@absolutejs/voice 0.0.22-beta.625 → 0.0.22-beta.627

package/dist/core/numberNormalizer.d.ts

@@ -0,0 +1 @@
+export declare const normalizeSpokenNumbers: (input: string) => string;

package/dist/core/types.d.ts

@@ -46,6 +46,11 @@ export type VoiceLanguageStrategy = {
     primaryLanguage?: string;
     secondaryLanguages: string[];
 };
+export type VoiceLanguageStrategyResolver<TContext = unknown> = (input: {
+    context: TContext;
+    scenarioId?: string;
+    sessionId: string;
+}) => Promise<VoiceLanguageStrategy | void> | VoiceLanguageStrategy | void;
 export type VoicePhraseHint = {
     text: string;
     aliases?: string[];
@@ -825,7 +830,7 @@ export type VoicePluginConfig<TContext = unknown, TSession extends VoiceSessionR
         }) => string | Promise<string>);
         maxReprompts?: number;
     };
-    languageStrategy?: VoiceLanguageStrategy;
+    languageStrategy?: VoiceLanguageStrategy | VoiceLanguageStrategyResolver<TContext>;
     lexicon?: VoiceLexiconEntry[] | VoiceLexiconResolver<TContext>;
     phraseHints?: VoicePhraseHint[] | VoicePhraseHintResolver<TContext>;
     preset?: VoiceRuntimePreset;
@@ -861,6 +866,7 @@ export type VoicePluginConfig<TContext = unknown, TSession extends VoiceSessionR
     defaultSilentTurnAck?: string;
     routeOnTurnTimeoutMs?: number;
     audioConditioning?: VoiceAudioConditioningConfig;
+    normalizeNumbers?: boolean;
     noiseSuppressor?: VoiceNoiseSuppressor;
     noiseSuppressorFormat?: AudioFormat;
     logger?: VoiceLogger;
@@ -1131,6 +1137,7 @@ export type CreateVoiceSessionOptions<TContext = unknown, TSession extends Voice
     sttLifecycle: VoiceSTTLifecycle;
     turnDetection: VoiceResolvedTurnDetectionConfig;
     audioConditioning?: VoiceResolvedAudioConditioningConfig;
+    normalizeNumbers?: boolean;
     noiseSuppressor?: VoiceNoiseSuppressor;
     noiseSuppressorFormat?: AudioFormat;
     handoff?: VoiceHandoffConfig<TContext, TSession, TResult>;

package/dist/index.d.ts

@@ -77,6 +77,7 @@ export { createVoiceMCPToolset } from "./core/mcpToolset";
 export type { CreateVoiceMCPToolsetOptions, MCPClientLike, MCPToolCallResult, MCPToolContentBlock, MCPToolDefinition, VoiceMCPToolResult, } from "./core/mcpToolset";
 export { createAIVoiceModel } from "./core/aiVoiceModel";
 export type { CreateAIVoiceModelOptions } from "./core/aiVoiceModel";
+export { normalizeSpokenNumbers } from "./core/numberNormalizer";
 export { createVoiceAIJudgeCompletion, createVoiceLLMJudge, } from "./core/llmJudge";
 export type { CreateVoiceAIJudgeCompletionOptions, CreateVoiceLLMJudgeOptions, VoiceLLMJudge, VoiceLLMJudgeCompletion, VoiceLLMJudgeCriterionVerdict, VoiceLLMJudgeInput, VoiceLLMJudgeRubric, VoiceLLMJudgeRubricCriterion, VoiceLLMJudgeVerdict, } from "./core/llmJudge";
 export { DEFAULT_VOICE_REDACTION_PATTERNS, createVoiceTranscriptRedactor, redactVoiceTranscript, } from "./core/redaction";

package/dist/index.js

@@ -3096,6 +3096,191 @@ var toVoiceSessionSummary = (session) => ({
 // src/core/session.ts
 import { Buffer as Buffer2 } from "buffer";
 
+// src/core/numberNormalizer.ts
+var ONES = {
+  eight: 8,
+  eighteen: 18,
+  eleven: 11,
+  fifteen: 15,
+  five: 5,
+  four: 4,
+  fourteen: 14,
+  nine: 9,
+  nineteen: 19,
+  one: 1,
+  seven: 7,
+  seventeen: 17,
+  six: 6,
+  sixteen: 16,
+  ten: 10,
+  thirteen: 13,
+  three: 3,
+  twelve: 12,
+  two: 2,
+  zero: 0
+};
+var TENS = {
+  eighty: 80,
+  fifty: 50,
+  forty: 40,
+  ninety: 90,
+  seventy: 70,
+  sixty: 60,
+  thirty: 30,
+  twenty: 20
+};
+var SCALES = {
+  billion: 1e9,
+  million: 1e6,
+  thousand: 1000,
+  trillion: 1000000000000
+};
+var MAGNITUDE_WORDS = [
+  [1000000000000, "trillion"],
+  [1e9, "billion"],
+  [1e6, "million"]
+];
+var FILLER = new Set(["and", "a", "an"]);
+var DECIMAL_PLACES = 3;
+var isValueWord = (word) => (word in ONES) || (word in TENS) || word === "hundred";
+var isScaleWord = (word) => (word in SCALES);
+var isNumberWord = (word) => isValueWord(word) || isScaleWord(word) || word === "point" || FILLER.has(word);
+var trimNumber = (value) => {
+  if (Number.isInteger(value))
+    return value.toLocaleString("en-US");
+  return String(Number(value.toFixed(DECIMAL_PLACES)));
+};
+var renderValue = (value, usedMagnitude) => {
+  if (usedMagnitude) {
+    for (const [scale, word] of MAGNITUDE_WORDS) {
+      if (value >= scale) {
+        const scaled = value / scale;
+        if (Number(scaled.toFixed(DECIMAL_PLACES)) === scaled) {
+          return `${trimNumber(scaled)} ${word}`;
+        }
+      }
+    }
+  }
+  return trimNumber(value);
+};
+var parseNumberWords = (words) => {
+  let total = 0;
+  let current = 0;
+  let usedMagnitude = false;
+  let sawNumber = false;
+  let decimal = null;
+  const foldDecimal = () => {
+    if (decimal && decimal.length > 0)
+      current += Number(`0.${decimal}`);
+    decimal = null;
+  };
+  for (const word of words) {
+    if (word === "point") {
+      decimal = "";
+      continue;
+    }
+    const one = ONES[word];
+    const ten = TENS[word];
+    const scale = SCALES[word];
+    if (decimal !== null) {
+      if (one !== undefined && one <= 9) {
+        decimal += String(one);
+        sawNumber = true;
+        continue;
+      }
+      foldDecimal();
+    }
+    if (FILLER.has(word))
+      continue;
+    if (one !== undefined) {
+      current += one;
+      sawNumber = true;
+    } else if (ten !== undefined) {
+      current += ten;
+      sawNumber = true;
+    } else if (word === "hundred") {
+      current = (current === 0 ? 1 : current) * 100;
+      sawNumber = true;
+    } else if (scale !== undefined) {
+      total += (current === 0 ? 1 : current) * scale;
+      current = 0;
+      sawNumber = true;
+      usedMagnitude = true;
+    }
+  }
+  foldDecimal();
+  if (!sawNumber)
+    return null;
+  return { usedMagnitude, value: total + current };
+};
+var PERCENT_RE = /^(per ?cent|percent|percentage)$/;
+var CURRENCY_RE = /^(dollars?|bucks?|usd)$/;
+var WORD_RE = /^[A-Za-z]+(?:-[A-Za-z]+)*$/;
+var wordsOf = (token) => token.toLowerCase().split("-");
+var isNumberToken = (token) => WORD_RE.test(token) && wordsOf(token).every(isNumberWord);
+var startsNumber = (token) => WORD_RE.test(token) && wordsOf(token).some(isValueWord);
+var isSpace = (token) => /^\s+$/.test(token);
+var normalizeSpokenNumbers = (input) => {
+  if (!input)
+    return input;
+  const parts = input.match(/[A-Za-z]+(?:-[A-Za-z]+)*|[^A-Za-z]+/g);
+  if (!parts)
+    return input;
+  const at = (idx) => parts[idx] ?? "";
+  const out = [];
+  let i = 0;
+  while (i < parts.length) {
+    const token = at(i);
+    const lower = token.toLowerCase();
+    if (lower === "a" || lower === "an") {
+      const nextWord = at(i + 2);
+      const nextHead = wordsOf(nextWord)[0] ?? "";
+      const nextIsScale = isSpace(at(i + 1)) && WORD_RE.test(nextWord) && (nextHead === "hundred" || isScaleWord(nextHead));
+      if (!nextIsScale) {
+        out.push(token);
+        i += 1;
+        continue;
+      }
+    } else if (!startsNumber(token)) {
+      out.push(token);
+      i += 1;
+      continue;
+    }
+    const spanIdx = [i];
+    let j = i + 1;
+    while (isSpace(at(j)) && isNumberToken(at(j + 1))) {
+      spanIdx.push(j + 1);
+      j += 2;
+    }
+    const words = spanIdx.flatMap((k) => wordsOf(at(k)));
+    while (words.length > 0 && FILLER.has(words[words.length - 1] ?? "")) {
+      words.pop();
+    }
+    const parsed = parseNumberWords(words);
+    if (!parsed) {
+      out.push(token);
+      i += 1;
+      continue;
+    }
+    let rendered = renderValue(parsed.value, parsed.usedMagnitude);
+    let lastIdx = spanIdx[spanIdx.length - 1] ?? i;
+    const unitWord = at(lastIdx + 2);
+    if (isSpace(at(lastIdx + 1)) && WORD_RE.test(unitWord)) {
+      const unit = unitWord.toLowerCase();
+      if (PERCENT_RE.test(unit)) {
+        rendered = `${rendered}%`;
+        lastIdx += 2;
+      } else if (CURRENCY_RE.test(unit)) {
+        rendered = rendered.startsWith("$") ? rendered : `$${rendered}`;
+        lastIdx += 2;
+      }
+    }
+    out.push(rendered);
+    i = lastIdx + 1;
+  }
+  return out.join("");
+};
+
 // src/core/debugTiming.ts
 var timingEnabled = () => process.env.ABSOLUTEJS_VOICE_TIMING === "1" || process.env.ABSOLUTEJS_VOICE_TIMING === "true";
 var emitTiming = (sessionId, stage, elapsedMs, detail) => {
@@ -5834,6 +6019,9 @@ var createVoiceSession = (options) => {
   };
   const completeTurn = async (session, turn) => {
     console.error(`[voice] completeTurn ENTER session=${options.id} turn=${turn.id} textLen=${turn.text?.length ?? 0}`);
+    if (options.normalizeNumbers && turn.text) {
+      turn.text = normalizeSpokenNumbers(turn.text);
+    }
     kickStuckCloseWatchdog();
     kickIdleRepromptWatchdog(true);
     const liveOpsControl = await options.liveOps?.getControl(options.id);
@@ -25430,6 +25618,16 @@ var createTwilioMediaStreamBridge = (socket, options) => {
     }
     return normalizePhraseHints(options.phraseHints);
   };
+  const resolveLanguageStrategy = async () => {
+    if (typeof options.languageStrategy === "function") {
+      return await options.languageStrategy({
+        context: options.context,
+        scenarioId: bridgeState.scenarioId ?? undefined,
+        sessionId: bridgeState.sessionId ?? ""
+      }) ?? undefined;
+    }
+    return options.languageStrategy;
+  };
   const ensureSession = async () => {
     if (sessionHandle) {
       return sessionHandle;
@@ -25437,6 +25635,7 @@ var createTwilioMediaStreamBridge = (socket, options) => {
     bridgeState.sessionId ??= `phone-${Date.now().toString(36)}`;
     const lexicon = await resolveLexicon();
     const phraseHints = await resolvePhraseHints();
+    const languageStrategy = await resolveLanguageStrategy();
     const normalizedOnTurn = normalizeOnTurn(options.onTurn);
     const route = {
       correctTurn: options.correctTurn,
@@ -25468,13 +25667,14 @@ var createTwilioMediaStreamBridge = (socket, options) => {
     const voiceSocket = createTwilioSocketAdapter(socket, () => bridgeState);
     sessionHandle = createVoiceSession({
       audioConditioning,
+      normalizeNumbers: options.normalizeNumbers,
       context: options.context,
       costAccountant: options.costAccountant,
       costTelemetry: options.costTelemetry,
       costTelephony: options.costTelephony,
       greeting: options.greeting,
       id: bridgeState.sessionId,
-      languageStrategy: options.languageStrategy,
+      languageStrategy,
       lexicon,
       logger,
       phraseHints,
@@ -39586,6 +39786,7 @@ var resolveSessionOptions = (config) => {
   const preset = resolveVoiceRuntimePreset(config.preset);
   return {
     audioConditioning: config.audioConditioning !== undefined ? resolveAudioConditioningConfig(config.audioConditioning) : preset.audioConditioning,
+    normalizeNumbers: config.normalizeNumbers,
     noiseSuppressor: config.noiseSuppressor,
     noiseSuppressorFormat: config.noiseSuppressorFormat,
     costTelemetry: config.costTelemetry,
@@ -39633,6 +39834,12 @@ var resolveLexicon = async (config, input) => {
   }
   return normalizeLexicon2(config.lexicon);
 };
+var resolveLanguageStrategy = async (config, input) => {
+  if (typeof config.languageStrategy === "function") {
+    return await config.languageStrategy(input) ?? undefined;
+  }
+  return config.languageStrategy;
+};
 var resolveProfileSwitchGuard = async (config, runtime, input) => {
   const guard = config.profileSwitchGuard;
   if (!guard || runtime.profileSwitchGuardedSessions.has(input.sessionId)) {
@@ -39797,8 +40004,14 @@ var voice = (config) => {
       scenarioId,
       sessionId
     });
+    const languageStrategy = await resolveLanguageStrategy(config, {
+      context,
+      scenarioId,
+      sessionId
+    });
     return createVoiceSession({
       audioConditioning: sessionOptions.audioConditioning,
+      normalizeNumbers: sessionOptions.normalizeNumbers,
       noiseSuppressor: sessionOptions.noiseSuppressor,
       noiseSuppressorFormat: sessionOptions.noiseSuppressorFormat,
       context,
@@ -39809,7 +40022,7 @@ var voice = (config) => {
       stuckCallClose: config.stuckCallClose,
       idleReprompt: config.idleReprompt,
       handoff: config.handoff,
-      languageStrategy: config.languageStrategy,
+      languageStrategy,
       lexicon,
       liveOps: config.liveOps,
       logger: sessionOptions.logger,
@@ -50973,7 +51186,7 @@ var DEFAULT_VOICE_CAMPAIGN_TEMPLATE_FILTERS = {
   ssml: (value) => escapeSpeech(String(value ?? "")),
   upper: (value) => String(value ?? "").toUpperCase()
 };
-var renderValue = (value, filters, filterChain) => {
+var renderValue2 = (value, filters, filterChain) => {
   let cursor = value;
   for (const segment of filterChain) {
     const parts2 = segment.split(":");
@@ -51018,7 +51231,7 @@ var resolveVoiceCampaignTemplate = (template, options) => {
       }
       return options.fallback ?? "";
     }
-    return renderValue(value, filters, segments.slice(1));
+    return renderValue2(value, filters, segments.slice(1));
   });
   return { missingVariables: Array.from(missing), output };
 };
@@ -53651,6 +53864,7 @@ export {
   parseVoiceSessionSnapshot,
   parseVoiceAIScorecardResponse,
   normalizeVoiceProofTrendReport,
+  normalizeSpokenNumbers,
   normalizePhoneNumber,
   muteVoiceMonitorIssue,
   matchesVoiceOpsTaskAssignmentRule,

package/dist/testing/index.js

@@ -5536,6 +5536,191 @@ var createVoiceMemoryStore = () => {
 // src/core/session.ts
 import { Buffer as Buffer2 } from "buffer";
 
+// src/core/numberNormalizer.ts
+var ONES = {
+  eight: 8,
+  eighteen: 18,
+  eleven: 11,
+  fifteen: 15,
+  five: 5,
+  four: 4,
+  fourteen: 14,
+  nine: 9,
+  nineteen: 19,
+  one: 1,
+  seven: 7,
+  seventeen: 17,
+  six: 6,
+  sixteen: 16,
+  ten: 10,
+  thirteen: 13,
+  three: 3,
+  twelve: 12,
+  two: 2,
+  zero: 0
+};
+var TENS = {
+  eighty: 80,
+  fifty: 50,
+  forty: 40,
+  ninety: 90,
+  seventy: 70,
+  sixty: 60,
+  thirty: 30,
+  twenty: 20
+};
+var SCALES = {
+  billion: 1e9,
+  million: 1e6,
+  thousand: 1000,
+  trillion: 1000000000000
+};
+var MAGNITUDE_WORDS = [
+  [1000000000000, "trillion"],
+  [1e9, "billion"],
+  [1e6, "million"]
+];
+var FILLER = new Set(["and", "a", "an"]);
+var DECIMAL_PLACES = 3;
+var isValueWord = (word) => (word in ONES) || (word in TENS) || word === "hundred";
+var isScaleWord = (word) => (word in SCALES);
+var isNumberWord = (word) => isValueWord(word) || isScaleWord(word) || word === "point" || FILLER.has(word);
+var trimNumber = (value) => {
+  if (Number.isInteger(value))
+    return value.toLocaleString("en-US");
+  return String(Number(value.toFixed(DECIMAL_PLACES)));
+};
+var renderValue = (value, usedMagnitude) => {
+  if (usedMagnitude) {
+    for (const [scale, word] of MAGNITUDE_WORDS) {
+      if (value >= scale) {
+        const scaled = value / scale;
+        if (Number(scaled.toFixed(DECIMAL_PLACES)) === scaled) {
+          return `${trimNumber(scaled)} ${word}`;
+        }
+      }
+    }
+  }
+  return trimNumber(value);
+};
+var parseNumberWords = (words) => {
+  let total = 0;
+  let current = 0;
+  let usedMagnitude = false;
+  let sawNumber = false;
+  let decimal = null;
+  const foldDecimal = () => {
+    if (decimal && decimal.length > 0)
+      current += Number(`0.${decimal}`);
+    decimal = null;
+  };
+  for (const word of words) {
+    if (word === "point") {
+      decimal = "";
+      continue;
+    }
+    const one = ONES[word];
+    const ten = TENS[word];
+    const scale = SCALES[word];
+    if (decimal !== null) {
+      if (one !== undefined && one <= 9) {
+        decimal += String(one);
+        sawNumber = true;
+        continue;
+      }
+      foldDecimal();
+    }
+    if (FILLER.has(word))
+      continue;
+    if (one !== undefined) {
+      current += one;
+      sawNumber = true;
+    } else if (ten !== undefined) {
+      current += ten;
+      sawNumber = true;
+    } else if (word === "hundred") {
+      current = (current === 0 ? 1 : current) * 100;
+      sawNumber = true;
+    } else if (scale !== undefined) {
+      total += (current === 0 ? 1 : current) * scale;
+      current = 0;
+      sawNumber = true;
+      usedMagnitude = true;
+    }
+  }
+  foldDecimal();
+  if (!sawNumber)
+    return null;
+  return { usedMagnitude, value: total + current };
+};
+var PERCENT_RE = /^(per ?cent|percent|percentage)$/;
+var CURRENCY_RE = /^(dollars?|bucks?|usd)$/;
+var WORD_RE = /^[A-Za-z]+(?:-[A-Za-z]+)*$/;
+var wordsOf = (token) => token.toLowerCase().split("-");
+var isNumberToken = (token) => WORD_RE.test(token) && wordsOf(token).every(isNumberWord);
+var startsNumber = (token) => WORD_RE.test(token) && wordsOf(token).some(isValueWord);
+var isSpace = (token) => /^\s+$/.test(token);
+var normalizeSpokenNumbers = (input) => {
+  if (!input)
+    return input;
+  const parts = input.match(/[A-Za-z]+(?:-[A-Za-z]+)*|[^A-Za-z]+/g);
+  if (!parts)
+    return input;
+  const at = (idx) => parts[idx] ?? "";
+  const out = [];
+  let i = 0;
+  while (i < parts.length) {
+    const token = at(i);
+    const lower = token.toLowerCase();
+    if (lower === "a" || lower === "an") {
+      const nextWord = at(i + 2);
+      const nextHead = wordsOf(nextWord)[0] ?? "";
+      const nextIsScale = isSpace(at(i + 1)) && WORD_RE.test(nextWord) && (nextHead === "hundred" || isScaleWord(nextHead));
+      if (!nextIsScale) {
+        out.push(token);
+        i += 1;
+        continue;
+      }
+    } else if (!startsNumber(token)) {
+      out.push(token);
+      i += 1;
+      continue;
+    }
+    const spanIdx = [i];
+    let j = i + 1;
+    while (isSpace(at(j)) && isNumberToken(at(j + 1))) {
+      spanIdx.push(j + 1);
+      j += 2;
+    }
+    const words = spanIdx.flatMap((k) => wordsOf(at(k)));
+    while (words.length > 0 && FILLER.has(words[words.length - 1] ?? "")) {
+      words.pop();
+    }
+    const parsed = parseNumberWords(words);
+    if (!parsed) {
+      out.push(token);
+      i += 1;
+      continue;
+    }
+    let rendered = renderValue(parsed.value, parsed.usedMagnitude);
+    let lastIdx = spanIdx[spanIdx.length - 1] ?? i;
+    const unitWord = at(lastIdx + 2);
+    if (isSpace(at(lastIdx + 1)) && WORD_RE.test(unitWord)) {
+      const unit = unitWord.toLowerCase();
+      if (PERCENT_RE.test(unit)) {
+        rendered = `${rendered}%`;
+        lastIdx += 2;
+      } else if (CURRENCY_RE.test(unit)) {
+        rendered = rendered.startsWith("$") ? rendered : `$${rendered}`;
+        lastIdx += 2;
+      }
+    }
+    out.push(rendered);
+    i = lastIdx + 1;
+  }
+  return out.join("");
+};
+
 // src/core/backchannel.ts
 var DEFAULT_CUES = [
   { text: "mm-hmm" },
@@ -8154,6 +8339,9 @@ var createVoiceSession = (options) => {
   };
   const completeTurn = async (session, turn) => {
     console.error(`[voice] completeTurn ENTER session=${options.id} turn=${turn.id} textLen=${turn.text?.length ?? 0}`);
+    if (options.normalizeNumbers && turn.text) {
+      turn.text = normalizeSpokenNumbers(turn.text);
+    }
     kickStuckCloseWatchdog();
     kickIdleRepromptWatchdog(true);
     const liveOpsControl = await options.liveOps?.getControl(options.id);
@@ -14396,6 +14584,16 @@ var createTwilioMediaStreamBridge = (socket, options) => {
     }
     return normalizePhraseHints(options.phraseHints);
   };
+  const resolveLanguageStrategy = async () => {
+    if (typeof options.languageStrategy === "function") {
+      return await options.languageStrategy({
+        context: options.context,
+        scenarioId: bridgeState.scenarioId ?? undefined,
+        sessionId: bridgeState.sessionId ?? ""
+      }) ?? undefined;
+    }
+    return options.languageStrategy;
+  };
   const ensureSession = async () => {
     if (sessionHandle) {
       return sessionHandle;
@@ -14403,6 +14601,7 @@ var createTwilioMediaStreamBridge = (socket, options) => {
     bridgeState.sessionId ??= `phone-${Date.now().toString(36)}`;
     const lexicon = await resolveLexicon();
     const phraseHints = await resolvePhraseHints();
+    const languageStrategy = await resolveLanguageStrategy();
     const normalizedOnTurn = normalizeOnTurn(options.onTurn);
     const route = {
       correctTurn: options.correctTurn,
@@ -14434,13 +14633,14 @@ var createTwilioMediaStreamBridge = (socket, options) => {
     const voiceSocket = createTwilioSocketAdapter(socket, () => bridgeState);
     sessionHandle = createVoiceSession({
       audioConditioning,
+      normalizeNumbers: options.normalizeNumbers,
       context: options.context,
       costAccountant: options.costAccountant,
       costTelemetry: options.costTelemetry,
       costTelephony: options.costTelephony,
       greeting: options.greeting,
       id: bridgeState.sessionId,
-      languageStrategy: options.languageStrategy,
+      languageStrategy,
       lexicon,
       logger,
       phraseHints,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@absolutejs/voice",
-  "version": "0.0.22-beta.625",
+  "version": "0.0.22-beta.627",
   "description": "Voice primitives and Elysia plugin for AbsoluteJS",
   "repository": {
     "type": "git",