opencharm 0.8.0 → 0.8.1

package/dist/main.mjs

@@ -997,7 +997,7 @@ function sendAdmin(socketPath, request) {
 }
 //#endregion
 //#region package.json
-var version = "0.8.0";
+var version = "0.8.1";
 function channelToCube(value) {
 	if (value < 48) return 0;
 	if (value < 115) return 1;
@@ -3414,6 +3414,7 @@ var TurnController = class {
 		this.#frames = [];
 		this.#face("listening");
 		this.#deps.agent.warm?.(this.#deps.sessionKey);
+		if (this.#deps.speakAloud?.() ?? true) this.#deps.voice.prime?.();
 	}
 	audio(frame) {
 		if (this.#state === "listening" && this.#frames.length < MAX_FRAMES) this.#frames.push(frame);
@@ -4378,21 +4379,25 @@ function audioOf(frame) {
 function voiceFor(language, voices) {
 	return voices[language] ?? voices.en ?? DEFAULT_VOICES.en;
 }
-function connect(text, language, voices, signal, options) {
-	const timeoutMs = options.timeoutMs ?? 15e3;
+function openSocket(options) {
 	const connection = randomUUID().replace(/-/g, "");
 	const url = `${options.url ?? SERVICE_URL}?TrustedClientToken=${TRUSTED_CLIENT_TOKEN}&Sec-MS-GEC=${secMsGec()}&Sec-MS-GEC-Version=${GEC_VERSION}&ConnectionId=${connection}`;
-	const packets = createChannel();
-	const reader = new WebmOpusReader();
-	let sent = 0;
-	let settled = false;
-	const socket = new WebSocket(url, {
+	return new WebSocket(url, {
 		headers: {
 			"User-Agent": USER_AGENT,
 			Origin: ORIGIN
 		},
-		handshakeTimeout: timeoutMs
+		handshakeTimeout: options.timeoutMs ?? 15e3
 	});
+}
+function connect(text, language, voices, signal, options, ready) {
+	const timeoutMs = options.timeoutMs ?? 15e3;
+	const connection = randomUUID().replace(/-/g, "");
+	const packets = createChannel();
+	const reader = new WebmOpusReader();
+	let sent = 0;
+	let settled = false;
+	const socket = ready ?? openSocket(options);
 	const finish = (error) => {
 		if (settled) return;
 		settled = true;
@@ -4407,7 +4412,7 @@ function connect(text, language, voices, signal, options) {
 	const timer = setTimeout(() => finish(/* @__PURE__ */ new Error("Microsoft's voice took too long")), timeoutMs);
 	if (signal.aborted) onAbort();
 	else signal.addEventListener("abort", onAbort, { once: true });
-	socket.on("open", () => {
+	const request = () => {
 		socket.send(message("speech.config", "application/json; charset=utf-8", JSON.stringify({ context: { synthesis: { audio: {
 			metadataoptions: {
 				sentenceBoundaryEnabled: "false",
@@ -4416,7 +4421,9 @@ function connect(text, language, voices, signal, options) {
 			outputFormat: OUTPUT_FORMAT
 		} } } })));
 		socket.send(message("ssml", "application/ssml+xml", ssml(text, voiceFor(language, voices)), connection));
-	});
+	};
+	if (socket.readyState === WebSocket.OPEN) request();
+	else socket.once("open", request);
 	socket.on("message", (data, isBinary) => {
 		const frame = Buffer.isBuffer(data) ? data : Buffer.concat(data);
 		if (!isBinary) {
@@ -4444,9 +4451,48 @@ function createMicrosoftSpeaker(options = {}) {
 		...DEFAULT_VOICES,
 		...options.voices
 	};
-	const stream = (text, signal, language = "en") => connect(text, language, voices, signal, options);
+	const spareMs = options.spareMs ?? 2e4;
+	let spare;
+	const takeSpare = () => {
+		const taken = spare;
+		spare = void 0;
+		if (!taken) return void 0;
+		if ((taken.socket.readyState === WebSocket.OPEN || taken.socket.readyState === WebSocket.CONNECTING) && Date.now() < taken.until) return taken.socket;
+		taken.socket.terminate();
+	};
+	const prime = () => {
+		if (spare) return;
+		const socket = openSocket(options);
+		socket.on("error", () => void 0);
+		spare = {
+			socket,
+			until: Date.now() + spareMs
+		};
+		setTimeout(() => {
+			if (spare?.socket !== socket) return;
+			spare = void 0;
+			socket.terminate();
+		}, spareMs).unref();
+	};
+	async function* stream(text, signal, language = "en") {
+		const ready = takeSpare();
+		if (ready) {
+			let spoke = false;
+			try {
+				for await (const packet of connect(text, language, voices, signal, options, ready)) {
+					spoke = true;
+					yield packet;
+				}
+				return;
+			} catch (error) {
+				if (spoke || signal.aborted) throw error;
+			}
+		}
+		yield* connect(text, language, voices, signal, options);
+	}
 	return {
 		name: "microsoft",
+		prime,
 		stream,
 		async synthesize(text, signal, language = "en") {
 			const packets = [];
@@ -4691,6 +4737,9 @@ function speakWithFallback(chain, options = {}) {
 		},
 		warm: async () => {
 			await Promise.allSettled(chain.map((speaker) => speaker.warm?.() ?? Promise.resolve()));
+		},
+		prime: () => {
+			if ((resting.get(first) ?? 0) <= now()) first.prime?.();
 		}
 	};
 }
@@ -4750,7 +4799,8 @@ function createVoiceFromConfig(config) {
 		language: config.voice.language,
 		warm: async () => {
 			await Promise.allSettled([listener.warm?.(), speaker.warm?.()]);
-		}
+		},
+		...speaker.prime ? { prime: speaker.prime } : {}
 	};
 }
 function charmToolsServers(config, tools) {
@@ -4969,7 +5019,7 @@ function formatIdentity(unit, version, commit) {
 let cached;
 function identity() {
 	if (cached) return cached;
-	const commit = "bf6c38c";
+	const commit = "e06b6db";
 	cached = {
 		version: `cli@${version}`,
 		commit,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "opencharm",
-  "version": "0.8.0",
+  "version": "0.8.1",
   "description": "An open-source body for the AI agent you already run: a face, a voice, one key. Claude Code, Codex, Gemini CLI, Hermes, OpenClaw or any ACP or OpenAI-compatible agent; on a small board, your Mac's notch or the browser.",
   "type": "module",
   "bin": {