.setPreferredVoice), Y = D(() => b(Je, Ye, M).reply, [
+		Je,
+		Ye,
+		M
+	]), X = Y.code, Ze = T((e) => y(e)?.tag ?? l(e), []), Qe = T((e) => y(e)?.rate ?? .9, []), $e = D(() => ee(q, F, Y.tag, ke), [
+		q,
+		F,
+		Y,
+		ke
+	]), et = D(() => ({
+		codeBlock: j("speech.codeBlock", "Code block"),
+		image: j("speech.image", "image"),
+		link: j("speech.link", "link"),
+		largeTable: j("speech.largeTable", "Table with {{rows}} rows and {{columns}} columns"),
+		truncated: j("speech.truncated", "The rest of this message is not read aloud"),
+		nestedItem: j("speech.nestedItem", "Sub-item {{ordinal}}"),
+		emailAt: j("speech.emailAt", "at"),
+		emailDot: j("speech.emailDot", "dot")
+	}), [j]), Z = T(() => {
+		let e = B.current;
+		if (e) try {
+			e.playback.cancel();
+		} finally {
+			e.notifyEnded();
+		}
+	}, []), Q = T(() => {
+		W(), Le.current = null, w(R.current, "stopped"), R.current = void 0, Z(), A.cancel(), m(Ie.current, "stopped"), P({ type: "stop" });
+	}, [
+		W,
+		Z,
+		A
+	]), tt = T((e) => {
+		let t = A.listVoices().find((t) => t.name === e) ?? null;
 		if (t) {
-			X();
+			Q();
 			try {
-				E.speak([v(Y)], {
+				A.speak([v(X)], {
 					voice: t,
-					lang: qe(Y),
+					lang: Ze(X),
 					fallback: !1
 				});
 			} catch {}
 		}
 	}, [
-		E,
-		Y,
+		A,
 		X,
-		qe
-	]), Qe = x((e) => Ae.get(e) ?? null, [Ae]), $e = x((e) => je.has(e) ? D("speech.deviceFallback", "Device voice used: the selected voice was unavailable.") : null, [je, D]), et = x((e, t, n, r, i) => {
-		U();
-		let o = !1, s = () => {}, ee = (a) => {
+		Q,
+		Ze
+	]), nt = T((e) => je.get(e) ?? null, [je]), rt = T((e) => Me.has(e) ? j("speech.deviceFallback", "Device voice used: the selected voice was unavailable.") : null, [Me, j]), it = T((e, t, n, r, i) => {
+		W();
+		let o = !1, s = () => {}, c = (a) => {
 			if (!o) {
-				if (o = !0, H.current = null, clearTimeout(c), s(), a.length === 0) {
-					ke(!0), m(e, "declined");
+				if (o = !0, U.current = null, clearTimeout(l), s(), a.length === 0) {
+					Ae(!0), m(e, "declined");
 					return;
 				}
-				M(a), Ie.current(e, t, a, n, r, i);
+				Oe(a), Be.current(e, t, a, n, r, i);
 			}
-		}, c = setTimeout(() => ee(E.listVoices()), a);
-		s = E.subscribeVoices(() => ee(E.listVoices())), H.current = () => {
-			o || (o = !0, clearTimeout(c), s(), H.current = null, m(e, "stopped"));
+		}, l = setTimeout(() => c(A.listVoices()), a);
+		s = A.subscribeVoices(() => c(A.listVoices())), U.current = () => {
+			o || (o = !0, clearTimeout(l), s(), U.current = null, m(e, "stopped"));
 		};
-	}, [U, E]), tt = x((e, t, n, r, a = !1, c) => {
-		let l = n;
-		if (l.length === 0) {
-			let e = E.listVoices();
-			e.length > 0 && (l = e, M(e));
+	}, [W, A]), at = T((e, t, n, r, a = !1, l, u = !1) => {
+		let d = n;
+		if (d.length === 0) {
+			let e = A.listVoices();
+			e.length > 0 && (d = e, Oe(e));
 		}
-		let u = qe(r), ne = ee(te(!0, l, r, N));
+		let te = Ze(r), ne = c(ee(!0, d, r, ke));
 		if (ne === "block") {
-			let t = D("speech.noVoiceForReply", "This reply is in {{language}}, and no {{language}} voice is installed on this device.", { language: o(u, O) });
-			P((n) => new Map(n).set(e, t)), m(e, "declined");
+			let t = j("speech.noVoiceForReply", "This reply is in {{language}}, and no {{language}} voice is installed on this device.", { language: o(te, M) });
+			I((n) => new Map(n).set(e, t)), m(e, "declined");
 			return;
 		}
+		if (u && (ne !== "speak" || !A.open)) return;
 		if (ne === "wait-for-voices") {
-			et(e, t, r, a, c);
+			it(e, t, r, a, l);
 			return;
 		}
-		let d = {
-			voice: s(l, u, q[r]),
-			lang: u,
-			rate: Je(r)
-		}, f = rt.current;
-		if (a && !f.autoSpeakEnabled && !f.sessionSpeaking && re(e)) {
-			if (E.wouldSpeakOnAvatar?.(d) !== !0) {
+		let p = {
+			voice: s(d, te, J[r]),
+			lang: te,
+			rate: Qe(r)
+		}, h = ct.current;
+		if (a && !h.autoSpeakEnabled && !h.sessionSpeaking && re(e)) {
+			if (A.wouldSpeakOnAvatar?.(p) !== !0) {
 				m(e, "declined");
 				return;
 			}
-			d.avatarOnly = !0;
+			p.avatarOnly = !0;
 		}
-		P((t) => {
+		I((t) => {
 			if (!t.has(e)) return t;
 			let n = new Map(t);
 			return n.delete(e), n;
 		});
-		let p = q[r], h = typeof p == "string" && i.some((e) => e.name === p) && !l.some((e) => e.name === p);
-		F((t) => {
-			if (t.has(e) === h) return t;
+		let g = J[r], _ = typeof g == "string" && i.some((e) => e.name === g) && !d.some((e) => e.name === g);
+		Ne((t) => {
+			if (t.has(e) === _) return t;
 			let n = new Set(t);
-			return h ? n.add(e) : n.delete(e), n;
-		}), U();
-		let g = I.current + 1;
-		I.current = g, L.current !== e && m(L.current, "stopped"), L.current = e, m(e, "speaking"), A({
+			return _ ? n.add(e) : n.delete(e), n;
+		}), W(), Z();
+		let v = L.current + 1;
+		L.current = v, Ie.current !== e && m(Ie.current, "stopped"), Ie.current = e, m(e, "speaking"), P({
 			type: "start",
 			messageId: e,
-			requestId: g
-		}), R.current = a ? g : null, a && he(e, "ttsStarted"), b(z.current, "stopped");
-		let _ = a ? void 0 : me(e);
-		z.current = _;
-		let v = c === void 0 ? ie(t) : ie(t, void 0, void 0, { firstChunkMaxChars: c });
-		Fe({
-			requestId: g,
-			chunks: v,
+			requestId: v
+		}), Le.current = a ? v : null, a && me(e, "ttsStarted"), w(R.current, "stopped");
+		let y = a ? void 0 : fe(e);
+		R.current = y;
+		let b = l === void 0 ? ie(t) : ie(t, void 0, void 0, { firstChunkMaxChars: l });
+		ze({
+			requestId: v,
+			chunks: b,
 			index: -1
-		}), V.current !== null && clearTimeout(V.current), V.current = setTimeout(() => {
-			Fe((e) => e?.requestId === g && e.index < 0 ? null : e);
-		}, Ce);
-		let ae = () => {
-			b(_, "failed"), I.current === g && V.current !== null && (clearTimeout(V.current), V.current = null), m(e, "failed"), A({
+		}), H.current !== null && clearTimeout(H.current), H.current = setTimeout(() => {
+			u || ze((e) => e?.requestId === v && e.index < 0 ? null : e);
+		}, Se);
+		let x = !1, S = () => {
+			x || (x = !0, !(!u || L.current !== v) && (B.current = null, f.audio(e, "ended")));
+		}, ae = () => {
+			w(y, "failed"), S(), L.current === v && H.current !== null && (clearTimeout(H.current), H.current = null), m(e, "failed"), P({
 				type: "failed",
-				requestId: g
+				requestId: v
 			});
-		}, y = !1;
+		}, C = !1;
 		try {
-			E.speak(v, d, {
+			let n = {
 				onStart: () => {
-					I.current === g && b(_, "answered"), I.current === g && R.current === g && he(e, "firstAudio");
+					L.current === v && w(y, "answered"), L.current === v && Le.current === v && me(e, "firstAudio");
 				},
-				onChunkStart: (e) => {
-					y = !0, I.current === g && V.current !== null && (clearTimeout(V.current), V.current = null), Fe((t) => t?.requestId === g ? {
-						...t,
-						index: e
-					} : I.current === g ? {
-						requestId: g,
-						chunks: v,
-						index: e
-					} : t);
+				onChunkStart: (t) => {
+					C = !0, u && L.current === v && (f.peek(e)?.final === !1 && pe(e, { spokeNarration: !0 }), f.audio(e, "started")), L.current === v && H.current !== null && (clearTimeout(H.current), H.current = null), ze((e) => e?.requestId === v ? {
+						...e,
+						index: t
+					} : L.current === v ? {
+						requestId: v,
+						chunks: b,
+						index: t
+					} : e);
+				},
+				onPause: () => {
+					u && L.current === v && f.audio(e, "paused");
 				},
 				onDeviceFallback: () => {
-					I.current === g && F((t) => t.has(e) ? t : new Set(t).add(e));
+					L.current === v && Ne((t) => t.has(e) ? t : new Set(t).add(e));
 				},
 				onEnd: () => {
-					b(_, "stopped"), I.current === g && R.current === g && he(e, "audioEnded"), m(e, "spoken"), A({
+					w(y, "stopped"), S(), L.current === v && Le.current === v && me(e, "audioEnded"), m(e, "spoken"), P({
 						type: "finished",
-						requestId: g
+						requestId: v
 					});
 				},
 				onError: (t) => {
-					b(_, t === "cancelled" ? "stopped" : "failed"), !y && I.current === g && F((t) => {
+					w(y, t === "cancelled" ? "stopped" : "failed"), S(), !C && L.current === v && Ne((t) => {
 						if (!t.has(e)) return t;
 						let n = new Set(t);
 						return n.delete(e), n;
-					}), t === "output-not-cleared" && I.current === g && Ne({ messageId: e }), m(e, t === "output-not-cleared" ? "not-cleared" : t === "output-left" ? y ? "cut" : "declined" : t === "cancelled" ? "stopped" : "failed"), A({
+					}), t === "output-not-cleared" && L.current === v && Fe({ messageId: e }), m(e, t === "output-not-cleared" ? "not-cleared" : t === "output-left" ? C ? "cut" : "declined" : t === "cancelled" ? "stopped" : "failed"), P({
 						type: "failed",
-						requestId: g
+						requestId: v
 					});
 				}
-			});
+			};
+			if (u && A.open) {
+				let i = A.open.bind(A);
+				f.audio(e, "queued");
+				let a = ue((e) => i(p, e), n, (e) => {
+					b = [...e], ze((e) => e?.requestId === v ? {
+						...e,
+						chunks: b
+					} : e);
+				}, () => pe(e, { speechDivergence: !0 }));
+				x ? a.cancel() : (B.current = {
+					messageId: e,
+					requestId: v,
+					locale: r,
+					playback: a,
+					notifyEnded: S
+				}, a.append(t));
+			} else A.speak(b, p, n);
 		} catch {
 			ae();
 		}
 	}, [
-		U,
-		E,
-		O,
-		q,
-		Je,
-		qe,
-		D,
-		N,
-		et
+		W,
+		Z,
+		A,
+		M,
+		J,
+		Qe,
+		Ze,
+		j,
+		ke,
+		it
 	]);
-	S(() => {
-		Ie.current = tt;
-	}, [tt]);
-	let Z = x((e, t, n, r) => {
-		if (!K) return;
-		let i = se(_(t, { labels: Se }), n ?? Y), a = _(t, { labels: oe(i, Xe) });
-		a && tt(e, a, j, i, r?.auto === !0, r?.firstChunkMaxChars);
+	E(() => {
+		Be.current = at;
+	}, [at]);
+	let ot = T((e, t, n, r) => {
+		if (!q) return;
+		let i = S(_(t, { labels: xe }), n ?? X), a = _(t, { labels: x(i, et) });
+		a && at(e, a, F, i, r?.auto === !0, r?.firstChunkMaxChars, r?.phraseStream);
 	}, [
-		K,
-		Xe,
-		Y,
-		tt,
-		j
-	]), nt = x((e, t, n) => {
-		if (le(k, e)) {
-			X();
+		q,
+		et,
+		X,
+		at,
+		F
+	]), st = T((e, t, n) => {
+		if (C(N, e)) {
+			Q();
 			return;
 		}
-		Z(e, t, n);
+		ot(e, t, n);
 	}, [
-		k,
-		Z,
-		X
+		N,
+		ot,
+		Q
 	]);
-	S(() => {
-		h || X();
-	}, [h, X]);
-	let rt = w({
-		autoSpeakEnabled: W,
-		sessionSpeaking: G,
+	E(() => {
+		h || Q();
+	}, [h, Q]);
+	let ct = O({
+		autoSpeakEnabled: G,
+		sessionSpeaking: K,
 		visible: h,
-		isSupported: K
+		isSupported: q
 	});
-	S(() => {
-		rt.current = {
-			autoSpeakEnabled: W,
-			sessionSpeaking: G,
+	E(() => {
+		ct.current = {
+			autoSpeakEnabled: G,
+			sessionSpeaking: K,
 			visible: h,
-			isSupported: K
+			isSupported: q
 		};
 	}, [
-		W,
 		G,
+		K,
 		h,
-		K
+		q
 	]);
-	let it = w(null), at = w(null), ot = w(() => {});
-	S(() => {
-		X(), P(/* @__PURE__ */ new Map()), F(/* @__PURE__ */ new Set()), p(n.getState().activeConversationId), it.current = Ve;
-		let e = at.current;
-		at.current = null, e && e.conversationId === Ve && ot.current(e.conversationId, e.messageId);
-	}, [Ve, X]), S(() => {
-		P((e) => e.size === 0 ? e : /* @__PURE__ */ new Map()), X();
-	}, [Y, X]), S(() => () => {
-		H.current?.(), H.current = null, b(z.current, "stopped"), z.current = void 0, E.cancel();
-	}, [E]), S(() => {
+	let lt = O(null), ut = O(null), dt = O(() => {});
+	E(() => {
+		Q(), V.current.clear(), I(/* @__PURE__ */ new Map()), Ne(/* @__PURE__ */ new Set()), p(n.getState().activeConversationId), lt.current = Ge;
+		let e = ut.current;
+		ut.current = null, e && e.conversationId === Ge && dt.current(e.conversationId, e.messageId);
+	}, [Ge, Q]), E(() => {
+		I((e) => e.size === 0 ? e : /* @__PURE__ */ new Map()), Q();
+	}, [X, Q]), E(() => () => {
+		U.current?.(), U.current = null, w(R.current, "stopped"), R.current = void 0, Z(), A.cancel();
+	}, [Z, A]), E(() => {
 		if (typeof window > "u") return;
-		let { settle: e, unregister: t } = ge(() => (b(z.current, "stopped"), z.current)), n = () => {
-			e(), z.current = void 0, E.cancel();
+		let { settle: e, unregister: t } = he(() => (w(R.current, "stopped"), R.current)), n = () => {
+			e(), R.current = void 0, Z(), A.cancel();
 		};
 		return window.addEventListener("pagehide", n), () => {
 			t(), window.removeEventListener("pagehide", n);
 		};
-	}, [E]), S(() => {
-		if (typeof document > "u" || !K) return;
+	}, [Z, A]), E(() => {
+		if (typeof document > "u" || !q) return;
 		let e = () => {
 			if (document.visibilityState !== "visible") return;
-			let e = E.listVoices();
-			e.length > 0 && U(), M(e), ke(!1), P((e) => e.size === 0 ? e : /* @__PURE__ */ new Map());
+			let e = A.listVoices();
+			e.length > 0 && W(), Oe(e), Ae(!1), I((e) => e.size === 0 ? e : /* @__PURE__ */ new Map());
 		};
 		return document.addEventListener("visibilitychange", e), window.addEventListener("focus", e), () => {
 			document.removeEventListener("visibilitychange", e), window.removeEventListener("focus", e);
 		};
 	}, [
-		U,
-		E,
-		K
+		W,
+		A,
+		q
 	]);
-	let st = w(Z);
-	S(() => {
-		st.current = Z;
-	}, [Z]);
-	let Q = x((e, t) => {
-		if (r(e) !== t || g(e.id) || B.current?.messageId === e.id || ze.current) return;
-		let { text: n } = pe(e.content);
+	let ft = O(ot);
+	E(() => {
+		ft.current = ot;
+	}, [ot]);
+	let $ = T((e, t) => {
+		if (r(e) !== t || g(e.id) || z.current?.messageId === e.id || Ue.current) return;
+		let { text: n } = le(e.content);
 		if (n.trim()) try {
-			st.current(e.id, n, e.metadata?.lang, { auto: !0 });
+			ft.current(e.id, n, e.metadata?.lang, { auto: !0 });
 		} catch {}
-	}, []), $ = x((e, t) => {
-		if (e !== it.current) {
-			n.getState().activeConversationId === e && (at.current = {
+	}, []), pt = T((e, t) => {
+		if (e !== lt.current) {
+			n.getState().activeConversationId === e && (ut.current = {
 				conversationId: e,
 				messageId: t
 			});
 			return;
 		}
-		let r = Ee(rt.current, t);
+		let r = Te(ct.current, t);
 		if (!r.autoSpeakEnabled && !r.sessionSpeaking || !r.visible || !r.isSupported) return;
 		let i = n.getState();
 		if (i.activeConversationId !== e) return;
 		let a = i.conversations.find((t) => t.id === e)?.messages.find((e) => e.id === t);
-		!a || a.role !== "assistant" || Q(a, r.sessionSpeaking);
-	}, [Q]);
-	S(() => {
-		ot.current = $;
-	}, [$]), S(() => {
-		if (!(!W && !G || !h || !K)) return n.subscribe((e, t) => {
+		!a || a.role !== "assistant" || $(a, r.sessionSpeaking);
+	}, [$]);
+	E(() => {
+		dt.current = pt;
+	}, [pt]), E(() => {
+		if (!(!G && !K || !h || !q)) return n.subscribe((e, t) => {
 			let n = e.activeConversationId;
 			if (!n || n !== t.activeConversationId) return;
 			let r = e.processingByConversationId[n], i = t.processingByConversationId[n];
-			if (r?.status !== "complete" || !fe(i?.status)) return;
+			if (r?.status !== "complete" || !ce(i?.status)) return;
 			let a = r.messageId ?? i?.messageId ?? null;
 			if (!a) return;
 			let o = e.conversations.find((e) => e.id === n)?.messages.find((e) => e.id === a);
-			!o || o.role !== "assistant" || Q(o, G);
+			!o || o.role !== "assistant" || $(o, K);
 		});
 	}, [
-		W,
 		G,
 		K,
+		q,
 		h,
-		Q
-	]), S(() => {
-		if (!(W || G || !Ue || !h || !K)) return n.subscribe((e, t) => {
-			let n = De(e, t);
-			!n || !re(n.id) || Q(n, !1);
+		$
+	]), E(() => {
+		if (!(G || K || !qe || !h || !q)) return n.subscribe((e, t) => {
+			let n = Ee(e, t);
+			!n || !re(n.id) || $(n, !1);
 		});
 	}, [
-		W,
 		G,
-		Ue,
 		K,
+		qe,
+		q,
 		h,
-		Q
+		$
 	]);
-	let ct = x((e) => {
+	let mt = T((e) => {
+		if (e.type === "audio") return;
+		if (e.type === "retract") {
+			B.current?.messageId === e.messageId && B.current.playback.retract();
+			return;
+		}
 		if (e.type === "abort") {
-			let t = B.current;
+			V.current.delete(e.messageId);
+			let t = z.current;
 			if (!t || t.messageId !== e.messageId) return;
-			B.current = null, (t.requestId === null ? t.waiting && H.current !== null : I.current === t.requestId) && X();
+			z.current = null, (t.requestId === null ? t.waiting && U.current !== null : L.current === t.requestId) && Q();
 			return;
 		}
-		let { messageId: t, item: r } = e, i = Ee(rt.current, t);
-		if (!i.autoSpeakEnabled && !i.sessionSpeaking || !i.visible || !i.isSupported || r.channel === "voice" !== i.sessionSpeaking || g(t) || ze.current || B.current?.messageId === t || n.getState().activeConversationId !== r.conversationId || it.current !== r.conversationId) return;
-		let { text: a } = pe(r.text);
-		if (!a.trim()) return;
-		let o = I.current;
+		let { messageId: t, item: r } = e;
+		if (e.type === "append" && B.current?.messageId !== t) {
+			if (V.current.has(t)) return;
+			V.current.add(t);
+		}
+		let i = Te(ct.current, t);
+		if (!i.autoSpeakEnabled && !i.sessionSpeaking || !i.visible || !i.isSupported || r.channel === "voice" !== i.sessionSpeaking || g(t) || Ue.current || n.getState().activeConversationId !== r.conversationId || lt.current !== r.conversationId) return;
+		let a = B.current;
+		if (a?.messageId === t && a.requestId === L.current) {
+			let t = _(e.type === "append" ? e.phrases.join(" ") : le(r.text).text, { labels: x(a.locale, et) });
+			e.type === "append" ? a.playback.append(t) : a.playback.finish(t);
+			return;
+		}
+		if (z.current?.messageId === t || e.type === "append" && (!de() || !A.open || r.channel !== "voice")) return;
+		let { text: o } = le(e.type === "append" ? e.phrases.join(" ") : r.text);
+		if (!o.trim()) return;
+		let s = L.current;
 		try {
-			st.current(t, a, r.lang, {
+			ft.current(t, o, r.lang, {
 				auto: !0,
-				firstChunkMaxChars: 110
+				firstChunkMaxChars: 110,
+				phraseStream: e.type === "append"
 			});
 		} catch {}
-		let s = I.current !== o;
-		B.current = {
+		let c = L.current !== s;
+		e.type === "append" && !c || (V.current.delete(t), z.current = {
 			messageId: t,
-			requestId: s ? I.current : null,
-			waiting: !s && H.current !== null
-		};
-	}, [X]);
-	S(() => d.subscribe(ct), [ct]);
-	let lt = C(() => l(j, J.tag), [j, J]), ut = C(() => s(j, J.tag, q[Y]), [
-		j,
+			requestId: c ? L.current : null,
+			waiting: !c && U.current !== null
+		});
+	}, [
+		Q,
+		A,
+		et
+	]);
+	E(() => f.subscribe(mt), [mt]);
+	let ht = D(() => u(F, Y.tag), [F, Y]), gt = D(() => s(F, Y.tag, J[X]), [
+		F,
+		Y,
 		J,
-		q,
-		Y
-	]), dt = C(() => ({
-		isSupported: K,
-		availability: Ye,
-		isBlocked: Ye === "no-voices",
-		disabledReason: Ye === "no-voices" ? D("speech.noVoicesOnDevice", "This device has no speech voices installed. Add one in your device speech or accessibility settings.") : null,
-		blockedFor: Qe,
-		deviceFallbackFor: $e,
-		voiceChoices: lt,
-		spokenChunks: ue(k, Pe),
-		activeVoiceName: ut?.name ?? null,
-		preferredVoiceName: q[Y] ?? null,
-		setVoice: (e) => Ke(Y, e),
-		speakingMessageId: k.status === "speaking" ? k.messageId : null,
-		outputNotCleared: Me,
-		isSpeaking: (e) => le(k, e),
-		toggle: nt,
-		stop: X,
-		autoSpeakEnabled: W,
-		setAutoSpeak: Le,
-		sessionSpeaking: G,
-		setSessionSpeaking: Re,
-		setSessionMicOpen: Be,
-		speakFinishedReply: $,
-		previewVoice: Ze
+		X
+	]), _t = D(() => ({
+		isSupported: q,
+		availability: $e,
+		isBlocked: $e === "no-voices",
+		disabledReason: $e === "no-voices" ? j("speech.noVoicesOnDevice", "This device has no speech voices installed. Add one in your device speech or accessibility settings.") : null,
+		blockedFor: nt,
+		deviceFallbackFor: rt,
+		voiceChoices: ht,
+		spokenChunks: oe(N, Re),
+		activeVoiceName: gt?.name ?? null,
+		preferredVoiceName: J[X] ?? null,
+		setVoice: (e) => Xe(X, e),
+		speakingMessageId: N.status === "speaking" ? N.messageId : null,
+		outputNotCleared: Pe,
+		isSpeaking: (e) => C(N, e),
+		toggle: st,
+		stop: Q,
+		autoSpeakEnabled: G,
+		setAutoSpeak: Ve,
+		sessionSpeaking: K,
+		setSessionSpeaking: He,
+		setSessionMicOpen: We,
+		speakFinishedReply: pt,
+		previewVoice: tt
 	}), [
-		K,
-		Ye,
-		Qe,
-		$e,
-		lt,
-		ut,
 		q,
-		Ke,
-		Y,
-		k,
-		Me,
+		$e,
 		nt,
+		rt,
+		ht,
+		gt,
+		J,
+		Xe,
 		X,
-		W,
-		Le,
+		N,
+		Pe,
+		st,
+		Q,
 		G,
-		Be,
-		$,
-		Ze,
-		D,
-		Pe
+		Ve,
+		K,
+		We,
+		pt,
+		tt,
+		j,
+		Re
 	]);
-	return /* @__PURE__ */ xe(Oe.Provider, {
-		value: dt,
-		children: f
+	return /* @__PURE__ */ be(De.Provider, {
+		value: _t,
+		children: ne
 	});
 };
 //#endregion
-export { O as AssistantSpeechProvider, D as useAssistantSpeech };
+export { M as AssistantSpeechProvider, j as useAssistantSpeech };