voice-router-dev 0.2.7 → 0.2.8

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/dist/index.d.mts CHANGED
@@ -6146,7 +6146,7 @@ interface Speaker {
6146
6146
  */
6147
6147
  interface Word {
6148
6148
  /** The transcribed word */
6149
- text: string;
6149
+ word: string;
6150
6150
  /** Start time in seconds */
6151
6151
  start: number;
6152
6152
  /** End time in seconds */
package/dist/index.d.ts CHANGED
@@ -6146,7 +6146,7 @@ interface Speaker {
6146
6146
  */
6147
6147
  interface Word {
6148
6148
  /** The transcribed word */
6149
- text: string;
6149
+ word: string;
6150
6150
  /** Start time in seconds */
6151
6151
  start: number;
6152
6152
  /** End time in seconds */
package/dist/index.js CHANGED
@@ -2048,7 +2048,7 @@ var GladiaAdapter = class extends BaseAdapter {
2048
2048
  }))
2049
2049
  );
2050
2050
  return extractWords(allWords, (item) => ({
2051
- text: item.word.word,
2051
+ word: item.word.word,
2052
2052
  start: item.word.start,
2053
2053
  end: item.word.end,
2054
2054
  confidence: item.word.confidence,
@@ -2068,11 +2068,11 @@ var GladiaAdapter = class extends BaseAdapter {
2068
2068
  end: utterance.end,
2069
2069
  speaker: utterance.speaker?.toString(),
2070
2070
  confidence: utterance.confidence,
2071
- words: utterance.words.map((word) => ({
2072
- text: word.word,
2073
- start: word.start,
2074
- end: word.end,
2075
- confidence: word.confidence
2071
+ words: utterance.words.map((w) => ({
2072
+ word: w.word,
2073
+ start: w.start,
2074
+ end: w.end,
2075
+ confidence: w.confidence
2076
2076
  }))
2077
2077
  }));
2078
2078
  }
@@ -2204,11 +2204,11 @@ var GladiaAdapter = class extends BaseAdapter {
2204
2204
  text: utterance.text,
2205
2205
  isFinal: messageData.is_final,
2206
2206
  confidence: utterance.confidence,
2207
- words: utterance.words.map((word) => ({
2208
- text: word.word,
2209
- start: word.start,
2210
- end: word.end,
2211
- confidence: word.confidence
2207
+ words: utterance.words.map((w) => ({
2208
+ word: w.word,
2209
+ start: w.start,
2210
+ end: w.end,
2211
+ confidence: w.confidence
2212
2212
  })),
2213
2213
  data: message
2214
2214
  });
@@ -2222,11 +2222,11 @@ var GladiaAdapter = class extends BaseAdapter {
2222
2222
  end: utterance.end,
2223
2223
  speaker: utterance.speaker?.toString(),
2224
2224
  confidence: utterance.confidence,
2225
- words: utterance.words.map((word) => ({
2226
- text: word.word,
2227
- start: word.start,
2228
- end: word.end,
2229
- confidence: word.confidence
2225
+ words: utterance.words.map((w) => ({
2226
+ word: w.word,
2227
+ start: w.start,
2228
+ end: w.end,
2229
+ confidence: w.confidence
2230
2230
  }))
2231
2231
  };
2232
2232
  callbacks?.onUtterance?.(utteranceData);
@@ -2931,14 +2931,14 @@ var AssemblyAIAdapter = class extends BaseAdapter {
2931
2931
  if (!transcript.words || transcript.words.length === 0) {
2932
2932
  return void 0;
2933
2933
  }
2934
- return transcript.words.map((word) => ({
2935
- text: word.text,
2936
- start: word.start / 1e3,
2934
+ return transcript.words.map((w) => ({
2935
+ word: w.text,
2936
+ start: w.start / 1e3,
2937
2937
  // Convert ms to seconds
2938
- end: word.end / 1e3,
2938
+ end: w.end / 1e3,
2939
2939
  // Convert ms to seconds
2940
- confidence: word.confidence,
2941
- speaker: word.speaker || void 0
2940
+ confidence: w.confidence,
2941
+ speaker: w.speaker || void 0
2942
2942
  }));
2943
2943
  }
2944
2944
  /**
@@ -2956,11 +2956,11 @@ var AssemblyAIAdapter = class extends BaseAdapter {
2956
2956
  // Convert ms to seconds
2957
2957
  speaker: utterance.speaker || void 0,
2958
2958
  confidence: utterance.confidence,
2959
- words: utterance.words.map((word) => ({
2960
- text: word.text,
2961
- start: word.start / 1e3,
2962
- end: word.end / 1e3,
2963
- confidence: word.confidence
2959
+ words: utterance.words.map((w) => ({
2960
+ word: w.text,
2961
+ start: w.start / 1e3,
2962
+ end: w.end / 1e3,
2963
+ confidence: w.confidence
2964
2964
  }))
2965
2965
  }));
2966
2966
  }
@@ -3053,12 +3053,12 @@ var AssemblyAIAdapter = class extends BaseAdapter {
3053
3053
  text: turnMsg.transcript,
3054
3054
  isFinal: turnMsg.end_of_turn,
3055
3055
  confidence: turnMsg.end_of_turn_confidence,
3056
- words: turnMsg.words.map((word) => ({
3057
- text: word.text,
3058
- start: word.start / 1e3,
3056
+ words: turnMsg.words.map((w) => ({
3057
+ word: w.text,
3058
+ start: w.start / 1e3,
3059
3059
  // Convert ms to seconds
3060
- end: word.end / 1e3,
3061
- confidence: word.confidence
3060
+ end: w.end / 1e3,
3061
+ confidence: w.confidence
3062
3062
  })),
3063
3063
  data: turnMsg
3064
3064
  });
@@ -3417,11 +3417,11 @@ var DeepgramAdapter = class extends BaseAdapter {
3417
3417
  return void 0;
3418
3418
  }
3419
3419
  return alternative.words.map(
3420
- (word) => ({
3421
- text: word.word || "",
3422
- start: word.start || 0,
3423
- end: word.end || 0,
3424
- confidence: word.confidence,
3420
+ (w) => ({
3421
+ word: w.word || "",
3422
+ start: w.start || 0,
3423
+ end: w.end || 0,
3424
+ confidence: w.confidence,
3425
3425
  speaker: void 0
3426
3426
  // Speaker info is at utterance level, not word level
3427
3427
  })
@@ -3441,11 +3441,11 @@ var DeepgramAdapter = class extends BaseAdapter {
3441
3441
  end: utterance.end || 0,
3442
3442
  speaker: utterance.speaker?.toString(),
3443
3443
  confidence: utterance.confidence,
3444
- words: utterance.words?.map((word) => ({
3445
- text: word.word || "",
3446
- start: word.start || 0,
3447
- end: word.end || 0,
3448
- confidence: word.confidence
3444
+ words: utterance.words?.map((w) => ({
3445
+ word: w.word || "",
3446
+ start: w.start || 0,
3447
+ end: w.end || 0,
3448
+ confidence: w.confidence
3449
3449
  }))
3450
3450
  }));
3451
3451
  }
@@ -3536,11 +3536,11 @@ var DeepgramAdapter = class extends BaseAdapter {
3536
3536
  if (channel) {
3537
3537
  const transcript = channel.transcript;
3538
3538
  const isFinal = message.is_final;
3539
- const words = channel.words?.map((word) => ({
3540
- text: word.word,
3541
- start: word.start,
3542
- end: word.end,
3543
- confidence: word.confidence
3539
+ const words = channel.words?.map((w) => ({
3540
+ word: w.word,
3541
+ start: w.start,
3542
+ end: w.end,
3543
+ confidence: w.confidence
3544
3544
  }));
3545
3545
  callbacks?.onTranscript?.({
3546
3546
  type: "transcript",
@@ -3881,12 +3881,12 @@ var AzureSTTAdapter = class extends BaseAdapter {
3881
3881
  const recognizedPhrases = transcriptionData.recognizedPhrases || [];
3882
3882
  const fullText = combinedPhrases.map((phrase) => phrase.display || phrase.lexical).join(" ") || "";
3883
3883
  const words = recognizedPhrases.flatMap(
3884
- (phrase) => (phrase.nBest?.[0]?.words || []).map((word) => ({
3885
- text: word.word,
3886
- start: word.offsetInTicks / 1e7,
3884
+ (phrase) => (phrase.nBest?.[0]?.words || []).map((w) => ({
3885
+ word: w.word,
3886
+ start: w.offsetInTicks / 1e7,
3887
3887
  // Convert ticks to seconds
3888
- end: (word.offsetInTicks + word.durationInTicks) / 1e7,
3889
- confidence: word.confidence,
3888
+ end: (w.offsetInTicks + w.durationInTicks) / 1e7,
3889
+ confidence: w.confidence,
3890
3890
  speaker: phrase.speaker !== void 0 ? phrase.speaker.toString() : void 0
3891
3891
  }))
3892
3892
  );
@@ -4167,10 +4167,10 @@ var OpenAIWhisperAdapter = class extends BaseAdapter {
4167
4167
  }
4168
4168
  if ("duration" in response && "language" in response) {
4169
4169
  const verboseResponse = response;
4170
- const words = verboseResponse.words?.map((word) => ({
4171
- text: word.word,
4172
- start: word.start,
4173
- end: word.end,
4170
+ const words = verboseResponse.words?.map((w) => ({
4171
+ word: w.word,
4172
+ start: w.start,
4173
+ end: w.end,
4174
4174
  confidence: void 0
4175
4175
  }));
4176
4176
  const requestId2 = `openai-${Date.now()}`;
@@ -4436,7 +4436,7 @@ var SpeechmaticsAdapter = class extends BaseAdapter {
4436
4436
  normalizeResponse(response) {
4437
4437
  const text = response.results.filter((r) => r.type === "word" && r.alternatives).map((r) => r.alternatives[0]?.content || "").join(" ");
4438
4438
  const words = response.results.filter((r) => r.type === "word" && r.start_time !== void 0 && r.end_time !== void 0).map((result) => ({
4439
- text: result.alternatives?.[0]?.content || "",
4439
+ word: result.alternatives?.[0]?.content || "",
4440
4440
  start: result.start_time,
4441
4441
  end: result.end_time,
4442
4442
  confidence: result.alternatives?.[0]?.confidence,
@@ -4587,12 +4587,12 @@ var GladiaWebhookHandler = class extends BaseWebhookHandler {
4587
4587
  /**
4588
4588
  * Convert Gladia WordDTO to unified Word type
4589
4589
  */
4590
- mapWord(word) {
4590
+ mapWord(w) {
4591
4591
  return {
4592
- text: word.word,
4593
- start: word.start,
4594
- end: word.end,
4595
- confidence: word.confidence
4592
+ word: w.word,
4593
+ start: w.start,
4594
+ end: w.end,
4595
+ confidence: w.confidence
4596
4596
  };
4597
4597
  }
4598
4598
  /**
@@ -4930,11 +4930,11 @@ var DeepgramWebhookHandler = class extends BaseWebhookHandler {
4930
4930
  raw: payload
4931
4931
  };
4932
4932
  }
4933
- const words = alternative.words && alternative.words.length > 0 ? alternative.words.map((word) => ({
4934
- text: word.word || "",
4935
- start: word.start || 0,
4936
- end: word.end || 0,
4937
- confidence: word.confidence
4933
+ const words = alternative.words && alternative.words.length > 0 ? alternative.words.map((w) => ({
4934
+ word: w.word || "",
4935
+ start: w.start || 0,
4936
+ end: w.end || 0,
4937
+ confidence: w.confidence
4938
4938
  })) : void 0;
4939
4939
  const speakers = response.results.utterances && response.results.utterances.length > 0 ? response.results.utterances.map((utterance) => ({
4940
4940
  id: utterance.speaker?.toString() || "unknown",
@@ -4948,11 +4948,11 @@ var DeepgramWebhookHandler = class extends BaseWebhookHandler {
4948
4948
  end: utterance.end || 0,
4949
4949
  speaker: utterance.speaker?.toString(),
4950
4950
  confidence: utterance.confidence,
4951
- words: utterance.words && utterance.words.length > 0 ? utterance.words.map((word) => ({
4952
- text: word.word || "",
4953
- start: word.start || 0,
4954
- end: word.end || 0,
4955
- confidence: word.confidence
4951
+ words: utterance.words && utterance.words.length > 0 ? utterance.words.map((w) => ({
4952
+ word: w.word || "",
4953
+ start: w.start || 0,
4954
+ end: w.end || 0,
4955
+ confidence: w.confidence
4956
4956
  })) : void 0
4957
4957
  })) : void 0;
4958
4958
  const summary = alternative.summaries?.[0]?.summary;