@elizaos/shared 2.0.0-beta.1 → 2.0.3-beta.3

package/voice/first-sentence-snip.js

@@ -0,0 +1,284 @@
+/**
+ * First-sentence snip helper for the TTS first-line cache.
+ *
+ * Pure, deterministic, no I/O. Both the local runtime and the Eliza Cloud
+ * `/api/v1/voice/tts` route call into this module so that byte-equal sentence
+ * text on either side maps to byte-equal cache keys.
+ *
+ * Algorithm summary (see R4 §2 in `.swarm/research/R4-tts-cache.md`):
+ *   1. Trim leading whitespace.
+ *   2. Walk char-by-char, tracking quote depth (skip terminators inside
+ *      `"..."`/`'...'`/`“...”`), decimal context (digit-`.`-digit isn't a
+ *      boundary), and abbreviation context (e.g. `Mr.`, `e.g.`, `U.S.`).
+ *   3. First surviving terminator (`.`/`!`/`?`/`…`/`。`/`！`/`？`) or `\n`
+ *      ends the sentence. Consume any contiguous run of terminators that
+ *      follow (so `"Wait..."` stays intact).
+ *   4. If no terminator and no newline → null (don't cache unterminated text).
+ *   5. Apply ≤10-word filter on the normalised form. If wordCount > 10,
+ *      return null (don't cache).
+ *
+ * `FIRST_SENTENCE_SNIP_VERSION` is part of the cache key — bumping it
+ * invalidates every existing cached entry.
+ */
+/**
+ * Cache-key algorithm version. Bump when the snip/normalise logic changes
+ * in a way that would produce different keys for the same input. Local and
+ * cloud caches re-key off this constant, so a bump rolls the entire cache.
+ */
+export const FIRST_SENTENCE_SNIP_VERSION = "1";
+/**
+ * Maximum word count (Unicode-aware) for a cacheable first-sentence snip.
+ * Beyond this the bytes won't repeat often enough to amortise the cache.
+ */
+export const FIRST_SENTENCE_MAX_WORDS = 10;
+const ABBREVIATIONS = new Set([
+    "mr",
+    "mrs",
+    "ms",
+    "mx",
+    "dr",
+    "st",
+    "jr",
+    "sr",
+    "prof",
+    "vs",
+    "etc",
+    "eg",
+    "ie",
+    "fig",
+    "no",
+    "vol",
+    "ch",
+    "pt",
+    "co",
+    "inc",
+    "ltd",
+    "us",
+    "uk",
+    "usa",
+]);
+/**
+ * Terminator chars that end a sentence. Includes CJK `。！？` so multilingual
+ * inputs aren't refused.
+ */
+const TERMINATOR_REGEX = /[.!?…。！？]/u;
+const QUOTE_PAIRS = new Map([
+    ['"', '"'],
+    ["'", "'"],
+    ["“", "”"],
+    ["‘", "’"],
+    ["«", "»"],
+    ["「", "」"],
+    ["『", "』"],
+]);
+const CLOSE_QUOTES = new Set(QUOTE_PAIRS.values());
+function isDigit(ch) {
+    return ch >= "0" && ch <= "9";
+}
+function isAlpha(ch) {
+    if (!ch)
+        return false;
+    const code = ch.charCodeAt(0);
+    return ((code >= 65 && code <= 90) || // A-Z
+        (code >= 97 && code <= 122) // a-z
+    );
+}
+/**
+ * Lower-case + strip dots from a token to compare against `ABBREVIATIONS`.
+ * Used to detect things like `Mr.` / `U.S.` / `e.g.` where the dot is part of
+ * an abbreviation rather than a sentence terminator.
+ */
+function isAbbrevToken(token) {
+    if (token.length === 0 || token.length > 6)
+        return false;
+    const lower = token.toLowerCase().replace(/\./g, "");
+    if (!lower)
+        return false;
+    return ABBREVIATIONS.has(lower);
+}
+/**
+ * Walk the input forwards looking for the first sentence-terminating position.
+ * Returns the *inclusive* end index of the terminator run (so the snip is
+ * `text.slice(start, endInclusive + 1)`), or -1 if no terminator was found.
+ *
+ * `start` is the position after any leading whitespace.
+ */
+function findTerminatorEnd(text, start) {
+    let i = start;
+    const quoteStack = [];
+    while (i < text.length) {
+        const ch = text[i] ?? "";
+        // Newline acts as a soft terminator (e.g. message body broken into lines).
+        if (ch === "\n") {
+            // Strip trailing whitespace from the run by returning index of \n.
+            return i;
+        }
+        // Open / close quotes for skipping terminators inside.
+        const closer = QUOTE_PAIRS.get(ch);
+        if (closer && closer !== ch) {
+            // Asymmetric pair → push the expected closer.
+            quoteStack.push(closer);
+            i++;
+            continue;
+        }
+        if (quoteStack.length > 0 && ch === quoteStack[quoteStack.length - 1]) {
+            quoteStack.pop();
+            i++;
+            continue;
+        }
+        // Symmetric ASCII quotes: toggle stack only when not already inside an
+        // asymmetric quote.
+        //
+        // Apostrophes (`'`) are NEVER treated as quote delimiters when they sit
+        // between two letters (contractions like `it's`, possessives like
+        // `Eliza's`) — otherwise a stray apostrophe would swallow the rest of
+        // the input and we'd never find a terminator.
+        if (ch === "'" && quoteStack.length === 0) {
+            const prev = i > 0 ? text[i - 1] : "";
+            const next = i + 1 < text.length ? text[i + 1] : "";
+            if (!(isAlpha(prev) && isAlpha(next))) {
+                quoteStack.push(ch);
+            }
+            i++;
+            continue;
+        }
+        if (ch === '"' && quoteStack.length === 0) {
+            quoteStack.push(ch);
+            i++;
+            continue;
+        }
+        if (quoteStack.length > 0 &&
+            CLOSE_QUOTES.has(quoteStack[quoteStack.length - 1] ?? "") &&
+            ch === quoteStack[quoteStack.length - 1]) {
+            quoteStack.pop();
+            i++;
+            continue;
+        }
+        // Skip everything inside quotes.
+        if (quoteStack.length > 0) {
+            i++;
+            continue;
+        }
+        if (!TERMINATOR_REGEX.test(ch)) {
+            i++;
+            continue;
+        }
+        // Candidate terminator. Apply guards.
+        if (ch === ".") {
+            // Decimal context: digit . digit
+            const prev = i > 0 ? text[i - 1] : "";
+            const next = i + 1 < text.length ? text[i + 1] : "";
+            if (isDigit(prev) && isDigit(next)) {
+                i++;
+                continue;
+            }
+            // Abbreviation context: look backwards for an alpha run, possibly
+            // with internal dots (e.g. `U.S.`). Bound the look-back to 6 chars.
+            let j = i - 1;
+            let token = "";
+            while (j >= 0 && j >= i - 6) {
+                const tj = text[j] ?? "";
+                if (isAlpha(tj) || tj === ".") {
+                    token = tj + token;
+                    j--;
+                    continue;
+                }
+                break;
+            }
+            if (token && isAbbrevToken(token)) {
+                i++;
+                continue;
+            }
+            // Also: avoid treating mid-word dot as terminator if followed
+            // immediately by alpha and no whitespace (e.g. "foo.bar"
+            // shouldn't terminate). Cache hit rate is more important than
+            // being clever here — conservative: skip.
+            if (isAlpha(next)) {
+                i++;
+                continue;
+            }
+        }
+        // Valid terminator. Consume any contiguous run of terminator chars so
+        // `Wait...` stays intact, then return the inclusive end.
+        let end = i;
+        while (end + 1 < text.length) {
+            const nxt = text[end + 1] ?? "";
+            if (TERMINATOR_REGEX.test(nxt)) {
+                end++;
+                continue;
+            }
+            break;
+        }
+        return end;
+    }
+    return -1;
+}
+/**
+ * Unicode-aware word count. Hyphenated words and apostrophe-internal words
+ * count as one (e.g. "twenty-three" → 1, "it's" → 1). Dotted acronyms with
+ * single-letter segments (e.g. "U.S.", "U.S.A.") count as one word. Decimal
+ * numbers like "3.14" count as one word.
+ */
+export function wordCount(s) {
+    if (!s)
+        return 0;
+    const matches = s.match(
+    // In priority order:
+    //   - decimal number (digits . digits, possibly multi-segment)
+    //   - dotted acronym (≥2 single-letter dot pairs)
+    //   - regular word with optional hyphen/apostrophe internal joiners
+    /\p{N}+(?:\.\p{N}+)+|(?:\p{L}\.){2,}|[\p{L}\p{N}]+(?:[’'-][\p{L}\p{N}]+)*/gu);
+    return matches ? matches.length : 0;
+}
+/**
+ * Normalise a snip for cache-key purposes. NFC + lower-case + collapse
+ * whitespace + strip trailing terminators/whitespace. Apostrophes preserved.
+ */
+export function normalizeForKey(snip) {
+    let s = snip.normalize("NFC");
+    // Trim leading whitespace including zero-width space (U+200B).
+    s = s.replace(/^[\s​]+/u, "").replace(/[\s​]+$/u, "");
+    s = s.toLowerCase();
+    s = s.replace(/\s+/gu, " ");
+    // Strip trailing terminator/whitespace run. CJK terminators included.
+    s = s.replace(/[\s​.!?…。！？]+$/gu, "");
+    return s;
+}
+/**
+ * Attempt to snip the first sentence from `text`. Returns `null` if:
+ *   - input is empty / whitespace only
+ *   - no sentence-terminator found
+ *   - normalised snip has > 10 words
+ *   - normalised snip is empty after stripping terminators
+ *
+ * Otherwise returns a `FirstSentenceSnipResult` carrying the raw and
+ * normalised forms plus the word count.
+ */
+export function firstSentenceSnip(text) {
+    if (typeof text !== "string" || text.length === 0)
+        return null;
+    // Find start after leading whitespace (incl. zero-width space).
+    const trimMatch = /^[\s​]+/u.exec(text);
+    const start = trimMatch ? trimMatch[0].length : 0;
+    if (start >= text.length)
+        return null;
+    const end = findTerminatorEnd(text, start);
+    if (end < 0)
+        return null;
+    const raw = text.slice(start, end + 1);
+    const normalized = normalizeForKey(raw);
+    if (!normalized)
+        return null;
+    const wc = wordCount(normalized);
+    if (wc === 0)
+        return null;
+    if (wc > FIRST_SENTENCE_MAX_WORDS)
+        return null;
+    return {
+        raw,
+        normalized,
+        wordCount: wc,
+        endOffset: end + 1,
+    };
+}
+//# sourceMappingURL=first-sentence-snip.js.map

package/voice/first-sentence-snip.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"first-sentence-snip.js","sourceRoot":"","sources":["../../src/voice/first-sentence-snip.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;GAqBG;AAEH;;;;GAIG;AACH,MAAM,CAAC,MAAM,2BAA2B,GAAG,GAAY,CAAC;AAGxD;;;GAGG;AACH,MAAM,CAAC,MAAM,wBAAwB,GAAG,EAAW,CAAC;AAEpD,MAAM,aAAa,GAAG,IAAI,GAAG,CAAS;IACpC,IAAI;IACJ,KAAK;IACL,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,MAAM;IACN,IAAI;IACJ,KAAK;IACL,IAAI;IACJ,IAAI;IACJ,KAAK;IACL,IAAI;IACJ,KAAK;IACL,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,KAAK;IACL,KAAK;IACL,IAAI;IACJ,IAAI;IACJ,KAAK;CACN,CAAC,CAAC;AAEH;;;GAGG;AACH,MAAM,gBAAgB,GAAG,YAAY,CAAC;AAEtC,MAAM,WAAW,GAAgC,IAAI,GAAG,CAAC;IACvD,CAAC,GAAG,EAAE,GAAG,CAAC;IACV,CAAC,GAAG,EAAE,GAAG,CAAC;IACV,CAAC,GAAG,EAAE,GAAG,CAAC;IACV,CAAC,GAAG,EAAE,GAAG,CAAC;IACV,CAAC,GAAG,EAAE,GAAG,CAAC;IACV,CAAC,GAAG,EAAE,GAAG,CAAC;IACV,CAAC,GAAG,EAAE,GAAG,CAAC;CACX,CAAC,CAAC;AAEH,MAAM,YAAY,GAAwB,IAAI,GAAG,CAAC,WAAW,CAAC,MAAM,EAAE,CAAC,CAAC;AAExE,SAAS,OAAO,CAAC,EAAU;IACzB,OAAO,EAAE,IAAI,GAAG,IAAI,EAAE,IAAI,GAAG,CAAC;AAChC,CAAC;AAED,SAAS,OAAO,CAAC,EAAU;IACzB,IAAI,CAAC,EAAE;QAAE,OAAO,KAAK,CAAC;IACtB,MAAM,IAAI,GAAG,EAAE,CAAC,UAAU,CAAC,CAAC,CAAC,CAAC;IAC9B,OAAO,CACL,CAAC,IAAI,IAAI,EAAE,IAAI,IAAI,IAAI,EAAE,CAAC,IAAI,MAAM;QACpC,CAAC,IAAI,IAAI,EAAE,IAAI,IAAI,IAAI,GAAG,CAAC,CAAC,MAAM;KACnC,CAAC;AACJ,CAAC;AAED;;;;GAIG;AACH,SAAS,aAAa,CAAC,KAAa;IAClC,IAAI,KAAK,CAAC,MAAM,KAAK,CAAC,IAAI,KAAK,CAAC,MAAM,GAAG,CAAC;QAAE,OAAO,KAAK,CAAC;IACzD,MAAM,KAAK,GAAG,KAAK,CAAC,WAAW,EAAE,CAAC,OAAO,CAAC,KAAK,EAAE,EAAE,CAAC,CAAC;IACrD,IAAI,CAAC,KAAK;QAAE,OAAO,KAAK,CAAC;IACzB,OAAO,aAAa,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;AAClC,CAAC;AAED;;;;;;GAMG;AACH,SAAS,iBAAiB,CAAC,IAAY,EAAE,KAAa;IACpD,IAAI,CAAC,GAAG,KAAK,CAAC;IACd,MAAM,UAAU,GAAa,EAAE,CAAC;IAEhC,OAAO,CAAC,GAAG,IAAI,CAAC,MAAM,EAAE,CAAC;QACvB,MAAM,EAAE,GAAG,IAAI,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC;QAEzB,2EAA2E;QAC3E,IAAI,EAAE,KAAK,IAAI,EAAE,CAAC;YAChB,mEAAmE;YACnE,OAAO,CAAC,CAAC;QACX,CAAC;QAED,uDAAuD;QACvD,MAAM,MAAM,GAAG,WAAW,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC;QACnC,IAAI,MAAM,IAAI,MAAM,KAAK,EAAE,EAAE,CAAC;YAC5B,8CAA8C;YAC9C,UAAU,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;YACxB,CAAC,EAAE,CAAC;YACJ,SAAS;QACX,CAAC;QACD,IAAI,UAAU,CAAC,MAAM,GAAG,CAAC,IAAI,EAAE,KAAK,UAAU,CAAC,UAAU,CAAC,MAAM,GAAG,CAAC,CAAC,EAAE,CAAC;YACtE,UAAU,CAAC,GAAG,EAAE,CAAC;YACjB,CAAC,EAAE,CAAC;YACJ,SAAS;QACX,CAAC;QACD,uEAAuE;QACvE,oBAAoB;QACpB,EAAE;QACF,wEAAwE;QACxE,kEAAkE;QAClE,sEAAsE;QACtE,8CAA8C;QAC9C,IAAI,EAAE,KAAK,GAAG,IAAI,UAAU,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YAC1C,MAAM,IAAI,GAAG,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC;YACtC,MAAM,IAAI,GAAG,CAAC,GAAG,CAAC,GAAG,IAAI,CAAC,MAAM,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC;YACpD,IAAI,CAAC,CAAC,OAAO,CAAC,IAAI,CAAC,IAAI,OAAO,CAAC,IAAI,CAAC,CAAC,EAAE,CAAC;gBACtC,UAAU,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;YACtB,CAAC;YACD,CAAC,EAAE,CAAC;YACJ,SAAS;QACX,CAAC;QACD,IAAI,EAAE,KAAK,GAAG,IAAI,UAAU,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YAC1C,UAAU,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;YACpB,CAAC,EAAE,CAAC;YACJ,SAAS;QACX,CAAC;QACD,IACE,UAAU,CAAC,MAAM,GAAG,CAAC;YACrB,YAAY,CAAC,GAAG,CAAC,UAAU,CAAC,UAAU,CAAC,MAAM,GAAG,CAAC,CAAC,IAAI,EAAE,CAAC;YACzD,EAAE,KAAK,UAAU,CAAC,UAAU,CAAC,MAAM,GAAG,CAAC,CAAC,EACxC,CAAC;YACD,UAAU,CAAC,GAAG,EAAE,CAAC;YACjB,CAAC,EAAE,CAAC;YACJ,SAAS;QACX,CAAC;QAED,iCAAiC;QACjC,IAAI,UAAU,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;YAC1B,CAAC,EAAE,CAAC;YACJ,SAAS;QACX,CAAC;QAED,IAAI,CAAC,gBAAgB,CAAC,IAAI,CAAC,EAAE,CAAC,EAAE,CAAC;YAC/B,CAAC,EAAE,CAAC;YACJ,SAAS;QACX,CAAC;QAED,sCAAsC;QACtC,IAAI,EAAE,KAAK,GAAG,EAAE,CAAC;YACf,iCAAiC;YACjC,MAAM,IAAI,GAAG,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC;YACtC,MAAM,IAAI,GAAG,CAAC,GAAG,CAAC,GAAG,IAAI,CAAC,MAAM,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC;YACpD,IAAI,OAAO,CAAC,IAAI,CAAC,IAAI,OAAO,CAAC,IAAI,CAAC,EAAE,CAAC;gBACnC,CAAC,EAAE,CAAC;gBACJ,SAAS;YACX,CAAC;YACD,kEAAkE;YAClE,oEAAoE;YACpE,IAAI,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC;YACd,IAAI,KAAK,GAAG,EAAE,CAAC;YACf,OAAO,CAAC,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC;gBAC5B,MAAM,EAAE,GAAG,IAAI,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC;gBACzB,IAAI,OAAO,CAAC,EAAE,CAAC,IAAI,EAAE,KAAK,GAAG,EAAE,CAAC;oBAC9B,KAAK,GAAG,EAAE,GAAG,KAAK,CAAC;oBACnB,CAAC,EAAE,CAAC;oBACJ,SAAS;gBACX,CAAC;gBACD,MAAM;YACR,CAAC;YACD,IAAI,KAAK,IAAI,aAAa,CAAC,KAAK,CAAC,EAAE,CAAC;gBAClC,CAAC,EAAE,CAAC;gBACJ,SAAS;YACX,CAAC;YACD,8DAA8D;YAC9D,yDAAyD;YACzD,8DAA8D;YAC9D,0CAA0C;YAC1C,IAAI,OAAO,CAAC,IAAI,CAAC,EAAE,CAAC;gBAClB,CAAC,EAAE,CAAC;gBACJ,SAAS;YACX,CAAC;QACH,CAAC;QAED,sEAAsE;QACtE,yDAAyD;QACzD,IAAI,GAAG,GAAG,CAAC,CAAC;QACZ,OAAO,GAAG,GAAG,CAAC,GAAG,IAAI,CAAC,MAAM,EAAE,CAAC;YAC7B,MAAM,GAAG,GAAG,IAAI,CAAC,GAAG,GAAG,CAAC,CAAC,IAAI,EAAE,CAAC;YAChC,IAAI,gBAAgB,CAAC,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC;gBAC/B,GAAG,EAAE,CAAC;gBACN,SAAS;YACX,CAAC;YACD,MAAM;QACR,CAAC;QACD,OAAO,GAAG,CAAC;IACb,CAAC;IACD,OAAO,CAAC,CAAC,CAAC;AACZ,CAAC;AAED;;;;;GAKG;AACH,MAAM,UAAU,SAAS,CAAC,CAAS;IACjC,IAAI,CAAC,CAAC;QAAE,OAAO,CAAC,CAAC;IACjB,MAAM,OAAO,GAAG,CAAC,CAAC,KAAK;IACrB,qBAAqB;IACrB,+DAA+D;IAC/D,kDAAkD;IAClD,oEAAoE;IACpE,4EAA4E,CAC7E,CAAC;IACF,OAAO,OAAO,CAAC,CAAC,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC;AACtC,CAAC;AAED;;;GAGG;AACH,MAAM,UAAU,eAAe,CAAC,IAAY;IAC1C,IAAI,CAAC,GAAG,IAAI,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC;IAC9B,+DAA+D;IAC/D,CAAC,GAAG,CAAC,CAAC,OAAO,CAAC,UAAU,EAAE,EAAE,CAAC,CAAC,OAAO,CAAC,UAAU,EAAE,EAAE,CAAC,CAAC;IACtD,CAAC,GAAG,CAAC,CAAC,WAAW,EAAE,CAAC;IACpB,CAAC,GAAG,CAAC,CAAC,OAAO,CAAC,OAAO,EAAE,GAAG,CAAC,CAAC;IAC5B,sEAAsE;IACtE,CAAC,GAAG,CAAC,CAAC,OAAO,CAAC,kBAAkB,EAAE,EAAE,CAAC,CAAC;IACtC,OAAO,CAAC,CAAC;AACX,CAAC;AAaD;;;;;;;;;GASG;AACH,MAAM,UAAU,iBAAiB,CAC/B,IAAY;IAEZ,IAAI,OAAO,IAAI,KAAK,QAAQ,IAAI,IAAI,CAAC,MAAM,KAAK,CAAC;QAAE,OAAO,IAAI,CAAC;IAE/D,gEAAgE;IAChE,MAAM,SAAS,GAAG,UAAU,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;IACxC,MAAM,KAAK,GAAG,SAAS,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC;IAClD,IAAI,KAAK,IAAI,IAAI,CAAC,MAAM;QAAE,OAAO,IAAI,CAAC;IAEtC,MAAM,GAAG,GAAG,iBAAiB,CAAC,IAAI,EAAE,KAAK,CAAC,CAAC;IAC3C,IAAI,GAAG,GAAG,CAAC;QAAE,OAAO,IAAI,CAAC;IAEzB,MAAM,GAAG,GAAG,IAAI,CAAC,KAAK,CAAC,KAAK,EAAE,GAAG,GAAG,CAAC,CAAC,CAAC;IACvC,MAAM,UAAU,GAAG,eAAe,CAAC,GAAG,CAAC,CAAC;IACxC,IAAI,CAAC,UAAU;QAAE,OAAO,IAAI,CAAC;IAE7B,MAAM,EAAE,GAAG,SAAS,CAAC,UAAU,CAAC,CAAC;IACjC,IAAI,EAAE,KAAK,CAAC;QAAE,OAAO,IAAI,CAAC;IAC1B,IAAI,EAAE,GAAG,wBAAwB;QAAE,OAAO,IAAI,CAAC;IAE/C,OAAO;QACL,GAAG;QACH,UAAU;QACV,SAAS,EAAE,EAAE;QACb,SAAS,EAAE,GAAG,GAAG,CAAC;KACnB,CAAC;AACJ,CAAC"}

package/voice/owner-inference.d.ts

@@ -0,0 +1,57 @@
+/**
+ * Owner-candidate inference for voice (#8785).
+ *
+ * "How does the agent know which speaker is the device OWNER?" Today the owner
+ * entity must be enrolled explicitly (the first-run voice flow). But when no
+ * owner is set, the agent should still be able to FORM a hypothesis from what it
+ * hears — the person who speaks to it most, most confidently, and clearly more
+ * than anyone else is the likely owner. This is the pure decision logic a
+ * provider/evaluator runs when the owner is unknown: it accumulates recognized
+ * voice turns and proposes a candidate only when the evidence is both sufficient
+ * (enough confident observations) and unambiguous (a clear lead over the
+ * runner-up). Otherwise it stays UNDECIDED — it never guesses an owner from one
+ * stray turn, because a wrong owner is a security and personalization hazard.
+ *
+ * Pure (no I/O, no models); the runtime feeds it the diarized/recognized
+ * observations and acts on a decided candidate (e.g. prompt to confirm, or set
+ * the owner setting). The Voice Workbench exercises the SAME function so the
+ * inference is benchmarked, not just shipped.
+ */
+/** One recognized voice turn: which enrolled/clustered speaker, how confident. */
+export interface OwnerObservation {
+    /** The entity/cluster the recognized voice resolved to (null = unrecognized). */
+    entityId: string | null;
+    /** Recognition confidence 0..1 (cosine-rescaled by the attribution pipeline). */
+    confidence: number;
+}
+export interface OwnerInferenceOptions {
+    /** Minimum qualifying observations before any candidate is proposed. */
+    minObservations?: number;
+    /** Confidence floor; observations below it don't count toward a candidate. */
+    minConfidence?: number;
+    /**
+     * Minimum lead (in confidence-weighted score) the top speaker must hold over
+     * the runner-up to be unambiguous. Prevents naming an owner in a two-equals
+     * household.
+     */
+    minMargin?: number;
+}
+export interface OwnerInferenceResult {
+    /** The proposed owner entity, or null when the evidence is insufficient. */
+    ownerEntityId: string | null;
+    /** Confidence-weighted share of the proposed owner (0..1), 0 when undecided. */
+    share: number;
+    /** Number of qualifying (confident, recognized) observations considered. */
+    qualifyingObservations: number;
+    /** Why the function decided / declined — surfaced for the provider's logs. */
+    reason: string;
+}
+/**
+ * Propose the most likely owner from recognized voice observations, or stay
+ * undecided. A candidate is returned only when there are at least
+ * `minObservations` confident, recognized turns AND the top speaker leads the
+ * runner-up by at least `minMargin` (confidence-weighted). Ties and thin
+ * evidence yield `ownerEntityId: null`.
+ */
+export declare function resolveOwnerCandidate(observations: ReadonlyArray<OwnerObservation>, options?: OwnerInferenceOptions): OwnerInferenceResult;
+//# sourceMappingURL=owner-inference.d.ts.map

package/voice/owner-inference.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"owner-inference.d.ts","sourceRoot":"","sources":["../../src/voice/owner-inference.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;GAkBG;AAEH,kFAAkF;AAClF,MAAM,WAAW,gBAAgB;IAC/B,iFAAiF;IACjF,QAAQ,EAAE,MAAM,GAAG,IAAI,CAAC;IACxB,iFAAiF;IACjF,UAAU,EAAE,MAAM,CAAC;CACpB;AAED,MAAM,WAAW,qBAAqB;IACpC,wEAAwE;IACxE,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,8EAA8E;IAC9E,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB;;;;OAIG;IACH,SAAS,CAAC,EAAE,MAAM,CAAC;CACpB;AAED,MAAM,WAAW,oBAAoB;IACnC,4EAA4E;IAC5E,aAAa,EAAE,MAAM,GAAG,IAAI,CAAC;IAC7B,gFAAgF;IAChF,KAAK,EAAE,MAAM,CAAC;IACd,4EAA4E;IAC5E,sBAAsB,EAAE,MAAM,CAAC;IAC/B,8EAA8E;IAC9E,MAAM,EAAE,MAAM,CAAC;CAChB;AAMD;;;;;;GAMG;AACH,wBAAgB,qBAAqB,CACnC,YAAY,EAAE,aAAa,CAAC,gBAAgB,CAAC,EAC7C,OAAO,GAAE,qBAA0B,GAClC,oBAAoB,CA4CtB"}

package/voice/owner-inference.js

@@ -0,0 +1,73 @@
+/**
+ * Owner-candidate inference for voice (#8785).
+ *
+ * "How does the agent know which speaker is the device OWNER?" Today the owner
+ * entity must be enrolled explicitly (the first-run voice flow). But when no
+ * owner is set, the agent should still be able to FORM a hypothesis from what it
+ * hears — the person who speaks to it most, most confidently, and clearly more
+ * than anyone else is the likely owner. This is the pure decision logic a
+ * provider/evaluator runs when the owner is unknown: it accumulates recognized
+ * voice turns and proposes a candidate only when the evidence is both sufficient
+ * (enough confident observations) and unambiguous (a clear lead over the
+ * runner-up). Otherwise it stays UNDECIDED — it never guesses an owner from one
+ * stray turn, because a wrong owner is a security and personalization hazard.
+ *
+ * Pure (no I/O, no models); the runtime feeds it the diarized/recognized
+ * observations and acts on a decided candidate (e.g. prompt to confirm, or set
+ * the owner setting). The Voice Workbench exercises the SAME function so the
+ * inference is benchmarked, not just shipped.
+ */
+const DEFAULT_MIN_OBSERVATIONS = 3;
+const DEFAULT_MIN_CONFIDENCE = 0.7;
+const DEFAULT_MIN_MARGIN = 1;
+/**
+ * Propose the most likely owner from recognized voice observations, or stay
+ * undecided. A candidate is returned only when there are at least
+ * `minObservations` confident, recognized turns AND the top speaker leads the
+ * runner-up by at least `minMargin` (confidence-weighted). Ties and thin
+ * evidence yield `ownerEntityId: null`.
+ */
+export function resolveOwnerCandidate(observations, options = {}) {
+    const minObservations = options.minObservations ?? DEFAULT_MIN_OBSERVATIONS;
+    const minConfidence = options.minConfidence ?? DEFAULT_MIN_CONFIDENCE;
+    const minMargin = options.minMargin ?? DEFAULT_MIN_MARGIN;
+    const scores = new Map();
+    let qualifying = 0;
+    let totalScore = 0;
+    for (const obs of observations) {
+        if (obs.entityId === null)
+            continue;
+        if (!(obs.confidence >= minConfidence))
+            continue;
+        qualifying += 1;
+        const next = (scores.get(obs.entityId) ?? 0) + obs.confidence;
+        scores.set(obs.entityId, next);
+        totalScore += obs.confidence;
+    }
+    if (qualifying < minObservations) {
+        return {
+            ownerEntityId: null,
+            share: 0,
+            qualifyingObservations: qualifying,
+            reason: `insufficient evidence (${qualifying}/${minObservations} confident observations)`,
+        };
+    }
+    const ranked = [...scores.entries()].sort((a, b) => b[1] - a[1]);
+    const [topId, topScore] = ranked[0];
+    const runnerUpScore = ranked[1]?.[1] ?? 0;
+    if (topScore - runnerUpScore < minMargin) {
+        return {
+            ownerEntityId: null,
+            share: totalScore > 0 ? topScore / totalScore : 0,
+            qualifyingObservations: qualifying,
+            reason: `ambiguous lead (top ${topScore.toFixed(2)} vs runner-up ${runnerUpScore.toFixed(2)}, margin < ${minMargin})`,
+        };
+    }
+    return {
+        ownerEntityId: topId,
+        share: totalScore > 0 ? topScore / totalScore : 0,
+        qualifyingObservations: qualifying,
+        reason: `dominant speaker (${topScore.toFixed(2)} of ${totalScore.toFixed(2)} over ${qualifying} observations)`,
+    };
+}
+//# sourceMappingURL=owner-inference.js.map

package/voice/owner-inference.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"owner-inference.js","sourceRoot":"","sources":["../../src/voice/owner-inference.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;GAkBG;AAkCH,MAAM,wBAAwB,GAAG,CAAC,CAAC;AACnC,MAAM,sBAAsB,GAAG,GAAG,CAAC;AACnC,MAAM,kBAAkB,GAAG,CAAC,CAAC;AAE7B;;;;;;GAMG;AACH,MAAM,UAAU,qBAAqB,CACnC,YAA6C,EAC7C,UAAiC,EAAE;IAEnC,MAAM,eAAe,GAAG,OAAO,CAAC,eAAe,IAAI,wBAAwB,CAAC;IAC5E,MAAM,aAAa,GAAG,OAAO,CAAC,aAAa,IAAI,sBAAsB,CAAC;IACtE,MAAM,SAAS,GAAG,OAAO,CAAC,SAAS,IAAI,kBAAkB,CAAC;IAE1D,MAAM,MAAM,GAAG,IAAI,GAAG,EAAkB,CAAC;IACzC,IAAI,UAAU,GAAG,CAAC,CAAC;IACnB,IAAI,UAAU,GAAG,CAAC,CAAC;IACnB,KAAK,MAAM,GAAG,IAAI,YAAY,EAAE,CAAC;QAC/B,IAAI,GAAG,CAAC,QAAQ,KAAK,IAAI;YAAE,SAAS;QACpC,IAAI,CAAC,CAAC,GAAG,CAAC,UAAU,IAAI,aAAa,CAAC;YAAE,SAAS;QACjD,UAAU,IAAI,CAAC,CAAC;QAChB,MAAM,IAAI,GAAG,CAAC,MAAM,CAAC,GAAG,CAAC,GAAG,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAC,GAAG,GAAG,CAAC,UAAU,CAAC;QAC9D,MAAM,CAAC,GAAG,CAAC,GAAG,CAAC,QAAQ,EAAE,IAAI,CAAC,CAAC;QAC/B,UAAU,IAAI,GAAG,CAAC,UAAU,CAAC;IAC/B,CAAC;IAED,IAAI,UAAU,GAAG,eAAe,EAAE,CAAC;QACjC,OAAO;YACL,aAAa,EAAE,IAAI;YACnB,KAAK,EAAE,CAAC;YACR,sBAAsB,EAAE,UAAU;YAClC,MAAM,EAAE,0BAA0B,UAAU,IAAI,eAAe,0BAA0B;SAC1F,CAAC;IACJ,CAAC;IAED,MAAM,MAAM,GAAG,CAAC,GAAG,MAAM,CAAC,OAAO,EAAE,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;IACjE,MAAM,CAAC,KAAK,EAAE,QAAQ,CAAC,GAAG,MAAM,CAAC,CAAC,CAAC,CAAC;IACpC,MAAM,aAAa,GAAG,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC;IAC1C,IAAI,QAAQ,GAAG,aAAa,GAAG,SAAS,EAAE,CAAC;QACzC,OAAO;YACL,aAAa,EAAE,IAAI;YACnB,KAAK,EAAE,UAAU,GAAG,CAAC,CAAC,CAAC,CAAC,QAAQ,GAAG,UAAU,CAAC,CAAC,CAAC,CAAC;YACjD,sBAAsB,EAAE,UAAU;YAClC,MAAM,EAAE,uBAAuB,QAAQ,CAAC,OAAO,CAAC,CAAC,CAAC,iBAAiB,aAAa,CAAC,OAAO,CAAC,CAAC,CAAC,cAAc,SAAS,GAAG;SACtH,CAAC;IACJ,CAAC;IAED,OAAO;QACL,aAAa,EAAE,KAAK;QACpB,KAAK,EAAE,UAAU,GAAG,CAAC,CAAC,CAAC,CAAC,QAAQ,GAAG,UAAU,CAAC,CAAC,CAAC,CAAC;QACjD,sBAAsB,EAAE,UAAU;QAClC,MAAM,EAAE,qBAAqB,QAAQ,CAAC,OAAO,CAAC,CAAC,CAAC,OAAO,UAAU,CAAC,OAAO,CAAC,CAAC,CAAC,SAAS,UAAU,gBAAgB;KAChH,CAAC;AACJ,CAAC"}

package/voice/respond-gate.d.ts

@@ -0,0 +1,88 @@
+/**
+ * Canonical client-side shouldRespond / echo-rejection gate for always-on voice.
+ *
+ * In always-on ("hands-free") mode the mic is open continuously, so the
+ * recognizer transcribes EVERYTHING it hears — the agent's own text-to-speech
+ * bleeding back into the mic, the speaker's disfluent thinking noises ("um…"),
+ * and other people in the room. Sending those as turns makes the agent reply to
+ * itself, to filler, or to a bystander — the "responding when it's not
+ * appropriate" problem.
+ *
+ * This is the SINGLE definition of that gate, shared by:
+ *   - the UI shell capture loop (`@elizaos/ui` `voice/should-respond` +
+ *     `voice/voice-turn-signal` re-export from here), and
+ *   - the Voice Workbench headless runner (the plugin's real-decision-logic
+ *     services adapter calls these so the benchmark exercises the SAME code the
+ *     client runs — no drift between what we test and what we ship).
+ *
+ * Pure (no DOM, no I/O); the only dependency is the canonical syntactic EOT
+ * heuristic in `@elizaos/shared/voice-eot`.
+ */
+/** How recent the agent's reply must be for the echo guard to apply. */
+export declare const ECHO_WINDOW_MS = 9000;
+/** Word-overlap fraction above which a turn is treated as TTS echo. */
+export declare const ECHO_OVERLAP_THRESHOLD = 0.7;
+export interface ShouldRespondContext {
+    /** The agent's most recent spoken reply, for the echo guard. */
+    recentAgentReply?: string;
+    /** Age of that reply in ms; the echo guard applies while it's recent. */
+    replyAgeMs?: number;
+    /**
+     * True while the agent is CURRENTLY speaking. Forces the echo guard on
+     * regardless of `replyAgeMs`, because a long reply's TTS is actively bleeding
+     * into the open mic even though its message was created many seconds ago (the
+     * age-only window would have already expired mid-speech).
+     */
+    agentSpeaking?: boolean;
+}
+/**
+ * Whether a transcribed voice turn should be sent to the agent (i.e. warrants a
+ * response). Returns false for pure disfluency and for near-verbatim echoes of
+ * the agent's recent speech.
+ */
+export declare function shouldRespondToVoiceTurn(transcript: string, context?: ShouldRespondContext): boolean;
+/** Mirrors the server-side VoiceTurnSignalMetadata shape the gate parses. */
+export interface VoiceTurnSignal {
+    endOfTurnProbability: number;
+    nextSpeaker: "agent" | "user" | "unknown";
+    agentShouldSpeak: boolean;
+    source: string;
+}
+/** Live speaker attribution from diarization (only where audio frames exist). */
+export interface VoiceTurnSpeakerAttribution {
+    /** Enrolled entity this turn was attributed to, or null when unknown. */
+    entityId: string | null;
+    /** Match confidence 0..1 (cosine-rescaled by the attribution pipeline). */
+    confidence: number;
+    /** True when attributed to the device owner / primary enrolled speaker. */
+    isOwner?: boolean;
+}
+export interface BuildVoiceTurnSignalContext extends ShouldRespondContext {
+    /** Speaker attribution for this turn (diarization; desktop/server only). */
+    speaker?: VoiceTurnSpeakerAttribution;
+    /** True when a wake word ("hey eliza") fired within the recent listen window. */
+    wakeWordActive?: boolean;
+    /** Entity ids the agent answers to without a wake word (owner + enrolled). */
+    knownSpeakerEntityIds?: readonly string[];
+    /**
+     * Cosine similarity (0..1) of THIS turn's speaker embedding against the
+     * agent's own TTS-voice imprint, when the speaker encoder is available. High =
+     * the agent is hearing ITSELF (its TTS bled back into the mic). This is the
+     * ACOUSTIC self-echo signal — it catches an echo the transcript word-overlap
+     * guard misses (a mis-transcribed echo whose words don't match the reply).
+     */
+    selfVoiceSimilarity?: number;
+}
+/** Server SUPPRESS threshold for EOT — below this reads as "user still talking". */
+export declare const SERVER_EOT_SUPPRESS_THRESHOLD = 0.4;
+/** Only a CONFIDENT bystander attribution is allowed to silence a turn. */
+export declare const BYSTANDER_SUPPRESS_CONFIDENCE = 0.7;
+/**
+ * Cosine at/above which an incoming turn is treated as the agent's OWN voice and
+ * hard-suppressed. WeSpeaker enrolls the device owner at ~0.78; the agent's own
+ * TTS imprint is a tighter, single-source match, so a slightly lower bar is
+ * safe and catches echo the transcript guard cannot.
+ */
+export declare const AGENT_SELF_VOICE_THRESHOLD = 0.7;
+export declare function buildVoiceTurnSignal(transcript: string, context?: BuildVoiceTurnSignalContext): VoiceTurnSignal;
+//# sourceMappingURL=respond-gate.d.ts.map

package/voice/respond-gate.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"respond-gate.d.ts","sourceRoot":"","sources":["../../src/voice/respond-gate.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;GAmBG;AAqBH,wEAAwE;AACxE,eAAO,MAAM,cAAc,OAAO,CAAC;AACnC,uEAAuE;AACvE,eAAO,MAAM,sBAAsB,MAAM,CAAC;AAU1C,MAAM,WAAW,oBAAoB;IACnC,gEAAgE;IAChE,gBAAgB,CAAC,EAAE,MAAM,CAAC;IAC1B,yEAAyE;IACzE,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB;;;;;OAKG;IACH,aAAa,CAAC,EAAE,OAAO,CAAC;CACzB;AAED;;;;GAIG;AACH,wBAAgB,wBAAwB,CACtC,UAAU,EAAE,MAAM,EAClB,OAAO,GAAE,oBAAyB,GACjC,OAAO,CAuBT;AAED,6EAA6E;AAC7E,MAAM,WAAW,eAAe;IAC9B,oBAAoB,EAAE,MAAM,CAAC;IAC7B,WAAW,EAAE,OAAO,GAAG,MAAM,GAAG,SAAS,CAAC;IAC1C,gBAAgB,EAAE,OAAO,CAAC;IAC1B,MAAM,EAAE,MAAM,CAAC;CAChB;AAED,iFAAiF;AACjF,MAAM,WAAW,2BAA2B;IAC1C,yEAAyE;IACzE,QAAQ,EAAE,MAAM,GAAG,IAAI,CAAC;IACxB,2EAA2E;IAC3E,UAAU,EAAE,MAAM,CAAC;IACnB,2EAA2E;IAC3E,OAAO,CAAC,EAAE,OAAO,CAAC;CACnB;AAED,MAAM,WAAW,2BAA4B,SAAQ,oBAAoB;IACvE,4EAA4E;IAC5E,OAAO,CAAC,EAAE,2BAA2B,CAAC;IACtC,iFAAiF;IACjF,cAAc,CAAC,EAAE,OAAO,CAAC;IACzB,8EAA8E;IAC9E,qBAAqB,CAAC,EAAE,SAAS,MAAM,EAAE,CAAC;IAC1C;;;;;;OAMG;IACH,mBAAmB,CAAC,EAAE,MAAM,CAAC;CAC9B;AAED,oFAAoF;AACpF,eAAO,MAAM,6BAA6B,MAAM,CAAC;AACjD,2EAA2E;AAC3E,eAAO,MAAM,6BAA6B,MAAM,CAAC;AACjD;;;;;GAKG;AACH,eAAO,MAAM,0BAA0B,MAAM,CAAC;AAE9C,wBAAgB,oBAAoB,CAClC,UAAU,EAAE,MAAM,EAClB,OAAO,GAAE,2BAAgC,GACxC,eAAe,CA0DjB"}

package/voice/respond-gate.js

@@ -0,0 +1,139 @@
+/**
+ * Canonical client-side shouldRespond / echo-rejection gate for always-on voice.
+ *
+ * In always-on ("hands-free") mode the mic is open continuously, so the
+ * recognizer transcribes EVERYTHING it hears — the agent's own text-to-speech
+ * bleeding back into the mic, the speaker's disfluent thinking noises ("um…"),
+ * and other people in the room. Sending those as turns makes the agent reply to
+ * itself, to filler, or to a bystander — the "responding when it's not
+ * appropriate" problem.
+ *
+ * This is the SINGLE definition of that gate, shared by:
+ *   - the UI shell capture loop (`@elizaos/ui` `voice/should-respond` +
+ *     `voice/voice-turn-signal` re-export from here), and
+ *   - the Voice Workbench headless runner (the plugin's real-decision-logic
+ *     services adapter calls these so the benchmark exercises the SAME code the
+ *     client runs — no drift between what we test and what we ship).
+ *
+ * Pure (no DOM, no I/O); the only dependency is the canonical syntactic EOT
+ * heuristic in `@elizaos/shared/voice-eot`.
+ */
+import { scoreEndOfTurnHeuristic } from "../voice-eot.js";
+/** Pure disfluencies — never a meaningful turn on their own. NOT answers. */
+const DISFLUENCIES = new Set([
+    "um",
+    "uh",
+    "uhh",
+    "uhm",
+    "umm",
+    "hmm",
+    "hm",
+    "mm",
+    "mmm",
+    "er",
+    "erm",
+    "ah",
+    "eh",
+]);
+/** How recent the agent's reply must be for the echo guard to apply. */
+export const ECHO_WINDOW_MS = 9000;
+/** Word-overlap fraction above which a turn is treated as TTS echo. */
+export const ECHO_OVERLAP_THRESHOLD = 0.7;
+function words(text) {
+    return text
+        .toLowerCase()
+        .replace(/[^a-z0-9'\s-]/gi, "")
+        .split(/\s+/)
+        .filter(Boolean);
+}
+/**
+ * Whether a transcribed voice turn should be sent to the agent (i.e. warrants a
+ * response). Returns false for pure disfluency and for near-verbatim echoes of
+ * the agent's recent speech.
+ */
+export function shouldRespondToVoiceTurn(transcript, context = {}) {
+    const w = words(transcript);
+    if (w.length === 0)
+        return false;
+    // Pure disfluency ("um", "uh huh"… with nothing substantive) → ignore.
+    if (w.every((word) => DISFLUENCIES.has(word)))
+        return false;
+    // Self-echo: the agent's own TTS heard back through the mic. Only consider it
+    // while the reply is recent, and only for multi-word turns (a one-word answer
+    // shouldn't be suppressed just because the word also appears in the reply).
+    const reply = context.recentAgentReply?.trim();
+    const age = context.replyAgeMs ?? Number.POSITIVE_INFINITY;
+    const echoActive = context.agentSpeaking === true || age <= ECHO_WINDOW_MS;
+    if (reply && echoActive && w.length >= 2) {
+        const replyWords = new Set(words(reply));
+        if (replyWords.size > 0) {
+            const overlap = w.filter((word) => replyWords.has(word)).length / w.length;
+            if (overlap >= ECHO_OVERLAP_THRESHOLD)
+                return false;
+        }
+    }
+    return true;
+}
+/** Server SUPPRESS threshold for EOT — below this reads as "user still talking". */
+export const SERVER_EOT_SUPPRESS_THRESHOLD = 0.4;
+/** Only a CONFIDENT bystander attribution is allowed to silence a turn. */
+export const BYSTANDER_SUPPRESS_CONFIDENCE = 0.7;
+/**
+ * Cosine at/above which an incoming turn is treated as the agent's OWN voice and
+ * hard-suppressed. WeSpeaker enrolls the device owner at ~0.78; the agent's own
+ * TTS imprint is a tighter, single-source match, so a slightly lower bar is
+ * safe and catches echo the transcript guard cannot.
+ */
+export const AGENT_SELF_VOICE_THRESHOLD = 0.7;
+export function buildVoiceTurnSignal(transcript, context = {}) {
+    const endOfTurnProbability = scoreEndOfTurnHeuristic(transcript);
+    // Transcript-level gate: the agent's own TTS echoed back through the mic, or
+    // pure thinking-noise ("um", "uh").
+    let agentShouldSpeak = shouldRespondToVoiceTurn(transcript, context);
+    // Audio-frame gate (only when diarization attributed the turn): a CONFIDENT
+    // bystander — someone who is neither the owner nor an enrolled speaker — who
+    // did NOT say the wake word is cross-talk, not a turn addressed to the agent.
+    // An uncertain attribution must never silence a real turn (fail open).
+    const speaker = context.speaker;
+    if (agentShouldSpeak && speaker && context.wakeWordActive !== true) {
+        const known = new Set(context.knownSpeakerEntityIds ?? []);
+        const enrolled = speaker.isOwner === true ||
+            (speaker.entityId !== null && known.has(speaker.entityId));
+        const confidentBystander = !enrolled &&
+            speaker.entityId !== null &&
+            speaker.confidence >= BYSTANDER_SUPPRESS_CONFIDENCE;
+        if (confidentBystander)
+            agentShouldSpeak = false;
+    }
+    // The wake word is an explicit address: it overrides bystander doubt and a
+    // soft echo/disfluency miss, because the user deliberately summoned the agent.
+    if (context.wakeWordActive === true)
+        agentShouldSpeak = true;
+    // Acoustic self-voice rejection: if the incoming voice MATCHES the agent's own
+    // TTS imprint while it is (or just was) speaking, it is the agent hearing
+    // itself — hard-suppress, even past the wake word. This is definitive (the
+    // owner cannot sound like the agent's synthetic voice) and catches the echo
+    // the transcript word-overlap guard misses (e.g. a mis-transcribed echo, or a
+    // "hey eliza" that the agent itself spoke).
+    const replyRecent = context.agentSpeaking === true ||
+        (context.replyAgeMs ?? Number.POSITIVE_INFINITY) <= ECHO_WINDOW_MS;
+    const isSelfVoice = context.selfVoiceSimilarity !== undefined &&
+        context.selfVoiceSimilarity >= AGENT_SELF_VOICE_THRESHOLD &&
+        replyRecent;
+    if (isSelfVoice)
+        agentShouldSpeak = false;
+    const nextSpeaker = !agentShouldSpeak
+        ? "user"
+        : endOfTurnProbability < SERVER_EOT_SUPPRESS_THRESHOLD
+            ? "user"
+            : "agent";
+    const source = isSelfVoice
+        ? "client-ambient+self-voice"
+        : context.wakeWordActive
+            ? "client-ambient+wakeword"
+            : speaker
+                ? "client-ambient+diarization"
+                : "client-ambient";
+    return { endOfTurnProbability, nextSpeaker, agentShouldSpeak, source };
+}
+//# sourceMappingURL=respond-gate.js.map