@burdenoff/website-sdk 2026.1105.3 → 2026.1106.1

package/dist/avatar-mic/index.mjs

@@ -1,5 +1,145 @@
 import { __publicField } from '../chunk-VNS5WDWO.mjs';
 
+// src/avatar-mic/speech-words.ts
+var FILLERS = /* @__PURE__ */ new Set([
+  "uh",
+  "um",
+  "uhm",
+  "umm",
+  "er",
+  "erm",
+  "eh",
+  "ah",
+  "hm",
+  "hmm",
+  "hmmm",
+  "mm",
+  "mmm",
+  "mhm",
+  "huh",
+  "oh",
+  "so",
+  "well",
+  "okay",
+  "ok",
+  "and",
+  "like",
+  "hey",
+  "hi",
+  "hello",
+  "actually"
+]);
+var OPEN_ENDINGS = /* @__PURE__ */ new Set([
+  "the",
+  "a",
+  "an",
+  "of",
+  "to",
+  "for",
+  "about",
+  "and",
+  "or",
+  "but",
+  "what",
+  "how",
+  "which",
+  "is",
+  "are",
+  "can"
+]);
+var PRONOUNS = /* @__PURE__ */ new Set([
+  "i",
+  "me",
+  "you",
+  "it",
+  "we",
+  "us",
+  "they",
+  "them",
+  "he",
+  "him",
+  "she",
+  "her",
+  "this",
+  "that",
+  "these",
+  "those"
+]);
+var ECHO_MIN_WORDS = 3;
+var SELF_CONTAINED_SCRIPT = /[\p{Script=Han}\p{Script=Hiragana}\p{Script=Katakana}\p{Script=Hangul}]/u;
+function tidyWords(text) {
+  return text.replace(/\s+/g, " ").trim();
+}
+function speechWords(text) {
+  const flat = text.toLowerCase().replace(/[^\p{L}\p{M}\p{N}]+/gu, " ").trim();
+  return flat === "" ? [] : flat.split(" ");
+}
+function startsWithWords(whole, head) {
+  return head.length <= whole.length && head.every((word, index) => whole[index] === word);
+}
+function endsWithWords(whole, tail) {
+  return tail.length <= whole.length && tail.every(
+    (word, index) => whole[whole.length - tail.length + index] === word
+  );
+}
+function startsWithWordsOf(text, head) {
+  return startsWithWords(speechWords(text), speechWords(head));
+}
+function mergeTurnPhrase(turn, phrase) {
+  const heard = speechWords(turn);
+  const incoming = speechWords(phrase);
+  if (incoming.length === 0) return turn;
+  if (heard.length === 0) return tidyWords(phrase);
+  if (startsWithWords(incoming, heard)) {
+    return incoming.length === heard.length ? turn : tidyWords(phrase);
+  }
+  if (incoming.length >= 2 && endsWithWords(heard, incoming)) return turn;
+  return tidyWords(`${turn} ${phrase}`);
+}
+function mergeInterim(turn, interim) {
+  const said = tidyWords(interim);
+  if (said === "") return turn;
+  const head = speechWords(turn);
+  if (head.length === 0) return said;
+  const tokens = said.split(" ");
+  let matched = 0;
+  for (let at = 0; at < tokens.length; at += 1) {
+    const words = speechWords(tokens[at] ?? "");
+    for (const word of words) {
+      if (word !== head[matched]) return tidyWords(`${turn} ${said}`);
+      matched += 1;
+      if (matched === head.length) {
+        return tidyWords(`${turn} ${tokens.slice(at + 1).join(" ")}`);
+      }
+    }
+  }
+  return tidyWords(`${turn} ${said}`);
+}
+function saidWords(words) {
+  return words.filter((word) => !FILLERS.has(word));
+}
+function isGenuineSpeech(text) {
+  const said = saidWords(speechWords(text));
+  if (said.length === 1 && PRONOUNS.has(said[0] ?? "")) return false;
+  const content = said.filter((word) => !OPEN_ENDINGS.has(word));
+  if (content.length === 0) return false;
+  const letters = [...content.join("")];
+  if (letters.length >= 2) return true;
+  return letters.length === 1 && SELF_CONTAINED_SCRIPT.test(letters[0] ?? "");
+}
+function waitsLonger(text) {
+  const words = speechWords(text);
+  const last = words[words.length - 1];
+  if (last === void 0) return false;
+  return saidWords(words).length <= 1 || OPEN_ENDINGS.has(last) || FILLERS.has(last);
+}
+function isEchoOf(text, spoken) {
+  if (spoken === null) return false;
+  const heard = speechWords(text);
+  if (heard.length < ECHO_MIN_WORDS) return false;
+  return ` ${speechWords(spoken).join(" ")} `.includes(` ${heard.join(" ")} `);
+}
+
 // src/avatar-mic/browser-press.ts
 var BROWSER_START_WAIT_MS = 15e3;
 var BROWSER_FINAL_WAIT_MS = 3e3;
@@ -9,6 +149,11 @@ var BROWSER_MAX_QUICK_ENDS = 8;
 var BROWSER_MAX_RESTARTS = 60;
 var BROWSER_PRIME_SETTLE_MS = 250;
 var BROWSER_GRANT_RETRY_MS = 350;
+var BROWSER_PAUSE_MIN_MS = 2e3;
+var BROWSER_PAUSE_MAX_MS = 2500;
+var BROWSER_PAUSE_AFTER_START_MS = 1500;
+var BROWSER_PAUSE_SHOWN_MS = 800;
+var BROWSER_ECHO_WINDOW_MS = 2500;
 var QUICK_END_MS = 1e3;
 var NETWORK_FAILURES_FOR_GOOD = 2;
 var IDLE = { name: "idle" };
@@ -80,8 +225,7 @@ function readRecognitionResults(results) {
       continue;
     }
     const last = finals.length > 0 ? finals[finals.length - 1] : void 0;
-    const lower = text.toLowerCase();
-    if (last !== void 0 && (lower === last.toLowerCase() || lower.startsWith(`${last.toLowerCase()} `))) {
+    if (last !== void 0 && startsWithWordsOf(text, last)) {
       finals[finals.length - 1] = text;
     } else {
       finals.push(text);
@@ -91,6 +235,11 @@ function readRecognitionResults(results) {
 }
 function createBrowserAvatarMic(options, seams = {}) {
   const { createRecognition, now, primeMicrophone, watchMicPermission, touch } = { ...BROWSER, ...seams };
+  const { endpoint } = options;
+  const pauseMs = endpoint === void 0 || !Number.isFinite(endpoint.pauseMs) ? BROWSER_PAUSE_MIN_MS : Math.min(
+    BROWSER_PAUSE_MAX_MS,
+    Math.max(BROWSER_PAUSE_MIN_MS, endpoint.pauseMs)
+  );
   let press = null;
   let epoch = 0;
   let state = IDLE;
@@ -104,14 +253,23 @@ function createBrowserAvatarMic(options, seams = {}) {
     state = next;
     options.onState(next);
   };
-  const heard = (p) => [...p.earlier, ...p.instance?.finals ?? [], p.instance?.interim ?? ""].filter((part) => part !== "").join(" ");
+  const turnOf = (earlier, instance) => {
+    let turn = earlier;
+    for (const final of instance?.finals ?? []) {
+      turn = mergeTurnPhrase(turn, final);
+    }
+    return mergeInterim(turn, instance?.interim ?? "");
+  };
+  const heard = (p) => turnOf(p.earlier, p.instance);
   const show = (p) => {
     if (p.stage === "listening") {
+      const text = heard(p);
       report({
         name: "listening",
         epoch: p.epoch,
         startedAt: p.startedAt,
-        text: heard(p)
+        text,
+        ...endpoint === void 0 ? {} : { genuine: isGenuineSpeech(text), pausing: p.pausing }
       });
     } else if (p.stage === "finishing") {
       report({
@@ -138,7 +296,9 @@ function createBrowserAvatarMic(options, seams = {}) {
       p.startTimer,
       p.limitTimer,
       p.finalTimer,
-      p.launchTimer
+      p.launchTimer,
+      p.pauseTimer,
+      p.pausingTimer
     ]) {
       if (timer !== null) clearTimeout(timer);
     }
@@ -146,6 +306,14 @@ function createBrowserAvatarMic(options, seams = {}) {
     p.limitTimer = null;
     p.finalTimer = null;
     p.launchTimer = null;
+    p.pauseTimer = null;
+    p.pausingTimer = null;
+  };
+  const clearPause = (p) => {
+    if (p.pauseTimer !== null) clearTimeout(p.pauseTimer);
+    if (p.pausingTimer !== null) clearTimeout(p.pausingTimer);
+    p.pauseTimer = null;
+    p.pausingTimer = null;
   };
   const settle = (p, next) => {
     if (p.stage === "ended") return;
@@ -160,23 +328,104 @@ function createBrowserAvatarMic(options, seams = {}) {
   const complete = (p) => {
     const text = heard(p);
     if (text !== "") networkFailures = 0;
-    settle(p, { name: "done", epoch: p.epoch, text, limit: p.limit });
+    if (endpoint === void 0) {
+      settle(p, { name: "done", epoch: p.epoch, text, limit: p.limit });
+      return;
+    }
+    const by = p.stage === "finishing" ? p.by ?? "tap" : p.stage === "starting" ? "silence" : "failed";
+    const echo = by === "pause" && p.speechAt !== null && p.speechAt - p.listenedAt <= BROWSER_ECHO_WINDOW_MS && isEchoOf(text, endpoint.echoOf());
+    settle(p, {
+      name: "done",
+      epoch: p.epoch,
+      text,
+      limit: p.limit,
+      by,
+      echo
+    });
   };
   const fail = (p, reason) => settle(p, { name: "failed", epoch: p.epoch, reason, text: heard(p) });
   const current = (p, instance) => press === p && p.stage !== "ended" && p.instance === instance;
+  const armPause = (p, turn, speech) => {
+    clearPause(p);
+    if (!speech) return;
+    const at = now();
+    if (!p.pausing) {
+      p.pausingTimer = setTimeout(
+        () => {
+          p.pausingTimer = null;
+          if (press !== p || p.stage !== "listening") return;
+          p.pausing = true;
+          show(p);
+        },
+        Math.max(0, p.wordsAt + BROWSER_PAUSE_SHOWN_MS - at)
+      );
+    }
+    const instance = p.instance;
+    if (instance === null || !instance.started) return;
+    const wait = waitsLonger(turn) ? 2 * pauseMs : pauseMs;
+    const due = Math.max(
+      p.wordsAt + wait,
+      instance.startedAt + BROWSER_PAUSE_AFTER_START_MS
+    );
+    p.pauseTimer = setTimeout(
+      () => {
+        p.pauseTimer = null;
+        if (press === p && p.stage === "listening" && p.instance === instance) {
+          stopRecording(p, "pause");
+        }
+      },
+      Math.max(0, due - at)
+    );
+  };
+  const listenOn = (p) => {
+    if (endpoint === void 0 || p.stage !== "listening") return;
+    const turn = heard(p);
+    const words = speechWords(turn).join(" ");
+    const at = now();
+    if (words !== p.words) {
+      p.words = words;
+      p.wordsAt = at;
+      p.pausing = false;
+    }
+    const speech = isGenuineSpeech(turn);
+    if (speech && p.speechAt === null) {
+      p.speechAt = at;
+      if (p.limitTimer !== null) clearTimeout(p.limitTimer);
+      p.limitTimer = setTimeout(() => {
+        p.limitTimer = null;
+        if (press === p) stopRecording(p, "limit");
+      }, options.maxPressMs);
+    }
+    armPause(p, turn, speech);
+  };
   const began = (p, instance) => {
     instance.started = true;
     instance.startedAt = now();
     hadMicrophone = true;
-    if (p.stage !== "starting") return;
-    if (p.startTimer !== null) clearTimeout(p.startTimer);
-    p.startTimer = null;
-    p.stage = "listening";
-    p.startedAt = Date.now();
-    p.limitTimer = setTimeout(() => {
-      p.limitTimer = null;
-      if (press === p) stopRecording(p, true);
-    }, options.maxPressMs);
+    if (p.stage === "starting") {
+      if (p.startTimer !== null) clearTimeout(p.startTimer);
+      p.startTimer = null;
+      p.stage = "listening";
+      p.startedAt = Date.now();
+      p.listenedAt = instance.startedAt;
+      p.limitTimer = setTimeout(() => {
+        p.limitTimer = null;
+        if (press !== p) return;
+        if (endpoint === void 0) {
+          stopRecording(p, "limit");
+        } else if (p.stage === "listening" && p.speechAt === null) {
+          settle(p, {
+            name: "done",
+            epoch: p.epoch,
+            text: "",
+            limit: false,
+            by: "silence",
+            echo: false
+          });
+        }
+      }, options.maxPressMs);
+    }
+    listenOn(p);
   };
   const launchIn = (p, delay) => {
     if (delay <= 0) {
@@ -202,9 +451,7 @@ function createBrowserAvatarMic(options, seams = {}) {
     );
   };
   const keep = (p, instance) => {
-    for (const text of [...instance.finals, instance.interim]) {
-      if (text !== "") p.earlier.push(text);
-    }
+    p.earlier = turnOf(p.earlier, instance);
   };
   const onError = (p, instance, code) => {
     if (code === "no-speech") {
@@ -270,6 +517,8 @@ function createBrowserAvatarMic(options, seams = {}) {
         else fail(p, "failed");
         return;
       case "listening": {
+        if (p.pauseTimer !== null) clearTimeout(p.pauseTimer);
+        p.pauseTimer = null;
         const quick = !instance.started || now() - instance.startedAt < QUICK_END_MS && instance.finals.length === 0 && instance.interim === "";
         restart(p, quick ? "quick" : "listened");
         return;
@@ -316,6 +565,7 @@ function createBrowserAvatarMic(options, seams = {}) {
       const read = readRecognitionResults(event.results);
       instance.finals = read.finals;
       instance.interim = read.interim;
+      listenOn(p);
       show(p);
     };
     recognition.onerror = (event) => {
@@ -359,14 +609,16 @@ function createBrowserAvatarMic(options, seams = {}) {
       }
     );
   }
-  function stopRecording(p, limit) {
+  function stopRecording(p, by) {
     if (p.stage !== "listening") return;
     if (p.limitTimer !== null) clearTimeout(p.limitTimer);
     if (p.launchTimer !== null) clearTimeout(p.launchTimer);
     p.limitTimer = null;
     p.launchTimer = null;
+    clearPause(p);
     p.stage = "finishing";
-    p.limit = limit;
+    p.limit = by === "limit";
+    p.by = by;
     const instance = p.instance;
     if (instance === null || !instance.started) {
       complete(p);
@@ -414,7 +666,7 @@ function createBrowserAvatarMic(options, seams = {}) {
       epoch,
       stage: "starting",
       instance: null,
-      earlier: [],
+      earlier: "",
       lang: options.lang,
       langRetried: false,
       restarts: 0,
@@ -427,7 +679,15 @@ function createBrowserAvatarMic(options, seams = {}) {
       startTimer: null,
       limitTimer: null,
       finalTimer: null,
-      launchTimer: null
+      launchTimer: null,
+      by: null,
+      words: "",
+      wordsAt: 0,
+      speechAt: null,
+      listenedAt: 0,
+      pausing: false,
+      pauseTimer: null,
+      pausingTimer: null
     };
     press = p;
     report({ name: "starting", epoch });
@@ -444,7 +704,7 @@ function createBrowserAvatarMic(options, seams = {}) {
       const p = press;
       if (p === null) return;
       if (p.stage === "starting") cancel();
-      else stopRecording(p, false);
+      else stopRecording(p, "tap");
     },
     cancel,
     state: () => state