@livx.cc/agentx 0.97.9 → 0.98.2

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/dist/cli.js CHANGED
@@ -1755,7 +1755,7 @@ function makeRealShellTool(options) {
1755
1755
  proc.stderr?.on("data", collect);
1756
1756
  proc.on("error", (err2) => {
1757
1757
  if (err2?.name === "AbortError" || ctl.signal.aborted) return finish(reasonFor(timedOut, timeoutMs, clean(out)));
1758
- log13.debug("shell spawn error", err2);
1758
+ log14.debug("shell spawn error", err2);
1759
1759
  finish(`[exit 1] ${err2?.message ?? err2}${out ? "\n" + clean(out) : ""}`);
1760
1760
  });
1761
1761
  proc.on("close", (code) => {
@@ -1817,7 +1817,7 @@ ${clean(out) || "(no output yet)"}`;
1817
1817
  }
1818
1818
  ];
1819
1819
  }
1820
- var log13, clean, DETACHED, SECRET_ENV_RE, _spawn, ShellJobRegistry, NO_JOB2;
1820
+ var log14, clean, DETACHED, SECRET_ENV_RE, _spawn, ShellJobRegistry, NO_JOB2;
1821
1821
  var init_tools_shell = __esm({
1822
1822
  "src/tools.shell.ts"() {
1823
1823
  "use strict";
@@ -1825,7 +1825,7 @@ var init_tools_shell = __esm({
1825
1825
  init_redact();
1826
1826
  init_logging();
1827
1827
  init_shell_sandbox();
1828
- log13 = forComponent("shell");
1828
+ log14 = forComponent("shell");
1829
1829
  clean = (s) => truncateOutput(redactSecrets(s.replace(/\n+$/, "")));
1830
1830
  DETACHED = { stdio: ["ignore", "pipe", "pipe"], detached: true };
1831
1831
  SECRET_ENV_RE = /(_API_KEY|_TOKEN|_SECRET|_PASSWORD|_PRIVATE_KEY|^AWS_|^GITHUB_TOKEN$|^OPENAI_|^ANTHROPIC_|^GOOGLE_|^GEMINI_|^GROQ_|^NPM_TOKEN$)/i;
@@ -4582,6 +4582,188 @@ function digestRun(messages, maxChars) {
4582
4582
  import { MemFilesystem as MemFilesystem2 } from "@livx.cc/wcli/core";
4583
4583
  init_logging();
4584
4584
 
4585
+ // src/voice/emotion.ts
4586
+ init_logging();
4587
+ var log8 = forComponent("Emotion");
4588
+ var EMOTIONS = [
4589
+ // primary (best results)
4590
+ "neutral",
4591
+ "angry",
4592
+ "excited",
4593
+ "content",
4594
+ "sad",
4595
+ "scared",
4596
+ // extended
4597
+ "happy",
4598
+ "enthusiastic",
4599
+ "elated",
4600
+ "triumphant",
4601
+ "amazed",
4602
+ "surprised",
4603
+ "flirtatious",
4604
+ "curious",
4605
+ "calm",
4606
+ "grateful",
4607
+ "affectionate",
4608
+ "sympathetic",
4609
+ "mysterious",
4610
+ "frustrated",
4611
+ "disgusted",
4612
+ "sarcastic",
4613
+ "disappointed",
4614
+ "hurt",
4615
+ "guilty",
4616
+ "bored",
4617
+ "tired",
4618
+ "nostalgic",
4619
+ "apologetic",
4620
+ "hesitant",
4621
+ "confused",
4622
+ "anxious",
4623
+ "panicked",
4624
+ "proud",
4625
+ "confident",
4626
+ "skeptical",
4627
+ "contemplative",
4628
+ "determined"
4629
+ ];
4630
+ var VALID = new Set(EMOTIONS);
4631
+ var ALIASES = {
4632
+ cheerful: "happy",
4633
+ joyful: "happy",
4634
+ joy: "happy",
4635
+ glad: "happy",
4636
+ pleased: "happy",
4637
+ warm: "affectionate",
4638
+ thrilled: "excited",
4639
+ eager: "enthusiastic",
4640
+ ecstatic: "elated",
4641
+ euphoric: "elated",
4642
+ mad: "angry",
4643
+ furious: "angry",
4644
+ annoyed: "frustrated",
4645
+ irritated: "frustrated",
4646
+ agitated: "frustrated",
4647
+ shocked: "surprised",
4648
+ astonished: "amazed",
4649
+ wonder: "amazed",
4650
+ worried: "anxious",
4651
+ nervous: "anxious",
4652
+ afraid: "scared",
4653
+ alarmed: "panicked",
4654
+ unsure: "hesitant",
4655
+ uncertain: "hesitant",
4656
+ doubtful: "skeptical",
4657
+ suspicious: "skeptical",
4658
+ thoughtful: "contemplative",
4659
+ focused: "determined",
4660
+ serious: "determined",
4661
+ playful: "flirtatious",
4662
+ teasing: "flirtatious",
4663
+ ironic: "sarcastic",
4664
+ cheeky: "sarcastic",
4665
+ thankful: "grateful",
4666
+ sorry: "apologetic",
4667
+ down: "sad",
4668
+ melancholic: "sad",
4669
+ gloomy: "sad",
4670
+ peaceful: "calm",
4671
+ serene: "calm",
4672
+ relaxed: "calm",
4673
+ sleepy: "tired"
4674
+ };
4675
+ var NONVERBAL = { laughter: "laughter", laughs: "laughter", laugh: "laughter", laughing: "laughter" };
4676
+ function normalizeEmotion(raw) {
4677
+ const k = raw.trim().toLowerCase();
4678
+ if (VALID.has(k)) return k;
4679
+ return ALIASES[k] ?? null;
4680
+ }
4681
+ function resolveTag(raw) {
4682
+ const k = raw.trim().toLowerCase();
4683
+ if (NONVERBAL[k]) return { kind: "nonverbal", value: NONVERBAL[k] };
4684
+ const e = normalizeEmotion(k);
4685
+ return e ? { kind: "emotion", value: e } : null;
4686
+ }
4687
+ var TAG_RE = /\[([a-zA-Z][a-zA-Z ]{0,24})\]/g;
4688
+ var PARTIAL_RE = /\[[a-zA-Z ]*$/;
4689
+ var cartesiaTag = (t) => t.kind === "nonverbal" ? `[${t.value}]` : `<emotion value="${t.value}"/>`;
4690
+ function renderEmotions(text, opts = { show: true }) {
4691
+ let speech = "", display = "", prose = "", last = 0;
4692
+ TAG_RE.lastIndex = 0;
4693
+ for (let m = TAG_RE.exec(text); m; m = TAG_RE.exec(text)) {
4694
+ const before = text.slice(last, m.index);
4695
+ speech += before;
4696
+ display += before;
4697
+ prose += before;
4698
+ const tag = resolveTag(m[1]);
4699
+ if (tag) {
4700
+ speech += cartesiaTag(tag);
4701
+ if (opts.show) display += m[0];
4702
+ } else {
4703
+ log8.debug(`dropping unknown emotion tag ${m[0]}`);
4704
+ }
4705
+ last = m.index + m[0].length;
4706
+ }
4707
+ const tail = text.slice(last);
4708
+ return { speech: speech + tail, display: display + tail, prose: prose + tail };
4709
+ }
4710
+ var EmotionStream = class {
4711
+ constructor(show = true) {
4712
+ this.show = show;
4713
+ }
4714
+ show;
4715
+ buf = "";
4716
+ pending = null;
4717
+ feed(delta) {
4718
+ this.buf += delta;
4719
+ return this.drain(false);
4720
+ }
4721
+ flush() {
4722
+ return this.drain(true);
4723
+ }
4724
+ drain(final) {
4725
+ let body = this.buf;
4726
+ if (!final) {
4727
+ const p = body.match(PARTIAL_RE);
4728
+ if (p) {
4729
+ this.buf = p[0];
4730
+ body = body.slice(0, body.length - p[0].length);
4731
+ } else this.buf = "";
4732
+ } else this.buf = "";
4733
+ let speech = "", display = "", prose = "", last = 0;
4734
+ TAG_RE.lastIndex = 0;
4735
+ for (let m = TAG_RE.exec(body); m; m = TAG_RE.exec(body)) {
4736
+ this.emit(body.slice(last, m.index), (s, d, p) => {
4737
+ speech += s;
4738
+ display += d;
4739
+ prose += p;
4740
+ });
4741
+ const tag = resolveTag(m[1]);
4742
+ if (tag) {
4743
+ this.pending = tag;
4744
+ if (this.show) display += m[0];
4745
+ } else log8.debug(`dropping unknown emotion tag ${m[0]}`);
4746
+ last = m.index + m[0].length;
4747
+ }
4748
+ this.emit(body.slice(last), (s, d, p) => {
4749
+ speech += s;
4750
+ display += d;
4751
+ prose += p;
4752
+ });
4753
+ return { speech, display, prose };
4754
+ }
4755
+ /** Emit a prose span, flushing any pending tag onto its FRONT (only once real words appear). */
4756
+ emit(text, sink) {
4757
+ if (!text) return;
4758
+ let speech = text;
4759
+ if (this.pending && /[\p{L}\p{N}]/u.test(text)) {
4760
+ speech = cartesiaTag(this.pending) + text;
4761
+ this.pending = null;
4762
+ }
4763
+ sink(speech, text, text);
4764
+ }
4765
+ };
4766
+
4585
4767
  // src/voice/spokenSplitter.ts
4586
4768
  var OPEN = "<spoken>";
4587
4769
  var CLOSE = "</spoken>";
@@ -4655,7 +4837,7 @@ var SpokenSplitter = class {
4655
4837
  };
4656
4838
 
4657
4839
  // src/duplex.ts
4658
- var log8 = forComponent("DuplexAgent");
4840
+ var log9 = forComponent("DuplexAgent");
4659
4841
  function describeCall(call) {
4660
4842
  const v = call.args && Object.values(call.args).find((x) => typeof x === "string" && x.trim());
4661
4843
  const hint = v ? ` (${String(v).replace(/\s+/g, " ").trim().slice(0, 48)})` : "";
@@ -4694,6 +4876,9 @@ var DuplexAgentOptions = class {
4694
4876
  /** Voice register: 'neutral' = clean spoken style; 'conversational' = human-like — fillers,
4695
4877
  * backchannels, impulsive first reactions before content (mimics real duplex conversation). */
4696
4878
  voiceStyle = "neutral";
4879
+ /** Teach the model to emit inline `[emotion]` tags for Cartesia emotion control. Only set when the
4880
+ * TTS actually speaks them — text-duplex (no TTS) would otherwise print literal tags. */
4881
+ emotionTags = false;
4697
4882
  /** Awaited BEFORE a worker spawns — open a per-task checkpoint frame, audit, etc.
4698
4883
  * (post-spawn would race the worker's first edits). */
4699
4884
  onTaskStart;
@@ -4726,6 +4911,7 @@ var VOICE_SYSTEM_PROMPT = 'You are a spoken voice assistant \u2014 the user HEAR
4726
4911
  var THINK_GUIDANCE = "\u2022 `Think` \u2014 your brain. A premium reasoning model, FAR more expensive than Act. Reserve it for open-ended architecture/design questions, or a problem Act already FAILED at. ALL implementation work \u2014 coding, refactoring, debugging, edge cases, tests \u2014 goes to Act; Act is highly capable. Never send the same work to both.";
4727
4912
  var THINK_DISABLED_GUIDANCE = "(Think tier is not available \u2014 use Act for all escalations.)";
4728
4913
  var VOICE_STYLE_CONVERSATIONAL = `Speak like a person in a live conversation, not an assistant reading a script. React first, then deliver: a quick impulsive beat ("oh nice", "hmm, hold on", "ah, got it") before the substance. Use contractions always. Vary sentence length \u2014 some very short. Light fillers and backchannels are fine ("mm-hm", "right", "let's see") but at most one per reply \u2014 never stack them. When you escalate to Act or Think, say it like a human would ("hang on, let me actually dig into that \u2014 gimme a minute") instead of announcing a task. When a result comes back, react to it like you just found out ("okay so \u2014 turns out\u2026"). Match the user's energy: a quick question gets a quick answer \u2014 a few words is a perfectly good turn. Prefer a short answer plus an offer ("want the details?") over covering everything. Never narrate your own mechanics (no "I will now act", no task ids out loud).`;
4914
+ var EMOTION_TAGS_GUIDANCE = `EMOTION: your voice is synthesized with emotion control. Prefix a sentence with an inline [emotion] tag, placed directly before the sentence it colors, to shape how it is spoken. Use it ONLY when the emotion genuinely fits the words (it amplifies real feeling, it cannot fake it) \u2014 do not tag every sentence; reserve it for moments that carry feeling, and vary which one you use. You may also drop [laughter] for a natural laugh. Available emotions: ${EMOTIONS.join(", ")}.`;
4729
4915
  var DuplexAgent = class _DuplexAgent {
4730
4916
  options;
4731
4917
  voice;
@@ -4751,6 +4937,8 @@ var DuplexAgent = class _DuplexAgent {
4751
4937
  // briefs dispatched this turn (detect identical re-dispatch)
4752
4938
  spokeThisTurn = false;
4753
4939
  // any non-empty text_delta streamed this turn
4940
+ heldThisTurn = false;
4941
+ // Hold called this turn → turn is INTENTIONALLY silent (suppress reflex text + no dead-air ack)
4754
4942
  nudging = false;
4755
4943
  // re-ack pass in flight: block ALL tools, prevent recursion
4756
4944
  reflexBuf = "";
@@ -4795,7 +4983,7 @@ var DuplexAgent = class _DuplexAgent {
4795
4983
  ...new Set(workerToolNames.filter((n) => n.startsWith("mcp__")).map((n) => n.slice(5).split("__")[0]))
4796
4984
  ];
4797
4985
  const workerMcp = mcpNames.length ? `, and it can use these MCP servers: ${[...new Set(mcpNames)].join(", ")}` + (mcpNames.some((n) => /browser/i.test(n)) ? ' \u2014 including driving a REAL browser (open tabs, navigate, click, screenshot), so answer "yes" if asked whether you can control/drive a browser and route an actual browse to Act' : "") : "";
4798
- const prompt = VOICE_SYSTEM_PROMPT.replace("{{MEMORY_SLOT}}", memSlot).replace("{{THINK_SLOT}}", thinkSlot).replace("{{WORKER_WEB}}", workerWeb + workerMcp) + (o.voiceStyle === "conversational" ? "\n" + VOICE_STYLE_CONVERSATIONAL : "") + `
4986
+ const prompt = VOICE_SYSTEM_PROMPT.replace("{{MEMORY_SLOT}}", memSlot).replace("{{THINK_SLOT}}", thinkSlot).replace("{{WORKER_WEB}}", workerWeb + workerMcp) + (o.voiceStyle === "conversational" ? "\n" + VOICE_STYLE_CONVERSATIONAL : "") + (o.emotionTags ? "\n" + EMOTION_TAGS_GUIDANCE : "") + `
4799
4987
  Today's date: ${(/* @__PURE__ */ new Date()).toDateString()}.`;
4800
4988
  const tools = [
4801
4989
  ...o.reflexOptions?.tools ?? [],
@@ -4813,13 +5001,14 @@ Today's date: ${(/* @__PURE__ */ new Date()).toDateString()}.`;
4813
5001
  confirm: host.confirm ? (p, m) => host.confirm(p, m) : void 0,
4814
5002
  notify: (ev) => {
4815
5003
  if (ev?.kind === "text_delta" && typeof ev.message === "string") {
5004
+ if (this.heldThisTurn) return;
4816
5005
  if (this.fabricationCut) return;
4817
5006
  const msg = ev.message;
4818
5007
  this.reflexBuf += msg;
4819
5008
  const m = this.reflexBuf.match(RESERVED_EVENT_MARKER) ?? this.reflexBuf.match(RESERVED_EVENT_OPENER);
4820
5009
  if (m) {
4821
5010
  this.fabricationCut = true;
4822
- log8.warn(`reflex fabricated a [task \u2026] event in its spoken stream \u2014 cutting it (kept ${m.index} chars)`);
5011
+ log9.warn(`reflex fabricated a [task \u2026] event in its spoken stream \u2014 cutting it (kept ${m.index} chars)`);
4823
5012
  const safe = this.reflexBuf.slice(this.reflexForwarded, m.index);
4824
5013
  if (!safe) return;
4825
5014
  if (safe.trim()) this.spokeThisTurn = true;
@@ -4882,6 +5071,7 @@ Today's date: ${(/* @__PURE__ */ new Date()).toDateString()}.`;
4882
5071
  this.turnDispatched = false;
4883
5072
  this.turnBriefs.clear();
4884
5073
  this.spokeThisTurn = false;
5074
+ this.heldThisTurn = false;
4885
5075
  this.reflexBuf = "";
4886
5076
  this.reflexForwarded = 0;
4887
5077
  this.fabricationCut = false;
@@ -4910,7 +5100,7 @@ Today's date: ${(/* @__PURE__ */ new Date()).toDateString()}.`;
4910
5100
  * voice) and emits an empty `final`, so no text_delta ever streams. Both ship silence; both repair.
4911
5101
  * Requires a host: without one there's no stream to detect speech on (and no one to speak to). */
4912
5102
  get silentTurn() {
4913
- return !!this.options.host && !this.spokeThisTurn;
5103
+ return !!this.options.host && !this.spokeThisTurn && !this.heldThisTurn;
4914
5104
  }
4915
5105
  /** A turn that voiced nothing is dead air. Re-prompt the reflex ONCE so the LLM itself voices a short
4916
5106
  * line (no template). If it STILL says nothing, fall back to a minimal line so silence never ships.
@@ -4921,7 +5111,7 @@ Today's date: ${(/* @__PURE__ */ new Date()).toDateString()}.`;
4921
5111
  try {
4922
5112
  await this.voice.send(fallback ? "[reminder] You said nothing to the user this turn. Tell them, in ONE short spoken sentence, what just happened \u2014 no tools." : dispatched ? "[reminder] You dispatched a task but said nothing to the user. Say ONE short spoken acknowledgement now \u2014 no tools." : "[reminder] You said nothing to the user this turn. Give your ONE short spoken reply now \u2014 no tools.");
4923
5113
  } catch (e) {
4924
- log8.warn(`ack nudge failed: ${e instanceof Error ? e.message : e}`);
5114
+ log9.warn(`ack nudge failed: ${e instanceof Error ? e.message : e}`);
4925
5115
  } finally {
4926
5116
  this.nudging = false;
4927
5117
  }
@@ -5013,7 +5203,7 @@ Today's date: ${(/* @__PURE__ */ new Date()).toDateString()}.`;
5013
5203
  buildBrief(brief, tier = "act", deliver = true) {
5014
5204
  const recent = this.voice.transcript.filter((m) => (m.role === "user" || m.role === "assistant") && contentText(m.content).trim()).slice(-this.options.excerptTurns).map((m) => `${m.role}: ${contentText(m.content)}`).join("\n");
5015
5205
  const verify = tier === "act" ? "\n\nBefore reporting done: re-read what you changed and check it against EVERY requirement above \u2014 fix any gap first. Your report is trusted without review." : "";
5016
- const deliverContract = deliver ? "\n\n## DELIVER (spoken delivery)\nYou are reporting back to a user who is LISTENING. Stream your work normally \u2014 your prose is the written work record and detail, and is NOT spoken. Wrap anything the user should HEAR in <spoken>\u2026</spoken> tags. LEAD WITH the actual content they asked for: if they asked for a specific piece of content \u2014 a value, a name, the actual lines, the writing itself \u2014 that content goes INSIDE the <spoken> tags, not a remark about it. Your FIRST <spoken> segment is substantive \u2014 never a greeting or an acknowledgement (the front-end has already acked; do not double-ack). Keep spoken text concise and natural for the ear: short sentences, no markdown." : "";
5206
+ const deliverContract = deliver ? "\n\n## DELIVER (spoken delivery)\nYou are reporting back to a user who is LISTENING. Stream your work normally \u2014 your prose is the written work record and detail, and is NOT spoken. Wrap anything the user should HEAR in <spoken>\u2026</spoken> tags. LEAD WITH the actual content they asked for: if they asked for a specific piece of content \u2014 a value, a name, the actual lines, the writing itself \u2014 that content goes INSIDE the <spoken> tags, not a remark about it. Your FIRST <spoken> segment is substantive \u2014 never a greeting or an acknowledgement (the front-end has already acked; do not double-ack). Keep spoken text concise and natural for the ear: short sentences, no markdown." + (this.options.emotionTags ? " Inside <spoken>, you may prefix a sentence with an inline [emotion] tag (e.g. [excited], [curious]) to color how it is voiced \u2014 only when it genuinely fits, and vary it; [laughter] gives a natural laugh." : "") : "";
5017
5207
  return (recent ? `${brief}
5018
5208
 
5019
5209
  ## Recent conversation (for context)
@@ -5129,7 +5319,7 @@ Another agent just implemented the above. Independently check the CURRENT state
5129
5319
  this.notify("task_verify", `task ${id}: verifying`, { id });
5130
5320
  const cres = await new Agent(checkerOpts).run(checkBrief);
5131
5321
  if (cres.finishReason !== "stop") {
5132
- log8.warn(`task ${id}: verify inconclusive (${cres.finishReason})`);
5322
+ log9.warn(`task ${id}: verify inconclusive (${cres.finishReason})`);
5133
5323
  this.notify("task_verify", `task ${id}: verify inconclusive (${cres.finishReason})`, { id, finishReason: cres.finishReason });
5134
5324
  }
5135
5325
  const sum = (a = 0, b = 0) => a + b;
@@ -5265,7 +5455,7 @@ Another agent just implemented the above. Independently check the CURRENT state
5265
5455
  rec.status = "done";
5266
5456
  rec.result = res.text;
5267
5457
  const incomplete = res.finishReason !== "stop";
5268
- log8.verbose(`task ${id} done (${res.steps} steps${incomplete ? `, INCOMPLETE: ${res.finishReason}` : ""})`);
5458
+ log9.verbose(`task ${id} done (${res.steps} steps${incomplete ? `, INCOMPLETE: ${res.finishReason}` : ""})`);
5269
5459
  this.notify("task_done", `task ${id} (${rec.label}) completed`, {
5270
5460
  id,
5271
5461
  text: res.text,
@@ -5291,7 +5481,7 @@ Another agent just implemented the above. Independently check the CURRENT state
5291
5481
  this.dropAsk(rec.id);
5292
5482
  rec.status = "error";
5293
5483
  rec.result = msg;
5294
- log8.warn(`task ${rec.id} failed: ${msg}`);
5484
+ log9.warn(`task ${rec.id} failed: ${msg}`);
5295
5485
  this.notify("task_error", `task ${rec.id} (${rec.label}) failed: ${msg}`);
5296
5486
  this.queueRevoice(this.integrationPrompt(rec, "error", msg, "error"), true);
5297
5487
  }
@@ -5471,6 +5661,7 @@ Another agent just implemented the above. Independently check the CURRENT state
5471
5661
  }
5472
5662
  },
5473
5663
  run: async ({ filler }) => {
5664
+ this.heldThisTurn = true;
5474
5665
  if (filler) this.notify("hold_filler", String(filler));
5475
5666
  return "Holding \u2014 listening for the rest of the user's thought. Do not respond further this turn.";
5476
5667
  }
@@ -5596,7 +5787,7 @@ init_logging();
5596
5787
 
5597
5788
  // src/voice/engine.ts
5598
5789
  init_logging();
5599
- var log9 = forComponent("VoiceEngine");
5790
+ var log10 = forComponent("VoiceEngine");
5600
5791
  var now = () => performance.now();
5601
5792
  var forSpeech = (t) => t.replace(/[*_`#]+/g, "").replace(/^[ \t]*[-•]\s+/gm, "").replace(/\s*[\u2013\u2014]\s*/g, ", ").replace(/[\u2010\u2011]/g, "-").replace(/\s*\|\s*/g, ", ").replace(/(\d)\s+%/g, "$1%").replace(/\.{3,}/g, ".");
5602
5793
  var VoiceEngineOptions = class {
@@ -5664,6 +5855,11 @@ var VoiceEngineOptions = class {
5664
5855
  * speech at all, an audible hiccup. Default OFF: the genuine-gated STT partial is the
5665
5856
  * mechanism-correct pause trigger; enable only if barge-in onset feels sluggish in a clean-AEC room. */
5666
5857
  overlapEnergyHold = false;
5858
+ /** Map inline `[emotion]` tags (emitted by the model, prompt-taught) into Cartesia inline emotion
5859
+ * tags in the spoken transcript (sonic-3 stitches the prosody). false = strip them silently. */
5860
+ emotions = true;
5861
+ /** Show the `[emotion]` tags in the on-screen echo (debug). false = hide (spoken-only). */
5862
+ showEmotions = false;
5667
5863
  };
5668
5864
  var VoiceEngine = class _VoiceEngine {
5669
5865
  options;
@@ -5709,6 +5905,9 @@ var VoiceEngine = class _VoiceEngine {
5709
5905
  // Central speech queue (above the TTS context): complete worker utterances serialize into ONE
5710
5906
  // playback stream, one-at-a-time, never splicing into the live reflex's open utterance.
5711
5907
  uttQueue = [];
5908
+ // Per-turn emotion-tag parser (reset on beginSpeech) — converts `[emotion]` → Cartesia inline tags
5909
+ // for TTS, tracks tag-free prose for echo discrimination, and surfaces display text for the screen.
5910
+ emo = null;
5712
5911
  constructor(options) {
5713
5912
  this.options = { ...new VoiceEngineOptions(), ...options };
5714
5913
  const o = this.options;
@@ -5726,7 +5925,7 @@ var VoiceEngine = class _VoiceEngine {
5726
5925
  this.stt.onLevel = (rms) => this.handleLevel(rms);
5727
5926
  await Promise.all([this.tts.connect(), this.stt.start()]);
5728
5927
  this.setState("listening");
5729
- log9.debug(`voice I/O up (${this.stt.usingAec ? "AEC" : "heuristic echo"} capture)`);
5928
+ log10.debug(`voice I/O up (${this.stt.usingAec ? "AEC" : "heuristic echo"} capture)`);
5730
5929
  }
5731
5930
  get usingAec() {
5732
5931
  return this.stt.usingAec;
@@ -5735,6 +5934,10 @@ var VoiceEngine = class _VoiceEngine {
5735
5934
  setBargeIn(on) {
5736
5935
  this.options.bargeIn = on;
5737
5936
  }
5937
+ /** Show/hide the `[emotion]` debug tags in the echo (next turn's stream picks it up). */
5938
+ setShowEmotions(on) {
5939
+ this.options.showEmotions = on;
5940
+ }
5738
5941
  idleWaiters = [];
5739
5942
  setState(s) {
5740
5943
  if (this.state === s) return;
@@ -5766,6 +5969,7 @@ var VoiceEngine = class _VoiceEngine {
5766
5969
  this.ctxOpen = true;
5767
5970
  this.spokeDeltas = false;
5768
5971
  this.reply = "";
5972
+ this.emo = this.options.emotions ? new EmotionStream(this.options.showEmotions) : null;
5769
5973
  this.echoWords = new Set(this.words(this.prevReply));
5770
5974
  this.tts.newContext();
5771
5975
  if (ack && this.options.ackPhrase) {
@@ -5776,21 +5980,31 @@ var VoiceEngine = class _VoiceEngine {
5776
5980
  if (!this.turnStartAt) this.turnStartAt = now();
5777
5981
  this.setState("thinking");
5778
5982
  }
5983
+ /** Feed a spoken delta. Returns the on-screen echo text (emotion tags shown/hidden per config) so the
5984
+ * host renders the SAME stream that was parsed for TTS — no second, state-doubling parse. */
5779
5985
  speakDelta(text) {
5780
- if (this.interrupted) return;
5986
+ if (this.interrupted) return "";
5781
5987
  if (!this.speaking || !this.ctxOpen) this.beginSpeech();
5782
- this.reply += text;
5988
+ const { speech, display, prose } = this.emo ? this.emo.feed(text) : { speech: text, display: text, prose: text };
5989
+ this.reply += prose;
5783
5990
  for (const w of this.words(this.reply)) this.echoWords.add(w);
5784
- this.tts.speak(forSpeech(text), true);
5785
- if (!this.spokeDeltas && this.turnStartAt) log9.debug(`ttft: ${Math.round(now() - this.turnStartAt)}ms`);
5991
+ this.tts.speak(forSpeech(speech), true);
5992
+ if (!this.spokeDeltas && this.turnStartAt) log10.debug(`ttft: ${Math.round(now() - this.turnStartAt)}ms`);
5786
5993
  this.spokeDeltas = true;
5787
5994
  this.setState("speaking");
5995
+ return display;
5788
5996
  }
5789
5997
  /** close the spoken turn (idempotent); stays audible until ALL audio arrived AND playback drains */
5790
5998
  endSpeech() {
5791
5999
  this.interrupted = false;
5792
6000
  if (!this.speaking) return;
5793
6001
  this.ctxOpen = false;
6002
+ if (this.emo) {
6003
+ const t = this.emo.flush();
6004
+ this.emo = null;
6005
+ if (t.prose) this.reply += t.prose;
6006
+ if (t.speech) this.tts.speak(forSpeech(t.speech), true);
6007
+ }
5794
6008
  if (this.reply) this.prevReply = this.reply;
5795
6009
  const settle = () => {
5796
6010
  if (this.ctxOpen) {
@@ -5803,7 +6017,7 @@ var VoiceEngine = class _VoiceEngine {
5803
6017
  }
5804
6018
  this.drainTimer = null;
5805
6019
  this.speaking = false;
5806
- if (this.turnStartAt) log9.debug(`turn: ${Math.round(now() - this.turnStartAt)}ms (incl. playback)`);
6020
+ if (this.turnStartAt) log10.debug(`turn: ${Math.round(now() - this.turnStartAt)}ms (incl. playback)`);
5807
6021
  this.echoUntil = now() + 2500;
5808
6022
  if (!this.usingAec) this.stt.reset();
5809
6023
  this.setState("listening");
@@ -5995,7 +6209,7 @@ var VoiceEngine = class _VoiceEngine {
5995
6209
  this.pendingUtt = this.mergeUtterance(this.pendingUtt, text);
5996
6210
  if (this.pendingTimer) clearTimeout(this.pendingTimer);
5997
6211
  if (this.options.incompleteMergeMs && this.looksIncomplete(this.pendingUtt)) {
5998
- log9.verbose(`hold: incomplete utterance "${this.pendingUtt.slice(-40)}"`);
6212
+ log10.verbose(`hold: incomplete utterance "${this.pendingUtt.slice(-40)}"`);
5999
6213
  this.options.onHold();
6000
6214
  if (this.options.holdFiller && !this.speaking) {
6001
6215
  this.beginSpeech();
@@ -6094,7 +6308,7 @@ async function resolveAuth(auth) {
6094
6308
  }
6095
6309
 
6096
6310
  // src/voice/soniox.ts
6097
- var log10 = forComponent("SonioxSTT");
6311
+ var log11 = forComponent("SonioxSTT");
6098
6312
  var now2 = () => performance.now();
6099
6313
  var SonioxSTTOptions = class {
6100
6314
  auth = "";
@@ -6163,9 +6377,9 @@ var SonioxSTT = class {
6163
6377
  this.ws.onmessage = (ev) => this.handle(JSON.parse(String(ev.data)));
6164
6378
  this.ws.onclose = (ev) => {
6165
6379
  if (this.stopped) return;
6166
- log10.warn(`soniox ws closed (${ev.code} ${ev.reason || ""}) \u2014 reconnecting`);
6380
+ log11.warn(`soniox ws closed (${ev.code} ${ev.reason || ""}) \u2014 reconnecting`);
6167
6381
  this.reset();
6168
- this.connectWs().catch((e) => log10.error(`soniox reconnect failed: ${e.message}`));
6382
+ this.connectWs().catch((e) => log11.error(`soniox reconnect failed: ${e.message}`));
6169
6383
  };
6170
6384
  }
6171
6385
  async start() {
@@ -6175,7 +6389,7 @@ var SonioxSTT = class {
6175
6389
  this.endpointTimer = setInterval(() => {
6176
6390
  const combined = (this.finalText + this.partialText).trim();
6177
6391
  if (!combined || now2() - this.lastChangeAt < this.options.silenceEndpointMs) return;
6178
- if (this.firstTokenAt) log10.debug(`stt: ${Math.round(now2() - this.firstTokenAt)}ms first-token\u2192silence-endpoint, "${combined.slice(0, 60)}"`);
6392
+ if (this.firstTokenAt) log11.debug(`stt: ${Math.round(now2() - this.firstTokenAt)}ms first-token\u2192silence-endpoint, "${combined.slice(0, 60)}"`);
6179
6393
  this.reset();
6180
6394
  this.onUtterance(combined, now2());
6181
6395
  }, 120);
@@ -6187,7 +6401,7 @@ var SonioxSTT = class {
6187
6401
  if (this.stopped) return;
6188
6402
  const ref = this.lastChunkAt || this.startedChunksAt;
6189
6403
  if (now2() - ref > noAudioMs) {
6190
- log10.error(`stt: no mic audio for >${Math.round(noAudioMs / 1e3)}s \u2014 capture device stopped delivering`);
6404
+ log11.error(`stt: no mic audio for >${Math.round(noAudioMs / 1e3)}s \u2014 capture device stopped delivering`);
6191
6405
  this.onFatal("microphone stopped delivering audio (try a different input device, e.g. AirPods, or check System Settings \u2192 Sound \u2192 Input)");
6192
6406
  this.stop();
6193
6407
  }
@@ -6207,7 +6421,7 @@ var SonioxSTT = class {
6207
6421
  });
6208
6422
  }
6209
6423
  handle(m) {
6210
- if (m.error_message) return log10.error(`soniox: ${m.error_message}`);
6424
+ if (m.error_message) return log11.error(`soniox: ${m.error_message}`);
6211
6425
  let endpoint = false;
6212
6426
  for (const t of m.tokens ?? []) {
6213
6427
  if (t.text === "<end>") endpoint = true;
@@ -6223,7 +6437,7 @@ var SonioxSTT = class {
6223
6437
  this.onPartial(combined);
6224
6438
  if (endpoint && this.finalText.trim()) {
6225
6439
  const utterance = this.finalText.trim();
6226
- if (this.firstTokenAt) log10.debug(`stt: ${Math.round(now2() - this.firstTokenAt)}ms first-token\u2192endpoint, "${utterance.slice(0, 60)}"`);
6440
+ if (this.firstTokenAt) log11.debug(`stt: ${Math.round(now2() - this.firstTokenAt)}ms first-token\u2192endpoint, "${utterance.slice(0, 60)}"`);
6227
6441
  this.reset();
6228
6442
  this.onUtterance(utterance, now2());
6229
6443
  }
@@ -6246,7 +6460,7 @@ var SonioxSTT = class {
6246
6460
 
6247
6461
  // src/voice/cartesia.ts
6248
6462
  init_logging();
6249
- var log11 = forComponent("CartesiaTTS");
6463
+ var log12 = forComponent("CartesiaTTS");
6250
6464
  var now3 = () => performance.now();
6251
6465
  var CartesiaTTSOptions = class {
6252
6466
  auth = "";
@@ -6296,9 +6510,9 @@ var CartesiaTTS = class _CartesiaTTS {
6296
6510
  this.ws.onerror = (e) => rej(new Error(`cartesia ws: ${e.message || "connect failed"}`));
6297
6511
  });
6298
6512
  this.ws.onclose = (ev) => {
6299
- log11.warn(`cartesia ws closed (${ev.code} ${ev.reason || ""})`);
6513
+ log12.warn(`cartesia ws closed (${ev.code} ${ev.reason || ""})`);
6300
6514
  if (!this.closed) {
6301
- this.connecting = this.doConnect().catch((e) => log11.error(`cartesia reconnect failed: ${e.message}`));
6515
+ this.connecting = this.doConnect().catch((e) => log12.error(`cartesia reconnect failed: ${e.message}`));
6302
6516
  }
6303
6517
  };
6304
6518
  this.ws.onmessage = (ev) => {
@@ -6320,11 +6534,11 @@ var CartesiaTTS = class _CartesiaTTS {
6320
6534
  this.down = true;
6321
6535
  this.downAt = now3();
6322
6536
  this.consecutiveOk = 0;
6323
- log11.warn(`TTS circuit breaker open \u2014 ${this.consecutiveErrors} consecutive errors, switching to text-only`);
6537
+ log12.warn(`TTS circuit breaker open \u2014 ${this.consecutiveErrors} consecutive errors, switching to text-only`);
6324
6538
  this.onDone();
6325
6539
  this.startProbe();
6326
6540
  } else if (!this.down) {
6327
- log11.warn(`cartesia: ${JSON.stringify(m)}`);
6541
+ log12.warn(`cartesia: ${JSON.stringify(m)}`);
6328
6542
  }
6329
6543
  }
6330
6544
  };
@@ -6338,7 +6552,7 @@ var CartesiaTTS = class _CartesiaTTS {
6338
6552
  this.consecutiveOk = 0;
6339
6553
  this.stopProbe();
6340
6554
  const downMs = this.downAt ? now3() - this.downAt : 0;
6341
- (downMs < 2e3 ? log11.debug : log11.info)(`TTS recovered${downMs ? ` (down ${downMs}ms)` : ""}`);
6555
+ (downMs < 2e3 ? log12.debug : log12.info)(`TTS recovered${downMs ? ` (down ${downMs}ms)` : ""}`);
6342
6556
  }
6343
6557
  /** Ensure the WS is open before sending — reconnects if idle-closed. */
6344
6558
  async ensureConnected() {
@@ -6418,7 +6632,7 @@ import { MemFilesystem as MemFilesystem3, IndexedDbFilesystem, CommandExecutor a
6418
6632
  init_logging();
6419
6633
  import { spawn } from "child_process";
6420
6634
  import { createHash } from "crypto";
6421
- var log12 = forComponent("mcp");
6635
+ var log13 = forComponent("mcp");
6422
6636
  var PROTOCOL_VERSION = "2025-06-18";
6423
6637
  var DEFAULT_TIMEOUT_MS = 3e4;
6424
6638
  var StdioTransport = class {
@@ -6437,7 +6651,7 @@ var StdioTransport = class {
6437
6651
  proc.stdout.setEncoding("utf8");
6438
6652
  proc.stdout.on("data", (chunk) => this.onData(chunk));
6439
6653
  proc.stderr.setEncoding("utf8");
6440
- proc.stderr.on("data", (chunk) => log12.debug(`[${command}] stderr:`, chunk.trimEnd()));
6654
+ proc.stderr.on("data", (chunk) => log13.debug(`[${command}] stderr:`, chunk.trimEnd()));
6441
6655
  proc.on("exit", (code) => this.failAll(new Error(`MCP server "${command}" exited (code ${code})`)));
6442
6656
  proc.on("error", (e) => this.failAll(e instanceof Error ? e : new Error(String(e))));
6443
6657
  }
@@ -6451,7 +6665,7 @@ var StdioTransport = class {
6451
6665
  try {
6452
6666
  this.dispatch(JSON.parse(line));
6453
6667
  } catch (e) {
6454
- log12.debug("dropping non-JSON line from MCP server:", line, e);
6668
+ log13.debug("dropping non-JSON line from MCP server:", line, e);
6455
6669
  }
6456
6670
  }
6457
6671
  }
@@ -6500,7 +6714,7 @@ var StdioTransport = class {
6500
6714
  try {
6501
6715
  this.proc?.stdin?.end();
6502
6716
  } catch (e) {
6503
- log12.debug("stdin end failed", e);
6717
+ log13.debug("stdin end failed", e);
6504
6718
  }
6505
6719
  this.proc?.kill();
6506
6720
  }
@@ -6569,7 +6783,7 @@ function parseSseResponse(body) {
6569
6783
  const obj = JSON.parse(trimmed.slice(5).trim());
6570
6784
  if (obj && (obj.result !== void 0 || obj.error !== void 0)) return obj;
6571
6785
  } catch (e) {
6572
- log12.debug("skipping unparseable SSE data line", e);
6786
+ log13.debug("skipping unparseable SSE data line", e);
6573
6787
  }
6574
6788
  }
6575
6789
  return {};
@@ -6637,7 +6851,7 @@ async function mountWithDeadline(name, cfg, mountTimeoutMs) {
6637
6851
  return { name, client, tools, specs, serverInfo: init?.serverInfo, config: cfg };
6638
6852
  })(), mountTimeoutMs, name);
6639
6853
  } catch (e) {
6640
- await client.close().catch((err2) => log12.debug(`close after failed mount of "${name}": ${err2}`));
6854
+ await client.close().catch((err2) => log13.debug(`close after failed mount of "${name}": ${err2}`));
6641
6855
  throw e;
6642
6856
  }
6643
6857
  }
@@ -6648,15 +6862,15 @@ function validEntries(servers) {
6648
6862
  return Object.entries(servers).filter(([name, cfg]) => {
6649
6863
  if (!cfg || cfg.disabled) return false;
6650
6864
  if (!cfg.command && !cfg.url) {
6651
- log12.warn(`MCP server "${name}" needs a command (stdio) or url (http) \u2014 skipping`);
6865
+ log13.warn(`MCP server "${name}" needs a command (stdio) or url (http) \u2014 skipping`);
6652
6866
  return false;
6653
6867
  }
6654
6868
  return true;
6655
6869
  });
6656
6870
  }
6657
6871
  function logMountFailure(name, e) {
6658
- if (e instanceof McpAuthError) log12.warn(`MCP "${name}" needs-auth: HTTP ${e.status} \u2014 set bearerToken or headers in its config; skipping`);
6659
- else log12.error(`MCP server "${name}" failed to mount: ${e?.message ?? e}`);
6872
+ if (e instanceof McpAuthError) log13.warn(`MCP "${name}" needs-auth: HTTP ${e.status} \u2014 set bearerToken or headers in its config; skipping`);
6873
+ else log13.error(`MCP server "${name}" failed to mount: ${e?.message ?? e}`);
6660
6874
  }
6661
6875
  async function mountMcpServers(servers = {}, opts = {}) {
6662
6876
  const entries = validEntries(servers);
@@ -6666,7 +6880,7 @@ async function mountMcpServers(servers = {}, opts = {}) {
6666
6880
  const name = entries[i][0];
6667
6881
  if (r.status === "fulfilled") {
6668
6882
  out.push(r.value);
6669
- log12.debug(`MCP "${name}" mounted \u2014 ${r.value.tools.length} tool(s)${r.value.serverInfo?.name ? ` from ${r.value.serverInfo.name}` : ""}`);
6883
+ log13.debug(`MCP "${name}" mounted \u2014 ${r.value.tools.length} tool(s)${r.value.serverInfo?.name ? ` from ${r.value.serverInfo.name}` : ""}`);
6670
6884
  } else logMountFailure(name, r.reason);
6671
6885
  });
6672
6886
  return out;
@@ -6706,7 +6920,7 @@ var McpPool = class {
6706
6920
  const prev = this.warm.get(key);
6707
6921
  if (prev) {
6708
6922
  clearTimeout(prev.timer);
6709
- if (prev.client !== client) void prev.client.close().catch((err2) => log12.debug(`warm-pool replace close failed: ${err2}`));
6923
+ if (prev.client !== client) void prev.client.close().catch((err2) => log13.debug(`warm-pool replace close failed: ${err2}`));
6710
6924
  }
6711
6925
  const e = { client, timer: void 0 };
6712
6926
  this.warm.set(key, e);
@@ -6723,7 +6937,7 @@ var McpPool = class {
6723
6937
  const e = this.warm.get(key);
6724
6938
  if (!e) return;
6725
6939
  this.warm.delete(key);
6726
- await e.client.close().catch((err2) => log12.debug(`warm-pool evict close failed: ${err2}`));
6940
+ await e.client.close().catch((err2) => log13.debug(`warm-pool evict close failed: ${err2}`));
6727
6941
  }
6728
6942
  async closeAll() {
6729
6943
  for (const e of this.warm.values()) {
@@ -6941,7 +7155,7 @@ init_tools_shell();
6941
7155
  // src/tools.notify.ts
6942
7156
  init_logging();
6943
7157
  import { execFile } from "child_process";
6944
- var log14 = forComponent("notify");
7158
+ var log15 = forComponent("notify");
6945
7159
  function makeNotifyTool(opts = {}) {
6946
7160
  const platform2 = opts.platform ?? process.platform;
6947
7161
  const run = opts.exec ?? execFile;
@@ -6965,7 +7179,7 @@ function makeNotifyTool(opts = {}) {
6965
7179
  return new Promise((resolve4) => {
6966
7180
  run(argv[0], argv[1], { timeout: 5e3 }, (e) => {
6967
7181
  if (e) {
6968
- log14.debug("notification failed", e);
7182
+ log15.debug("notification failed", e);
6969
7183
  resolve4(`Notification failed: ${e.message}`);
6970
7184
  } else resolve4("Notification shown.");
6971
7185
  });
@@ -6981,7 +7195,7 @@ import { BodDB as BodDB2 } from "@bod.ee/db";
6981
7195
  init_logging();
6982
7196
  import { existsSync as existsSync2, readFileSync as readFileSync3 } from "fs";
6983
7197
  import { homedir } from "os";
6984
- var log15 = forComponent("cli-util");
7198
+ var log16 = forComponent("cli-util");
6985
7199
  function dotDirs(base, sub, opts = {}) {
6986
7200
  const home = opts.home ?? homedir();
6987
7201
  const dirs = [`${base}/.agent/${sub}`, `${base}/.claude/${sub}`, `${home}/.agent/${sub}`, `${home}/.claude/${sub}`];
@@ -6998,7 +7212,7 @@ function parseJson(text, fallback, what = "json") {
6998
7212
  try {
6999
7213
  return JSON.parse(text);
7000
7214
  } catch (e) {
7001
- log15.debug(`parseJson(${what}) failed: ${e.message}`);
7215
+ log16.debug(`parseJson(${what}) failed: ${e.message}`);
7002
7216
  return fallback;
7003
7217
  }
7004
7218
  }
@@ -7008,7 +7222,7 @@ function readJsonFile(path, fallback) {
7008
7222
  try {
7009
7223
  text = readFileSync3(path, "utf8");
7010
7224
  } catch (e) {
7011
- log15.debug(`readJsonFile(${path}) unreadable: ${e.message}`);
7225
+ log16.debug(`readJsonFile(${path}) unreadable: ${e.message}`);
7012
7226
  return fallback;
7013
7227
  }
7014
7228
  return parseJson(text, fallback, path);
@@ -7261,7 +7475,7 @@ import { existsSync as existsSync4, mkdirSync as mkdirSync5, readFileSync as rea
7261
7475
  import { homedir as homedir3 } from "os";
7262
7476
  import { dirname as dirname3, join as join6 } from "path";
7263
7477
  import { fileURLToPath } from "url";
7264
- var log16 = forComponent("VoiceIO");
7478
+ var log17 = forComponent("VoiceIO");
7265
7479
  var now4 = () => performance.now();
7266
7480
  var Player = class {
7267
7481
  proc = null;
@@ -7275,7 +7489,7 @@ var Player = class {
7275
7489
  ["-loglevel", "quiet", "-nodisp", "-fflags", "nobuffer", "-flags", "low_delay", "-probesize", "32", "-f", "s16le", "-ar", String(TTS_SAMPLE_RATE), "-ch_layout", "mono", "-i", "-"],
7276
7490
  { stdio: ["pipe", "ignore", "ignore"] }
7277
7491
  );
7278
- this.proc.on("error", (e) => log16.warn(`ffplay error: ${e.message}`));
7492
+ this.proc.on("error", (e) => log17.warn(`ffplay error: ${e.message}`));
7279
7493
  this.proc.stdin.on("error", () => {
7280
7494
  });
7281
7495
  this.bytesWritten = 0;
@@ -7313,7 +7527,7 @@ function detectFfmpegMic() {
7313
7527
  const devices = [...audio.matchAll(/\[(\d+)\] (.+)/g)].map(([, idx, name]) => ({ idx, name: name.trim() }));
7314
7528
  const mic = devices.find((d) => /microphone|built-in/i.test(d.name) && !/teams|blackhole|loopback/i.test(d.name)) ?? devices[0];
7315
7529
  if (!mic) throw new Error("no audio input device found");
7316
- log16.debug(`ffmpeg mic: [${mic.idx}] ${mic.name}`);
7530
+ log17.debug(`ffmpeg mic: [${mic.idx}] ${mic.name}`);
7317
7531
  return `:${mic.idx}`;
7318
7532
  }
7319
7533
  function detectedInputDevice() {
@@ -7349,15 +7563,15 @@ function resolveAecBinary() {
7349
7563
  if (existsSync4(bin) && statSync3(bin).mtimeMs >= statSync3(src).mtimeMs) return bin;
7350
7564
  if (spawnSync2("which", ["swiftc"]).status !== 0) return null;
7351
7565
  mkdirSync5(cacheDir, { recursive: true });
7352
- log16.info("compiling AEC mic helper (first run)\u2026");
7566
+ log17.info("compiling AEC mic helper (first run)\u2026");
7353
7567
  const build = spawnSync2("swiftc", ["-O", "-o", bin, src, "-Xlinker", "-sectcreate", "-Xlinker", "__TEXT", "-Xlinker", "__info_plist", "-Xlinker", plist], { encoding: "utf8" });
7354
7568
  if (build.status !== 0) {
7355
- log16.warn(`AEC build failed: ${build.stderr?.slice(0, 400)}`);
7569
+ log17.warn(`AEC build failed: ${build.stderr?.slice(0, 400)}`);
7356
7570
  return null;
7357
7571
  }
7358
7572
  const sign = spawnSync2("codesign", ["-fs", "-", bin], { encoding: "utf8" });
7359
7573
  if (sign.status !== 0) {
7360
- log16.warn(`codesign failed: ${sign.stderr?.slice(0, 200)}`);
7574
+ log17.warn(`codesign failed: ${sign.stderr?.slice(0, 200)}`);
7361
7575
  return null;
7362
7576
  }
7363
7577
  return bin;
@@ -7399,16 +7613,16 @@ var NodeMicSource = class {
7399
7613
  this.proc = spawn2(this.bin, [], { stdio: ["ignore", "pipe", "ignore"] });
7400
7614
  } else {
7401
7615
  if (spawnSync2("which", ["ffmpeg"]).status !== 0) throw new Error("voice I/O unavailable: no AEC helper and no ffmpeg on PATH");
7402
- log16.info("mic: raw capture (no AEC) \u2014 echo handled heuristically; headphones recommended");
7616
+ log17.info("mic: raw capture (no AEC) \u2014 echo handled heuristically; headphones recommended");
7403
7617
  this.proc = spawn2(
7404
7618
  "ffmpeg",
7405
7619
  ["-loglevel", "error", "-f", "avfoundation", "-i", detectFfmpegMic(), "-ar", String(STT_SAMPLE_RATE), "-ac", "1", "-f", "s16le", "-"],
7406
7620
  { stdio: ["ignore", "pipe", "pipe"] }
7407
7621
  );
7408
- this.proc.stderr.on("data", (d) => log16.warn(`ffmpeg: ${String(d).trim()}`));
7622
+ this.proc.stderr.on("data", (d) => log17.warn(`ffmpeg: ${String(d).trim()}`));
7409
7623
  }
7410
7624
  this.proc.on("exit", (c) => {
7411
- if (c && !this.stopped) log16.error(`mic capture exited (${c}) \u2014 check mic permission / MIC_DEVICE / MIC_AEC=0`);
7625
+ if (c && !this.stopped) log17.error(`mic capture exited (${c}) \u2014 check mic permission / MIC_DEVICE / MIC_AEC=0`);
7412
7626
  });
7413
7627
  this.proc.stdout.on("data", (chunk) => onChunk(chunk));
7414
7628
  }
@@ -7474,7 +7688,7 @@ var AecDuplexAudio = class {
7474
7688
  this.proc.stdin.on("error", () => {
7475
7689
  });
7476
7690
  this.proc.on("exit", (c) => {
7477
- if (c && !this.stopped) log16.error(`aec duplex audio exited (${c}) \u2014 check mic permission / MIC_AEC=0`);
7691
+ if (c && !this.stopped) log17.error(`aec duplex audio exited (${c}) \u2014 check mic permission / MIC_AEC=0`);
7478
7692
  });
7479
7693
  this.proc.stdout.on("data", (chunk) => {
7480
7694
  this.gotChunk = true;
@@ -7490,7 +7704,7 @@ var AecDuplexAudio = class {
7490
7704
  openMicSettings();
7491
7705
  this.onFatal?.("microphone permission denied \u2014 enable it in System Settings \u2192 Privacy & Security \u2192 Microphone for your terminal, then restart it");
7492
7706
  }
7493
- } else log16.debug(`mic-aec: ${s}`);
7707
+ } else log17.debug(`mic-aec: ${s}`);
7494
7708
  }
7495
7709
  });
7496
7710
  if (!this.noVpio && !this.triedFallback) {
@@ -7499,7 +7713,7 @@ var AecDuplexAudio = class {
7499
7713
  this.triedFallback = true;
7500
7714
  this.noVpio = true;
7501
7715
  this._aec = false;
7502
- log16.warn("mic-aec: VPIO delivered no audio in 2.5s \u2014 falling back to non-VPIO capture (no AEC \u2192 half-duplex, no barge-in)");
7716
+ log17.warn("mic-aec: VPIO delivered no audio in 2.5s \u2014 falling back to non-VPIO capture (no AEC \u2192 half-duplex, no barge-in)");
7503
7717
  this.onDegrade?.();
7504
7718
  this.killProc();
7505
7719
  this.spawnHelper();
@@ -7607,6 +7821,10 @@ var VoiceIOOptions = class extends VoiceEngineOptions {
7607
7821
  sonioxApiKey = process.env.SONIOX_API_KEY ?? "";
7608
7822
  cartesiaApiKey = process.env.CARTESIA_API_KEY ?? "";
7609
7823
  cartesiaVoiceId = process.env.CARTESIA_VOICE_ID ?? "";
7824
+ emotions = process.env.VOICE_EMOTIONS !== "0";
7825
+ // Cartesia inline emotion tags (sonic-3)
7826
+ showEmotions = process.env.VOICE_SHOW_EMOTIONS === "1";
7827
+ // surface the tags in the on-screen echo (debug; opt-in)
7610
7828
  };
7611
7829
  var VoiceIO = class extends VoiceEngine {
7612
7830
  duplexSource;
@@ -7750,7 +7968,7 @@ async function loadConfig(cwd) {
7750
7968
 
7751
7969
  // cli/hooks-config.ts
7752
7970
  import { spawnSync as spawnSync3 } from "child_process";
7753
- var log17 = forComponent("hooks");
7971
+ var log18 = forComponent("hooks");
7754
7972
  var escapeRegex = (s) => s.replace(/[.+^${}()|[\]\\]/g, "\\$&");
7755
7973
  function ruleMatches(rule, toolName) {
7756
7974
  if (!rule.tool || rule.tool === "*") return true;
@@ -7767,7 +7985,7 @@ function runCmd(rule, env) {
7767
7985
  });
7768
7986
  return { code: r.status ?? 1, out: ((r.stdout ?? "") + (r.stderr ?? "")).trim() };
7769
7987
  } catch (e) {
7770
- log17.debug(`hook command failed: ${rule.command}`, e);
7988
+ log18.debug(`hook command failed: ${rule.command}`, e);
7771
7989
  return { code: 1, out: String(e?.message ?? e) };
7772
7990
  }
7773
7991
  }
@@ -7874,7 +8092,7 @@ function formatDiff(ops, opts = {}) {
7874
8092
  import { existsSync as existsSync6, mkdirSync as mkdirSync6, readFileSync as readFileSync6, writeFileSync as writeFileSync4, readdirSync, renameSync, symlinkSync as symlinkSync2, unlinkSync, readlinkSync } from "fs";
7875
8093
  import { homedir as homedir5 } from "os";
7876
8094
  import { join as join8 } from "path";
7877
- var log18 = forComponent("session");
8095
+ var log19 = forComponent("session");
7878
8096
  var globalDir = () => join8(homedir5(), ".agent", "sessions");
7879
8097
  var SessionStore = class {
7880
8098
  dir;
@@ -7919,7 +8137,7 @@ var SessionStore = class {
7919
8137
  }
7920
8138
  load(id) {
7921
8139
  if (!this.safeId(id)) {
7922
- log18.debug(`rejecting unsafe session id: ${id}`);
8140
+ log19.debug(`rejecting unsafe session id: ${id}`);
7923
8141
  return void 0;
7924
8142
  }
7925
8143
  const path = join8(this.dir, `${id}.json`);
@@ -7927,7 +8145,7 @@ var SessionStore = class {
7927
8145
  try {
7928
8146
  return JSON.parse(readFileSync6(path, "utf8"));
7929
8147
  } catch (e) {
7930
- log18.debug(`unreadable session ${id} \u2014 ignoring`, e);
8148
+ log19.debug(`unreadable session ${id} \u2014 ignoring`, e);
7931
8149
  return void 0;
7932
8150
  }
7933
8151
  }
@@ -7940,7 +8158,7 @@ var SessionStore = class {
7940
8158
  try {
7941
8159
  metas.push(JSON.parse(readFileSync6(join8(this.dir, f), "utf8")).meta);
7942
8160
  } catch (e) {
7943
- log18.debug(`skipping unreadable session file ${f}`, e);
8161
+ log19.debug(`skipping unreadable session file ${f}`, e);
7944
8162
  }
7945
8163
  }
7946
8164
  return metas.sort((a, b) => b.updated - a.updated);
@@ -8108,7 +8326,7 @@ import { execFile as execFile3 } from "child_process";
8108
8326
  import { promisify } from "util";
8109
8327
  import { writeFileSync as writeFileSync5, mkdirSync as mkdirSync7 } from "fs";
8110
8328
  import { join as join9, resolve as resolve2, sep as sep2 } from "path";
8111
- var log19 = forComponent("checkpoints");
8329
+ var log20 = forComponent("checkpoints");
8112
8330
  var exec = promisify(execFile3);
8113
8331
  var DEFAULT_EXCLUDE = [".agent/", ".git/", "node_modules/", "dist/", "build/", ".next/", "target/", ".venv/", "__pycache__/", "*.log"];
8114
8332
  var ShadowRepo = class {
@@ -8146,7 +8364,7 @@ var ShadowRepo = class {
8146
8364
  writeFileSync5(join9(this.gitDir, "info", "exclude"), this.exclude.join("\n") + "\n");
8147
8365
  this.ready = true;
8148
8366
  } catch (e) {
8149
- log19.debug(`git checkpoints unavailable for ${this.workTree}`, e);
8367
+ log20.debug(`git checkpoints unavailable for ${this.workTree}`, e);
8150
8368
  this.ready = false;
8151
8369
  }
8152
8370
  return this.ready;
@@ -8157,7 +8375,7 @@ var ShadowRepo = class {
8157
8375
  }
8158
8376
  async commit(label, forced = []) {
8159
8377
  await this.run("add", "-A");
8160
- for (const p of forced) await this.run("add", "-f", "--", p).catch((e) => log19.debug(`force-add failed: ${p}`, e));
8378
+ for (const p of forced) await this.run("add", "-f", "--", p).catch((e) => log20.debug(`force-add failed: ${p}`, e));
8161
8379
  await this.run("commit", "--allow-empty", "-q", "-m", label);
8162
8380
  }
8163
8381
  /** Inject the CURRENT (pre-edit) content of `paths` into the turn-open restore point by amending it.
@@ -8165,8 +8383,8 @@ var ShadowRepo = class {
8165
8383
  * turn-boundary `add -A` would never have captured it. Amend (vs a new commit) keeps one restore
8166
8384
  * point per turn, so the REPL's turn↔frame mapping stays intact. */
8167
8385
  async amendForced(paths) {
8168
- for (const p of paths) await this.run("add", "-f", "--", p).catch((e) => log19.debug(`force-capture failed: ${p}`, e));
8169
- await this.run("commit", "--amend", "--no-edit", "-q", "--allow-empty").catch((e) => log19.debug("amend failed", e));
8386
+ for (const p of paths) await this.run("add", "-f", "--", p).catch((e) => log20.debug(`force-capture failed: ${p}`, e));
8387
+ await this.run("commit", "--amend", "--no-edit", "-q", "--allow-empty").catch((e) => log20.debug("amend failed", e));
8170
8388
  }
8171
8389
  /** Commits on `ref`, oldest-first (canonical index space). */
8172
8390
  async log(ref) {
@@ -8226,7 +8444,7 @@ var ShadowRepo = class {
8226
8444
  await this.run("gc", "--auto", "-q").catch(() => {
8227
8445
  });
8228
8446
  } catch (e) {
8229
- log19.debug("checkpoint prune failed", e);
8447
+ log20.debug("checkpoint prune failed", e);
8230
8448
  }
8231
8449
  }
8232
8450
  };
@@ -8285,7 +8503,7 @@ var GitCheckpoints = class {
8285
8503
  use(sessionId) {
8286
8504
  if (sessionId === this.session) return;
8287
8505
  this.session = sessionId;
8288
- if (this.started) for (const r of this.repos) void r.point(this.ref()).catch((e) => log19.debug("re-point failed", e));
8506
+ if (this.started) for (const r of this.repos) void r.point(this.ref()).catch((e) => log20.debug("re-point failed", e));
8289
8507
  }
8290
8508
  async begin(label) {
8291
8509
  if (!await this.start()) return;
@@ -8297,7 +8515,7 @@ var GitCheckpoints = class {
8297
8515
  try {
8298
8516
  await this.repos[i].commit(msg, forced);
8299
8517
  } catch (e) {
8300
- log19.debug("checkpoint commit failed", e);
8518
+ log20.debug("checkpoint commit failed", e);
8301
8519
  }
8302
8520
  }
8303
8521
  if (slow) clearTimeout(slow);
@@ -8327,7 +8545,7 @@ var GitCheckpoints = class {
8327
8545
  if (this.forced.has(abs)) continue;
8328
8546
  this.forced.add(abs);
8329
8547
  const i = this.repoIndexFor(abs);
8330
- if (i >= 0) await this.repos[i].amendForced([abs]).catch((e) => log19.debug("amendForced failed", e));
8548
+ if (i >= 0) await this.repos[i].amendForced([abs]).catch((e) => log20.debug("amendForced failed", e));
8331
8549
  }
8332
8550
  }
8333
8551
  };
@@ -10033,7 +10251,7 @@ import { spawnSync as spawnSync5 } from "child_process";
10033
10251
  import { writeFileSync as writeFileSync8, mkdirSync as mkdirSync9, readdirSync as readdirSync2, unlinkSync as unlinkSync3, chmodSync, existsSync as existsSync7 } from "fs";
10034
10252
  import { homedir as homedir7 } from "os";
10035
10253
  import { join as join12 } from "path";
10036
- var log20 = forComponent("os-sched");
10254
+ var log21 = forComponent("os-sched");
10037
10255
  var OsScheduler = class {
10038
10256
  options;
10039
10257
  constructor(options) {
@@ -10095,7 +10313,7 @@ var OsScheduler = class {
10095
10313
  }
10096
10314
  }
10097
10315
  } catch (e) {
10098
- log20.debug(`cancel ${id}`, e);
10316
+ log21.debug(`cancel ${id}`, e);
10099
10317
  }
10100
10318
  for (const f of [`${id}.json`, `${id}.sh`]) {
10101
10319
  try {
@@ -10212,7 +10430,7 @@ import { spawn as spawn3 } from "child_process";
10212
10430
  import { existsSync as existsSync8, mkdirSync as mkdirSync10, unlinkSync as unlinkSync4, readdirSync as readdirSync3 } from "fs";
10213
10431
  import { homedir as homedir8 } from "os";
10214
10432
  import { join as join13 } from "path";
10215
- var log21 = forComponent("remote-trigger");
10433
+ var log22 = forComponent("remote-trigger");
10216
10434
  var TRIGGER_DIR = () => join13(homedir8(), ".agent", "triggers");
10217
10435
  var sockPath = (sessionId, dir = TRIGGER_DIR()) => join13(dir, `${sessionId}.sock`);
10218
10436
  var TriggerServer = class {
@@ -10250,13 +10468,13 @@ var TriggerServer = class {
10250
10468
  conn.end(JSON.stringify({ ok: false, error: String(e) }) + "\n");
10251
10469
  }
10252
10470
  });
10253
- conn.on("error", (e) => log21.debug("trigger conn error", e));
10471
+ conn.on("error", (e) => log22.debug("trigger conn error", e));
10254
10472
  });
10255
- this.server.on("error", (e) => log21.debug("trigger server error", e));
10473
+ this.server.on("error", (e) => log22.debug("trigger server error", e));
10256
10474
  this.server.listen(p);
10257
10475
  this.path = p;
10258
10476
  } catch (e) {
10259
- log21.debug("trigger server unavailable", e);
10477
+ log22.debug("trigger server unavailable", e);
10260
10478
  }
10261
10479
  }
10262
10480
  /** Re-bind on /resume (the session id changed). */
@@ -10377,7 +10595,7 @@ var italic = C("3");
10377
10595
  var strike = C("9");
10378
10596
  var link = (text, url) => useColor ? `\x1B]8;;${url}\x1B\\${cyan(text)}\x1B]8;;\x1B\\` : `${text} (${url})`;
10379
10597
  var err = (s) => process.stderr.write(s);
10380
- var log22 = forComponent("cli");
10598
+ var log23 = forComponent("cli");
10381
10599
  var VERSION = (() => {
10382
10600
  try {
10383
10601
  return JSON.parse(readFileSync8(new URL("../package.json", import.meta.url), "utf8")).version ?? "?";
@@ -11016,7 +11234,7 @@ function costOf(pricing, promptTokens = 0, completionTokens = 0, cacheCreationTo
11016
11234
  function turnCost(model, usage) {
11017
11235
  return costOf(getModelInfo(model)?.pricing, usage?.promptTokens ?? 0, usage?.completionTokens ?? 0, usage?.cacheCreationTokens ?? 0, usage?.cacheReadTokens ?? 0, model);
11018
11236
  }
11019
- async function evaluateGoal(ai, condition, transcript, log23) {
11237
+ async function evaluateGoal(ai, condition, transcript, log24) {
11020
11238
  const recent = transcript.filter((m) => m.role === "assistant").slice(-8).map((m) => {
11021
11239
  const text = typeof m.content === "string" ? m.content : m.content.filter((p) => p.type === "text").map((p) => p.text).join(" ");
11022
11240
  return text.slice(0, 600);
@@ -11036,7 +11254,7 @@ ${recent}` }
11036
11254
  const match = r.content.match(/\{[\s\S]*\}/);
11037
11255
  if (match) return JSON.parse(match[0]);
11038
11256
  } catch (e) {
11039
- log23(dim(` (goal evaluator error: ${e?.message ?? e})
11257
+ log24(dim(` (goal evaluator error: ${e?.message ?? e})
11040
11258
  `));
11041
11259
  }
11042
11260
  return { met: false, reason: "evaluation unclear" };
@@ -11248,7 +11466,7 @@ function mcpAgentTools(mounted, opts) {
11248
11466
  return tools;
11249
11467
  }
11250
11468
  async function closeMcp(mounted) {
11251
- await Promise.all(mounted.map((m) => m.client.close().catch((e) => log22.debug("mcp close failed", e))));
11469
+ await Promise.all(mounted.map((m) => m.client.close().catch((e) => log23.debug("mcp close failed", e))));
11252
11470
  }
11253
11471
  var IMG_EXT = { ".png": "image/png", ".jpg": "image/jpeg", ".jpeg": "image/jpeg", ".gif": "image/gif", ".webp": "image/webp" };
11254
11472
  function mentionRefs(line) {
@@ -11298,7 +11516,7 @@ async function expandMentions(fs, line) {
11298
11516
  if (loaded.includes(ref) || missing.includes(ref)) continue;
11299
11517
  if (ref.includes(":") && mcpMentionResolver) {
11300
11518
  const body = await mcpMentionResolver(ref).catch((e) => {
11301
- log22.debug("mcp mention resolve failed", e);
11519
+ log23.debug("mcp mention resolve failed", e);
11302
11520
  return null;
11303
11521
  });
11304
11522
  if (body != null) {
@@ -11381,7 +11599,7 @@ async function runTurn(agent, store, session, task, cp, cwd = process.cwd(), sen
11381
11599
  try {
11382
11600
  store.save(session);
11383
11601
  } catch (ex) {
11384
- log22.debug("mid-turn session flush failed", ex);
11602
+ log23.debug("mid-turn session flush failed", ex);
11385
11603
  }
11386
11604
  }
11387
11605
  return origNotify(e);
@@ -11528,14 +11746,14 @@ var isCancelTeardown = (e) => {
11528
11746
  function installCancelGuards(mounted) {
11529
11747
  process.on("unhandledRejection", (e) => {
11530
11748
  if (isCancelTeardown(e)) {
11531
- log22.debug("suppressed unhandledRejection (cursor stream cancel)", e);
11749
+ log23.debug("suppressed unhandledRejection (cursor stream cancel)", e);
11532
11750
  return;
11533
11751
  }
11534
- log22.error("unhandledRejection", e);
11752
+ log23.error("unhandledRejection", e);
11535
11753
  });
11536
11754
  process.on("uncaughtException", (e) => {
11537
11755
  if (isCancelTeardown(e)) {
11538
- log22.debug("suppressed uncaughtException (cursor stream cancel)", e);
11756
+ log23.debug("suppressed uncaughtException (cursor stream cancel)", e);
11539
11757
  return;
11540
11758
  }
11541
11759
  console.error(e);
@@ -11595,6 +11813,8 @@ async function repl(args, ai, cfg, cwd) {
11595
11813
  let dx;
11596
11814
  let voiceIO;
11597
11815
  let voiceLineOpen = false;
11816
+ const emotionsOn = process.env.VOICE_EMOTIONS !== "0";
11817
+ let showEmotions = process.env.VOICE_SHOW_EMOTIONS === "1";
11598
11818
  const voiceEcho = (text) => {
11599
11819
  const s = forSpeech(text);
11600
11820
  if (!s) return;
@@ -11669,7 +11889,7 @@ async function repl(args, ai, cfg, cwd) {
11669
11889
  spinner.stop();
11670
11890
  voiceIO.enqueueUtterance(e.message);
11671
11891
  editorRef?.suspend();
11672
- voiceEcho(e.message);
11892
+ voiceEcho(emotionsOn ? renderEmotions(e.message, { show: showEmotions }).display : e.message);
11673
11893
  voiceEchoEnd();
11674
11894
  editorRef?.resume();
11675
11895
  editorRef?.redrawNow();
@@ -11683,9 +11903,9 @@ async function repl(args, ai, cfg, cwd) {
11683
11903
  }
11684
11904
  if (e.kind === "text_delta" && voiceIO) {
11685
11905
  spinner.stop();
11686
- voiceIO.speakDelta(e.message);
11906
+ const echo = voiceIO.speakDelta(e.message);
11687
11907
  editorRef?.suspend();
11688
- voiceEcho(e.message);
11908
+ voiceEcho(echo);
11689
11909
  return;
11690
11910
  } else if (e.kind === "text_delta" && stashText()) {
11691
11911
  process.stdout.write("\r\x1B[K");
@@ -11759,8 +11979,8 @@ async function repl(args, ai, cfg, cwd) {
11759
11979
  providerOptionsFor: (m) => cursorProviderOptions(m, cwd, cfg.mcpServers),
11760
11980
  ...(args.thinkModel ?? cfg.thinkModel) !== void 0 ? { thinkModel: (args.thinkModel ?? cfg.thinkModel) === false ? false : resolveModelOrNewest(String(args.thinkModel ?? cfg.thinkModel)) } : {},
11761
11981
  host,
11762
- ...args.voice ? { voiceStyle: "conversational", progressUpdates: true, askRelay: true } : {},
11763
- // voice: progress asides + worker questions relayed through the conversation
11982
+ ...args.voice ? { voiceStyle: "conversational", progressUpdates: true, askRelay: true, emotionTags: emotionsOn } : {},
11983
+ // voice: progress asides + worker questions relayed through the conversation; emotion tags taught only with a TTS that speaks them
11764
11984
  // Per-TASK checkpoint frames (the natural undo unit in duplex = one delegation): opened BEFORE
11765
11985
  // the worker spawns (post-spawn would race its first edits). `checkpoints` is bound below.
11766
11986
  onTaskStart: async (_id, label) => {
@@ -12012,7 +12232,7 @@ Added entries are loadable now via the Skill/SlashCommand tools; removed ones ar
12012
12232
  mkdirSync11(join14(cwd, ".agent"), { recursive: true });
12013
12233
  appendFileSync(histPath, line + "\n");
12014
12234
  } catch (e) {
12015
- log22.debug("history write failed", e);
12235
+ log23.debug("history write failed", e);
12016
12236
  }
12017
12237
  };
12018
12238
  const ago = (t) => {
@@ -12083,7 +12303,7 @@ Added entries are loadable now via the Skill/SlashCommand tools; removed ones ar
12083
12303
  try {
12084
12304
  store.save(session);
12085
12305
  } catch (e) {
12086
- log22.debug("session save after rewind failed", e);
12306
+ log23.debug("session save after rewind failed", e);
12087
12307
  }
12088
12308
  err(green(" \u27F2 jumped back") + dim(` \u2014 ${face.transcript.length} message(s) kept; edit + resend
12089
12309
  `));
@@ -12117,7 +12337,7 @@ ${task}`;
12117
12337
  bangContext.length = 0;
12118
12338
  }
12119
12339
  const delta = await refreshCatalogs().catch((e) => {
12120
- log22.debug("catalog refresh failed", e);
12340
+ log23.debug("catalog refresh failed", e);
12121
12341
  return "";
12122
12342
  });
12123
12343
  if (delta) {
@@ -12380,7 +12600,7 @@ ${task}`;
12380
12600
  desc: "rescan skills/commands dirs and rebuild the system prompt (one cache miss) \u2014 picks up entries created mid-session",
12381
12601
  run: async () => {
12382
12602
  await refreshCatalogs().catch((e) => {
12383
- log22.debug("catalog refresh failed", e);
12603
+ log23.debug("catalog refresh failed", e);
12384
12604
  });
12385
12605
  face.reprepare();
12386
12606
  err(green(` \u2713 reloaded \u2014 ${skills.length} skill(s), ${cmds.length} command(s); system prompt rebuilds on next message
@@ -12450,6 +12670,24 @@ ${task}`;
12450
12670
  }
12451
12671
  await toggleVoice();
12452
12672
  }
12673
+ }, "voice-emotions": {
12674
+ desc: "show/hide the [emotion] debug tags in the echo \u2014 /voice-emotions <on|off> (TTS emotion control stays on regardless)",
12675
+ run: async (a) => {
12676
+ if (!emotionsOn) {
12677
+ err(dim(" (emotion control is off \u2014 VOICE_EMOTIONS=0)\n"));
12678
+ return;
12679
+ }
12680
+ const v = a[0]?.toLowerCase();
12681
+ if (v === "on" || v === "off") {
12682
+ showEmotions = v === "on";
12683
+ voiceIO?.setShowEmotions(showEmotions);
12684
+ err(green(` \u2713 emotion tags ${showEmotions ? "shown" : "hidden"} in echo
12685
+ `));
12686
+ return;
12687
+ }
12688
+ err(dim(` emotion tags in echo: ${showEmotions ? "shown" : "hidden"} (use /voice-emotions on|off)
12689
+ `));
12690
+ }
12453
12691
  }, "voice-model": {
12454
12692
  desc: "switch the reflex (voice) model \u2014 /voice-model <id>, or alone for a picker",
12455
12693
  run: async (a) => {
@@ -12842,7 +13080,7 @@ ${task}`;
12842
13080
  try {
12843
13081
  for (const def of (await loadAgents(fs2, d)).agents) if (!seen.has(def.name)) seen.set(def.name, { def, from: d });
12844
13082
  } catch (e) {
12845
- log22.debug(`loadAgents(${d}) failed`, e);
13083
+ log23.debug(`loadAgents(${d}) failed`, e);
12846
13084
  }
12847
13085
  }
12848
13086
  if (!seen.size) {
@@ -12930,7 +13168,7 @@ ${task}`;
12930
13168
  }
12931
13169
  if (idx >= 0) {
12932
13170
  const old = mounted.splice(idx, 1)[0];
12933
- await old.client.close().catch((e) => log22.debug("mcp close failed", e));
13171
+ await old.client.close().catch((e) => log23.debug("mcp close failed", e));
12934
13172
  }
12935
13173
  try {
12936
13174
  const m = await mountMcpServer(name, conf);
@@ -12958,7 +13196,7 @@ ${task}`;
12958
13196
  }
12959
13197
  const m = mounted.splice(idx, 1)[0];
12960
13198
  remountMcpTools();
12961
- await m.client.close().catch((e) => log22.debug("mcp close failed", e));
13199
+ await m.client.close().catch((e) => log23.debug("mcp close failed", e));
12962
13200
  err(dim(` removed "${name}"
12963
13201
  `));
12964
13202
  return;
@@ -13178,7 +13416,7 @@ ${task}`;
13178
13416
  try {
13179
13417
  return readdirSync4(join14(cwd, absDir.replace(/^\/+/, "")), { withFileTypes: true }).map((d) => ({ name: d.name, dir: d.isDirectory() }));
13180
13418
  } catch (e) {
13181
- log22.debug("completion readdir failed", absDir, e);
13419
+ log23.debug("completion readdir failed", absDir, e);
13182
13420
  return null;
13183
13421
  }
13184
13422
  };
@@ -13305,7 +13543,7 @@ ${out}
13305
13543
  return;
13306
13544
  }
13307
13545
  await refreshCatalogs().catch((e) => {
13308
- log22.debug("catalog refresh failed", e);
13546
+ log23.debug("catalog refresh failed", e);
13309
13547
  });
13310
13548
  const sk = skills.find((x) => x.name === name);
13311
13549
  if (sk) {
@@ -13348,6 +13586,9 @@ ${out}
13348
13586
  const fakeVoice = process.env.AGENTX_VOICE_FAKE ? fakeVoiceParts(process.env.AGENTX_VOICE_FAKE) : null;
13349
13587
  voiceIO = new VoiceIO({
13350
13588
  ...fakeVoice ?? {},
13589
+ emotions: emotionsOn,
13590
+ showEmotions,
13591
+ // local is authoritative (a /voice-emotions before mic-on still applies)
13351
13592
  // No ack phrase by default: a fixed "Mm-hm," every turn reads robotic, Haiku's TTFT doesn't
13352
13593
  // need masking (~0.7-1.2s full turns), and the conversational register already opens with a
13353
13594
  // natural reaction. The mechanism (+ echo-leak guard) stays for slower voice models.