opencharm 0.4.0 → 0.5.0

package/README.md

@@ -20,6 +20,7 @@ npx opencharm
 | `opencharm pair <code>`                                                                               | Pair the charm showing that code; you choose its PIN (`--name <name>`)                                                                                                                                                                                                                                                                 |
 | `opencharm status`                                                                                    | What's running (charmd, each connected charm's build, the starter commit of the workspace; spec 015), then the paired charms and whether they are connected, locked or blocked                                                                                                                                                         |
 | `opencharm lock` / `unlock` / `revoke <charm>`                                                        | Lock it now; clear a block and unlock; forget it                                                                                                                                                                                                                                                                                       |
+| `opencharm replies on\|off`                                                                           | The desktop charm speaks its replies, or shows them as text, from the next one; a charm on a board always speaks. To keep it, set `"speakReplies": false` in `opencharm.json`                                                                                                                                                          |
 | `opencharm look`                                                                                      | Show the charm's look, or change it on every unlocked charm until charmd restarts (`--colour <colour>`, `--greeting <text>`, `--motion full\|calm`); to keep it, set `charm` in `opencharm.json`                                                                                                                                       |
 | `opencharm dev face <charm> <state> [text]` / `dev say <charm> <text>` / `dev ask <charm> <question>` | Show a face, speak, or ask a question (hold = yes, press = no) on a connected charm (testing)                                                                                                                                                                                                                                          |
 | `opencharm voice` / `voice install`                                                                   | The local voice's models (Parakeet listening, Supertonic speaking): whether they're here, and download them now (charmd does it on its first start otherwise)                                                                                                                                                                          |

package/dist/main.mjs

@@ -828,6 +828,7 @@ const schema = z.object({
 	agent: agentSchema.default({ adapter: "fake" }),
 	turnTimeoutSeconds: z.number().int().min(5).max(600).default(60),
 	logTranscripts: z.boolean().default(false),
+	speakReplies: z.boolean().default(true),
 	charm: charmSchema.prefault({})
 }).strict();
 function defaultAdminSocket(statePath) {
@@ -961,7 +962,7 @@ function sendAdmin(socketPath, request) {
 }
 //#endregion
 //#region package.json
-var version = "0.4.0";
+var version = "0.5.0";
 function channelToCube(value) {
 	if (value < 48) return 0;
 	if (value < 115) return 1;
@@ -1225,6 +1226,19 @@ async function runAdminCommand(ctx, command, args, deps) {
 			await look(ctx, args, deps);
 			return;
 		}
+		if (command === "replies") {
+			const choice = firstPositional(args);
+			if (choice !== "on" && choice !== "off") {
+				fail$2(ctx, "Usage: opencharm replies on|off   (spoken, or text on the desktop charm)");
+				return;
+			}
+			await deps.send(deps.socket, {
+				cmd: "replies",
+				speak: choice === "on"
+			});
+			ctx.out.write(choice === "on" ? "Replies are spoken, from the next one.\n" : "Replies show as text on the desktop charm, from the next one (a charm on a board still speaks).\n");
+			return;
+		}
 		if (command === "status") {
 			const answer = statusSchema.safeParse(await deps.send(deps.socket, { cmd: "status" }));
 			if (!answer.success) {
@@ -1452,6 +1466,7 @@ function runHelp(ctx) {
     opencharm lock <charm>          lock a charm now
     opencharm unlock <charm>        clear a block and unlock
     opencharm revoke <charm>        forget a charm (it must pair again)
+    opencharm replies on|off        the desktop charm speaks its replies, or shows them as text
     opencharm look                  show or change the charm's look until charmd restarts   [--colour <c>] [--greeting <text>] [--motion full|calm]
     opencharm dev face <charm> <state> [text]   show a face (testing)
     opencharm dev say <charm> <text>            speak through it (testing)
@@ -1796,6 +1811,7 @@ async function serveMcp(input, output, deps) {
 }
 //#endregion
 //#region ../charmd/src/admin/admin-actions.ts
+const repliesInput = z.object({ speak: z.boolean() }).strict();
 const pairInput = z.object({
 	code: z.string().regex(/^\d{6}$/, "the code has 6 digits"),
 	pin: z.string().regex(/^\d{4,12}$/, "the PIN is 4 to 12 digits"),
@@ -1829,7 +1845,7 @@ function parseInput(schema, input) {
 	if (!result.success) throw new Error(result.error.issues.map((i) => i.message).join("; "));
 	return result.data;
 }
-function createAdminActions({ store, pairing, sessions, look, agentCanChangeLook, identity, starter, voice }) {
+function createAdminActions({ store, pairing, sessions, look, agentCanChangeLook, identity, starter, voice, replies }) {
 	function sessionsOf(charmId) {
 		return [...sessions()].filter((s) => s.charmId === charmId && s.state !== "closed");
 	}
@@ -1934,6 +1950,14 @@ function createAdminActions({ store, pairing, sessions, look, agentCanChangeLook
 				};
 			});
 		},
+		replies(input) {
+			return settled(() => {
+				const { speak } = parseInput(repliesInput, input);
+				if (!replies) throw new Error("This charmd can't change how replies are given.");
+				replies.speak = speak;
+				return { speak };
+			});
+		},
 		status() {
 			return {
 				charmd: identity ?? "unknown",
@@ -1967,7 +1991,8 @@ const COMMANDS = [
 	"devFace",
 	"devSay",
 	"devAsk",
-	"look"
+	"look",
+	"replies"
 ];
 function isCommand(value) {
 	return typeof value === "string" && COMMANDS.includes(value);
@@ -2248,7 +2273,8 @@ var Session = class {
 			sendAudio: (packet) => {
 				if (this.#state === "unlocked") this.#deps.outbound.sendAudio?.(packet);
 			},
-			ask: (text, options) => this.ask(text, options)
+			ask: (text, options) => this.ask(text, options),
+			kind: () => this.#build?.kind
 		});
 		return this.#turn;
 	}
@@ -3220,6 +3246,10 @@ var SentenceSplitter = class {
 	#buffer = "";
 	#pending = "";
 	#started = false;
+	#firstClause;
+	constructor(options = {}) {
+		this.#firstClause = options.firstClause ?? true;
+	}
 	#take(piece) {
 		const text = piece.replace(/\s+/g, " ").trim();
 		if (!text) return [];
@@ -3234,7 +3264,7 @@ var SentenceSplitter = class {
 		this.#buffer += chunk;
 		const out = [];
 		let start = 0;
-		for (const match of this.#buffer.matchAll(this.#started ? BOUNDARY : FIRST_BOUNDARY)) {
+		for (const match of this.#buffer.matchAll(this.#started || !this.#firstClause ? BOUNDARY : FIRST_BOUNDARY)) {
 			const end = match.index + match[0].length;
 			if (/^[,;:]/.test(match[0]) && (this.#started || this.#pending.length + end - start < FIRST_CLAUSE_CHARS)) continue;
 			const lastWord = this.#buffer.slice(start, end).trim().split(/\s+/).at(-1)?.toLowerCase() ?? "";
@@ -3259,6 +3289,8 @@ function cleanForSpeech(text) {
 //#endregion
 //#region ../charmd/src/turn/turn.ts
 const MAX_FRAMES = 1e3;
+const READING_WORDS_PER_SECOND = 3;
+const MIN_READING_MS = 2e3;
 const LOOKAHEAD = 2;
 const AGENT_LABELS = { ...Object.fromEntries(Object.entries(ACP_AGENTS).map(([name, agent]) => [name, agent.label])) };
 var TurnTimeout = class extends Error {};
@@ -3269,6 +3301,10 @@ var StageError = class extends Error {
 		this.stage = stage;
 	}
 };
+function readingMs(text) {
+	const words = text.split(/\s+/).filter(Boolean).length;
+	return Math.max(MIN_READING_MS, words / READING_WORDS_PER_SECOND * 1e3);
+}
 function sleepMs(ms) {
 	return new Promise((resolve) => setTimeout(resolve, ms));
 }
@@ -3374,6 +3410,7 @@ var TurnController = class {
 		let timer = startClock();
 		let speakingStarted = false;
 		let asking = false;
+		let reading = false;
 		const restartClock = () => {
 			clearTimeout(timer);
 			timer = startClock();
@@ -3384,7 +3421,15 @@ var TurnController = class {
 				restartClock();
 			},
 			touch: () => {
-				if (speakingStarted && !asking) restartClock();
+				if (speakingStarted && !asking && !reading) restartClock();
+			},
+			pause: () => {
+				reading = true;
+				clearTimeout(timer);
+			},
+			resume: () => {
+				reading = false;
+				restartClock();
 			}
 		};
 		const timeoutGuard = new Promise((_, reject) => signal.addEventListener("abort", () => {
@@ -3429,7 +3474,7 @@ var TurnController = class {
 				signal,
 				...ask ? { ask: (action) => this.ask(`${label} wants to ${action}.`) } : {}
 			});
-			said = await this.#speak(reply, signal, live, timings, started, timeoutGuard, guessLanguage(heard) ?? this.#deps.voice.language ?? "en", clock);
+			said = await this.#speak(reply, signal, live, timings, started, timeoutGuard, guessLanguage(heard) ?? this.#deps.voice.language ?? "en", clock, this.#deps.speakAloud?.() ?? true);
 			if (!live()) return;
 			if (this.#speaking) this.#deps.send({
 				type: "tts",
@@ -3471,7 +3516,7 @@ var TurnController = class {
 			});
 		}
 	}
-	async #speak(source, signal, live, timings, started, timeoutGuard, language, clock) {
+	async #speak(source, signal, live, timings, started, timeoutGuard, language, clock, aloud) {
 		const now = this.#deps.now ?? Date.now;
 		const going = () => live() && !signal.aborted;
 		let speaking = language;
@@ -3497,11 +3542,11 @@ var TurnController = class {
 				text,
 				language: sentenceLanguage
 			});
-			startUpTo(playing + LOOKAHEAD);
+			if (aloud) startUpTo(playing + LOOKAHEAD);
 			wake();
 		};
 		const produce = (async () => {
-			const splitter = new SentenceSplitter();
+			const splitter = new SentenceSplitter({ firstClause: aloud });
 			await inStage("agent", async () => {
 				for await (const chunk of source) {
 					if (!going()) return;
@@ -3524,6 +3569,11 @@ var TurnController = class {
 					await new Promise((resolve) => wake = resolve);
 					continue;
 				}
+				if (!aloud) {
+					await this.#showForReading(queue[playing].text, signal, going, clock, timings, started);
+					playing += 1;
+					continue;
+				}
 				startUpTo(playing + LOOKAHEAD);
 				const { text, packets } = queue[playing];
 				if (!packets) throw new Error("A sentence wasn't started");
@@ -3563,9 +3613,20 @@ var TurnController = class {
 			clearTimeout(timer);
 			timer = startClock();
 		};
+		let reading = false;
 		const clock = {
 			started: restartClock,
-			touch: restartClock
+			touch: () => {
+				if (!reading) restartClock();
+			},
+			pause: () => {
+				reading = true;
+				clearTimeout(timer);
+			},
+			resume: () => {
+				reading = false;
+				restartClock();
+			}
 		};
 		const timeoutGuard = new Promise((_, reject) => controller.signal.addEventListener("abort", () => {
 			if (controller.signal.reason instanceof TurnTimeout) reject(controller.signal.reason);
@@ -3575,7 +3636,7 @@ var TurnController = class {
 			yield await Promise.resolve(text);
 		}
 		try {
-			await this.#speak(once(), controller.signal, live, {}, Date.now(), timeoutGuard, guessLanguage(text) ?? this.#deps.voice.language ?? "en", clock);
+			await this.#speak(once(), controller.signal, live, {}, Date.now(), timeoutGuard, guessLanguage(text) ?? this.#deps.voice.language ?? "en", clock, this.#deps.speakAloud?.() ?? true);
 			if (!live()) return;
 			if (this.#speaking) this.#deps.send({
 				type: "tts",
@@ -3598,6 +3659,28 @@ var TurnController = class {
 			clearTimeout(timer);
 		}
 	}
+	async #showForReading(text, signal, going, clock, timings, started) {
+		if (!going()) return;
+		if (!this.#speaking) {
+			this.#speaking = true;
+			this.#state = "speaking";
+			timings.firstTextMs = (this.#deps.now ?? Date.now)() - started;
+			clock.started();
+			this.#deps.send({
+				type: "tts",
+				state: "start"
+			});
+		}
+		this.#deps.send({
+			type: "tts",
+			state: "sentence_start",
+			text
+		});
+		const sleep = this.#deps.sleep ?? sleepMs;
+		clock.pause();
+		await Promise.race([sleep(readingMs(text)), new Promise((resolve) => signal.addEventListener("abort", () => resolve(), { once: true }))]);
+		clock.resume();
+	}
 	#failureLine(error) {
 		if (error instanceof TurnTimeout) return "That took too long.";
 		if (error instanceof StageError && error.cause instanceof VoiceNotReady) return error.cause.message;
@@ -4670,6 +4753,7 @@ async function startDaemon(config, options = {}) {
 	});
 	const sessions = /* @__PURE__ */ new Map();
 	const voice = createVoiceFromConfig(config);
+	const replies = { speak: config.speakReplies };
 	voice.warm?.().catch(() => void 0);
 	const log = options.log ?? ((entry) => {
 		if (!options.quiet) console.log(JSON.stringify({
@@ -4691,6 +4775,7 @@ async function startDaemon(config, options = {}) {
 		agentCanChangeLook,
 		identity: options.identity,
 		starter: options.starter,
+		replies,
 		voice: () => {
 			const downloading = ["parakeet", "supertonic"].flatMap((id) => {
 				const state = modelState(id);
@@ -4749,7 +4834,8 @@ async function startDaemon(config, options = {}) {
 				agent,
 				timeoutMs: config.turnTimeoutSeconds * 1e3,
 				logTranscripts: config.logTranscripts,
-				log
+				log,
+				speakAloud: () => io.kind() !== "desktop" || replies.speak
 			})
 		});
 		sessions.set(session.connectionId, session);
@@ -4831,7 +4917,7 @@ function formatIdentity(unit, version, commit) {
 let cached;
 function identity() {
 	if (cached) return cached;
-	const commit = "42da0c7";
+	const commit = "385e52a";
 	cached = {
 		version: `cli@${version}`,
 		commit,
@@ -5311,6 +5397,7 @@ async function main(argv) {
 		case "status":
 		case "dev":
 		case "look":
+		case "replies":
 			await runAdminCommand(ctx, command, rest, {
 				socket: loadConfig(flagValue(rest, "--config")).adminSocket,
 				send: sendAdmin,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "opencharm",
-  "version": "0.4.0",
+  "version": "0.5.0",
   "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": {