@qvac/tts-ggml 0.13.0 → 0.14.1

package/index.js

@@ -34,8 +34,6 @@ var __importStar = (this && this.__importStar) || (function () {
 })();
 /* eslint-disable @typescript-eslint/no-require-imports -- Bare modules and @qvac/logging expose CommonJS export shapes. */
 const bareOs = require("bare-os");
-const path = require("bare-path");
-const fs = require("bare-fs");
 const loggingModule = require("@qvac/logging");
 // Published logging releases expose a CJS constructor; the workspace exposes
 // an ESM default. Bare require(ESM) returns a namespace without __esModule.
@@ -44,1092 +42,24 @@ const QvacLogger = typeof loggingExport === "function" ? loggingExport : logging
 /* eslint-enable @typescript-eslint/no-require-imports */
 const infer_base_1 = require("@qvac/infer-base");
 const tts_1 = require("./tts");
+const addonEvents_1 = require("./lib/addonEvents");
 const errorModule = __importStar(require("./lib/error"));
-const pocketConfig_1 = require("./lib/pocketConfig");
 const fit_1 = require("./lib/fit");
-const textChunker_1 = require("./lib/textChunker");
 const textStreamAccumulator_1 = require("./lib/textStreamAccumulator");
+const engineIds_1 = require("./lib/engineIds");
+const engineParams_1 = require("./lib/engineParams");
+const audio8_1 = require("./lib/engines/audio8");
+const parler_1 = require("./lib/engines/parler");
+const exclusiveQueue_1 = require("./lib/exclusiveQueue");
+const jobFields_1 = require("./lib/jobFields");
+const pocketLifecycle_1 = require("./lib/pocketLifecycle");
+const reload_1 = require("./lib/reload");
+const sentenceStream_1 = require("./lib/sentenceStream");
+const setup_1 = require("./lib/setup");
+const streamingSupport_1 = require("./lib/streamingSupport");
+const voiceControls_1 = require("./lib/voiceControls");
 const { platform } = bareOs;
 const { ERR_CODES, QvacErrorAddonTTSGgml } = errorModule;
-const ENGINE_CHATTERBOX = "chatterbox";
-const ENGINE_SUPERTONIC = "supertonic";
-// CosyVoice3 (Fun-CosyVoice3-0.5B / 1.5B). Ships as a small set of GGUFs
-// (cosyvoice3-{llm,flow,hift}-*.gguf) plus voice.gguf, vocab.json and
-// merges.txt; a modelDir holding them (or an explicit
-// `files.cosyvoiceModelDir`) routes here.
-const ENGINE_COSYVOICE3 = "cosyvoice3";
-const ENGINE_PARLER = "parler";
-const ENGINE_AUDIO8 = "audio8";
-const ENGINE_MOSS = "moss";
-const ENGINE_MOSS_SFX = "moss-sfx";
-const ENGINE_MOSS_SPEECH = "moss-speech";
-const ENGINE_POCKET = "pocket";
-const MIN_OUTPUT_SAMPLE_RATE = 8000;
-const MAX_OUTPUT_SAMPLE_RATE = 192000;
-const CHATTERBOX_T3_TURBO = "chatterbox-t3-turbo.gguf";
-const CHATTERBOX_T3_MTL = "chatterbox-t3-mtl.gguf";
-const CHATTERBOX_S3GEN_DEFAULT = "chatterbox-s3gen.gguf";
-const CHATTERBOX_S3GEN_MTL = "chatterbox-s3gen-mtl.gguf";
-const SUPERTONIC_DEFAULT = "supertonic.gguf";
-const SUPERTONIC_MTL = "supertonic2.gguf";
-const SUPERTONIC_V3_RE = /^supertonic3(-[a-z0-9_]+)?\.gguf$/i;
-const SUPERTONIC_V3_QUANT_ORDER = [
-    "f16",
-    "f32",
-    "q8_0",
-    "q4_0",
-];
-// The LLM sub-model is the tell for CosyVoice3 modelDir auto-detection.
-const COSYVOICE3_LLM_RE = /^cosyvoice3-llm(-[a-z0-9_]+)?\.gguf$/i;
-// CosyVoice3 instruct2 control vocabulary (cosyvoice/utils/common.py
-// instruct_list). The structured `instruct` option renders to the exact
-// trained instruction string; the native engine wraps it as
-// "You are a helpful assistant. " + <instruction> + "<|endofprompt|>" and
-// drops the LM prompt speech tokens. One control applies per synthesis
-// (the model is trained on single instructions).
-const COSYVOICE_DIALECTS = {
-    cantonese: "广东话",
-    northeastern: "东北话",
-    gansu: "甘肃话",
-    guizhou: "贵州话",
-    henan: "河南话",
-    hubei: "湖北话",
-    hunan: "湖南话",
-    jiangxi: "江西话",
-    minnan: "闽南话",
-    ningxia: "宁夏话",
-    shanxi: "山西话",
-    shaanxi: "陕西话",
-    shandong: "山东话",
-    shanghai: "上海话",
-    sichuan: "四川话",
-    tianjin: "天津话",
-    yunnan: "云南话",
-};
-const COSYVOICE_VOLUMES = {
-    loud: "Please say a sentence as loudly as possible.",
-    soft: "Please say a sentence in a very soft voice.",
-};
-const COSYVOICE_STYLES = {
-    peppa: "我想体验一下小猪佩奇风格，可以吗？",
-    robot: "你可以尝试用机器人的方式解答吗？",
-};
-/**
- * Look up a structured-instruct control value, throwing a clear error for an
- * invalid key instead of letting an `undefined` render into the instruction
- * string.
- */
-function cosyvoiceInstructValue(map, key, kind) {
-    const value = map[key];
-    if (value == null) {
-        throw new Error(`Invalid CosyVoice instruct ${kind} "${key}". Valid ${kind}s: ${Object.keys(map).join(", ")}.`);
-    }
-    return value;
-}
-/**
- * Render a CosyVoice3 `instruct` option to the trained instruction string.
- * A raw string passes through (trimmed); the structured form emits exactly one
- * control by precedence dialect > volume > style. Returns "" for no
- * instruction. An invalid structured key throws.
- */
-function renderCosyvoiceInstruct(instruct) {
-    // Only `undefined` means "omitted" (zero-shot). An explicit `null` is a
-    // malformed value, not an omission, so it is rejected below like any other
-    // non-object.
-    if (instruct === undefined)
-        return "";
-    if (typeof instruct === "string")
-        return instruct.trim();
-    // A JS caller can pass values the type forbids (null, a number, an array, a
-    // Date, a class instance). Check defensively as `unknown`: require an ordinary
-    // object literal so exotic objects can't slip past the presence checks below
-    // and degrade to zero-shot, and so control fields are read from the object's
-    // own properties rather than an inherited prototype.
-    const control = instruct;
-    if (control === null || typeof control !== "object" || Array.isArray(control)) {
-        throw new Error("Invalid CosyVoice instruct: expected a string or a plain control " +
-            "object such as { dialect: 'cantonese' }.");
-    }
-    const proto = Object.getPrototypeOf(control);
-    if (proto !== Object.prototype && proto !== null) {
-        throw new Error("Invalid CosyVoice instruct: expected a plain control object such as " +
-            "{ dialect: 'cantonese' }, not an instance of another type.");
-    }
-    // Reject unknown structured keys (typos like `{ dialekt: 'cantonese' }`)
-    // before the precedence chain, which would otherwise fall through to "".
-    const supportedControls = ["dialect", "volume", "style"];
-    const unknownKeys = Object.keys(instruct).filter((key) => !supportedControls.includes(key));
-    if (unknownKeys.length > 0) {
-        const hint = unknownKeys.some((key) => key === "emotion" || key === "speed")
-            ? " Emotion and speaking rate moved to the top-level `emotion` / `pace` options."
-            : "";
-        throw new Error(`Invalid CosyVoice instruct key(s): ${unknownKeys.join(", ")}. ` +
-            `Valid keys: dialect, volume, style.${hint}`);
-    }
-    // Presence, not truthiness: a control that is set to a malformed value
-    // (`{ dialect: '' }`, `{ dialect: null }`) must raise via
-    // cosyvoiceInstructValue rather than fall through to zero-shot. `undefined`
-    // means the field was not set, so it is skipped. Precedence: dialect >
-    // volume > style.
-    if (instruct.dialect !== undefined) {
-        return `请用${cosyvoiceInstructValue(COSYVOICE_DIALECTS, instruct.dialect, "dialect")}表达。`;
-    }
-    if (instruct.volume !== undefined) {
-        return cosyvoiceInstructValue(COSYVOICE_VOLUMES, instruct.volume, "volume");
-    }
-    if (instruct.style !== undefined) {
-        return cosyvoiceInstructValue(COSYVOICE_STYLES, instruct.style, "style");
-    }
-    return "";
-}
-// Parler GGUFs ship per quant tier with the quant in the filename
-// (`parler-mini-v1-q8_0.gguf`); a bare `parler.gguf` deliberately does not
-// match, keeping ambiguous files out of the modelDir auto-detect path.
-const PARLER_RE = /^parler-(mini|large|indic)(-v\d+)?(-[a-z0-9_]+)?\.gguf$/i;
-const PARLER_VARIANT_ORDER = ["mini", "large", "indic"];
-const PARLER_QUANT_ORDER = ["q8_0", "q6_k", "f16", "f32"];
-const PARLER_DESCRIPTION_KEYS = ["description", "voiceDescription"];
-// What native Parler reads as template fields. `emotion` / `pace` stay listed
-// here because build_description() still renders them, even though they are
-// cross-engine options at the JS surface.
-const PARLER_TEMPLATE_KEYS = [
-    "voice",
-    "emotion",
-    "pitch",
-    "pace",
-    "expressivity",
-    "noise",
-    "reverb",
-    "quality",
-];
-// The subset that stays Parler-only now that emotion/pace are cross-engine.
-const PARLER_ONLY_TEMPLATE_KEYS = [
-    "voice",
-    "pitch",
-    "expressivity",
-    "noise",
-    "reverb",
-    "quality",
-];
-const PARLER_FIELD_KEYS = [
-    ...PARLER_DESCRIPTION_KEYS,
-    ...PARLER_TEMPLATE_KEYS,
-];
-const PARLER_ONLY_KEYS = [
-    ...PARLER_DESCRIPTION_KEYS,
-    ...PARLER_ONLY_TEMPLATE_KEYS,
-];
-// Canonical cross-engine conditioning vocabulary. This is a MIRROR, not the
-// source of truth: tts-cpp owns it in include/tts-cpp/voice_controls.h, and
-// test/unit/voice-controls.test.js pins this copy against the native lists.
-const EMOTIONS = [
-    "command",
-    "anger",
-    "narration",
-    "conversation",
-    "disgust",
-    "fear",
-    "happy",
-    "neutral",
-    "proper noun",
-    "news",
-    "sad",
-    "surprise",
-];
-const PACES = ["slow", "moderate", "fast"];
-const CONDITIONING_KEYS = ["emotion", "pace"];
-// Audio8 ships three GGUFs per quant tier, with the tier in the filename
-// (`audio8-lm-q8_0.gguf`); a bare `audio8-lm.gguf` matches too, as
-// forward-compat with a single-tier release.
-const AUDIO8_LM_RE = /^audio8-lm(-[a-z0-9_]+)?\.gguf$/i;
-const AUDIO8_DECODER_RE = /^audio8-codec-decoder(-[a-z0-9_]+)?\.gguf$/i;
-const AUDIO8_ENCODER_RE = /^audio8-codec-encoder(-[a-z0-9_]+)?\.gguf$/i;
-const AUDIO8_QUANT_ORDER = ["q8_0", "f16", "q4_0", "f32"];
-const AUDIO8_VOICE_KEYS = ["referenceAudio", "referenceText"];
-const MOSS_BACKBONE_RE = /^moss-tts-delay(-[a-z0-9_]+)?\.gguf$/i;
-const MOSS_DIALOGUE_BACKBONE_RE = /^moss-ttsd(-[a-z0-9_]+)?\.gguf$/i;
-const MOSS_DECODER_RE = /^moss-codec-decoder(-[a-z0-9_]+)?\.gguf$/i;
-const MOSS_ENCODER_RE = /^moss-codec-encoder(-[a-z0-9_]+)?\.gguf$/i;
-const MOSS_NATIVE_SAMPLE_RATE = 24000;
-const MOSS_FRAMES_PER_SECOND = 12.5;
-const MOSS_MAX_NEW_TOKENS = 2048;
-const MOSS_MAX_CHANNELS = 32;
-const MOSS_TERMINATION_ROWS = 2;
-const MOSS_MAX_DURATION_TOKENS = MOSS_MAX_NEW_TOKENS - (MOSS_MAX_CHANNELS - 1) - MOSS_TERMINATION_ROWS;
-const MOSS_INSTANCE_VOICE_KEYS = ["referenceAudio", "dialogueReferences"];
-const MOSS_SFX_RE = /^moss-sfx(-[a-z0-9_]+)*\.gguf$/i;
-const MOSS_SFX_NATIVE_SAMPLE_RATE = 48000;
-const MOSS_SFX_MAX_SECONDS = 30;
-const MOSS_SFX_MAX_STEPS = 1000;
-const MOSS_SFX_MAX_GUIDANCE = 50;
-const MOSS_SFX_MAX_SHIFT = 100;
-const MOSS_SFX_CALL_KEYS = [
-    "seconds",
-    "negativePrompt",
-    "steps",
-    "guidance",
-    "shift",
-];
-const MOSS_SPEECH_RE = /^moss-speech(-(?!codec)[a-z0-9_]+)?\.gguf$/i;
-const MOSS_SPEECH_CODEC_RE = /^moss-speech-codec(-[a-z0-9_]+)?\.gguf$/i;
-const MOSS_SPEECH_NATIVE_SAMPLE_RATE = 24000;
-const MOSS_SPEECH_MIN_SAMPLE_RATE = 8000;
-const MOSS_SPEECH_MAX_SAMPLE_RATE = 192000;
-const MOSS_SPEECH_MAX_NEW_TOKENS = 4096;
-const MOSS_SPEECH_MAX_REPLY_SECONDS = 3600;
-const MOSS_SPEECH_MAX_TEMPERATURE = 10;
-const MOSS_SPEECH_MAX_TOP_P = 1;
-const MOSS_SPEECH_ROLES = ["system", "user", "assistant"];
-const MOSS_SPEECH_ONLY_KEYS = [
-    "audio",
-    "sampleRate",
-    "messages",
-    "systemPrompt",
-    "replyVoice",
-    "replyVoiceSampleRate",
-    "textReply",
-    "maxReplySeconds",
-    "maxNewTokens",
-];
-const MOSS_SPEECH_CONTROL_KEYS = [
-    "textReply",
-    "maxReplySeconds",
-    "maxNewTokens",
-    "greedy",
-    "temperature",
-    "topP",
-    "topK",
-];
-/** Engines that draw tokens, and so accept temperature / topK / topP / maxFrames. */
-const SAMPLING_ENGINES = [ENGINE_PARLER, ENGINE_AUDIO8];
-// Per-engine supported subsets, mirroring controls::supported_emotions() /
-// supported_paces(). An empty list means the engine has no such control.
-const ENGINE_EMOTIONS = {
-    [ENGINE_PARLER]: EMOTIONS,
-    // Only the emotions Fun-CosyVoice3 has a trained instruction for; the rest,
-    // and the upstream spelling "angry", are rejected rather than paraphrased.
-    [ENGINE_COSYVOICE3]: ["anger", "happy", "neutral", "sad"],
-    [ENGINE_SUPERTONIC]: [],
-    [ENGINE_CHATTERBOX]: [],
-    [ENGINE_POCKET]: [],
-    [ENGINE_AUDIO8]: [],
-    [ENGINE_MOSS]: [],
-    [ENGINE_MOSS_SFX]: [],
-    [ENGINE_MOSS_SPEECH]: [],
-};
-const ENGINE_PACES = {
-    [ENGINE_PARLER]: PACES, // rendered into the training caption
-    [ENGINE_COSYVOICE3]: PACES, // slow/fast -> instruct; moderate -> none
-    [ENGINE_SUPERTONIC]: PACES, // mapped onto the duration multiplier
-    [ENGINE_CHATTERBOX]: [], // time-stretch only; use `speed`
-    [ENGINE_POCKET]: [],
-    [ENGINE_AUDIO8]: [], // no rate control at all
-    [ENGINE_MOSS]: [],
-    [ENGINE_MOSS_SFX]: [],
-    [ENGINE_MOSS_SPEECH]: [],
-};
-// Which channels an engine can change per call. Supertonic takes its pace
-// through EngineOptions when the engine is built, and tts-cpp exposes no
-// per-call surface for it, so it only moves at construction or reload().
-// tts-cpp engine names (tts-cpp/voice_controls.h engine_name) that differ from
-// this package's engine types.
-const FABRIC_ENGINE_TYPES = {
-    cosyvoice: ENGINE_COSYVOICE3,
-};
-const ENGINE_PER_CALL_CONDITIONING = {
-    [ENGINE_PARLER]: CONDITIONING_KEYS,
-    [ENGINE_COSYVOICE3]: CONDITIONING_KEYS,
-    [ENGINE_SUPERTONIC]: [],
-    [ENGINE_CHATTERBOX]: [],
-    [ENGINE_POCKET]: [],
-    [ENGINE_AUDIO8]: [],
-    [ENGINE_MOSS]: [],
-    [ENGINE_MOSS_SFX]: [],
-    [ENGINE_MOSS_SPEECH]: [],
-};
-function normalizeError(error) {
-    return typeof error === "string"
-        ? error
-        : error instanceof Error
-            ? error
-            : new Error("Unknown TTS error");
-}
-function firstNonEmpty(...candidates) {
-    for (const value of candidates) {
-        if (value != null && value !== "")
-            return value;
-    }
-    return undefined;
-}
-function fileExistsSafe(filePath) {
-    if (!filePath)
-        return false;
-    try {
-        return fs.existsSync(filePath);
-    }
-    catch {
-        return false;
-    }
-}
-function findSupertonicV3InDir(modelDir) {
-    if (!modelDir)
-        return undefined;
-    let entries;
-    try {
-        entries = fs.readdirSync(modelDir);
-    }
-    catch {
-        return undefined;
-    }
-    const matches = entries.filter((name) => SUPERTONIC_V3_RE.test(name));
-    if (matches.length === 0)
-        return undefined;
-    function rank(name) {
-        if (/^supertonic3\.gguf$/i.test(name))
-            return 0;
-        const match = name.match(/^supertonic3-(.+)\.gguf$/i);
-        const index = match
-            ? SUPERTONIC_V3_QUANT_ORDER.indexOf(match[1].toLowerCase())
-            : -1;
-        return index === -1
-            ? SUPERTONIC_V3_QUANT_ORDER.length + 1
-            : index + 1;
-    }
-    matches.sort((left, right) => rank(left) - rank(right));
-    return path.join(modelDir, matches[0]);
-}
-/**
- * True when `modelDir` contains a CosyVoice3 LLM GGUF (the sub-model that
- * unambiguously identifies a CosyVoice3 model directory).
- */
-function dirHasCosyvoice3(modelDir) {
-    if (!modelDir)
-        return false;
-    let entries;
-    try {
-        entries = fs.readdirSync(modelDir);
-    }
-    catch {
-        return false;
-    }
-    return entries.some((name) => COSYVOICE3_LLM_RE.test(name));
-}
-/**
- * Find a Parler GGUF in `modelDir`, ranked by variant (mini before large
- * before indic) and, within a variant, by quant tier (q8_0 > q6_k > f16 > f32;
- * a bare `parler-<variant>.gguf` wins as forward-compat).
- */
-function findParlerInDir(modelDir) {
-    if (!modelDir)
-        return undefined;
-    let entries;
-    try {
-        entries = fs.readdirSync(modelDir);
-    }
-    catch {
-        return undefined;
-    }
-    const matches = entries.filter((name) => PARLER_RE.test(name));
-    if (matches.length === 0)
-        return undefined;
-    function rank(name) {
-        const match = name.match(PARLER_RE);
-        if (!match)
-            return Number.MAX_SAFE_INTEGER;
-        const variant = PARLER_VARIANT_ORDER.indexOf(match[1].toLowerCase());
-        let quantRank = 0;
-        if (match[3]) {
-            const index = PARLER_QUANT_ORDER.indexOf(match[3].slice(1).toLowerCase());
-            quantRank =
-                index === -1 ? PARLER_QUANT_ORDER.length + 2 : index + 1;
-        }
-        return variant * 100 + quantRank;
-    }
-    matches.sort((left, right) => rank(left) - rank(right));
-    return path.join(modelDir, matches[0]);
-}
-/** Names of the entries that carry a value, for "wrong engine" messages. */
-function setOptionNames(options) {
-    return Object.entries(options)
-        .filter(([, value]) => value != null)
-        .map(([name]) => name);
-}
-function readDirSafe(modelDir) {
-    if (!modelDir)
-        return [];
-    try {
-        return fs.readdirSync(modelDir);
-    }
-    catch {
-        return [];
-    }
-}
-function ggufQuantRank(name, pattern) {
-    const match = name.match(pattern);
-    if (!match)
-        return Number.MAX_SAFE_INTEGER;
-    if (!match[1])
-        return 0;
-    const index = AUDIO8_QUANT_ORDER.indexOf(match[1].slice(1).toLowerCase());
-    return index === -1 ? AUDIO8_QUANT_ORDER.length + 1 : index + 1;
-}
-/**
- * Find one Audio8 or MOSS GGUF in `modelDir`, preferring the quant tier that
- * reads best for its size (q8_0 > f16 > q4_0 > f32).
- */
-function findQuantRankedGguf(modelDir, pattern) {
-    if (!modelDir)
-        return undefined;
-    const matches = readDirSafe(modelDir).filter((name) => pattern.test(name));
-    if (matches.length === 0)
-        return undefined;
-    matches.sort((left, right) => ggufQuantRank(left, pattern) - ggufQuantRank(right, pattern));
-    return path.join(modelDir, matches[0]);
-}
-function assertSoundEffectRange(name, value, accepts, range) {
-    if (typeof value === "number" && Number.isFinite(value) && accepts(value))
-        return;
-    throw new Error(`tts-ggml: moss-sfx ${name} must be ${range} (got ${String(value)})`);
-}
-function assertSoundEffectFields(fields) {
-    if (fields.seconds !== undefined) {
-        assertSoundEffectRange("seconds", fields.seconds, (n) => n > 0 && n <= MOSS_SFX_MAX_SECONDS, `in (0, ${MOSS_SFX_MAX_SECONDS}]`);
-    }
-    if (fields.steps !== undefined) {
-        assertSoundEffectRange("steps", fields.steps, (n) => Number.isInteger(n) && n >= 1 && n <= MOSS_SFX_MAX_STEPS, `an integer in [1, ${MOSS_SFX_MAX_STEPS}]`);
-    }
-    if (fields.guidance !== undefined) {
-        assertSoundEffectRange("guidance", fields.guidance, (n) => n >= 1 && n <= MOSS_SFX_MAX_GUIDANCE, `in [1, ${MOSS_SFX_MAX_GUIDANCE}]`);
-    }
-    if (fields.shift !== undefined) {
-        assertSoundEffectRange("shift", fields.shift, (n) => n > 0 && n <= MOSS_SFX_MAX_SHIFT, `in (0, ${MOSS_SFX_MAX_SHIFT}]`);
-    }
-    if (fields.negativePrompt !== undefined && typeof fields.negativePrompt !== "string") {
-        throw new Error("tts-ggml: moss-sfx negativePrompt must be a string");
-    }
-}
-function isSpeechPcm(value) {
-    return value instanceof Int16Array || value instanceof Float32Array;
-}
-function assertSpeechSampleRate(value, name, where) {
-    if (typeof value === "number" &&
-        Number.isInteger(value) &&
-        value >= MOSS_SPEECH_MIN_SAMPLE_RATE &&
-        value <= MOSS_SPEECH_MAX_SAMPLE_RATE) {
-        return;
-    }
-    throw new Error(`tts-ggml: ${where}: moss-speech ${name} must be an integer in ` +
-        `[${MOSS_SPEECH_MIN_SAMPLE_RATE}, ${MOSS_SPEECH_MAX_SAMPLE_RATE}] (got ${String(value)})`);
-}
-function assertSpeechAudio(audio, rate, names, where) {
-    if (!isSpeechPcm(audio) || audio.length === 0) {
-        throw new Error(`tts-ggml: ${where}: moss-speech ${names[0]} must be a non-empty Int16Array or Float32Array`);
-    }
-    assertSpeechSampleRate(rate, names[1], where);
-}
-function assertSpeechMessage(message, index, where) {
-    const label = `messages[${index}]`;
-    if (message == null || typeof message !== "object") {
-        throw new Error(`tts-ggml: ${where}: moss-speech ${label} must be an object`);
-    }
-    const turn = message;
-    if (!MOSS_SPEECH_ROLES.includes(turn.role)) {
-        throw new Error(`tts-ggml: ${where}: moss-speech ${label}.role must be one of ${MOSS_SPEECH_ROLES.join(", ")}`);
-    }
-    const hasText = typeof turn.text === "string" && turn.text.length > 0;
-    const hasAudio = turn.audio !== undefined;
-    if (hasText === hasAudio) {
-        throw new Error(`tts-ggml: ${where}: moss-speech ${label} carries exactly one of text or audio`);
-    }
-    if (hasAudio) {
-        assertSpeechAudio(turn.audio, turn.sampleRate, [`${label}.audio`, `${label}.sampleRate`], where);
-    }
-}
-function assertSpeechMessages(messages, where) {
-    if (messages === undefined)
-        return;
-    if (!Array.isArray(messages)) {
-        throw new Error(`tts-ggml: ${where}: moss-speech messages must be an array`);
-    }
-    assertEachSpeechMessage(messages, where);
-}
-function assertEachSpeechMessage(messages, where) {
-    messages.forEach((message, index) => assertSpeechMessage(message, index, where));
-}
-function assertSpeechUserTurn(fields, where) {
-    const hasText = typeof fields.input === "string" && fields.input.trim().length > 0;
-    const hasAudio = fields.audio !== undefined;
-    if (hasText && hasAudio) {
-        throw new Error(`tts-ggml: ${where}: moss-speech takes the user turn as audio or input text, not both`);
-    }
-    if (hasAudio) {
-        assertSpeechAudio(fields.audio, fields.sampleRate, ["audio", "sampleRate"], where);
-        return;
-    }
-    if (!hasText) {
-        throw new Error(`tts-ggml: ${where}: moss-speech needs the user turn as audio (with sampleRate) or input text`);
-    }
-}
-function assertSpeechRange(name, value, accepts, range, where) {
-    if (value === undefined)
-        return;
-    if (typeof value === "number" && Number.isFinite(value) && accepts(value))
-        return;
-    const got = typeof value === "number" ? String(value) : typeof value;
-    throw new Error(`tts-ggml: ${where}: moss-speech ${name} must be ${range} (got ${got})`);
-}
-function assertSpeechControls(fields, where) {
-    assertSpeechRange("maxReplySeconds", fields.maxReplySeconds, (n) => n >= 0 && n <= MOSS_SPEECH_MAX_REPLY_SECONDS, `in [0, ${MOSS_SPEECH_MAX_REPLY_SECONDS}]`, where);
-    assertSpeechRange("maxNewTokens", fields.maxNewTokens, (n) => Number.isInteger(n) && n >= 1 && n <= MOSS_SPEECH_MAX_NEW_TOKENS, `an integer in [1, ${MOSS_SPEECH_MAX_NEW_TOKENS}]`, where);
-    assertSpeechRange("temperature", fields.temperature, (n) => n > 0 && n <= MOSS_SPEECH_MAX_TEMPERATURE, `in (0, ${MOSS_SPEECH_MAX_TEMPERATURE}]`, where);
-    assertSpeechRange("topP", fields.topP, (n) => n > 0 && n <= MOSS_SPEECH_MAX_TOP_P, `in (0, ${MOSS_SPEECH_MAX_TOP_P}]`, where);
-    assertSpeechRange("topK", fields.topK, (n) => Number.isInteger(n) && n >= 0, "an integer >= 0", where);
-    assertSpeechBooleans(fields, where);
-    if (fields.systemPrompt !== undefined && typeof fields.systemPrompt !== "string") {
-        throw new Error(`tts-ggml: ${where}: moss-speech systemPrompt must be a string`);
-    }
-}
-function assertSpeechBooleans(fields, where) {
-    for (const flag of ["textReply", "greedy"]) {
-        if (fields[flag] !== undefined && typeof fields[flag] !== "boolean") {
-            throw new Error(`tts-ggml: ${where}: moss-speech ${flag} must be a boolean`);
-        }
-    }
-}
-function assertSpeechRateHasAudio(audio, rate, names, where) {
-    if (audio !== undefined || rate === undefined)
-        return;
-    throw new Error(`tts-ggml: ${where}: moss-speech ${names[1]} needs ${names[0]}`);
-}
-function assertSpeechCall(fields, where) {
-    assertSpeechUserTurn(fields, where);
-    assertSpeechRateHasAudio(fields.audio, fields.sampleRate, ["audio", "sampleRate"], where);
-    assertSpeechRateHasAudio(fields.replyVoice, fields.replyVoiceSampleRate, ["replyVoice", "replyVoiceSampleRate"], where);
-    assertSpeechMessages(fields.messages, where);
-    if (fields.replyVoice !== undefined) {
-        assertSpeechAudio(fields.replyVoice, fields.replyVoiceSampleRate, ["replyVoice", "replyVoiceSampleRate"], where);
-    }
-    assertSpeechControls(fields, where);
-}
-function speechUserTurn(fields) {
-    if (fields.audio !== undefined) {
-        return { role: "user", audio: fields.audio, sampleRate: fields.sampleRate };
-    }
-    return { role: "user", text: fields.input };
-}
-function speechMessages(fields) {
-    const system = fields.systemPrompt ? [{ role: "system", text: fields.systemPrompt }] : [];
-    return [...system, ...(fields.messages ?? []), speechUserTurn(fields)];
-}
-function speechReplyVoice(fields) {
-    if (fields.replyVoice === undefined)
-        return {};
-    return { replyVoice: fields.replyVoice, replyVoiceSampleRate: fields.replyVoiceSampleRate };
-}
-function speechControls(fields) {
-    const present = MOSS_SPEECH_CONTROL_KEYS.filter((key) => fields[key] !== undefined);
-    return Object.fromEntries(present.map((key) => [key, fields[key]]));
-}
-/** Collect the fields only MOSS-Speech reads that are present on `source`; undefined when none are set. */
-function pickSpeechFields(source) {
-    if (source == null || typeof source !== "object")
-        return undefined;
-    const present = MOSS_SPEECH_ONLY_KEYS.filter((key) => source[key] !== undefined);
-    if (present.length === 0)
-        return undefined;
-    return Object.fromEntries(present.map((key) => [key, source[key]]));
-}
-/**
- * Collect the MOSS-SoundEffect generation controls present on `source`.
- * Returns undefined when none are set.
- */
-function pickSoundEffectFields(source) {
-    if (source == null || typeof source !== "object")
-        return undefined;
-    const present = MOSS_SFX_CALL_KEYS.filter((key) => source[key] !== undefined);
-    if (present.length === 0)
-        return undefined;
-    return Object.fromEntries(present.map((key) => [key, source[key]]));
-}
-/**
- * Collect the Audio8 voice-cloning properties present on `source` (a run
- * input, streaming options, or constructor options). Returns undefined when
- * none are set.
- */
-function pickAudio8VoiceFields(source) {
-    if (source == null || typeof source !== "object")
-        return undefined;
-    const out = {};
-    let any = false;
-    for (const key of AUDIO8_VOICE_KEYS) {
-        const value = source[key];
-        if (value != null && value !== "") {
-            out[key] = String(value);
-            any = true;
-        }
-    }
-    return any ? out : undefined;
-}
-/**
- * Collect the Parler description/template properties present on `source`
- * (a run input, streaming options, or constructor options). Returns undefined
- * when none are set.
- */
-function pickParlerDescFields(source) {
-    if (source == null || typeof source !== "object")
-        return undefined;
-    const out = {};
-    let any = false;
-    for (const key of PARLER_FIELD_KEYS) {
-        const value = source[key];
-        if (value != null && value !== "") {
-            out[key] = String(value);
-            any = true;
-        }
-    }
-    return any ? out : undefined;
-}
-/**
- * Same-level conflict check: a free-text description cannot be merged with
- * template fields, so setting both together is an error.
- */
-function assertParlerDescFieldsConsistent(fields, where) {
-    if (!fields)
-        return;
-    const hasDescription = PARLER_DESCRIPTION_KEYS.some((key) => fields[key] != null);
-    const templateKeys = PARLER_TEMPLATE_KEYS.filter((key) => fields[key] != null);
-    if (hasDescription && templateKeys.length > 0) {
-        throw new Error(`tts-ggml: ${where}: 'description' is mutually exclusive with the ` +
-            `voice-template options (got ${templateKeys.join(", ")})`);
-    }
-}
-/** Which of `keys` are set on `source`. Used to name stray options in errors. */
-function keysPresent(source, keys) {
-    if (source == null || typeof source !== "object")
-        return [];
-    const record = source;
-    return keys.filter((key) => record[key] != null && record[key] !== "");
-}
-/** Collect the cross-engine conditioning properties present on `source`. */
-function pickConditioningFields(source) {
-    if (source == null || typeof source !== "object")
-        return undefined;
-    const out = {};
-    let any = false;
-    for (const key of CONDITIONING_KEYS) {
-        const value = source[key];
-        if (value != null && value !== "") {
-            out[key] = value;
-            any = true;
-        }
-    }
-    return any ? out : undefined;
-}
-function assertCanonicalValue(value, canonical, kind) {
-    if (canonical.includes(value.toLowerCase()))
-        return;
-    throw new Error(`tts-ggml: invalid ${kind} "${value}". ` +
-        `Valid ${kind}s: ${canonical.join(", ")}.`);
-}
-function supportedValues(engine, kind) {
-    return kind === "emotion" ? ENGINE_EMOTIONS[engine] : ENGINE_PACES[engine];
-}
-function assertEngineSupports(engine, kind, value) {
-    const supported = supportedValues(engine, kind);
-    if (supported.length === 0) {
-        const hint = kind === "pace"
-            ? "; use `speed` (an exact multiplier) instead"
-            : "";
-        throw new Error(`tts-ggml: the ${engine} engine does not support \`${kind}\`${hint}`);
-    }
-    if (supported.includes(value.toLowerCase()))
-        return;
-    throw new Error(`tts-ggml: ${kind} "${value}" is not supported by the ${engine} engine. ` +
-        `${engine} supports: ${supported.join(", ")}.`);
-}
-/**
- * Validate emotion/pace against the canonical vocabulary first (so a typo reads
- * clearly), then against what this engine actually supports.
- */
-function validateConditioning(engine, fields) {
-    if (!fields)
-        return;
-    if (fields.emotion != null) {
-        assertCanonicalValue(fields.emotion, EMOTIONS, "emotion");
-        assertEngineSupports(engine, "emotion", fields.emotion);
-    }
-    if (fields.pace != null) {
-        assertCanonicalValue(fields.pace, PACES, "pace");
-        assertEngineSupports(engine, "pace", fields.pace);
-    }
-}
-/**
- * Reject a channel the engine supports but cannot change per call, so it is
- * never accepted here and then dropped on the way to the engine. Runs after
- * validateConditioning so an engine with no such control keeps its own message.
- */
-function assertPerCallConditioning(engine, fields, where) {
-    if (!fields)
-        return;
-    const perCall = ENGINE_PER_CALL_CONDITIONING[engine];
-    const fixed = CONDITIONING_KEYS.filter((key) => fields[key] != null && !perCall.includes(key));
-    if (fixed.length === 0)
-        return;
-    throw new Error(`tts-ggml: ${where}: ${engine} applies ` +
-        `${fixed.map((key) => `\`${key}\``).join(", ")} when the engine is ` +
-        `built, so it cannot change per call; set it in the constructor or ` +
-        `reload({ ${fixed.join(", ")} })`);
-}
-// CosyVoice3 is trained on one instruction per synthesis. Only `pace: moderate`
-// resolves to no instruction, so it is the one value that never conflicts; every
-// emotion carries a trained instruction, `neutral` included.
-const COSYVOICE_DISENGAGED = {
-    pace: "moderate",
-};
-function engagedCosyvoiceControls(fields, instruct) {
-    const engaged = [];
-    for (const key of CONDITIONING_KEYS) {
-        const value = fields?.[key];
-        if (value != null && value.toLowerCase() !== COSYVOICE_DISENGAGED[key]) {
-            engaged.push(`${key}="${value}"`);
-        }
-    }
-    if (instruct != null && instruct !== "")
-        engaged.push("instruct");
-    return engaged;
-}
-function assertSingleCosyvoiceControl(fields, instruct, where) {
-    const engaged = engagedCosyvoiceControls(fields, instruct);
-    if (engaged.length <= 1)
-        return;
-    throw new Error(`tts-ggml: ${where}: conflicting conditioning controls ` +
-        `(${engaged.join(", ")}); cosyvoice3 is trained on one instruction per ` +
-        `synthesis -- set exactly one (pace="moderate" disengages a channel)`);
-}
-function normalizeGgmlFiles(files) {
-    if (files == null || typeof files !== "object")
-        return {};
-    return {
-        modelDir: firstNonEmpty(files.modelDir),
-        pocketFlowModel: firstNonEmpty(files.pocketFlowModel),
-        pocketMimiModel: firstNonEmpty(files.pocketMimiModel),
-        pocketFrontend: firstNonEmpty(files.pocketFrontend),
-        pocketVoice: firstNonEmpty(files.pocketVoice),
-        t3Model: firstNonEmpty(files.t3Model, files.t3ModelPath, files.t3),
-        s3genModel: firstNonEmpty(files.s3genModel, files.s3genModelPath, files.s3gen),
-        supertonicModel: firstNonEmpty(files.supertonicModel, files.supertonicModelPath, files.supertonic),
-        // CosyVoice3: either a dedicated modelDir of cosyvoice3-*.gguf files, or
-        // explicit per-component paths. Falls back to the shared `modelDir`.
-        cosyvoiceModelDir: firstNonEmpty(files.cosyvoiceModelDir),
-        cosyvoiceLlmModel: firstNonEmpty(files.cosyvoiceLlmModel, files.cosyvoiceLlmModelPath),
-        cosyvoiceFlowModel: firstNonEmpty(files.cosyvoiceFlowModel, files.cosyvoiceFlowModelPath),
-        cosyvoiceHiftModel: firstNonEmpty(files.cosyvoiceHiftModel, files.cosyvoiceHiftModelPath),
-        cosyvoiceS3tokModel: firstNonEmpty(files.cosyvoiceS3tokModel, files.cosyvoiceS3tokModelPath),
-        cosyvoiceCampplusModel: firstNonEmpty(files.cosyvoiceCampplusModel, files.cosyvoiceCampplusModelPath),
-        cosyvoiceVocab: firstNonEmpty(files.cosyvoiceVocab, files.cosyvoiceVocabPath),
-        cosyvoiceMerges: firstNonEmpty(files.cosyvoiceMerges, files.cosyvoiceMergesPath),
-        cosyvoiceVoiceModel: firstNonEmpty(files.cosyvoiceVoiceModel, files.cosyvoiceVoiceModelPath),
-        parlerModel: firstNonEmpty(files.parlerModel, files.parlerModelPath, files.parler),
-        audio8Lm: firstNonEmpty(files.audio8Lm, files.audio8LmPath),
-        audio8CodecDecoder: firstNonEmpty(files.audio8CodecDecoder, files.audio8CodecDecoderPath),
-        audio8CodecEncoder: firstNonEmpty(files.audio8CodecEncoder, files.audio8CodecEncoderPath),
-        mossBackbone: firstNonEmpty(files.mossBackbone, files.mossBackbonePath),
-        mossCodecDecoder: firstNonEmpty(files.mossCodecDecoder, files.mossCodecDecoderPath),
-        mossCodecEncoder: firstNonEmpty(files.mossCodecEncoder, files.mossCodecEncoderPath),
-        mossSoundEffect: firstNonEmpty(files.mossSoundEffect, files.mossSoundEffectPath),
-        mossSpeechModel: firstNonEmpty(files.mossSpeechModel, files.mossSpeechModelPath),
-        mossSpeechCodec: firstNonEmpty(files.mossSpeechCodec, files.mossSpeechCodecPath),
-        voicesDir: firstNonEmpty(files.voicesDir),
-        lavasrEnhancer: firstNonEmpty(files.lavasrEnhancer),
-        lavasrDenoiser: firstNonEmpty(files.lavasrDenoiser),
-        mecabDictDir: firstNonEmpty(files.mecabDictDir, files.mecabDictPath),
-        cangjieTsvPath: firstNonEmpty(files.cangjieTsvPath, files.cangjieTsv),
-    };
-}
-function detectEngineType(engine, files) {
-    if (engine === ENGINE_CHATTERBOX ||
-        engine === ENGINE_SUPERTONIC ||
-        engine === ENGINE_COSYVOICE3 ||
-        engine === ENGINE_PARLER ||
-        engine === ENGINE_AUDIO8 ||
-        engine === ENGINE_MOSS ||
-        engine === ENGINE_MOSS_SFX ||
-        engine === ENGINE_MOSS_SPEECH ||
-        engine === ENGINE_POCKET) {
-        return engine;
-    }
-    if (engine != null && engine !== "") {
-        throw new Error("tts-ggml: 'engine' option must be 'chatterbox', 'supertonic', " +
-            `'cosyvoice3', 'parler', 'audio8', 'moss', 'moss-sfx', 'moss-speech' or 'pocket' (got '${String(engine)}')`);
-    }
-    if (files.pocketFlowModel || files.pocketMimiModel)
-        return ENGINE_POCKET;
-    // Explicit CosyVoice3 files/dir take precedence over shared-modelDir sniffing.
-    if (files.cosyvoiceModelDir || files.cosyvoiceLlmModel) {
-        return ENGINE_COSYVOICE3;
-    }
-    if (files.t3Model || files.s3genModel)
-        return ENGINE_CHATTERBOX;
-    if (files.supertonicModel)
-        return ENGINE_SUPERTONIC;
-    if (files.parlerModel)
-        return ENGINE_PARLER;
-    if (files.audio8Lm || files.audio8CodecDecoder)
-        return ENGINE_AUDIO8;
-    if (files.mossBackbone || files.mossCodecDecoder)
-        return ENGINE_MOSS;
-    if (files.mossSoundEffect)
-        return ENGINE_MOSS_SFX;
-    if (files.mossSpeechModel || files.mossSpeechCodec)
-        return ENGINE_MOSS_SPEECH;
-    if (files.modelDir) {
-        if (dirHasCosyvoice3(files.modelDir))
-            return ENGINE_COSYVOICE3;
-        const hasChatterbox = fileExistsSafe(path.join(files.modelDir, CHATTERBOX_T3_TURBO)) ||
-            fileExistsSafe(path.join(files.modelDir, CHATTERBOX_T3_MTL));
-        const hasSupertonic = fileExistsSafe(path.join(files.modelDir, SUPERTONIC_DEFAULT)) ||
-            fileExistsSafe(path.join(files.modelDir, SUPERTONIC_MTL)) ||
-            !!findSupertonicV3InDir(files.modelDir);
-        if (hasChatterbox)
-            return ENGINE_CHATTERBOX;
-        if (hasSupertonic)
-            return ENGINE_SUPERTONIC;
-        if (findParlerInDir(files.modelDir))
-            return ENGINE_PARLER;
-        if (findQuantRankedGguf(files.modelDir, AUDIO8_LM_RE))
-            return ENGINE_AUDIO8;
-        if (findQuantRankedGguf(files.modelDir, MOSS_BACKBONE_RE))
-            return ENGINE_MOSS;
-        if (findQuantRankedGguf(files.modelDir, MOSS_DIALOGUE_BACKBONE_RE)) {
-            return ENGINE_MOSS;
-        }
-        if (findQuantRankedGguf(files.modelDir, MOSS_SFX_RE))
-            return ENGINE_MOSS_SFX;
-        if (findQuantRankedGguf(files.modelDir, MOSS_SPEECH_RE))
-            return ENGINE_MOSS_SPEECH;
-        if (fileExistsSafe(path.join(files.modelDir, "flow-lm.gguf")))
-            return ENGINE_POCKET;
-    }
-    return ENGINE_CHATTERBOX;
-}
-function resolveSupertonicModelDirPath(modelDir) {
-    const english = path.join(modelDir, SUPERTONIC_DEFAULT);
-    const multilingual = path.join(modelDir, SUPERTONIC_MTL);
-    const versionThree = findSupertonicV3InDir(modelDir);
-    if (fileExistsSafe(english))
-        return english;
-    if (fileExistsSafe(multilingual))
-        return multilingual;
-    if (versionThree)
-        return versionThree;
-    return english;
-}
-function resolveChatterboxModelDirPaths(modelDir) {
-    const turboT3 = path.join(modelDir, CHATTERBOX_T3_TURBO);
-    const multilingualT3 = path.join(modelDir, CHATTERBOX_T3_MTL);
-    const defaultS3 = path.join(modelDir, CHATTERBOX_S3GEN_DEFAULT);
-    const multilingualS3 = path.join(modelDir, CHATTERBOX_S3GEN_MTL);
-    if (fileExistsSafe(multilingualT3) &&
-        !fileExistsSafe(turboT3)) {
-        return {
-            t3: multilingualT3,
-            s3: fileExistsSafe(multilingualS3)
-                ? multilingualS3
-                : defaultS3,
-        };
-    }
-    return { t3: turboT3, s3: defaultS3 };
-}
-function defaultAccumulateSentencesForStreamInput(textStream) {
-    if (textStream == null ||
-        typeof textStream === "string" ||
-        Array.isArray(textStream)) {
-        return false;
-    }
-    return (typeof textStream[Symbol.asyncIterator] === "function");
-}
-function ttsOutputDebugString(data) {
-    if (!data)
-        return "";
-    if (typeof data === "string")
-        return data;
-    if (typeof data === "number" ||
-        typeof data === "boolean" ||
-        typeof data === "bigint") {
-        return data.toString();
-    }
-    if (typeof data === "symbol")
-        return data.description || "";
-    if (typeof data === "function")
-        return data.name;
-    const value = data;
-    const summary = {};
-    if (value.sampleRate != null)
-        summary.sampleRate = value.sampleRate;
-    if (value.chunkIndex != null)
-        summary.chunkIndex = value.chunkIndex;
-    if (value.isLast != null)
-        summary.isLast = value.isLast;
-    if (value.sentenceChunk != null) {
-        summary.sentenceChunk = value.sentenceChunk;
-    }
-    if (value.outputArray &&
-        typeof value.outputArray.length === "number") {
-        summary.outputArrayLen = value.outputArray.length;
-    }
-    return JSON.stringify(summary);
-}
-function resolveDefaultLazySessionLoading(lazySessionLoading) {
-    if (lazySessionLoading != null)
-        return lazySessionLoading;
-    return platform() === "ios" || platform() === "android";
-}
-function validateOutputSampleRate(outputSampleRate) {
-    if (outputSampleRate == null)
-        return null;
-    if (outputSampleRate < MIN_OUTPUT_SAMPLE_RATE ||
-        outputSampleRate > MAX_OUTPUT_SAMPLE_RATE) {
-        throw new Error(`outputSampleRate must be between ${MIN_OUTPUT_SAMPLE_RATE} and ` +
-            `${MAX_OUTPUT_SAMPLE_RATE}, got ${outputSampleRate}`);
-    }
-    return outputSampleRate;
-}
-function resolveEnhancerGgufPath(files, enhancer) {
-    if (enhancer != null && enhancer.type !== "lavasr") {
-        throw new Error(`tts-ggml: unknown enhancer.type '${String(enhancer.type)}', expected 'lavasr'.`);
-    }
-    return firstNonEmpty(files.lavasrEnhancer, enhancer?.enhancerPath);
-}
-function resolveDenoiserGgufPath(files, denoiser) {
-    if (denoiser != null && denoiser.type !== "lavasr") {
-        throw new Error(`tts-ggml: unknown denoiser.type '${String(denoiser.type)}', expected 'lavasr'.`);
-    }
-    return firstNonEmpty(files.lavasrDenoiser, denoiser?.denoiserPath);
-}
-function assertGpuIntentConsistent(useGPU, nGpuLayers) {
-    if (typeof useGPU !== "boolean" || nGpuLayers == null)
-        return;
-    if (useGPU === (nGpuLayers !== 0))
-        return;
-    throw new Error(`tts-ggml: useGPU=${String(useGPU)} conflicts with ` +
-        `nGpuLayers=${nGpuLayers}. Either drop one of the two, or make ` +
-        "them agree (useGPU:true + nGpuLayers!=0, or " +
-        "useGPU:false + nGpuLayers=0).");
-}
-const AUDIO8_NUMERIC_SAMPLING_FIELDS = [
-    "temperature",
-    "topK",
-    "topP",
-    "maxFrames",
-    "seed",
-    "threads",
-];
-function assertAudio8SamplingFinite(sampling, where) {
-    for (const field of AUDIO8_NUMERIC_SAMPLING_FIELDS) {
-        const value = sampling[field];
-        if (value === undefined || Number.isFinite(value))
-            continue;
-        throw new Error(`tts-ggml: ${where}: ${field} must be a finite number`);
-    }
-}
-function isValidDurationTokens(value) {
-    return Number.isInteger(value) && value >= 0 && value <= MOSS_MAX_DURATION_TOKENS;
-}
-function assertMossDurationTokens(value, where) {
-    if (value == null || isValidDurationTokens(value))
-        return;
-    throw new Error(`tts-ggml: ${where}: durationTokens must be an integer from 0 to ` +
-        `${MOSS_MAX_DURATION_TOKENS} (0 = free length)`);
-}
-function copyDialogueReferences(references) {
-    if (references == null)
-        return undefined;
-    return Array.isArray(references)
-        ? [...references]
-        : references;
-}
-function isNonEmptyString(value) {
-    return typeof value === "string" && value !== "";
-}
-function assertDialogueReferenceList(references) {
-    if (Array.isArray(references) &&
-        references.length > 0 &&
-        references.every(isNonEmptyString)) {
-        return;
-    }
-    throw new Error("tts-ggml: dialogueReferences must be a non-empty array of WAV paths, " +
-        "one per speaker");
-}
-function assertNotNegative(value, name, zeroMeans, where) {
-    if (value === undefined || value >= 0)
-        return;
-    throw new Error(`tts-ggml: ${where}: ${name} must be >= 0 (0 = ${zeroMeans})`);
-}
-function assertAudio8TopP(value, where) {
-    if (value === undefined || (value > 0 && value <= 1))
-        return;
-    throw new Error(`tts-ggml: ${where}: topP must be in (0, 1]`);
-}
-/**
- * Mirrors Audio8Model::validateSampling, which stays the source of truth.
- * Reload checks the merged knobs here because native only sees them when the
- * instance is rebuilt: without this the refused value is already on the JS
- * object by then, and the two would describe different samplers.
- */
-function assertAudio8SamplingValid(sampling, where) {
-    assertAudio8SamplingFinite(sampling, where);
-    assertNotNegative(sampling.temperature, "temperature", "greedy", where);
-    assertNotNegative(sampling.topK, "topK", "no top-k cutoff", where);
-    assertNotNegative(sampling.maxFrames, "maxFrames", "engine default", where);
-    assertAudio8TopP(sampling.topP, where);
-}
-function isAudioOutputEvent(data) {
-    return (data != null &&
-        typeof data === "object" &&
-        "outputArray" in data &&
-        data.outputArray != null);
-}
-function isStatsEvent(data) {
-    return (data != null &&
-        typeof data === "object" &&
-        ("totalTime" in data ||
-            "audioDurationMs" in data ||
-            "totalSamples" in data));
-}
-/**
- * What `tokensPerSecond` counts. Audio8 paces on the codec frame grid and
- * reports frames per second in batch synthesis; the text-paced engines report
- * characters per second. Streaming has to keep the same unit as `run()` or the
- * number silently changes meaning between the two entry points.
- */
-function emptyStreamAccumulator() {
-    return {
-        totalTime: 0,
-        audioDurationMs: 0,
-        totalSamples: 0,
-        generatedFrames: 0,
-    };
-}
-function streamedTokenCount(chunks, accumulator, countsFrames) {
-    return countsFrames
-        ? accumulator.generatedFrames
-        : chunks.join("").length;
-}
-function computeSentenceStreamStats(chunks, accumulator, countsFrames) {
-    const { generatedFrames, ...totals } = accumulator;
-    const produced = streamedTokenCount(chunks, accumulator, countsFrames);
-    const stats = {
-        ...totals,
-        tokensPerSecond: accumulator.totalTime > 0 ? produced / accumulator.totalTime : 0,
-        realTimeFactor: accumulator.audioDurationMs > 0
-            ? (accumulator.totalTime * 1000) /
-                accumulator.audioDurationMs
-            : 0,
-    };
-    if (countsFrames)
-        stats.generatedFrames = generatedFrames;
-    return stats;
-}
 /**
  * GGML-backed TTS via the `tts-cpp` library. Wraps the chatterbox,
  * supertonic, parler, cosyvoice3, audio8 and moss engines behind a single
@@ -1144,15 +74,15 @@ class TTSGgml {
     static inferenceManagerConfig = {
         noAdditionalDownload: true,
     };
-    static ENGINE_CHATTERBOX = ENGINE_CHATTERBOX;
-    static ENGINE_SUPERTONIC = ENGINE_SUPERTONIC;
-    static ENGINE_COSYVOICE3 = ENGINE_COSYVOICE3;
-    static ENGINE_PARLER = ENGINE_PARLER;
-    static ENGINE_AUDIO8 = ENGINE_AUDIO8;
-    static ENGINE_MOSS = ENGINE_MOSS;
-    static ENGINE_MOSS_SFX = ENGINE_MOSS_SFX;
-    static ENGINE_MOSS_SPEECH = ENGINE_MOSS_SPEECH;
-    static ENGINE_POCKET = ENGINE_POCKET;
+    static ENGINE_CHATTERBOX = engineIds_1.ENGINE_CHATTERBOX;
+    static ENGINE_SUPERTONIC = engineIds_1.ENGINE_SUPERTONIC;
+    static ENGINE_COSYVOICE3 = engineIds_1.ENGINE_COSYVOICE3;
+    static ENGINE_PARLER = engineIds_1.ENGINE_PARLER;
+    static ENGINE_AUDIO8 = engineIds_1.ENGINE_AUDIO8;
+    static ENGINE_MOSS = engineIds_1.ENGINE_MOSS;
+    static ENGINE_MOSS_SFX = engineIds_1.ENGINE_MOSS_SFX;
+    static ENGINE_MOSS_SPEECH = engineIds_1.ENGINE_MOSS_SPEECH;
+    static ENGINE_POCKET = engineIds_1.ENGINE_POCKET;
     opts;
     exclusiveRun;
     logger;
@@ -1160,98 +90,14 @@ class TTSGgml {
     addon;
     _job;
     _runExclusive;
-    _ttsInferenceQueueWaiter;
-    _sentenceStreamCtx;
-    _config;
-    _pocketJobPending = null;
-    _pocketCancelPromise = null;
-    _pocketLifecycleInProgress = false;
-    _pocketParams = null;
-    _pocketOptions = {};
-    _pocketFiles = {};
+    _exclusiveStreams = new exclusiveQueue_1.ExclusiveResponseQueue();
+    /** Everything resolved from the constructor options; reload rewrites parts of it. */
+    _setup;
+    _sentenceStream;
+    _events;
+    /** The Pocket execution rules, or null for every other engine. */
+    _pocket;
     _lazySessionLoading;
-    _outputSampleRate;
-    _engineType;
-    _voicesDir;
-    _supertonicModelPath;
-    _t3ModelPath;
-    _s3genModelPath;
-    _cosyvoiceModelDir;
-    _cosyvoiceLlmModelPath;
-    _cosyvoiceFlowModelPath;
-    _cosyvoiceHiftModelPath;
-    _cosyvoiceS3tokModelPath;
-    _cosyvoiceCampplusModelPath;
-    _mecabDictPath;
-    _cangjieTsvPath;
-    _referenceAudio;
-    _voiceDir;
-    _seed;
-    _nGpuLayers;
-    _nCtx;
-    _kvCacheType;
-    _threads;
-    _streamChunkTokens;
-    _streamFirstChunkTokens;
-    _streamLeftContextTokens;
-    _cfmSteps;
-    _cfgRate;
-    _batchCfmSteps;
-    _nPredict;
-    _maxSentenceChars;
-    _crossfadeMs;
-    _repeatPenalty;
-    _exaggeration;
-    _cfgWeight;
-    _minP;
-    _streamChunkTolerancePct;
-    _streamMinChunkTokens;
-    _voiceJsonPath;
-    _prewarmText;
-    _vulkanDevice;
-    _flowCutPrompt;
-    _cosyvoiceVocabPath;
-    _cosyvoiceMergesPath;
-    _cosyvoiceVoiceModelPath;
-    _promptText;
-    _instruct;
-    _voice;
-    _steps;
-    _speed;
-    _noiseNpyPath;
-    _enhancerGgufPath;
-    _denoiserGgufPath;
-    _backendsDir;
-    _openclCacheDir;
-    _vulkanCacheDir;
-    _parlerModelPath;
-    _audio8LmPath;
-    _audio8CodecDecoderPath;
-    _audio8CodecEncoderPath;
-    _mossBackbonePath;
-    _mossCodecDecoderPath;
-    _mossCodecEncoderPath;
-    _mossSoundEffectPath;
-    _mossSpeechModelPath;
-    _mossSpeechCodecPath;
-    _dialogueReferences;
-    _durationTokens;
-    _referenceText;
-    _greedy;
-    _description;
-    _emotion;
-    _pitch;
-    _pace;
-    _expressivity;
-    _noise;
-    _reverb;
-    _quality;
-    _temperature;
-    _topK;
-    _topP;
-    _maxFrames;
-    _minNewTokens;
-    _normalizeNumbers;
     constructor(options = {}) {
         this.opts = options.opts || {};
         this.exclusiveRun = !!options.exclusiveRun;
@@ -1262,720 +108,47 @@ class TTSGgml {
             destroyed: false,
         };
         this.addon = null;
-        this._sentenceStreamCtx = null;
-        this._ttsInferenceQueueWaiter = Promise.resolve();
         this._job = (0, infer_base_1.createJobHandler)({
-            cancel: () => this._engineType === ENGINE_POCKET ? this.cancel() : this._optionalAddon()?.cancel(),
+            cancel: () => this._pocket ? this.cancel() : this._optionalAddon()?.cancel(),
         });
-        this._runExclusive = this.exclusiveRun
-            ? (0, infer_base_1.exclusiveRunQueue)()
-            : async function runNow(callback) {
-                return callback();
-            };
-        const normalizedFiles = normalizeGgmlFiles(options.files || {});
-        this._config = { ...(options.config || {}) };
-        this._lazySessionLoading = resolveDefaultLazySessionLoading(options.lazySessionLoading);
-        this._outputSampleRate = validateOutputSampleRate(this._config.outputSampleRate);
-        this._engineType = detectEngineType(options.engine, normalizedFiles);
-        this._assignMossVoiceOptions(options);
-        if (this._engineType === ENGINE_POCKET) {
-            if (normalizedFiles.lavasrEnhancer || normalizedFiles.lavasrDenoiser) {
-                throw new Error("Pocket does not support LavaSR enhancement or denoising");
-            }
-            this._pocketFiles = normalizedFiles;
-            this._pocketOptions = { ...options };
-            this._pocketParams = (0, pocketConfig_1.buildPocketParams)(normalizedFiles, options, this._config);
-        }
-        this._resolveEngineAndModelPaths(normalizedFiles);
-        this._mecabDictPath = firstNonEmpty(options.mecabDictPath, options.mecabDictDir, normalizedFiles.mecabDictDir);
-        this._cangjieTsvPath = firstNonEmpty(options.cangjieTsvPath, normalizedFiles.cangjieTsvPath);
-        this._assignSynthesisOptions(options);
-        this._enhancerGgufPath = resolveEnhancerGgufPath(normalizedFiles, options.enhancer);
-        this._denoiserGgufPath = resolveDenoiserGgufPath(normalizedFiles, options.denoiser);
-        this._backendsDir = firstNonEmpty(options.backendsDir, this._config.backendsDir);
-        this._openclCacheDir = firstNonEmpty(options.openclCacheDir, this._config.openclCacheDir);
-        this._vulkanCacheDir = firstNonEmpty(options.vulkanCacheDir, this._config.vulkanCacheDir);
-        assertGpuIntentConsistent(this._config.useGPU, this._nGpuLayers);
-        this._assertEngineStreamingSupport();
-        if (this._config.useGPU === undefined &&
-            this._nGpuLayers == null) {
-            this._config.useGPU = false;
-        }
-    }
-    _resolveEngineAndModelPaths(files) {
-        this._voicesDir = files.voicesDir;
-        if (this._engineType === ENGINE_POCKET)
-            return;
-        if (this._engineType === ENGINE_COSYVOICE3) {
-            // CosyVoice3 discovers its sub-model GGUFs from a model directory; the
-            // native engine resolves the individual components. Explicit
-            // per-component paths win over the directory.
-            this._cosyvoiceModelDir = firstNonEmpty(files.cosyvoiceModelDir, files.modelDir);
-            this._cosyvoiceLlmModelPath = files.cosyvoiceLlmModel;
-            this._cosyvoiceFlowModelPath = files.cosyvoiceFlowModel;
-            this._cosyvoiceHiftModelPath = files.cosyvoiceHiftModel;
-            this._cosyvoiceS3tokModelPath = files.cosyvoiceS3tokModel;
-            this._cosyvoiceCampplusModelPath = files.cosyvoiceCampplusModel;
-            this._cosyvoiceVocabPath = files.cosyvoiceVocab;
-            this._cosyvoiceMergesPath = files.cosyvoiceMerges;
-            this._cosyvoiceVoiceModelPath = files.cosyvoiceVoiceModel;
-            return;
-        }
-        if (this._engineType === ENGINE_SUPERTONIC) {
-            this._supertonicModelPath = firstNonEmpty(files.supertonicModel, files.modelDir
-                ? resolveSupertonicModelDirPath(files.modelDir)
-                : undefined);
-            return;
-        }
-        if (this._engineType === ENGINE_PARLER) {
-            this._parlerModelPath = firstNonEmpty(files.parlerModel, files.modelDir ? findParlerInDir(files.modelDir) : undefined);
-            return;
-        }
-        if (this._engineType === ENGINE_AUDIO8) {
-            this._resolveAudio8ModelPaths(files);
-            return;
-        }
-        if (this._engineType === ENGINE_MOSS) {
-            this._resolveMossModelPaths(files);
-            return;
-        }
-        if (this._engineType === ENGINE_MOSS_SFX) {
-            this._mossSoundEffectPath = firstNonEmpty(files.mossSoundEffect, findQuantRankedGguf(files.modelDir, MOSS_SFX_RE));
-            return;
-        }
-        if (this._engineType === ENGINE_MOSS_SPEECH) {
-            this._mossSpeechModelPath = firstNonEmpty(files.mossSpeechModel, findQuantRankedGguf(files.modelDir, MOSS_SPEECH_RE));
-            this._mossSpeechCodecPath = firstNonEmpty(files.mossSpeechCodec, findQuantRankedGguf(files.modelDir, MOSS_SPEECH_CODEC_RE));
-            return;
-        }
-        if (files.modelDir) {
-            const resolved = resolveChatterboxModelDirPaths(files.modelDir);
-            this._t3ModelPath = firstNonEmpty(files.t3Model, resolved.t3);
-            this._s3genModelPath = firstNonEmpty(files.s3genModel, resolved.s3);
-        }
-        else {
-            this._t3ModelPath = files.t3Model;
-            this._s3genModelPath = files.s3genModel;
-        }
-    }
-    _resolveAudio8ModelPaths(files) {
-        this._audio8LmPath = firstNonEmpty(files.audio8Lm, findQuantRankedGguf(files.modelDir, AUDIO8_LM_RE));
-        this._audio8CodecDecoderPath = firstNonEmpty(files.audio8CodecDecoder, findQuantRankedGguf(files.modelDir, AUDIO8_DECODER_RE));
-        this._audio8CodecEncoderPath = firstNonEmpty(files.audio8CodecEncoder, findQuantRankedGguf(files.modelDir, AUDIO8_ENCODER_RE));
-    }
-    _mossBackbonePatterns() {
-        return this._dialogueReferences
-            ? [MOSS_DIALOGUE_BACKBONE_RE]
-            : [MOSS_BACKBONE_RE, MOSS_DIALOGUE_BACKBONE_RE];
-    }
-    _findMossBackbone(modelDir) {
-        return firstNonEmpty(...this._mossBackbonePatterns().map((pattern) => findQuantRankedGguf(modelDir, pattern)));
-    }
-    _resolveMossModelPaths(files) {
-        this._mossBackbonePath = firstNonEmpty(files.mossBackbone, this._findMossBackbone(files.modelDir));
-        this._mossCodecDecoderPath = firstNonEmpty(files.mossCodecDecoder, findQuantRankedGguf(files.modelDir, MOSS_DECODER_RE));
-        this._mossCodecEncoderPath = firstNonEmpty(files.mossCodecEncoder, findQuantRankedGguf(files.modelDir, MOSS_ENCODER_RE));
-    }
-    _assignMossVoiceOptions(options) {
-        this._dialogueReferences = copyDialogueReferences(options.dialogueReferences);
-        this._durationTokens = options.durationTokens ?? undefined;
-    }
-    _assignSynthesisOptions(options) {
-        this._referenceAudio = options.referenceAudio;
-        this._referenceText = options.referenceText;
-        this._greedy = options.greedy;
-        this._voiceDir = options.voiceDir;
-        this._seed = options.seed;
-        this._nGpuLayers = options.nGpuLayers;
-        this._nCtx = options.nCtx;
-        this._kvCacheType = options.kvCacheType;
-        this._threads = options.threads;
-        this._streamChunkTokens = options.streamChunkTokens;
-        this._streamFirstChunkTokens = options.streamFirstChunkTokens;
-        this._streamChunkTolerancePct = options.streamChunkTolerancePct;
-        this._streamMinChunkTokens = options.streamMinChunkTokens;
-        this._streamLeftContextTokens = options.streamLeftContextTokens;
-        this._cfmSteps = options.cfmSteps;
-        this._cfgRate = options.cfgRate;
-        this._batchCfmSteps = options.batchCfmSteps;
-        this._nPredict = options.nPredict;
-        this._maxSentenceChars = options.maxSentenceChars;
-        this._crossfadeMs = options.crossfadeMs;
-        this._repeatPenalty = options.repeatPenalty;
-        this._exaggeration = options.exaggeration;
-        this._cfgWeight = options.cfgWeight;
-        this._minP = options.minP;
-        this._voiceJsonPath = options.voiceJsonPath;
-        this._prewarmText = options.prewarmText;
-        this._vulkanDevice = options.vulkanDevice;
-        this._flowCutPrompt = options.flowCutPrompt;
-        // CosyVoice3-only: transcript of the reference audio for zero-shot cloning.
-        this._promptText = options.promptText;
-        // CosyVoice3-only: render the structured/raw instruct2 control to its string.
-        this._instruct = renderCosyvoiceInstruct(options.instruct) || undefined;
-        this._voice = firstNonEmpty(options.voice, options.voiceName);
-        this._steps = firstNonEmpty(options.steps, options.numInferenceSteps);
-        this._speed = options.speed;
-        this._noiseNpyPath = options.noiseNpyPath;
-        // Parler voice-description surface (all optional; the all-defaults render
-        // is the models' recommended fallback caption).
-        this._description = firstNonEmpty(options.description, options.voiceDescription);
-        this._emotion = options.emotion;
-        this._pitch = options.pitch;
-        this._pace = options.pace;
-        this._expressivity = options.expressivity;
-        this._noise = options.noise;
-        this._reverb = options.reverb;
-        this._quality = options.quality;
-        this._temperature = options.temperature;
-        this._topK = options.topK;
-        this._topP = options.topP;
-        this._maxFrames = options.maxFrames;
-        this._minNewTokens = options.minNewTokens;
-        this._normalizeNumbers = options.normalizeNumbers;
-    }
-    _assertEngineStreamingSupport() {
-        // Checked ahead of the shared denoiser guard below: Supertonic rejects both
-        // LavaSR stages for one reason (tts-cpp has only one-shot enhance /
-        // denoise, and the engine reports no native rate before synthesis).
-        if (this._engineType === ENGINE_SUPERTONIC &&
-            (this._enhancerGgufPath || this._denoiserGgufPath) &&
-            this._requestsChunkStreaming()) {
-            throw new Error("tts-ggml: the LavaSR enhancer/denoiser are not supported with " +
-                "supertonic native chunk streaming (streamChunkTokens > 0). Use " +
-                "batch synthesis or sentence-level streaming via runStream() / " +
-                "runStreaming() / run({ streamOutput: true }) for enhanced or " +
-                "denoised output.");
-        }
-        if (this._engineType === ENGINE_AUDIO8 &&
-            (this._streamChunkTokens != null ||
-                this._streamFirstChunkTokens != null)) {
-            throw new Error("tts-ggml: streamChunkTokens / streamFirstChunkTokens are not " +
-                "supported by the audio8 engine, which synthesizes a whole " +
-                "utterance at a time. Use sentence-level streaming via " +
-                "runStream() / runStreaming() / run({ streamOutput: true }).");
-        }
-        if ((this._engineType === ENGINE_MOSS_SFX ||
-            this._engineType === ENGINE_MOSS_SPEECH) &&
-            (this._streamChunkTokens != null ||
-                this._streamFirstChunkTokens != null)) {
-            throw new Error("tts-ggml: streamChunkTokens / streamFirstChunkTokens are not " +
-                `supported by the ${this._engineType} engine, which returns a ` +
-                "whole result at a time. Use run().");
-        }
-        if (this._engineType === ENGINE_MOSS &&
-            this._streamFirstChunkTokens != null) {
-            throw new Error("tts-ggml: streamFirstChunkTokens is not supported by the moss " +
-                "engine, which streams fixed-size chunks; set streamChunkTokens " +
-                `(codec frames per chunk, ${MOSS_FRAMES_PER_SECOND} per second) only.`);
-        }
-        // Runs before the denoiser guard so a Parler description/template conflict
-        // is reported ahead of the engine-agnostic streaming constraints.
-        this._assertParlerOptionConsistency();
-        this._assertEngineScopedOptions();
-        this._assertCosyvoiceOptionConsistency();
-        this._assertCosyvoiceCloneConsistent();
-        this._assertAudio8OptionConsistency();
-        this._assertMossOptionConsistency();
-        this._assertMossSoundEffectOptionConsistency();
-        this._assertMossSpeechOptionConsistency();
-        this._assertConditioningConsistency("constructor");
-        if (this._denoiserGgufPath && this._requestsChunkStreaming()) {
-            throw new Error("tts-ggml: the LavaSR denoiser is not yet supported with " +
-                "native chunk streaming (streamChunkTokens > 0). Use batch " +
-                "synthesis, or drop the denoiser for streaming. Streaming " +
-                "denoise is a planned follow-up (needs a stateful streaming " +
-                "denoiser).");
-        }
-    }
-    // Every streaming engine starts native chunk streaming on
-    // streamChunkTokens > 0 alone: a count of 0 means batch, and
-    // streamFirstChunkTokens only sizes the first chunk once streaming is on.
-    _requestsChunkStreaming() {
-        return (this._streamChunkTokens ?? 0) > 0;
-    }
-    _assertParlerOptionConsistency() {
-        if (this._engineType === ENGINE_PARLER) {
-            assertParlerDescFieldsConsistent(pickParlerDescFields({
-                description: this._description,
-                voice: this._voice,
-                emotion: this._emotion,
-                pitch: this._pitch,
-                pace: this._pace,
-                expressivity: this._expressivity,
-                noise: this._noise,
-                reverb: this._reverb,
-                quality: this._quality,
-            }), "constructor");
-            return;
-        }
-        const parlerOnly = [];
-        if (this._description != null) {
-            parlerOnly.push("description/voiceDescription");
-        }
-        // emotion / pace are absent because they are cross-engine, validated by
-        // _assertConditioningConsistency; temperature / topK / topP / maxFrames are
-        // absent because audio8 samples too, so SAMPLING_ENGINES decides them.
-        parlerOnly.push(...setOptionNames({
-            pitch: this._pitch,
-            expressivity: this._expressivity,
-            noise: this._noise,
-            reverb: this._reverb,
-            quality: this._quality,
-            minNewTokens: this._minNewTokens,
-            normalizeNumbers: this._normalizeNumbers,
-        }));
-        if (parlerOnly.length > 0) {
-            throw new Error(`tts-ggml: ${parlerOnly.join(", ")} are parler-only options ` +
-                `(engine is ${this._engineType})`);
-        }
-        this._assertSamplerOptionSupport();
-    }
-    _assertSamplerOptionSupport() {
-        if (this._engineType === ENGINE_POCKET || SAMPLING_ENGINES.includes(this._engineType))
-            return;
-        // Chatterbox's T3 samples too, but its length cap is nPredict, not maxFrames.
-        const tokenSampling = this._engineType === ENGINE_CHATTERBOX
-            ? {}
-            : {
-                temperature: this._temperature,
-                topK: this._topK,
-                topP: this._topP,
-            };
-        const sampling = setOptionNames({
-            ...tokenSampling,
-            maxFrames: this._maxFrames,
+        this._runExclusive = (0, exclusiveQueue_1.resolveExclusiveRunner)(this.exclusiveRun);
+        this._lazySessionLoading = (0, setup_1.resolveDefaultLazySessionLoading)(options.lazySessionLoading);
+        this._setup = (0, setup_1.resolveEngineSetup)(options);
+        this._sentenceStream = new sentenceStream_1.SentenceStreamer({
+            startJob: () => this._job.start(),
+            dispatchJob: (job) => this._dispatchJob(job),
+            endJob: (stats) => this._events.endJob(stats),
+            failJob: (error) => this._job.fail(error),
+            // Audio8 and MOSS report `tokensPerSecond` as codec frames per second,
+            // so the streaming aggregate counts frames too rather than characters.
+            countsFrames: () => this._setup.engineType === engineIds_1.ENGINE_AUDIO8 ||
+                this._setup.engineType === engineIds_1.ENGINE_MOSS,
         });
-        if (sampling.length === 0)
-            return;
-        throw new Error(`tts-ggml: ${sampling.join(", ")} are parler/audio8-only options ` +
-            `(engine is ${this._engineType})`);
-    }
-    /**
-     * Options only some engines read. Set on any other engine they would be
-     * dropped without effect, so they are rejected by name instead.
-     */
-    _assertEngineScopedOptions() {
-        const scoped = [
-            [
-                [ENGINE_CHATTERBOX],
-                {
-                    batchCfmSteps: this._batchCfmSteps,
-                    nPredict: this._nPredict,
-                    maxSentenceChars: this._maxSentenceChars,
-                    crossfadeMs: this._crossfadeMs,
-                    repeatPenalty: this._repeatPenalty,
-                    exaggeration: this._exaggeration,
-                    cfgWeight: this._cfgWeight,
-                    minP: this._minP,
-                },
-            ],
-            [
-                [ENGINE_CHATTERBOX, ENGINE_COSYVOICE3],
-                { streamLeftContextTokens: this._streamLeftContextTokens },
-            ],
-            [
-                [ENGINE_SUPERTONIC],
-                {
-                    voiceJsonPath: this._voiceJsonPath,
-                    prewarmText: this._prewarmText,
-                    streamChunkTolerancePct: this._streamChunkTolerancePct,
-                    streamMinChunkTokens: this._streamMinChunkTokens,
-                },
-            ],
-            [
-                [ENGINE_SUPERTONIC, ENGINE_COSYVOICE3],
-                { vulkanDevice: this._vulkanDevice },
-            ],
-            [[ENGINE_COSYVOICE3], { flowCutPrompt: this._flowCutPrompt }],
-        ];
-        for (const [engines, fields] of scoped) {
-            if (engines.includes(this._engineType))
-                continue;
-            const names = setOptionNames(fields);
-            if (names.length === 0)
-                continue;
-            throw new Error(`tts-ggml: ${names.join(", ")} are ${engines.join("/")}-only ` +
-                `options (engine is ${this._engineType})`);
-        }
-    }
-    _assertAudio8OptionConsistency() {
-        if (this._engineType === ENGINE_AUDIO8) {
-            if (this._enhancerGgufPath || this._denoiserGgufPath) {
-                throw new Error("tts-ggml: the LavaSR enhancer/denoiser are not supported with " +
-                    "the audio8 engine (native 44.1 kHz output needs no bandwidth " +
-                    "extension). Drop lavasrEnhancer / lavasrDenoiser.");
-            }
-            this._assertAudio8VoiceConsistent({
-                referenceAudio: this._referenceAudio,
-                referenceText: this._referenceText,
-            }, "constructor");
-            return;
-        }
-        const audio8Only = setOptionNames({
-            referenceText: this._referenceText,
-            greedy: this._greedy,
+        this._events = new addonEvents_1.AddonEventRouter({
+            job: this._job,
+            stream: this._sentenceStream,
+            logger: () => this._getLogger(),
+            reportStats: () => !!this.opts.stats,
         });
-        if (audio8Only.length > 0) {
-            throw new Error(`tts-ggml: ${audio8Only.join(", ")} are audio8-only options ` +
-                `(engine is ${this._engineType})`);
-        }
-    }
-    _assertMossSoundEffectOptionConsistency() {
-        if (this._engineType !== ENGINE_MOSS_SFX)
-            return;
-        if (this._enhancerGgufPath || this._denoiserGgufPath) {
-            throw new Error("tts-ggml: the LavaSR enhancer/denoiser are not supported with " +
-                "the moss-sfx engine. Drop lavasrEnhancer / lavasrDenoiser.");
-        }
-        if (this._referenceAudio) {
-            throw new Error("tts-ggml: referenceAudio is not supported by the moss-sfx engine, " +
-                "which generates sound effects from a text prompt");
-        }
-        if (this._steps != null) {
-            throw new Error("tts-ggml: the moss-sfx engine takes steps per call " +
-                "(run({ input, steps })), not in the constructor");
-        }
-        this._assertMossSoundEffectOutputRate();
-        if (!this._mossSoundEffectPath) {
-            throw new Error("tts-ggml: the moss-sfx engine needs its GGUF: stage " +
-                "moss-sfx-*.gguf in modelDir or set files.mossSoundEffect");
-        }
-    }
-    _assertMossSpeechOptionConsistency() {
-        if (this._engineType !== ENGINE_MOSS_SPEECH)
-            return;
-        if (this._enhancerGgufPath || this._denoiserGgufPath) {
-            throw new Error("tts-ggml: the LavaSR enhancer/denoiser are not supported with " +
-                "the moss-speech engine. Drop lavasrEnhancer / lavasrDenoiser.");
-        }
-        if (this._referenceAudio) {
-            throw new Error("tts-ggml: the moss-speech engine takes its reply voice per call " +
-                "(run({ replyVoice, replyVoiceSampleRate })), not referenceAudio");
-        }
-        this._assertMossSpeechOutputRate();
-        if (!this._mossSpeechModelPath || !this._mossSpeechCodecPath) {
-            throw new Error("tts-ggml: the moss-speech engine needs its model and codec GGUFs: " +
-                "stage moss-speech-*.gguf and moss-speech-codec-*.gguf in modelDir " +
-                "or set files.mossSpeechModel and files.mossSpeechCodec");
-        }
-    }
-    _assertMossSpeechOutputRate() {
-        if (this._outputSampleRate == null ||
-            this._outputSampleRate === MOSS_SPEECH_NATIVE_SAMPLE_RATE) {
-            return;
-        }
-        throw new Error(`tts-ggml: the moss-speech engine outputs ${MOSS_SPEECH_NATIVE_SAMPLE_RATE} Hz ` +
-            `audio and cannot resample (outputSampleRate=${this._outputSampleRate})`);
-    }
-    _assertMossSoundEffectOutputRate() {
-        if (this._outputSampleRate == null ||
-            this._outputSampleRate === MOSS_SFX_NATIVE_SAMPLE_RATE) {
-            return;
-        }
-        throw new Error(`tts-ggml: the moss-sfx engine outputs ${MOSS_SFX_NATIVE_SAMPLE_RATE} Hz ` +
-            `audio and cannot resample (outputSampleRate=${this._outputSampleRate})`);
-    }
-    _assertMossOptionConsistency() {
-        if (this._engineType !== ENGINE_MOSS) {
-            this._assertNoMossOnlyOptions();
-            return;
-        }
-        if (this._enhancerGgufPath || this._denoiserGgufPath) {
-            throw new Error("tts-ggml: the LavaSR enhancer/denoiser are not supported with " +
-                "the moss engine. Drop lavasrEnhancer / lavasrDenoiser.");
-        }
-        this._assertMossOutputRate();
-        this._assertMossVoiceConsistent();
-        assertMossDurationTokens(this._durationTokens, "constructor");
-        this._assertMossDialogueReferences();
-    }
-    _assertNoMossOnlyOptions() {
-        const mossOnly = setOptionNames({
-            durationTokens: this._durationTokens,
-            dialogueReferences: this._dialogueReferences !== undefined || undefined,
-        });
-        if (mossOnly.length === 0)
-            return;
-        throw new Error(`tts-ggml: ${mossOnly.join(", ")} are moss-only options ` +
-            `(engine is ${this._engineType})`);
-    }
-    _assertMossDialogueReferences() {
-        if (this._dialogueReferences === undefined)
-            return;
-        assertDialogueReferenceList(this._dialogueReferences);
-        if (this._referenceAudio) {
-            throw new Error("tts-ggml: referenceAudio and dialogueReferences are exclusive; " +
-                "pass one recording per speaker in dialogueReferences");
-        }
-        if (!this._mossCodecEncoderPath) {
-            throw new Error("tts-ggml: dialogue synthesis with the moss engine needs the codec " +
-                "encoder GGUF (files.mossCodecEncoder)");
-        }
-        if (!this._mossBackbonePath) {
-            throw new Error("tts-ggml: dialogue synthesis with the moss engine needs the MOSS-TTSD " +
-                "backbone: stage moss-ttsd-*.gguf in modelDir or set files.mossBackbone");
-        }
-    }
-    _assertSentenceStreamingAllowed(where) {
-        if (this._engineType === ENGINE_MOSS_SPEECH) {
-            throw new Error(`tts-ggml: ${where}: the moss-speech engine answers one turn at a ` +
-                "time and does not stream; use run()");
-        }
-        if (this._engineType === ENGINE_MOSS_SFX) {
-            throw new Error(`tts-ggml: ${where}: the moss-sfx engine generates a whole sound ` +
-                "effect from one prompt and does not stream; use run()");
-        }
-        if (this._dialogueReferences === undefined)
-            return;
-        throw new Error(`tts-ggml: ${where}: MOSS dialogue cannot be split into sentences, ` +
-            "because every job must open with the reference transcripts; use " +
-            "run() or native streaming (streamChunkTokens)");
-    }
-    _assertMossOutputRate() {
-        if (this._outputSampleRate == null ||
-            this._outputSampleRate === MOSS_NATIVE_SAMPLE_RATE) {
-            return;
-        }
-        throw new Error(`tts-ggml: the moss engine emits ${MOSS_NATIVE_SAMPLE_RATE} Hz only; ` +
-            `drop outputSampleRate (got ${this._outputSampleRate}).`);
-    }
-    _assertMossVoiceConsistent() {
-        if (!this._referenceAudio || this._mossCodecEncoderPath)
-            return;
-        throw new Error("tts-ggml: voice cloning with the moss engine needs the codec " +
-            "encoder GGUF (files.mossCodecEncoder)");
-    }
-    /**
-     * A recording without its transcript is accepted by the model but degrades
-     * the clone silently, and a transcript alone has nothing to attach to, so
-     * both halves have to arrive together.
-     */
-    _assertAudio8VoiceConsistent(voice, where) {
-        if (!voice.referenceAudio && !voice.referenceText)
-            return;
-        if (!voice.referenceAudio) {
-            throw new Error(`tts-ggml: ${where}: referenceText needs a referenceAudio to ` +
-                "transcribe");
-        }
-        if (!voice.referenceText) {
-            throw new Error(`tts-ggml: ${where}: referenceAudio needs a referenceText saying ` +
-                "what the recording says");
-        }
-        if (!this._audio8CodecEncoderPath) {
-            throw new Error(`tts-ggml: ${where}: voice cloning needs the audio8 codec encoder ` +
-                "GGUF (files.audio8CodecEncoder)");
-        }
-    }
-    /**
-     * `promptText` is deliberately not required: its absence selects
-     * cross-lingual mode. With a model dir present but no cloning GGUFs in it,
-     * the native load fail-closes instead.
-     */
-    _assertCosyvoiceCloneConsistent() {
-        if (this._engineType !== ENGINE_COSYVOICE3)
-            return;
-        if (!this._referenceAudio)
-            return;
-        const haveExplicit = this._cosyvoiceS3tokModelPath && this._cosyvoiceCampplusModelPath;
-        if (!haveExplicit && !this._cosyvoiceModelDir) {
-            throw new Error("tts-ggml: referenceAudio (CosyVoice3 voice cloning) needs the " +
-                "speech-tokenizer and speaker-encoder GGUFs: set " +
-                "files.cosyvoiceS3tokModel + files.cosyvoiceCampplusModel, or a " +
-                "files.cosyvoiceModelDir containing cosyvoice3-s3tok*.gguf and " +
-                "cosyvoice3-campplus*.gguf");
-        }
-    }
-    _assertCosyvoiceOptionConsistency() {
-        if (this._engineType === ENGINE_COSYVOICE3)
-            return;
-        const cosyvoiceOnly = [];
-        const cosyvoiceOnlyFields = {
-            instruct: this._instruct,
-            promptText: this._promptText,
-        };
-        for (const [key, value] of Object.entries(cosyvoiceOnlyFields)) {
-            if (value != null)
-                cosyvoiceOnly.push(key);
-        }
-        if (cosyvoiceOnly.length > 0) {
-            throw new Error(`tts-ggml: ${cosyvoiceOnly.join(", ")} are cosyvoice3-only options ` +
-                `(engine is ${this._engineType})`);
-        }
-    }
-    /**
-     * Validate the cross-engine emotion/pace surface against this engine, and
-     * enforce CosyVoice3's one-instruction-per-synthesis rule.
-     */
-    _assertConditioningConsistency(where) {
-        const fields = pickConditioningFields({
-            emotion: this._emotion,
-            pace: this._pace,
-        });
-        validateConditioning(this._engineType, fields);
-        if (this._engineType === ENGINE_COSYVOICE3) {
-            assertSingleCosyvoiceControl(fields, this._instruct, where);
-        }
-    }
-    /**
-     * The per-call surface of a non-Parler engine: only the cross-engine
-     * emotion/pace it declares support for, never Parler's template fields.
-     */
-    _resolveConditioningJobFields(source, where, instruct) {
-        const parlerOnly = keysPresent(source, PARLER_ONLY_KEYS);
-        if (parlerOnly.length > 0) {
-            throw new Error(`tts-ggml: ${where}: per-call description/voice-template options ` +
-                `are parler-only (engine is ${this._engineType}; got ` +
-                `${parlerOnly.join(", ")})`);
-        }
-        const fields = pickConditioningFields(source);
-        if (!fields)
-            return undefined;
-        validateConditioning(this._engineType, fields);
-        assertPerCallConditioning(this._engineType, fields, where);
-        if (this._engineType === ENGINE_COSYVOICE3) {
-            // A per-call control replaces the constructor's, so only the per-call
-            // fields participate in the one-instruction count.
-            assertSingleCosyvoiceControl(fields, instruct, where);
-        }
-        return fields;
-    }
-    /**
-     * CosyVoice3's per-call instruction, rendered like the constructor's. The
-     * native layer reads it next to emotion/pace, and any of the three replaces
-     * the configured conditioning for that synthesis.
-     */
-    _resolveJobInstruct(source, where) {
-        if (source?.instruct === undefined)
-            return undefined;
-        if (this._engineType !== ENGINE_COSYVOICE3) {
-            throw new Error(`tts-ggml: ${where}: per-call instruct is cosyvoice3-only ` +
-                `(engine is ${this._engineType})`);
-        }
-        return renderCosyvoiceInstruct(source.instruct) || undefined;
-    }
-    /**
-     * Parler's per-call surface: description/template fields, where a per-call
-     * template cannot be merged with a constructor-level free-text description.
-     */
-    _resolveParlerJobFields(source, where) {
-        const fields = pickParlerDescFields(source);
-        if (!fields)
-            return undefined;
-        assertParlerDescFieldsConsistent(fields, where);
-        const conditioning = pickConditioningFields(source);
-        validateConditioning(this._engineType, conditioning);
-        assertPerCallConditioning(this._engineType, conditioning, where);
-        const hasDescription = PARLER_DESCRIPTION_KEYS.some((key) => fields[key] != null);
-        if (!hasDescription && this._description != null) {
-            throw new Error(`tts-ggml: ${where}: per-call template options cannot be combined ` +
-                "with a constructor-level description; pass a per-call " +
-                "description instead");
-        }
-        return fields;
-    }
-    /**
-     * The voice a per-call override actually synthesizes with. Mirrors
-     * Audio8Model::resolveVoice: a per-call recording replaces both halves, so
-     * it cannot inherit the configured transcript, which describes a different
-     * recording; a per-call transcript alone corrects the configured one.
-     */
-    _mergeAudio8Voice(fields) {
-        if (fields.referenceAudio) {
-            return {
-                referenceAudio: fields.referenceAudio,
-                referenceText: fields.referenceText,
-            };
-        }
-        return {
-            referenceAudio: this._referenceAudio,
-            referenceText: fields.referenceText ?? this._referenceText,
-        };
-    }
-    /**
-     * Extract + validate the per-call Audio8 voice fields from a run input or
-     * streaming options. Returns undefined when none are present.
-     */
-    _resolveAudio8JobFields(source, where) {
-        const fields = pickAudio8VoiceFields(source);
-        if (!fields)
-            return undefined;
-        if (this._engineType !== ENGINE_AUDIO8) {
-            throw new Error(`tts-ggml: ${where}: per-call referenceAudio/referenceText are ` +
-                `audio8-only (engine is ${this._engineType})`);
-        }
-        this._assertAudio8VoiceConsistent(this._mergeAudio8Voice(fields), where);
-        return fields;
-    }
-    _assertNoSpeechFields(source, where) {
-        const fields = pickSpeechFields(source);
-        if (!fields)
-            return;
-        throw new Error(`tts-ggml: ${where}: ${Object.keys(fields).join(", ")} are ` +
-            `moss-speech-only options (engine is ${this._engineType})`);
-    }
-    _assertNoOtherEngineFields(source, where) {
-        this._resolveConditioningJobFields(source, where, this._resolveJobInstruct(source, where));
-        this._resolveAudio8JobFields(source, where);
-        this._resolveSoundEffectJobFields(source, where);
-    }
-    /**
-     * Validate the MOSS-Speech call and fold the new user turn (spoken `audio`
-     * or text `input`) and `systemPrompt` into the native `messages` list.
-     */
-    _resolveSpeechJobFields(source, where) {
-        const fields = source ?? {};
-        assertSpeechCall(fields, where);
-        return {
-            messages: speechMessages(fields),
-            ...speechReplyVoice(fields),
-            ...speechControls(fields),
-        };
-    }
-    /**
-     * Extract + validate the per-call MOSS-SoundEffect controls from a run
-     * input. Returns undefined when none are present.
-     */
-    _resolveSoundEffectJobFields(source, where) {
-        const fields = pickSoundEffectFields(source);
-        if (!fields)
-            return undefined;
-        if (this._engineType !== ENGINE_MOSS_SFX) {
-            throw new Error(`tts-ggml: ${where}: ${Object.keys(fields).join(", ")} are ` +
-                `moss-sfx-only options (engine is ${this._engineType})`);
-        }
-        assertSoundEffectFields(fields);
-        return fields;
-    }
-    /**
-     * The per-call fields of whichever engine is loaded, if any are set. Parler
-     * takes the full description/template surface, Audio8 its voice override,
-     * and every engine the cross-engine conditioning it supports.
-     */
-    _resolveJobFields(source, where) {
-        if (this._engineType === ENGINE_MOSS_SPEECH) {
-            this._assertNoOtherEngineFields(source, where);
-            return this._resolveSpeechJobFields(source, where);
-        }
-        const instruct = this._resolveJobInstruct(source, where);
-        const engineFields = this._engineType === ENGINE_PARLER
-            ? this._resolveParlerJobFields(source, where)
-            : this._resolveConditioningJobFields(source, where, instruct);
-        const audio8 = this._resolveAudio8JobFields(source, where);
-        const soundEffect = this._resolveSoundEffectJobFields(source, where);
-        this._assertNoSpeechFields(source, where);
-        if (!engineFields && !audio8 && !soundEffect && instruct === undefined) {
-            return undefined;
-        }
-        return {
-            ...(engineFields ?? {}),
-            ...(audio8 ?? {}),
-            ...(soundEffect ?? {}),
-            ...(instruct === undefined ? {} : { instruct }),
-        };
+        this._pocket =
+            this._setup.engineType === engineIds_1.ENGINE_POCKET
+                ? new pocketLifecycle_1.PocketLifecycle({
+                    setup: this._setup,
+                    state: this.state,
+                    logger: () => this._getLogger(),
+                    optionalAddon: () => this._optionalAddon(),
+                    setAddon: (addon) => {
+                        this.addon = addon;
+                    },
+                    createAddon: (params) => this._spawnAddon(params),
+                    unloadModel: () => this._unloadModel(),
+                    activeJob: () => this._job.active,
+                    failStream: (reason) => this._sentenceStream.fail(reason),
+                })
+                : null;
     }
     getEngineType() {
-        return this._engineType;
+        return this._setup.engineType;
     }
     getApiDefinition() {
         const api = (0, infer_base_1.getApiDefinition)();
@@ -2005,25 +178,14 @@ class TTSGgml {
         this.state.weightsLoaded = true;
     }
     async run(input) {
-        await this._waitPocketCancel();
-        this._checkPocketReload();
-        if (input?.streamOutput === true && this._engineType !== ENGINE_POCKET) {
-            this._assertSentenceStreamingAllowed("run with streamOutput");
-            if (typeof input.input !== "string" ||
-                input.input.trim().length === 0) {
-                throw new QvacErrorAddonTTSGgml({
-                    code: ERR_CODES.FAILED_TO_APPEND,
-                    adds: "run with streamOutput: non-empty string `input` is required",
-                });
-            }
-            const jobFields = this._resolveJobFields(input, "run");
-            const runStream = () => this._runStreamOrchestrator(input.input, {
-                locale: input.locale,
-                maxChunkScalars: input.maxChunkScalars,
-            }, jobFields);
-            return this.exclusiveRun
-                ? this._enqueueExclusiveTtsResponse(runStream)
-                : runStream();
+        await this._pocket?.waitCancel();
+        this._pocket?.checkReload();
+        if (input?.streamOutput === true && this._setup.engineType !== engineIds_1.ENGINE_POCKET) {
+            (0, streamingSupport_1.assertSentenceStreamingAllowed)(this._setup, "run with streamOutput");
+            const text = (0, sentenceStream_1.requireStreamOutputText)(input.input);
+            const jobFields = (0, jobFields_1.resolveJobFields)(this._setup, input, "run");
+            const { locale, maxChunkScalars } = input;
+            return this._startStream(() => this._runStreamOrchestrator(text, { locale, maxChunkScalars }, jobFields));
         }
         return this._runExclusive(() => this._runInternal(input));
     }
@@ -2032,20 +194,19 @@ class TTSGgml {
      * `run({ input: text, streamOutput: true, ...options })`.
      */
     async runStream(text, options = {}) {
-        await this._waitPocketCancel();
-        this._checkPocketReload();
+        await this._pocket?.waitCancel();
+        this._pocket?.checkReload();
         const normalized = options == null || typeof options !== "object" ? {} : options;
-        if (this._engineType === ENGINE_POCKET) {
-            const run = () => this._runStreamOrchestrator(text, normalized, this._resolveJobFields(normalized, "runStream"));
-            return this.exclusiveRun ? this._enqueueExclusiveTtsResponse(run) : run();
+        if (this._setup.engineType === engineIds_1.ENGINE_POCKET) {
+            return this._startStream(() => this._runStreamOrchestrator(text, normalized, (0, jobFields_1.resolveJobFields)(this._setup, normalized, "runStream")));
         }
         return this.run({
             input: text,
             streamOutput: true,
             locale: normalized.locale,
             maxChunkScalars: normalized.maxChunkScalars,
-            ...(pickParlerDescFields(normalized) ?? {}),
-            ...(pickAudio8VoiceFields(normalized) ?? {}),
+            ...((0, parler_1.pickParlerDescFields)(normalized) ?? {}),
+            ...((0, audio8_1.pickAudio8VoiceFields)(normalized) ?? {}),
             ...(normalized.instruct === undefined
                 ? {}
                 : { instruct: normalized.instruct }),
@@ -2059,239 +220,39 @@ class TTSGgml {
      * small streamed fragments are coalesced.
      */
     async runStreaming(textStream, options = {}) {
-        await this._waitPocketCancel();
-        this._checkPocketReload();
-        this._assertSentenceStreamingAllowed("runStreaming");
-        const jobFields = this._resolveJobFields(options, "runStreaming");
-        const streamOptions = this._resolveRunStreamingOptions(textStream, options);
-        let normalized = this._normalizeTextStream(textStream);
+        await this._pocket?.waitCancel();
+        this._pocket?.checkReload();
+        (0, streamingSupport_1.assertSentenceStreamingAllowed)(this._setup, "runStreaming");
+        const jobFields = (0, jobFields_1.resolveJobFields)(this._setup, options, "runStreaming");
+        const streamOptions = (0, sentenceStream_1.resolveRunStreamingOptions)(textStream, options);
+        let normalized = (0, sentenceStream_1.normalizeTextStream)(textStream);
         if (streamOptions.accumulateSentences) {
             normalized = (0, textStreamAccumulator_1.accumulateTextStream)(normalized, {
                 sentenceDelimiterPreset: streamOptions.sentenceDelimiterPreset,
                 maxBufferScalars: streamOptions.maxBufferScalars,
                 flushAfterMs: streamOptions.flushAfterMs,
                 sentenceDelimiter: streamOptions.sentenceDelimiter,
-                language: this._config.language,
-            });
-        }
-        const runStream = () => this._runTextStreamOrchestrator(normalized, jobFields);
-        return this.exclusiveRun
-            ? this._enqueueExclusiveTtsResponse(runStream)
-            : runStream();
-    }
-    async _enqueueExclusiveTtsResponse(run) {
-        const previous = this._ttsInferenceQueueWaiter;
-        let releaseSlot = () => { };
-        this._ttsInferenceQueueWaiter = new Promise((resolve) => {
-            releaseSlot = resolve;
-        });
-        await previous;
-        let response;
-        try {
-            response = run();
-        }
-        catch (error) {
-            releaseSlot();
-            throw error;
-        }
-        void response
-            .await()
-            .finally(releaseSlot)
-            .catch(() => { });
-        return response;
-    }
-    _resolveRunStreamingOptions(textStream, options) {
-        const normalized = options == null || typeof options !== "object" ? {} : options;
-        let accumulateSentences = normalized.accumulateSentences;
-        if (accumulateSentences === undefined) {
-            accumulateSentences =
-                defaultAccumulateSentencesForStreamInput(textStream);
-        }
-        return {
-            accumulateSentences: !!accumulateSentences,
-            sentenceDelimiterPreset: normalized.sentenceDelimiterPreset === "latin" ||
-                normalized.sentenceDelimiterPreset === "cjk" ||
-                normalized.sentenceDelimiterPreset === "multilingual"
-                ? normalized.sentenceDelimiterPreset
-                : "multilingual",
-            maxBufferScalars: normalized.maxBufferScalars,
-            flushAfterMs: normalized.flushAfterMs ?? textStreamAccumulator_1.DEFAULT_FLUSH_AFTER_MS,
-            sentenceDelimiter: normalized.sentenceDelimiter instanceof RegExp
-                ? normalized.sentenceDelimiter
-                : undefined,
-        };
-    }
-    _normalizeTextStream(textStream) {
-        if (textStream == null) {
-            throw new QvacErrorAddonTTSGgml({
-                code: ERR_CODES.FAILED_TO_APPEND,
-                adds: "runStreaming: text stream is required",
+                language: this._setup.config.language,
             });
         }
-        if (typeof textStream === "string") {
-            // eslint-disable-next-line @typescript-eslint/require-await -- async iterable shape is required by the public API.
-            return (async function* oneString() {
-                yield textStream;
-            })();
-        }
-        if (typeof textStream[Symbol.asyncIterator] === "function") {
-            return textStream;
-        }
-        if (Array.isArray(textStream) ||
-            typeof textStream[Symbol.iterator] === "function") {
-            // eslint-disable-next-line @typescript-eslint/require-await -- adapts synchronous iterables to the async iterable contract.
-            return (async function* fromIterable() {
-                for (const value of textStream) {
-                    yield value;
-                }
-            })();
-        }
-        throw new QvacErrorAddonTTSGgml({
-            code: ERR_CODES.FAILED_TO_APPEND,
-            adds: "runStreaming: expected string, array of strings, Iterable, or AsyncIterable",
+        return this._startStream(() => {
+            this._pocket?.checkRequest();
+            return this._sentenceStream.streamText(normalized, jobFields);
         });
     }
-    _runTextStreamOrchestrator(source, jobFields) {
-        this._checkPocketRequest();
-        const response = this._job.start();
-        this._sentenceStreamCtx = {
-            textStreamMode: true,
-            asyncTextSource: source,
-            chunks: [],
-            chunkIdx: 0,
-            acc: emptyStreamAccumulator(),
-            chunkResolver: null,
-            jobFields,
-        };
-        const context = this._sentenceStreamCtx;
-        void this._sentenceStreamTextIterableDrive().catch((error) => {
-            if (context !== this._sentenceStreamCtx)
-                return;
-            this._rejectActiveChunk(error);
-            this._sentenceStreamCtx = null;
-            this._job.fail(normalizeError(error));
-        });
-        return response;
-    }
-    async _sentenceStreamTextIterableDrive() {
-        const context = this._sentenceStreamCtx;
-        if (!context ||
-            !context.textStreamMode ||
-            !context.asyncTextSource) {
-            return;
-        }
-        try {
-            for await (const piece of context.asyncTextSource) {
-                if (context !== this._sentenceStreamCtx)
-                    return;
-                const text = String(piece).trim();
-                if (text.length === 0)
-                    continue;
-                context.chunks.push(text);
-                context.chunkIdx = context.chunks.length - 1;
-                const done = new Promise((resolve, reject) => {
-                    context.chunkResolver = { resolve, reject };
-                });
-                // Dispatch may fail before the completion promise is awaited.
-                void done.catch(() => { });
-                await this._dispatchJob({
-                    type: "text",
-                    input: text,
-                    ...(context.jobFields ?? {}),
-                });
-                await done;
-            }
-        }
-        catch (error) {
-            if (context !== this._sentenceStreamCtx)
-                return;
-            this._rejectActiveChunk(error);
-            this._sentenceStreamCtx = null;
-            this._job.fail(normalizeError(error));
-            return;
-        }
-        if (context !== this._sentenceStreamCtx)
-            return;
-        const current = this._sentenceStreamCtx;
-        const chunks = current?.chunks || [];
-        const accumulator = current?.acc || emptyStreamAccumulator();
-        this._sentenceStreamCtx = null;
-        this._endJobWithStats(chunks.length === 0
-            ? {
-                totalTime: 0,
-                tokensPerSecond: 0,
-                realTimeFactor: 0,
-                audioDurationMs: 0,
-                totalSamples: 0,
-            }
-            : computeSentenceStreamStats(chunks, accumulator, this._pacesOnFrames()));
-    }
-    /**
-     * Audio8 and MOSS report `tokensPerSecond` as codec frames per second, so
-     * the streaming aggregate has to count frames too rather than characters.
-     */
-    _pacesOnFrames() {
-        return (this._engineType === ENGINE_AUDIO8 || this._engineType === ENGINE_MOSS);
+    /** Streams line up behind each other under `exclusiveRun`; otherwise they start at once. */
+    _startStream(run) {
+        return this.exclusiveRun ? this._exclusiveStreams.enqueue(run) : run();
     }
     _runStreamOrchestrator(text, options, jobFields) {
-        this._checkPocketRequest();
-        const chunks = (0, textChunker_1.splitTtsText)(String(text), {
-            language: this._config.language,
-            locale: options.locale,
-            maxScalars: options.maxChunkScalars,
-        });
-        if (chunks.length === 0) {
-            throw new QvacErrorAddonTTSGgml({
-                code: ERR_CODES.FAILED_TO_APPEND,
-                adds: "chunked synthesis: text produced no chunks after split",
-            });
-        }
-        this._checkPocketRequest();
-        const response = this._job.start();
-        this._sentenceStreamCtx = {
-            chunks,
-            chunkIdx: 0,
-            acc: emptyStreamAccumulator(),
-            chunkResolver: null,
-            jobFields,
-        };
-        const context = this._sentenceStreamCtx;
-        void this._sentenceStreamDriveBody().catch((error) => {
-            if (context !== this._sentenceStreamCtx)
-                return;
-            this._rejectActiveChunk(error);
-            this._sentenceStreamCtx = null;
-            this._job.fail(normalizeError(error));
-        });
-        return response;
-    }
-    async _sentenceStreamDriveBody() {
-        const context = this._sentenceStreamCtx;
-        if (!context || context.textStreamMode)
-            return;
-        for (let index = 0; index < context.chunks.length; index++) {
-            if (context !== this._sentenceStreamCtx)
-                return;
-            context.chunkIdx = index;
-            const done = new Promise((resolve, reject) => {
-                context.chunkResolver = { resolve, reject };
-            });
-            // Dispatch may fail before the completion promise is awaited.
-            void done.catch(() => { });
-            await this._dispatchJob({
-                type: "text",
-                input: context.chunks[index],
-                ...(context.jobFields ?? {}),
-            });
-            await done;
-        }
-        if (context !== this._sentenceStreamCtx)
-            return;
-        this._sentenceStreamCtx = null;
+        this._pocket?.checkRequest();
+        const chunks = (0, sentenceStream_1.splitSentenceChunks)(text, options, this._setup.config.language);
+        this._pocket?.checkRequest();
+        return this._sentenceStream.streamChunks(chunks, jobFields);
     }
     async _load() {
-        this._getLogger().info("[TTSGgml] Language:", this._config.language || "en");
-        const addon = this._createAddon(this._buildTtsParams(), this._addonOutputCallback.bind(this));
+        this._getLogger().info("[TTSGgml] Language:", this._setup.config.language || "en");
+        const addon = this._spawnAddon(this._buildTtsParams());
         this.addon = addon;
         try {
             await addon.activate();
@@ -2307,407 +268,11 @@ class TTSGgml {
         }
     }
     _buildTtsParams() {
-        if (this._pocketParams)
-            return { ...this._pocketParams };
-        if (this._engineType === ENGINE_SUPERTONIC) {
-            return this._buildSupertonicParams();
-        }
-        if (this._engineType === ENGINE_COSYVOICE3) {
-            return this._buildCosyvoiceParams();
-        }
-        if (this._engineType === ENGINE_PARLER) {
-            return this._buildParlerParams();
-        }
-        if (this._engineType === ENGINE_AUDIO8) {
-            return this._buildAudio8Params();
-        }
-        if (this._engineType === ENGINE_MOSS) {
-            return this._buildMossParams();
-        }
-        if (this._engineType === ENGINE_MOSS_SFX) {
-            return this._buildMossSoundEffectParams();
-        }
-        if (this._engineType === ENGINE_MOSS_SPEECH) {
-            return this._buildMossSpeechParams();
-        }
-        return this._buildChatterboxParams();
-    }
-    _buildCosyvoiceParams() {
-        const parameters = {
-            engineType: ENGINE_COSYVOICE3,
-            cosyvoiceModelDir: this._cosyvoiceModelDir || "",
-            language: this._config.language || "en",
-        };
-        if (this._cosyvoiceLlmModelPath) {
-            parameters.cosyvoiceLlmModelPath = this._cosyvoiceLlmModelPath;
-        }
-        if (this._cosyvoiceFlowModelPath) {
-            parameters.cosyvoiceFlowModelPath = this._cosyvoiceFlowModelPath;
-        }
-        if (this._cosyvoiceHiftModelPath) {
-            parameters.cosyvoiceHiftModelPath = this._cosyvoiceHiftModelPath;
-        }
-        if (this._cosyvoiceS3tokModelPath) {
-            parameters.cosyvoiceS3tokModelPath = this._cosyvoiceS3tokModelPath;
-        }
-        if (this._cosyvoiceCampplusModelPath) {
-            parameters.cosyvoiceCampplusModelPath = this._cosyvoiceCampplusModelPath;
-        }
-        if (this._cosyvoiceVocabPath) {
-            parameters.cosyvoiceVocabPath = this._cosyvoiceVocabPath;
-        }
-        if (this._cosyvoiceMergesPath) {
-            parameters.cosyvoiceMergesPath = this._cosyvoiceMergesPath;
-        }
-        if (this._cosyvoiceVoiceModelPath) {
-            parameters.cosyvoiceVoiceModelPath = this._cosyvoiceVoiceModelPath;
-        }
-        if (this._vulkanDevice != null) {
-            parameters.vulkanDevice = this._vulkanDevice | 0;
-        }
-        if (this._flowCutPrompt != null) {
-            parameters.flowCutPrompt = !!this._flowCutPrompt;
-        }
-        if (this._referenceAudio != null) {
-            parameters.referenceAudio = this._referenceAudio;
-        }
-        if (this._promptText != null) {
-            parameters.promptText = String(this._promptText);
-        }
-        if (this._instruct)
-            parameters.instruct = this._instruct;
-        // Canonical values; the native layer maps them to the trained instructions.
-        if (this._emotion != null)
-            parameters.emotion = String(this._emotion);
-        if (this._pace != null)
-            parameters.pace = String(this._pace);
-        if (this._voice)
-            parameters.voice = this._voice;
-        this._assignCommonNativeParams(parameters);
-        if (this._cfmSteps != null)
-            parameters.cfmSteps = this._cfmSteps | 0;
-        if (this._streamChunkTokens != null) {
-            parameters.streamChunkTokens = this._streamChunkTokens | 0;
-        }
-        if (this._streamFirstChunkTokens != null) {
-            parameters.streamFirstChunkTokens = this._streamFirstChunkTokens | 0;
-        }
-        if (this._streamLeftContextTokens != null) {
-            parameters.streamLeftContextTokens = this._streamLeftContextTokens | 0;
-        }
-        return parameters;
-    }
-    _buildChatterboxParams() {
-        const parameters = {
-            engineType: ENGINE_CHATTERBOX,
-            t3ModelPath: this._t3ModelPath || "",
-            s3genModelPath: this._s3genModelPath || "",
-            language: this._config.language || "en",
-        };
-        this._assignCommonNativeParams(parameters);
-        if (this._referenceAudio != null) {
-            parameters.referenceAudio = this._referenceAudio;
-        }
-        if (this._voiceDir != null)
-            parameters.voiceDir = this._voiceDir;
-        if (this._nCtx != null)
-            parameters.nCtx = this._nCtx | 0;
-        if (this._kvCacheType != null) {
-            parameters.kvCacheType = String(this._kvCacheType);
-        }
-        if (this._streamChunkTokens != null) {
-            parameters.streamChunkTokens = this._streamChunkTokens | 0;
-        }
-        if (this._streamFirstChunkTokens != null) {
-            parameters.streamFirstChunkTokens =
-                this._streamFirstChunkTokens | 0;
-        }
-        if (this._cfmSteps != null) {
-            parameters.cfmSteps = this._cfmSteps | 0;
-        }
-        if (this._streamLeftContextTokens != null) {
-            parameters.streamLeftContextTokens = this._streamLeftContextTokens | 0;
-        }
-        if (this._batchCfmSteps != null) {
-            parameters.batchCfmSteps = this._batchCfmSteps | 0;
-        }
-        if (this._cfgRate != null) {
-            parameters.cfgRate = Number(this._cfgRate);
-        }
-        if (this._nPredict != null)
-            parameters.nPredict = this._nPredict | 0;
-        if (this._maxSentenceChars != null) {
-            parameters.maxSentenceChars = this._maxSentenceChars | 0;
-        }
-        if (this._crossfadeMs != null) {
-            parameters.crossfadeMs = this._crossfadeMs | 0;
-        }
-        this._assignSamplingParams(parameters);
-        if (this._repeatPenalty != null) {
-            parameters.repeatPenalty = Number(this._repeatPenalty);
-        }
-        if (this._exaggeration != null) {
-            parameters.exaggeration = Number(this._exaggeration);
-        }
-        if (this._cfgWeight != null) {
-            parameters.cfgWeight = Number(this._cfgWeight);
-        }
-        if (this._minP != null)
-            parameters.minP = Number(this._minP);
-        if (this._speed != null)
-            parameters.speed = Number(this._speed);
-        if (this._mecabDictPath) {
-            parameters.mecabDictPath = this._mecabDictPath;
-        }
-        if (this._cangjieTsvPath) {
-            parameters.cangjieTsvPath = this._cangjieTsvPath;
-        }
-        return parameters;
-    }
-    _buildSupertonicParams() {
-        const parameters = {
-            engineType: ENGINE_SUPERTONIC,
-            supertonicModelPath: this._supertonicModelPath || "",
-            language: this._config.language || "en",
-        };
-        this._assignCommonNativeParams(parameters);
-        if (this._voice)
-            parameters.voice = this._voice;
-        if (this._steps != null)
-            parameters.steps = this._steps | 0;
-        if (this._speed != null)
-            parameters.speed = Number(this._speed);
-        // The native layer maps the step onto a multiplier relative to the GGUF's
-        // own default_speed, and rejects pace + speed together.
-        if (this._pace != null)
-            parameters.pace = String(this._pace);
-        if (this._noiseNpyPath) {
-            parameters.noiseNpyPath = this._noiseNpyPath;
-        }
-        if (this._vulkanCacheDir) {
-            parameters.vulkanCacheDir = this._vulkanCacheDir;
-        }
-        if (this._voiceJsonPath)
-            parameters.voiceJsonPath = this._voiceJsonPath;
-        if (this._prewarmText)
-            parameters.prewarmText = this._prewarmText;
-        if (this._vulkanDevice != null) {
-            parameters.vulkanDevice = this._vulkanDevice | 0;
-        }
-        if (this._streamChunkTokens != null) {
-            parameters.streamChunkTokens = this._streamChunkTokens | 0;
-        }
-        if (this._streamFirstChunkTokens != null) {
-            parameters.streamFirstChunkTokens = this._streamFirstChunkTokens | 0;
-        }
-        if (this._streamChunkTolerancePct != null) {
-            parameters.streamChunkTolerancePct = this._streamChunkTolerancePct | 0;
-        }
-        if (this._streamMinChunkTokens != null) {
-            parameters.streamMinChunkTokens = this._streamMinChunkTokens | 0;
-        }
-        return parameters;
-    }
-    _buildParlerParams() {
-        // Re-checked here (not only in the constructor) so reload({...}) cannot
-        // smuggle in a description/template conflict.
-        if (this._description != null) {
-            const templateSet = [
-                this._voice,
-                this._emotion,
-                this._pitch,
-                this._pace,
-                this._expressivity,
-                this._noise,
-                this._reverb,
-                this._quality,
-            ].some((value) => value != null && value !== "");
-            if (templateSet) {
-                throw new Error("tts-ggml: 'description' is mutually exclusive with the " +
-                    "voice-template options");
-            }
-        }
-        const parameters = {
-            engineType: ENGINE_PARLER,
-            parlerModelPath: this._parlerModelPath || "",
-        };
-        if (this._description != null) {
-            parameters.description = String(this._description);
-        }
-        if (this._voice)
-            parameters.voice = String(this._voice);
-        if (this._emotion != null) {
-            parameters.emotion = String(this._emotion);
-        }
-        if (this._pitch != null)
-            parameters.pitch = String(this._pitch);
-        if (this._pace != null)
-            parameters.pace = String(this._pace);
-        if (this._expressivity != null) {
-            parameters.expressivity = String(this._expressivity);
-        }
-        if (this._noise != null)
-            parameters.noise = String(this._noise);
-        if (this._reverb != null)
-            parameters.reverb = String(this._reverb);
-        if (this._quality != null) {
-            parameters.quality = String(this._quality);
-        }
-        this._assignSamplingParams(parameters);
-        if (this._minNewTokens != null) {
-            parameters.minNewTokens = this._minNewTokens | 0;
-        }
-        if (this._streamChunkTokens != null) {
-            parameters.streamChunkTokens = this._streamChunkTokens | 0;
-        }
-        if (this._streamFirstChunkTokens != null) {
-            parameters.streamFirstChunkTokens =
-                this._streamFirstChunkTokens | 0;
-        }
-        if (this._normalizeNumbers != null) {
-            parameters.normalizeNumbers = !!this._normalizeNumbers;
-        }
-        this._assignBackendParams(parameters);
-        return parameters;
-    }
-    _buildAudio8Params() {
-        // Backstop for callers that reach the params build by another route than
-        // the constructor or _applyAudio8Reload, both of which check first.
-        this._assertAudio8VoiceConsistent({
-            referenceAudio: this._referenceAudio,
-            referenceText: this._referenceText,
-        }, "reload");
-        const parameters = {
-            engineType: ENGINE_AUDIO8,
-            audio8LmPath: this._audio8LmPath || "",
-            audio8CodecDecoderPath: this._audio8CodecDecoderPath || "",
-        };
-        if (this._audio8CodecEncoderPath) {
-            parameters.audio8CodecEncoderPath = this._audio8CodecEncoderPath;
-        }
-        if (this._referenceAudio != null) {
-            parameters.referenceAudio = this._referenceAudio;
-        }
-        if (this._referenceText != null) {
-            parameters.referenceText = this._referenceText;
-        }
-        if (this._greedy != null)
-            parameters.greedy = !!this._greedy;
-        this._assignSamplingParams(parameters);
-        this._assignBackendParams(parameters);
-        return parameters;
-    }
-    _buildMossParams() {
-        this._assertMossOutputRate();
-        const parameters = {
-            engineType: ENGINE_MOSS,
-            mossBackbonePath: this._mossBackbonePath || "",
-            mossCodecDecoderPath: this._mossCodecDecoderPath || "",
-            language: this._config.language || "en",
-        };
-        if (this._mossCodecEncoderPath) {
-            parameters.mossCodecEncoderPath = this._mossCodecEncoderPath;
-        }
-        if (this._referenceAudio != null) {
-            parameters.referenceAudio = this._referenceAudio;
-        }
-        if (this._dialogueReferences) {
-            parameters.dialogueReferences = [...this._dialogueReferences];
-        }
-        if (this._durationTokens != null) {
-            parameters.durationTokens = this._durationTokens | 0;
-        }
-        if (this._streamChunkTokens != null) {
-            parameters.streamChunkTokens = this._streamChunkTokens | 0;
-        }
-        this._assignBackendParams(parameters);
-        return parameters;
-    }
-    _buildMossSoundEffectParams() {
-        this._assertMossSoundEffectOutputRate();
-        const parameters = {
-            engineType: ENGINE_MOSS_SFX,
-            mossSoundEffectPath: this._mossSoundEffectPath || "",
-        };
-        this._assignBackendParams(parameters);
-        return parameters;
-    }
-    _buildMossSpeechParams() {
-        this._assertMossSpeechOutputRate();
-        const parameters = {
-            engineType: ENGINE_MOSS_SPEECH,
-            mossSpeechModelPath: this._mossSpeechModelPath || "",
-            mossSpeechCodecPath: this._mossSpeechCodecPath || "",
-        };
-        this._assignBackendParams(parameters);
-        return parameters;
-    }
-    /** The knobs every engine in SAMPLING_ENGINES reads. */
-    _assignSamplingParams(parameters) {
-        if (this._temperature != null) {
-            parameters.temperature = Number(this._temperature);
-        }
-        if (this._topK != null)
-            parameters.topK = this._topK | 0;
-        if (this._topP != null)
-            parameters.topP = Number(this._topP);
-        if (this._maxFrames != null) {
-            parameters.maxFrames = this._maxFrames | 0;
-        }
-    }
-    /**
-     * Backend and output plumbing for the engines that take no `language`,
-     * where `_assignCommonNativeParams` would be the wrong shape.
-     */
-    _assignBackendParams(parameters) {
-        if (this._seed != null)
-            parameters.seed = this._seed | 0;
-        if (this._threads != null)
-            parameters.threads = this._threads | 0;
-        if (this._outputSampleRate != null) {
-            parameters.outputSampleRate = this._outputSampleRate | 0;
-        }
-        if (this._nGpuLayers != null) {
-            parameters.nGpuLayers = this._nGpuLayers | 0;
-        }
-        if (this._config.useGPU != null) {
-            parameters.useGPU = !!this._config.useGPU;
-        }
-        this._assignLavasrParams(parameters);
-        if (this._backendsDir) {
-            parameters.backendsDir = this._backendsDir;
-        }
-    }
-    _assignCommonNativeParams(parameters) {
-        if (this._seed != null)
-            parameters.seed = this._seed | 0;
-        if (this._threads != null)
-            parameters.threads = this._threads | 0;
-        if (this._nGpuLayers != null) {
-            parameters.nGpuLayers = this._nGpuLayers | 0;
-        }
-        if (this._outputSampleRate != null) {
-            parameters.outputSampleRate = this._outputSampleRate | 0;
-        }
-        if (this._config.useGPU != null) {
-            parameters.useGPU = !!this._config.useGPU;
-        }
-        this._assignLavasrParams(parameters);
-        if (this._backendsDir) {
-            parameters.backendsDir = this._backendsDir;
-        }
-        if (this._openclCacheDir) {
-            parameters.openclCacheDir = this._openclCacheDir;
-        }
+        return (0, engineParams_1.buildNativeParams)(this._setup);
     }
-    /** LavaSR post-processing paths, shared by every engine that supports them. */
-    _assignLavasrParams(parameters) {
-        if (this._enhancerGgufPath) {
-            parameters.lavasrEnhancerPath = this._enhancerGgufPath;
-        }
-        if (this._denoiserGgufPath) {
-            parameters.lavasrDenoiserPath = this._denoiserGgufPath;
-        }
+    /** A native instance that reports through this wrapper's output callback. */
+    _spawnAddon(configuration) {
+        return this._createAddon(configuration, this._addonOutputCallback.bind(this));
     }
     _createAddon(configuration, outputCallback) {
         // eslint-disable-next-line @typescript-eslint/no-require-imports -- native binding is resolved lazily from package prebuilds.
@@ -2719,7 +284,7 @@ class TTSGgml {
     }
     async _unloadModel() {
         await this.cancel();
-        this._failAndClearActiveResponse("Model was unloaded");
+        this._sentenceStream.fail("Model was unloaded");
         const addon = this._optionalAddon();
         try {
             if (addon)
@@ -2739,36 +304,24 @@ class TTSGgml {
             this.state.destroyed = true;
         });
     }
+    async _withPocketLifecycle(action) {
+        return this._pocket ? this._pocket.withLifecycle(action) : action();
+    }
     async _runInternal(input) {
-        await this._waitPocketCancel();
-        this._checkPocketRequest();
-        const jobFields = this._resolveJobFields(input, "run");
+        await this._pocket?.waitCancel();
+        this._pocket?.checkRequest();
+        const jobFields = (0, jobFields_1.resolveJobFields)(this._setup, input, "run");
         const signal = input?.signal;
-        const pocketSignal = this._engineType === ENGINE_POCKET && signal && !signal.aborted ? signal : undefined;
+        // Pocket keeps the signal to itself: the abort fails the response with
+        // the caller's reason and stops native work behind the completion barrier.
+        const pocketSignal = this._pocket && signal && !signal.aborted ? signal : undefined;
         const response = this._job.start({
             signal: pocketSignal ? undefined : signal,
         });
         if (signal?.aborted)
             return response;
-        if (pocketSignal) {
-            const onAbort = () => {
-                if (this._job.active !== response)
-                    return;
-                const reason = pocketSignal.reason;
-                const error = reason instanceof Error ? reason : new Error(typeof reason === "string" ? `Aborted: ${reason}` : "Aborted");
-                // Preserve the caller's reason while stopping native work. Cancellation
-                // retains the completion barrier until its terminal callback arrives.
-                try {
-                    response.failed(error);
-                }
-                finally {
-                    void this.cancel().catch((error) => this._getLogger().error("Pocket abort cancellation failed", error));
-                }
-            };
-            pocketSignal.addEventListener("abort", onAbort, { once: true });
-            const detach = () => pocketSignal.removeEventListener("abort", onAbort);
-            void response.await().then(detach, detach);
-        }
+        if (pocketSignal)
+            this._pocket?.attachAbort(response, pocketSignal);
         try {
             await this._dispatchJob({
                 type: input.type || "text",
@@ -2777,495 +330,42 @@ class TTSGgml {
             });
         }
         catch (error) {
-            this._job.fail(normalizeError(error));
+            this._job.fail((0, sentenceStream_1.normalizeError)(error));
             throw error;
         }
         return response;
     }
-    _mergeSentenceStreamStats(accumulator, data) {
-        if (accumulator.firstAudioMs === undefined &&
-            typeof data.firstAudioMs === "number" &&
-            Number.isFinite(data.firstAudioMs) &&
-            data.firstAudioMs >= 0) {
-            accumulator.firstAudioMs = data.firstAudioMs;
-        }
-        accumulator.totalTime +=
-            typeof data.totalTime === "number" ? data.totalTime : 0;
-        accumulator.audioDurationMs +=
-            typeof data.audioDurationMs === "number"
-                ? data.audioDurationMs
-                : 0;
-        accumulator.totalSamples +=
-            typeof data.totalSamples === "number" ? data.totalSamples : 0;
-        accumulator.generatedFrames +=
-            typeof data.generatedFrames === "number" ? data.generatedFrames : 0;
-        if (typeof data.codecSidecarLoaded === "number") {
-            accumulator.codecSidecarLoaded = data.codecSidecarLoaded;
-        }
-        if (typeof data.codecOnCoreml === "number") {
-            accumulator.codecOnCoreml = data.codecOnCoreml;
-        }
-    }
-    _rejectActiveChunk(error) {
-        const resolver = this._sentenceStreamCtx?.chunkResolver;
-        if (!resolver)
-            return;
-        this._sentenceStreamCtx.chunkResolver = null;
-        resolver.reject(error);
-    }
-    _endJobWithStats(stats) {
-        if (this.opts.stats)
-            this._job.end(stats);
-        else
-            this._job.end();
-    }
     _addonOutputCallback(_addon, event, data, error) {
-        if (this._pocketJobPending && ((typeof error === "string" && error.length > 0) || isStatsEvent(data))) {
-            const pending = this._pocketJobPending;
-            this._pocketJobPending = null;
-            pending.resolve();
-        }
-        if (typeof error === "string" && error.length > 0) {
-            this._handleAddonError(error);
-        }
-        else if (isAudioOutputEvent(data)) {
-            this._handleAddonOutput(data);
-        }
-        else if (isStatsEvent(data)) {
-            this._handleAddonStats(data);
-        }
-        else {
-            this._getLogger().debug(`Received TTS event: ${String(event)}`);
-        }
-    }
-    _handleAddonError(error) {
-        this._getLogger().error(`TTS job failed with error: ${error}`);
-        this._rejectActiveChunk(new Error(error));
-        this._job.fail(error);
-    }
-    _handleAddonOutput(data) {
-        try {
-            this._getLogger().debug(`TTS job produced output: ${ttsOutputDebugString(data)}`);
-        }
-        catch (error) {
-            if (error instanceof RangeError) {
-                this._getLogger().debug("TTS job produced output: [data too large]");
-            }
-            else {
-                throw error;
-            }
-        }
-        this._job.output(this._sentenceStreamCtx
-            ? this._enrichStreamChunk(data)
-            : data);
-    }
-    _enrichStreamChunk(data) {
-        const context = this._sentenceStreamCtx;
-        if (!context) {
-            return data;
-        }
-        const index = context.chunkIdx;
-        const enriched = {
-            outputArray: data.outputArray,
-            chunkIndex: index,
-            sentenceChunk: context.chunks[index] || "",
-        };
-        if (data.sampleRate != null) {
-            enriched.sampleRate = data.sampleRate;
-        }
-        if (!context.textStreamMode) {
-            enriched.isLast = index >= context.chunks.length - 1 && data.isLast !== false;
-        }
-        return enriched;
-    }
-    _handleAddonStats(data) {
-        this._getLogger().info(`TTS job completed. Stats: ${JSON.stringify(data)}`);
-        const context = this._sentenceStreamCtx;
-        if (!context) {
-            this._endJobWithStats(data);
-            return;
-        }
-        this._mergeSentenceStreamStats(context.acc, data);
-        if (context.chunkResolver) {
-            context.chunkResolver.resolve();
-            context.chunkResolver = null;
-        }
-        if (!context.textStreamMode &&
-            context.chunkIdx >= context.chunks.length - 1) {
-            this._endJobWithStats(computeSentenceStreamStats(context.chunks, context.acc, this._pacesOnFrames()));
-        }
-    }
-    async _waitPocketCancel() {
-        if (this._engineType === ENGINE_POCKET && this._pocketCancelPromise)
-            await this._pocketCancelPromise;
-    }
-    async _withPocketLifecycle(action) {
-        if (this._engineType !== ENGINE_POCKET)
-            return action();
-        await this._waitPocketCancel();
-        this._checkPocketReload();
-        this._pocketLifecycleInProgress = true;
-        try {
-            await action();
-        }
-        finally {
-            this._pocketLifecycleInProgress = false;
-        }
-    }
-    _checkPocketReload() {
-        if (this._pocketLifecycleInProgress)
-            throw new Error("Pocket lifecycle operation is already in progress");
-    }
-    _checkPocketRequest() {
-        if (this._engineType === ENGINE_POCKET) {
-            if (this.state.destroyed)
-                throw new Error("Pocket instance was destroyed");
-            if (!this.state.weightsLoaded || !this._optionalAddon())
-                throw new Error("Pocket model is not loaded");
-        }
-        if (this._engineType === ENGINE_POCKET && (this._job.active || this._pocketJobPending)) {
-            throw new Error("Pocket synthesis is already in progress");
-        }
-        this._checkPocketReload();
+        this._pocket?.onNativeEvent(data, error);
+        this._events.route(event, data, error);
     }
     async _dispatchJob(input) {
         const addon = this._requireAddon();
-        if (this._engineType !== ENGINE_POCKET)
-            return addon.runJob(input);
-        if (this._pocketJobPending)
-            throw new Error("Pocket native job is already in progress");
-        let resolve = () => { };
-        const promise = new Promise((done) => { resolve = done; });
-        const pending = { promise, resolve };
-        this._pocketJobPending = pending;
-        try {
-            await addon.runJob(input);
-        }
-        catch (error) {
-            if (this._pocketJobPending === pending)
-                this._pocketJobPending = null;
-            resolve();
-            throw error;
+        if (this._setup.engineType === engineIds_1.ENGINE_AUDIO8 && this._sentenceStream.ctx) {
+            return addon.runJob({ ...input, streamOutput: true });
         }
+        return this._pocket ? this._pocket.runJob(addon, input) : addon.runJob(input);
     }
     async cancel() {
+        if (this._pocket)
+            return this._pocket.cancel();
         const addon = this._optionalAddon();
-        if (this._engineType !== ENGINE_POCKET) {
-            if (addon?.cancel)
-                await addon.cancel();
-            return;
-        }
-        if (this._pocketCancelPromise)
-            return this._pocketCancelPromise;
-        const pending = this._pocketJobPending;
-        this._pocketCancelPromise = (async () => {
-            this._failAndClearActiveResponse("Synthesis cancelled");
-            if (addon?.cancel)
-                await addon.cancel();
-            // Joining the worker can precede delivery of its terminal callback.
-            if (pending)
-                await pending.promise;
-        })();
-        try {
-            await this._pocketCancelPromise;
-        }
-        finally {
-            this._pocketCancelPromise = null;
-        }
-    }
-    _failAndClearActiveResponse(reason) {
-        this._rejectActiveChunk(reason instanceof Error ? reason : new Error(reason));
-        this._sentenceStreamCtx = null;
-        this._job.fail(reason);
-    }
-    /** Everything reload() may overwrite, so a rejected reload can undo itself. */
-    _captureReloadableState() {
-        return {
-            language: this._config.language,
-            useGPU: this._config.useGPU,
-            outputSampleRate: this._outputSampleRate,
-            emotion: this._emotion,
-            pace: this._pace,
-            description: this._description,
-            voice: this._voice,
-            pitch: this._pitch,
-            expressivity: this._expressivity,
-            noise: this._noise,
-            reverb: this._reverb,
-            quality: this._quality,
-            temperature: this._temperature,
-            topK: this._topK,
-            topP: this._topP,
-            maxFrames: this._maxFrames,
-            minNewTokens: this._minNewTokens,
-            normalizeNumbers: this._normalizeNumbers,
-            seed: this._seed,
-            durationTokens: this._durationTokens,
-        };
-    }
-    _restoreReloadableState(state) {
-        this._config.language = state.language;
-        this._config.useGPU = state.useGPU;
-        this._outputSampleRate = state.outputSampleRate;
-        this._emotion = state.emotion;
-        this._pace = state.pace;
-        this._description = state.description;
-        this._voice = state.voice;
-        this._pitch = state.pitch;
-        this._expressivity = state.expressivity;
-        this._noise = state.noise;
-        this._reverb = state.reverb;
-        this._quality = state.quality;
-        this._temperature = state.temperature;
-        this._topK = state.topK;
-        this._topP = state.topP;
-        this._maxFrames = state.maxFrames;
-        this._minNewTokens = state.minNewTokens;
-        this._normalizeNumbers = state.normalizeNumbers;
-        this._seed = state.seed;
-        this._durationTokens = state.durationTokens;
-    }
-    _applyReloadableRuntimeConfig(newConfig) {
-        this._assertMossReloadKeepsVoice(newConfig);
-        const runtimeConfig = newConfig;
-        if (runtimeConfig.language !== undefined) {
-            this._config.language = runtimeConfig.language;
-        }
-        if (runtimeConfig.useGPU !== undefined) {
-            this._config.useGPU = runtimeConfig.useGPU;
-        }
-        if (runtimeConfig.outputSampleRate !== undefined) {
-            this._outputSampleRate = validateOutputSampleRate(runtimeConfig.outputSampleRate);
-        }
-        this._applyAudio8Reload(newConfig);
-        this._applyMossReload(newConfig);
-    }
-    _assertMossReloadKeepsVoice(newConfig) {
-        if (this._engineType !== ENGINE_MOSS)
-            return;
-        const voiceKeys = MOSS_INSTANCE_VOICE_KEYS.filter((key) => newConfig[key] !== undefined);
-        if (voiceKeys.length === 0)
-            return;
-        throw new Error(`tts-ggml: reload: the moss engine encodes ${voiceKeys.join(", ")} ` +
-            "once per instance; create a new instance to clone different recordings");
-    }
-    _applyMossReload(newConfig) {
-        if (this._engineType !== ENGINE_MOSS)
-            return;
-        const durationTokens = newConfig
-            .durationTokens;
-        if (durationTokens === undefined)
-            return;
-        assertMossDurationTokens(durationTokens, "reload");
-        this._durationTokens = durationTokens ?? undefined;
-    }
-    // Cross-engine conditioning is reloadable on every engine that supports it,
-    // so both families change emotion the same way.
-    _applyReloadableConditioning(newConfig) {
-        const conditioning = newConfig;
-        if (conditioning.emotion === undefined &&
-            conditioning.pace === undefined) {
-            return;
-        }
-        if (conditioning.emotion !== undefined) {
-            this._emotion = conditioning.emotion;
-        }
-        if (conditioning.pace !== undefined) {
-            this._pace = conditioning.pace;
-        }
-        this._assertConditioningConsistency("reload");
-    }
-    // Parler description/template + sampling knobs are reloadable; they rebuild
-    // the engine's default description / sampler. _buildParlerParams re-validates
-    // so a wrong-engine reload still throws.
-    _applyReloadableParlerConfig(newConfig) {
-        if (this._engineType !== ENGINE_PARLER)
-            return;
-        const parlerConfig = newConfig;
-        if (parlerConfig.description !== undefined ||
-            parlerConfig.voiceDescription !== undefined) {
-            this._description = firstNonEmpty(parlerConfig.description, parlerConfig.voiceDescription);
-        }
-        if (parlerConfig.voice !== undefined) {
-            this._voice = parlerConfig.voice;
-        }
-        if (parlerConfig.pitch !== undefined) {
-            this._pitch = parlerConfig.pitch;
-        }
-        if (parlerConfig.expressivity !== undefined) {
-            this._expressivity = parlerConfig.expressivity;
-        }
-        if (parlerConfig.noise !== undefined) {
-            this._noise = parlerConfig.noise;
-        }
-        if (parlerConfig.reverb !== undefined) {
-            this._reverb = parlerConfig.reverb;
-        }
-        if (parlerConfig.quality !== undefined) {
-            this._quality = parlerConfig.quality;
-        }
-        if (parlerConfig.temperature !== undefined) {
-            this._temperature = parlerConfig.temperature;
-        }
-        if (parlerConfig.topK !== undefined) {
-            this._topK = parlerConfig.topK;
-        }
-        if (parlerConfig.topP !== undefined) {
-            this._topP = parlerConfig.topP;
-        }
-        if (parlerConfig.maxFrames !== undefined) {
-            this._maxFrames = parlerConfig.maxFrames;
-        }
-        if (parlerConfig.minNewTokens !== undefined) {
-            this._minNewTokens = parlerConfig.minNewTokens;
-        }
-        if (parlerConfig.normalizeNumbers !== undefined) {
-            this._normalizeNumbers = parlerConfig.normalizeNumbers;
-        }
-        if (parlerConfig.seed !== undefined) {
-            this._seed = parlerConfig.seed;
-        }
-    }
-    /**
-     * Apply the new configuration and build the native parameters from it. A
-     * rejected value leaves the instance exactly as it was, so a later partial
-     * reload is not validated against state the caller never accepted.
-     */
-    _applyReloadableConfig(newConfig) {
-        const snapshot = this._captureReloadableState();
-        try {
-            this._applyReloadableRuntimeConfig(newConfig);
-            this._applyReloadableConditioning(newConfig);
-            this._applyReloadableParlerConfig(newConfig);
-            return this._buildTtsParams();
-        }
-        catch (error) {
-            this._restoreReloadableState(snapshot);
-            throw error;
-        }
-    }
-    async _reloadPocket(newConfig) {
-        await this._waitPocketCancel();
-        this._checkPocketReload();
-        if (this.state.destroyed)
-            throw new Error("Pocket instance was destroyed");
-        const options = { ...this._pocketOptions, ...newConfig };
-        const config = { ...this._config, ...newConfig };
-        const params = (0, pocketConfig_1.buildPocketParams)(this._pocketFiles, options, config);
-        this._pocketLifecycleInProgress = true;
-        let replacement = null;
-        try {
-            // Drain native work and release its model before allocating another one.
-            // Failed activation leaves the instance unloaded with its last good config.
-            await this._unloadModel();
-            replacement = this._createAddon(params, this._addonOutputCallback.bind(this));
-            await replacement.activate();
-            this.addon = replacement;
-            replacement = null;
-            this._pocketParams = params;
-            this._pocketOptions = options;
-            this._config.language = "en";
-            this._config.useGPU = false;
-            this._outputSampleRate = typeof params.outputSampleRate === "number" ? params.outputSampleRate : null;
-            this._config.outputSampleRate = this._outputSampleRate ?? undefined;
-            this.state.configLoaded = true;
-            this.state.weightsLoaded = true;
-        }
-        finally {
-            try {
-                if (replacement)
-                    await replacement.destroyInstance();
-            }
-            finally {
-                this._pocketLifecycleInProgress = false;
-            }
-        }
+        if (addon?.cancel)
+            await addon.cancel();
     }
     async reload(newConfig = {}) {
         this._getLogger().debug("Reloading addon with new configuration", newConfig);
-        if (this._engineType === ENGINE_POCKET)
-            return this._reloadPocket(newConfig);
-        const parameters = this._applyReloadableConfig(newConfig);
+        if (this._pocket)
+            return this._pocket.reload(newConfig);
+        const parameters = (0, reload_1.applyReloadableConfig)(this._setup, newConfig);
         await this.cancel();
-        this._failAndClearActiveResponse("Model was reloaded");
+        this._sentenceStream.fail("Model was reloaded");
         const existingAddon = this._optionalAddon();
         if (existingAddon)
             await existingAddon.destroyInstance();
-        this.addon = this._createAddon(parameters, this._addonOutputCallback.bind(this));
+        this.addon = this._spawnAddon(parameters);
         await this._requireAddon().activate();
     }
-    /**
-     * The voice a reload lands on. Same rule as _mergeAudio8Voice and
-     * Audio8Model::resolveVoice: a new recording replaces both halves, because
-     * the configured transcript describes the recording being replaced. Reload
-     * reads `undefined` as "not supplied", so an explicit empty string reaches
-     * the guard instead of being ignored.
-     */
-    _mergeAudio8ReloadVoice(config) {
-        if (config.referenceAudio !== undefined) {
-            return {
-                referenceAudio: config.referenceAudio,
-                referenceText: config.referenceText,
-            };
-        }
-        return {
-            referenceAudio: this._referenceAudio,
-            referenceText: config.referenceText !== undefined
-                ? config.referenceText
-                : this._referenceText,
-        };
-    }
-    /**
-     * The knobs a reload lands on, merged the same way as the voice so the
-     * whole set can be checked before any of it is written.
-     */
-    _mergeAudio8ReloadSampling(config) {
-        return {
-            greedy: config.greedy ?? this._greedy,
-            temperature: config.temperature ?? this._temperature,
-            topK: config.topK ?? this._topK,
-            topP: config.topP ?? this._topP,
-            maxFrames: config.maxFrames ?? this._maxFrames,
-            seed: config.seed ?? this._seed,
-            threads: config.threads ?? this._threads,
-        };
-    }
-    _applyAudio8Sampling(config) {
-        if (config.greedy !== undefined)
-            this._greedy = config.greedy;
-        if (config.temperature !== undefined) {
-            this._temperature = config.temperature;
-        }
-        if (config.topK !== undefined)
-            this._topK = config.topK;
-        if (config.topP !== undefined)
-            this._topP = config.topP;
-        if (config.maxFrames !== undefined) {
-            this._maxFrames = config.maxFrames;
-        }
-        if (config.seed !== undefined)
-            this._seed = config.seed;
-        if (config.threads !== undefined)
-            this._threads = config.threads;
-    }
-    /**
-     * Audio8 voice + sampling knobs are reloadable; they rebuild the engine's
-     * sampler and speaker history. Both merges are checked before either is
-     * written, so a rejected reload leaves the instance exactly as it was
-     * rather than half-moved onto the configuration that was refused.
-     */
-    _applyAudio8Reload(newConfig) {
-        if (this._engineType !== ENGINE_AUDIO8)
-            return;
-        const config = newConfig;
-        const voice = this._mergeAudio8ReloadVoice(config);
-        const sampling = this._mergeAudio8ReloadSampling(config);
-        this._assertAudio8VoiceConsistent(voice, "reload");
-        assertAudio8SamplingValid(sampling, "reload");
-        this._referenceAudio = voice.referenceAudio;
-        this._referenceText = voice.referenceText;
-        this._applyAudio8Sampling(sampling);
-    }
     static getModelKey(_params) {
         void _params;
         return "tts-ggml";
@@ -3279,19 +379,7 @@ class TTSGgml {
     static getVoiceControls() {
         // eslint-disable-next-line @typescript-eslint/no-require-imports -- native binding is resolved lazily from package prebuilds.
         const binding = require("./binding");
-        const native = binding.getVoiceControls();
-        const engines = {};
-        for (const [name, controls] of Object.entries(native.engines)) {
-            engines[FABRIC_ENGINE_TYPES[name] ?? name] = {
-                emotions: [...controls.emotions],
-                paces: [...controls.paces],
-            };
-        }
-        return {
-            emotions: [...native.emotions],
-            paces: [...native.paces],
-            engines,
-        };
+        return (0, voiceControls_1.toVoiceControlsCatalog)(binding.getVoiceControls());
     }
     _requireAddon() {
         const addon = this._optionalAddon();