@livx.cc/agentx 0.99.112 → 0.99.113

package/dist/cli.js

@@ -9328,14 +9328,40 @@ var ACK = new Set("yeah yea yes yep yup ya ok okay k right sure true totally exa
 var LEAD = /* @__PURE__ */ new Set(["and", "so", "um", "uh", "er", "well", "oh"]);
 var LEADER = /* @__PURE__ */ new Set(["yeah", "yea", "yes", "yep", "yup", "ok", "okay", "right", "sure", "alright", "mm", "mhm", "hmm", "uh-huh", "totally", "exactly", "true"]);
 var PHRASES = ["got it", "i see", "makes sense", "of course", "for sure", "all right", "that's right", "that's true", "thats right", "thats true", "love it", "i know", "me too", "uh huh"];
+var CONTINUE = /^(?:(?:and|so|oh) )?(?:and|so|really|seriously|then what|and then|go on|keep going|ואז|באמת|נו)$/;
+var AGREE = new Set("yeah yea yes yep yup right sure true totally exactly definitely absolutely agreed correct indeed \u05DB\u05DF \u05E0\u05DB\u05D5\u05DF \u05D1\u05D3\u05D9\u05D5\u05E7 \u05D1\u05E8\u05D5\u05E8 \u05D1\u05D8\u05D7 \u05D0\u05DB\u05DF \u05DC\u05D2\u05DE\u05E8\u05D9 \u05DE\u05E1\u05DB\u05D9\u05DD \u05DE\u05E1\u05DB\u05D9\u05DE\u05D4 \u05DE\u05DE\u05E9".split(" "));
+var AGREE_PHRASES = ["of course", "for sure", "that's right", "that's true", "thats right", "thats true", "makes sense", "i agree"];
+var HE_ACK = new Set("\u05DB\u05DF \u05E0\u05DB\u05D5\u05DF \u05D1\u05D3\u05D9\u05D5\u05E7 \u05D1\u05E8\u05D5\u05E8 \u05D1\u05D8\u05D7 \u05D0\u05DB\u05DF \u05DC\u05D2\u05DE\u05E8\u05D9 \u05DE\u05DE\u05E9 \u05D0\u05D5\u05E7\u05D9\u05D9 \u05D0\u05D5\u05E7\u05D9 \u05E1\u05D1\u05D1\u05D4 \u05D9\u05E4\u05D4 \u05DE\u05E2\u05D5\u05DC\u05D4 \u05D5\u05D5\u05D0\u05D5 \u05D0\u05D4\u05D4 \u05D0\u05D4 \u05D4\u05DE\u05DE \u05DE\u05DE\u05DE".split(" "));
+function cueKind(text) {
+  const t = norm2(text);
+  if (!t) return null;
+  if (CONTINUE.test(t)) return "continue";
+  let r = t;
+  for (const p of AGREE_PHRASES) if (r.includes(p)) return "agree";
+  r = r.split(/[\s-]+/).filter((w) => !LEAD.has(w)).join(" ");
+  return r.split(" ").some((w) => AGREE.has(w)) ? "agree" : null;
+}
+var ASIDES = {
+  en: { agree: ["Oh yeah, definitely.", "Yeah, exactly.", "Right, yeah."], continue: ["Yeah, so\u2026", "So, yeah\u2026"] },
+  he: { agree: ["\u05DB\u05DF, \u05D1\u05D3\u05D9\u05D5\u05E7.", "\u05DB\u05DF, \u05DE\u05DE\u05E9.", "\u05E0\u05DB\u05D5\u05DF, \u05DB\u05DF."], continue: ["\u05DB\u05DF, \u05D0\u05D6\u2026", "\u05D0\u05D6 \u05DB\u05DF\u2026"] }
+};
+function overlapAside(text, lang, avoid = "", pick = (n) => Math.floor(Math.random() * n)) {
+  const kind = cueKind(text);
+  const l = new RegExp("\\p{Script=Hebrew}", "u").test(text) ? "he" : /[a-z]/i.test(text) ? "en" : lang;
+  const pool = kind && l ? ASIDES[l]?.[kind] : void 0;
+  if (!pool?.length) return null;
+  const fresh = pool.filter((x) => x !== avoid);
+  return fresh[pick(fresh.length)] ?? null;
+}
 function overlapRule(text) {
+  if (CONTINUE.test(norm2(text))) return "keep";
   if (/\?/.test(text)) return "cut";
   let t = norm2(text);
   if (!t) return void 0;
   if (HARD.test(t)) return "cut";
   for (const p of PHRASES) t = t.split(p).join(" ");
   const words4 = t.split(/[\s-]+/).filter(Boolean);
-  const ack = (w) => ACK.has(w) || /^(?:ha)+h?$|^(?:he)+h?$|^m+(?:hm+)?$|^h?m+$/.test(w);
+  const ack = (w) => ACK.has(w) || HE_ACK.has(w) || /^(?:ha)+h?$|^(?:he)+h?$|^m+(?:hm+)?$|^h?m+$/.test(w);
   if (words4.every(ack)) return "keep";
   const lead2 = words4.findIndex((w) => !LEAD.has(w));
   if (lead2 >= 0 && LEADER.has(words4[lead2]) && words4.slice(lead2).filter((w) => !ack(w) && !LEAD.has(w)).length <= 1) return "keep";
@@ -13616,6 +13642,8 @@ var VoiceCallOptions = class {
   onMetric = () => {
   };
   now = () => performance.now();
+  /** An aside (aside()) needs this long since the last one — a cue answered every time sounds scripted. */
+  asideGapMs = 8e3;
 };
 var VoiceCall = class {
   constructor(options) {
@@ -13659,6 +13687,7 @@ var VoiceCall = class {
   judge;
   cur;
   lang = null;
+  lastAside = { text: "", at: -Infinity };
   lastAudibleAt = -Infinity;
   firstFinalAt = null;
   // the first final of the turn being formed (hold/merge)
@@ -13767,6 +13796,19 @@ var VoiceCall = class {
     if (d.verdict === "cut") this.barge();
     return d.verdict;
   }
+  /** The fill channel's spoken aside for an overlap judged 'keep' ("yeah" → "Oh yeah, definitely."): the host plays it
+   *  while the reply is held, then resumes the same reply (bod-voice decideOverlap's aside). ≤4 canned words, never a
+   *  claim, the cue's language; null = stay quiet (a plain listener cue, too soon after the last, or no line). It is a
+   *  fill line: never MAIN history. */
+  aside(text) {
+    const now4 = this.o.now();
+    if (now4 - this.lastAside.at < this.o.asideGapMs) return null;
+    const line = overlapAside(text, this.lang, this.lastAside.text);
+    if (!line) return null;
+    this.lastAside = { text: line, at: now4 };
+    this.o.onMetric({ name: "fill", kind: "aside", text: line, lang: this.lang });
+    return line;
+  }
   /** Did the last barge cut OUR fill line `text`? (hosts strip it from the cut reply — it is not MAIN's speech) */
   spoke(text) {
     return this.fill.spoke(text);
@@ -15355,7 +15397,9 @@ function bindVoiceCall(v, call, live, decideMs = 900) {
     if (!live() || ev.kind !== "overlap_pause") return;
     const heard = String(ev.text ?? "");
     void call.overlap(heard).then((verdict) => {
-      if (live()) v.decideOverlap(verdict === "keep" ? "keep_talking" : "yield", heard);
+      if (!live()) return;
+      if (verdict === "keep") v.decideOverlap("keep_talking", heard, call.aside(heard) ?? void 0);
+      else v.decideOverlap("yield", heard);
     });
   };
   Object.defineProperty(call.fill.options, "aec", { get: () => v.usingAec, set: () => {
@@ -15457,6 +15501,8 @@ var VoiceIO = class extends VoiceEngine {
       ...o,
       stt: o.stt ?? new SonioxSTT({ auth: o.sonioxApiKey, source: duplex ?? new NodeMicSource(), terms: o.sttTerms }),
       tts: o.tts ?? new CartesiaTTS({ auth: o.cartesiaApiKey, voiceId: o.cartesiaVoiceId }),
+      asideTts: o.tts ? null : new CartesiaTTS({ auth: o.cartesiaApiKey, voiceId: o.cartesiaVoiceId }),
+      // asides (VoiceCall.aside): a 2nd client — the reply's context may still stream
       player: o.player ?? duplex ?? new Player(),
       bargeRmsMult: Number(process.env.BARGE_RMS_MULT || o.bargeRmsMult),
       bargeRmsFloor: Number(process.env.BARGE_RMS_FLOOR || o.bargeRmsFloor),
@@ -22467,11 +22513,14 @@ function channelOptions(ai, model) {
     onEvent: (e) => log40.debug(`channel ${e.trigger}: rule=${e.rule} choice=${e.choice ?? "-"} \u2192 ${e.acted}${e.text ? ` "${e.text}"` : ""}${e.cue ? " (planned cue)" : ""} ${e.ms}ms`)
   };
 }
-function overlapResponder(ai, mode, out) {
+function overlapResponder(ai, mode, out, aside) {
   const judge = softBarge(mode) ? new OverlapJudge({ model: overlapModelFor(ai) }) : void 0;
   return (m) => {
     const text = String(m.text ?? "");
-    void (judge ? judge.judge(text) : Promise.resolve({ verdict: "cut", by: "mode", ms: 0 })).then((d) => out({ type: "overlap", ...m.id !== void 0 ? { id: m.id } : {}, verdict: d.verdict, by: d.by, ms: d.ms }));
+    void (judge ? judge.judge(text) : Promise.resolve({ verdict: "cut", by: "mode", ms: 0 })).then((d) => {
+      const line = d.verdict === "keep" ? aside?.(text) : null;
+      out({ type: "overlap", ...m.id !== void 0 ? { id: m.id } : {}, verdict: d.verdict, by: d.by, ms: d.ms, ...line ? { aside: line } : {} });
+    });
   };
 }
 async function duplexServe(args, ai, cfg, cwd) {
@@ -22607,7 +22656,7 @@ async function channelServe(args, ai, cfg) {
   const overlap2 = overlapResponder(ai, args.channelMode, (o) => {
     if (o.verdict === "cut") call.barge();
     out(o);
-  });
+  }, (t) => call.aside(t));
   out({ type: "ready", model, channelMode: args.channelMode, softBarge: softBarge(args.channelMode), call: true, ...notes.length ? { notes } : {} });
   const rl = createInterface({ input: process.stdin });
   rl.on("line", (line) => {