@livekit/agents-plugin-cartesia 1.4.8 → 1.4.11
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.cjs +1 -1
- package/dist/index.js +1 -1
- package/dist/models.cjs.map +1 -1
- package/dist/models.d.cts +3 -1
- package/dist/models.d.ts +3 -1
- package/dist/models.d.ts.map +1 -1
- package/dist/models.js.map +1 -1
- package/dist/stt.cjs +19 -6
- package/dist/stt.cjs.map +1 -1
- package/dist/stt.d.cts +2 -0
- package/dist/stt.d.ts +2 -0
- package/dist/stt.d.ts.map +1 -1
- package/dist/stt.js +21 -6
- package/dist/stt.js.map +1 -1
- package/package.json +6 -6
- package/src/models.ts +3 -1
- package/src/stt.ts +39 -4
package/dist/index.cjs
CHANGED
package/dist/index.js
CHANGED
package/dist/models.cjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/models.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\n/**\n * LiveKit uses this encoding for all audio\n */\nexport const AUDIO_ENCODING = 'pcm_s16le';\n\n// ============================================================================\n// TTS\n// ============================================================================\n\n/**\n * See [the docs](https://docs.cartesia.ai/build-with-cartesia/tts-models/latest) for all options.\n */\nexport type TTSModels =\n | 'sonic'\n | 'sonic-2'\n | 'sonic-3'\n | 'sonic-lite'\n | 'sonic-preview'\n | 'sonic-turbo';\n\n/**\n * See [the docs](https://docs.cartesia.ai/build-with-cartesia/tts-models/latest) for all options.\n */\nexport type TTSLanguages = 'en' | 'es' | 'fr' | 'de' | 'pt' | 'zh' | 'ja';\n\nexport const TTSDefaultVoiceId = 'f786b574-daa5-4673-aa0c-cbe3e8534c02';\n\nexport const isSonic3 = (model: string): boolean => model.startsWith('sonic-3');\n\nexport type TTSVoiceSpeed = 'fastest' | 'fast' | 'normal' | 'slow' | 'slowest';\n\nexport type TTSVoiceEmotion =\n | 'anger:lowest'\n | 'anger:low'\n | 'anger'\n | 'anger:high'\n | 'anger:highest'\n | 'positivity:lowest'\n | 'positivity:low'\n | 'positivity'\n | 'positivity:high'\n | 'positivity:highest'\n | 'surprise:lowest'\n | 'surprise:low'\n | 'surprise'\n | 'surprise:high'\n | 'surprise:highest'\n | 'sadness:lowest'\n | 'sadness:low'\n | 'sadness'\n | 'sadness:high'\n | 'sadness:highest'\n | 'curiosity:lowest'\n | 'curiosity:low'\n | 'curiosity'\n | 'curiosity:high'\n | 'curiosity:highest';\n\n/**\n * @deprecated Encoding should not be parameterized. Only `pcm_s16le`is allowed. Prefer using {@link AUDIO_ENCODING}.\n */\nexport type TTSEncoding = 'pcm_s16le';\n\n// ============================================================================\n// STT\n// ============================================================================\n\n/**\n * See [the docs](https://docs.cartesia.ai/build-with-cartesia/stt-models/latest) for all options.\n */\nexport type STTModel = 'ink-2';\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAOO,MAAM,iBAAiB;AAsBvB,MAAM,oBAAoB;AAE1B,MAAM,WAAW,CAAC,UAA2B,MAAM,WAAW,SAAS;","names":[]}
|
|
1
|
+
{"version":3,"sources":["../src/models.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\n/**\n * LiveKit uses this encoding for all audio\n */\nexport const AUDIO_ENCODING = 'pcm_s16le';\n\n// ============================================================================\n// TTS\n// ============================================================================\n\n/**\n * See [the docs](https://docs.cartesia.ai/build-with-cartesia/tts-models/latest) for all options.\n */\nexport type TTSModels =\n | 'sonic'\n | 'sonic-2'\n | 'sonic-3'\n | 'sonic-lite'\n | 'sonic-preview'\n | 'sonic-turbo';\n\n/**\n * See [the docs](https://docs.cartesia.ai/build-with-cartesia/tts-models/latest) for all options.\n */\nexport type TTSLanguages = 'en' | 'es' | 'fr' | 'de' | 'pt' | 'zh' | 'ja';\n\nexport const TTSDefaultVoiceId = 'f786b574-daa5-4673-aa0c-cbe3e8534c02';\n\nexport const isSonic3 = (model: string): boolean => model.startsWith('sonic-3');\n\nexport type TTSVoiceSpeed = 'fastest' | 'fast' | 'normal' | 'slow' | 'slowest';\n\nexport type TTSVoiceEmotion =\n | 'anger:lowest'\n | 'anger:low'\n | 'anger'\n | 'anger:high'\n | 'anger:highest'\n | 'positivity:lowest'\n | 'positivity:low'\n | 'positivity'\n | 'positivity:high'\n | 'positivity:highest'\n | 'surprise:lowest'\n | 'surprise:low'\n | 'surprise'\n | 'surprise:high'\n | 'surprise:highest'\n | 'sadness:lowest'\n | 'sadness:low'\n | 'sadness'\n | 'sadness:high'\n | 'sadness:highest'\n | 'curiosity:lowest'\n | 'curiosity:low'\n | 'curiosity'\n | 'curiosity:high'\n | 'curiosity:highest';\n\n/**\n * @deprecated Encoding should not be parameterized. Only `pcm_s16le`is allowed. Prefer using {@link AUDIO_ENCODING}.\n */\nexport type TTSEncoding = 'pcm_s16le';\n\n// ============================================================================\n// STT\n// ============================================================================\n\n/**\n * See [the docs](https://docs.cartesia.ai/build-with-cartesia/stt-models/latest) for all options.\n *\n * `ink-2` currently supports English only; `ink-whisper` is multilingual.\n */\nexport type STTModel = 'ink-2' | 'ink-whisper';\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAOO,MAAM,iBAAiB;AAsBvB,MAAM,oBAAoB;AAE1B,MAAM,WAAW,CAAC,UAA2B,MAAM,WAAW,SAAS;","names":[]}
|
package/dist/models.d.cts
CHANGED
|
@@ -20,6 +20,8 @@ export type TTSVoiceEmotion = 'anger:lowest' | 'anger:low' | 'anger' | 'anger:hi
|
|
|
20
20
|
export type TTSEncoding = 'pcm_s16le';
|
|
21
21
|
/**
|
|
22
22
|
* See [the docs](https://docs.cartesia.ai/build-with-cartesia/stt-models/latest) for all options.
|
|
23
|
+
*
|
|
24
|
+
* `ink-2` currently supports English only; `ink-whisper` is multilingual.
|
|
23
25
|
*/
|
|
24
|
-
export type STTModel = 'ink-2';
|
|
26
|
+
export type STTModel = 'ink-2' | 'ink-whisper';
|
|
25
27
|
//# sourceMappingURL=models.d.ts.map
|
package/dist/models.d.ts
CHANGED
|
@@ -20,6 +20,8 @@ export type TTSVoiceEmotion = 'anger:lowest' | 'anger:low' | 'anger' | 'anger:hi
|
|
|
20
20
|
export type TTSEncoding = 'pcm_s16le';
|
|
21
21
|
/**
|
|
22
22
|
* See [the docs](https://docs.cartesia.ai/build-with-cartesia/stt-models/latest) for all options.
|
|
23
|
+
*
|
|
24
|
+
* `ink-2` currently supports English only; `ink-whisper` is multilingual.
|
|
23
25
|
*/
|
|
24
|
-
export type STTModel = 'ink-2';
|
|
26
|
+
export type STTModel = 'ink-2' | 'ink-whisper';
|
|
25
27
|
//# sourceMappingURL=models.d.ts.map
|
package/dist/models.d.ts.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"models.d.ts","sourceRoot":"","sources":["../src/models.ts"],"names":[],"mappings":"AAIA;;GAEG;AACH,eAAO,MAAM,cAAc,cAAc,CAAC;AAM1C;;GAEG;AACH,MAAM,MAAM,SAAS,GACjB,OAAO,GACP,SAAS,GACT,SAAS,GACT,YAAY,GACZ,eAAe,GACf,aAAa,CAAC;AAElB;;GAEG;AACH,MAAM,MAAM,YAAY,GAAG,IAAI,GAAG,IAAI,GAAG,IAAI,GAAG,IAAI,GAAG,IAAI,GAAG,IAAI,GAAG,IAAI,CAAC;AAE1E,eAAO,MAAM,iBAAiB,yCAAyC,CAAC;AAExE,eAAO,MAAM,QAAQ,GAAI,OAAO,MAAM,KAAG,OAAsC,CAAC;AAEhF,MAAM,MAAM,aAAa,GAAG,SAAS,GAAG,MAAM,GAAG,QAAQ,GAAG,MAAM,GAAG,SAAS,CAAC;AAE/E,MAAM,MAAM,eAAe,GACvB,cAAc,GACd,WAAW,GACX,OAAO,GACP,YAAY,GACZ,eAAe,GACf,mBAAmB,GACnB,gBAAgB,GAChB,YAAY,GACZ,iBAAiB,GACjB,oBAAoB,GACpB,iBAAiB,GACjB,cAAc,GACd,UAAU,GACV,eAAe,GACf,kBAAkB,GAClB,gBAAgB,GAChB,aAAa,GACb,SAAS,GACT,cAAc,GACd,iBAAiB,GACjB,kBAAkB,GAClB,eAAe,GACf,WAAW,GACX,gBAAgB,GAChB,mBAAmB,CAAC;AAExB;;GAEG;AACH,MAAM,MAAM,WAAW,GAAG,WAAW,CAAC;AAMtC
|
|
1
|
+
{"version":3,"file":"models.d.ts","sourceRoot":"","sources":["../src/models.ts"],"names":[],"mappings":"AAIA;;GAEG;AACH,eAAO,MAAM,cAAc,cAAc,CAAC;AAM1C;;GAEG;AACH,MAAM,MAAM,SAAS,GACjB,OAAO,GACP,SAAS,GACT,SAAS,GACT,YAAY,GACZ,eAAe,GACf,aAAa,CAAC;AAElB;;GAEG;AACH,MAAM,MAAM,YAAY,GAAG,IAAI,GAAG,IAAI,GAAG,IAAI,GAAG,IAAI,GAAG,IAAI,GAAG,IAAI,GAAG,IAAI,CAAC;AAE1E,eAAO,MAAM,iBAAiB,yCAAyC,CAAC;AAExE,eAAO,MAAM,QAAQ,GAAI,OAAO,MAAM,KAAG,OAAsC,CAAC;AAEhF,MAAM,MAAM,aAAa,GAAG,SAAS,GAAG,MAAM,GAAG,QAAQ,GAAG,MAAM,GAAG,SAAS,CAAC;AAE/E,MAAM,MAAM,eAAe,GACvB,cAAc,GACd,WAAW,GACX,OAAO,GACP,YAAY,GACZ,eAAe,GACf,mBAAmB,GACnB,gBAAgB,GAChB,YAAY,GACZ,iBAAiB,GACjB,oBAAoB,GACpB,iBAAiB,GACjB,cAAc,GACd,UAAU,GACV,eAAe,GACf,kBAAkB,GAClB,gBAAgB,GAChB,aAAa,GACb,SAAS,GACT,cAAc,GACd,iBAAiB,GACjB,kBAAkB,GAClB,eAAe,GACf,WAAW,GACX,gBAAgB,GAChB,mBAAmB,CAAC;AAExB;;GAEG;AACH,MAAM,MAAM,WAAW,GAAG,WAAW,CAAC;AAMtC;;;;GAIG;AACH,MAAM,MAAM,QAAQ,GAAG,OAAO,GAAG,aAAa,CAAC"}
|
package/dist/models.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/models.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\n/**\n * LiveKit uses this encoding for all audio\n */\nexport const AUDIO_ENCODING = 'pcm_s16le';\n\n// ============================================================================\n// TTS\n// ============================================================================\n\n/**\n * See [the docs](https://docs.cartesia.ai/build-with-cartesia/tts-models/latest) for all options.\n */\nexport type TTSModels =\n | 'sonic'\n | 'sonic-2'\n | 'sonic-3'\n | 'sonic-lite'\n | 'sonic-preview'\n | 'sonic-turbo';\n\n/**\n * See [the docs](https://docs.cartesia.ai/build-with-cartesia/tts-models/latest) for all options.\n */\nexport type TTSLanguages = 'en' | 'es' | 'fr' | 'de' | 'pt' | 'zh' | 'ja';\n\nexport const TTSDefaultVoiceId = 'f786b574-daa5-4673-aa0c-cbe3e8534c02';\n\nexport const isSonic3 = (model: string): boolean => model.startsWith('sonic-3');\n\nexport type TTSVoiceSpeed = 'fastest' | 'fast' | 'normal' | 'slow' | 'slowest';\n\nexport type TTSVoiceEmotion =\n | 'anger:lowest'\n | 'anger:low'\n | 'anger'\n | 'anger:high'\n | 'anger:highest'\n | 'positivity:lowest'\n | 'positivity:low'\n | 'positivity'\n | 'positivity:high'\n | 'positivity:highest'\n | 'surprise:lowest'\n | 'surprise:low'\n | 'surprise'\n | 'surprise:high'\n | 'surprise:highest'\n | 'sadness:lowest'\n | 'sadness:low'\n | 'sadness'\n | 'sadness:high'\n | 'sadness:highest'\n | 'curiosity:lowest'\n | 'curiosity:low'\n | 'curiosity'\n | 'curiosity:high'\n | 'curiosity:highest';\n\n/**\n * @deprecated Encoding should not be parameterized. Only `pcm_s16le`is allowed. Prefer using {@link AUDIO_ENCODING}.\n */\nexport type TTSEncoding = 'pcm_s16le';\n\n// ============================================================================\n// STT\n// ============================================================================\n\n/**\n * See [the docs](https://docs.cartesia.ai/build-with-cartesia/stt-models/latest) for all options.\n */\nexport type STTModel = 'ink-2';\n"],"mappings":"AAOO,MAAM,iBAAiB;AAsBvB,MAAM,oBAAoB;AAE1B,MAAM,WAAW,CAAC,UAA2B,MAAM,WAAW,SAAS;","names":[]}
|
|
1
|
+
{"version":3,"sources":["../src/models.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\n/**\n * LiveKit uses this encoding for all audio\n */\nexport const AUDIO_ENCODING = 'pcm_s16le';\n\n// ============================================================================\n// TTS\n// ============================================================================\n\n/**\n * See [the docs](https://docs.cartesia.ai/build-with-cartesia/tts-models/latest) for all options.\n */\nexport type TTSModels =\n | 'sonic'\n | 'sonic-2'\n | 'sonic-3'\n | 'sonic-lite'\n | 'sonic-preview'\n | 'sonic-turbo';\n\n/**\n * See [the docs](https://docs.cartesia.ai/build-with-cartesia/tts-models/latest) for all options.\n */\nexport type TTSLanguages = 'en' | 'es' | 'fr' | 'de' | 'pt' | 'zh' | 'ja';\n\nexport const TTSDefaultVoiceId = 'f786b574-daa5-4673-aa0c-cbe3e8534c02';\n\nexport const isSonic3 = (model: string): boolean => model.startsWith('sonic-3');\n\nexport type TTSVoiceSpeed = 'fastest' | 'fast' | 'normal' | 'slow' | 'slowest';\n\nexport type TTSVoiceEmotion =\n | 'anger:lowest'\n | 'anger:low'\n | 'anger'\n | 'anger:high'\n | 'anger:highest'\n | 'positivity:lowest'\n | 'positivity:low'\n | 'positivity'\n | 'positivity:high'\n | 'positivity:highest'\n | 'surprise:lowest'\n | 'surprise:low'\n | 'surprise'\n | 'surprise:high'\n | 'surprise:highest'\n | 'sadness:lowest'\n | 'sadness:low'\n | 'sadness'\n | 'sadness:high'\n | 'sadness:highest'\n | 'curiosity:lowest'\n | 'curiosity:low'\n | 'curiosity'\n | 'curiosity:high'\n | 'curiosity:highest';\n\n/**\n * @deprecated Encoding should not be parameterized. Only `pcm_s16le`is allowed. Prefer using {@link AUDIO_ENCODING}.\n */\nexport type TTSEncoding = 'pcm_s16le';\n\n// ============================================================================\n// STT\n// ============================================================================\n\n/**\n * See [the docs](https://docs.cartesia.ai/build-with-cartesia/stt-models/latest) for all options.\n *\n * `ink-2` currently supports English only; `ink-whisper` is multilingual.\n */\nexport type STTModel = 'ink-2' | 'ink-whisper';\n"],"mappings":"AAOO,MAAM,iBAAiB;AAsBvB,MAAM,oBAAoB;AAE1B,MAAM,WAAW,CAAC,UAA2B,MAAM,WAAW,SAAS;","names":[]}
|
package/dist/stt.cjs
CHANGED
|
@@ -37,7 +37,8 @@ const defaultSTTOptions = {
|
|
|
37
37
|
sampleRate: 16e3,
|
|
38
38
|
/** recommended default */
|
|
39
39
|
audioChunkDurationMS: 160,
|
|
40
|
-
baseUrl: "https://api.cartesia.ai"
|
|
40
|
+
baseUrl: "https://api.cartesia.ai",
|
|
41
|
+
language: "en"
|
|
41
42
|
};
|
|
42
43
|
function mergeSTTOptions(base, override) {
|
|
43
44
|
return {
|
|
@@ -45,9 +46,13 @@ function mergeSTTOptions(base, override) {
|
|
|
45
46
|
baseUrl: override.baseUrl ?? base.baseUrl,
|
|
46
47
|
model: override.model ?? base.model,
|
|
47
48
|
sampleRate: override.sampleRate ?? base.sampleRate,
|
|
48
|
-
audioChunkDurationMS: override.audioChunkDurationMS ?? base.audioChunkDurationMS
|
|
49
|
+
audioChunkDurationMS: override.audioChunkDurationMS ?? base.audioChunkDurationMS,
|
|
50
|
+
language: override.language !== void 0 ? (0, import_agents.normalizeLanguage)(override.language) : base.language
|
|
49
51
|
};
|
|
50
52
|
}
|
|
53
|
+
function resolveSTTModel(language) {
|
|
54
|
+
return (0, import_agents.getBaseLanguage)(language) === "en" ? "ink-2" : "ink-whisper";
|
|
55
|
+
}
|
|
51
56
|
class STT extends import_agents.stt.STT {
|
|
52
57
|
#opts;
|
|
53
58
|
constructor(opts = {}) {
|
|
@@ -64,6 +69,9 @@ class STT extends import_agents.stt.STT {
|
|
|
64
69
|
);
|
|
65
70
|
}
|
|
66
71
|
this.#opts = mergeSTTOptions({ ...defaultSTTOptions, apiKey }, opts);
|
|
72
|
+
if (opts.model === void 0) {
|
|
73
|
+
this.#opts.model = resolveSTTModel(this.#opts.language);
|
|
74
|
+
}
|
|
67
75
|
}
|
|
68
76
|
get label() {
|
|
69
77
|
return "cartesia.STT";
|
|
@@ -80,6 +88,12 @@ class STT extends import_agents.stt.STT {
|
|
|
80
88
|
stream(options) {
|
|
81
89
|
return new SpeechStream(this, this.#opts, options == null ? void 0 : options.connOptions);
|
|
82
90
|
}
|
|
91
|
+
updateOptions(opts) {
|
|
92
|
+
this.#opts = mergeSTTOptions(this.#opts, opts);
|
|
93
|
+
if (opts.language !== void 0 && opts.model === void 0) {
|
|
94
|
+
this.#opts.model = resolveSTTModel(this.#opts.language);
|
|
95
|
+
}
|
|
96
|
+
}
|
|
83
97
|
}
|
|
84
98
|
class SpeechStream extends import_agents.stt.SpeechStream {
|
|
85
99
|
#opts;
|
|
@@ -333,7 +347,7 @@ class SpeechStream extends import_agents.stt.SpeechStream {
|
|
|
333
347
|
}
|
|
334
348
|
case "turn.eager_end": {
|
|
335
349
|
if (!this.#speaking) return;
|
|
336
|
-
const transcript = data.transcript;
|
|
350
|
+
const transcript = data.transcript || this.#currentTranscript;
|
|
337
351
|
if (!transcript) return;
|
|
338
352
|
this.#currentTranscript = transcript;
|
|
339
353
|
this.#sendTranscriptEvent(import_agents.stt.SpeechEventType.PREFLIGHT_TRANSCRIPT, transcript);
|
|
@@ -345,7 +359,7 @@ class SpeechStream extends import_agents.stt.SpeechStream {
|
|
|
345
359
|
return;
|
|
346
360
|
case "turn.end": {
|
|
347
361
|
if (!this.#speaking) return;
|
|
348
|
-
const transcript = data.transcript;
|
|
362
|
+
const transcript = data.transcript || this.#currentTranscript;
|
|
349
363
|
this.#sendTranscriptEvent(import_agents.stt.SpeechEventType.FINAL_TRANSCRIPT, transcript);
|
|
350
364
|
this.#speaking = false;
|
|
351
365
|
if (!this.queue.closed)
|
|
@@ -376,8 +390,7 @@ class SpeechStream extends import_agents.stt.SpeechStream {
|
|
|
376
390
|
requestId: this.#requestId,
|
|
377
391
|
alternatives: [
|
|
378
392
|
{
|
|
379
|
-
|
|
380
|
-
language: (0, import_agents.asLanguageCode)("en"),
|
|
393
|
+
language: (0, import_agents.asLanguageCode)(this.#opts.language),
|
|
381
394
|
text: transcript,
|
|
382
395
|
startTime: 0,
|
|
383
396
|
endTime: 0,
|
package/dist/stt.cjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2026 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n APIConnectionError,\n AudioByteStream,\n asLanguageCode,\n calculateAudioDurationSeconds,\n log,\n stt,\n waitForAbort,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { IncomingMessage } from 'node:http';\nimport { WebSocket } from 'ws';\nimport { AUDIO_ENCODING, type STTModel } from './models.js';\n\nconst AUTHORIZATION_HEADER = 'X-API-Key';\nconst VERSION_HEADER = 'Cartesia-Version';\nconst REQUEST_ID_HEADER = 'cartesia-request-id';\nconst API_VERSION = '2026-03-01';\nconst DRAIN_TIMEOUT_MS = 5000;\nconst KEEPALIVE_INTERVAL_MS = 30000;\n\n/**\n * Fires once when the WebSocket connection is established.\n *\n * You do not need to wait for this event before sending audio.\n */\ntype STTConnectedEvent = {\n type: 'connected';\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\n/**\n * Model predicts {@link stt.SpeechEventType.START_OF_SPEECH}.\n */\ntype STTTurnStartEvent = {\n type: 'turn.start';\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\n/**\n * Fires repeatedly as the model transcribes the current user turn.\n *\n * Can be used for {@link stt.SpeechEventType.INTERIM_TRANSCRIPT}.\n */\ntype STTTurnUpdateEvent = {\n type: 'turn.update';\n /**\n * Cumulative text for the current turn, i.e. the full text transcribed so far in this turn, not a delta.\n */\n transcript: string;\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\n/**\n * Fires when the model predicts that the user might be done speaking.\n *\n * Can be used for {@link stt.SpeechEventType.PREFLIGHT_TRANSCRIPT}.\n */\ntype STTTurnEagerEndEvent = {\n type: 'turn.eager_end';\n /**\n * Cumulative text for the current turn, i.e. the full text transcribed so far in this turn, not a delta.\n */\n transcript: string;\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\n/**\n * Fires after {@link STTTurnEagerEndEvent} if the user turn has not actually ended.\n */\ntype STTTurnResumeEvent = {\n type: 'turn.resume';\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\n/**\n * Marks the end of a user turn.\n *\n * This is used for {@link stt.SpeechEventType.END_OF_SPEECH} and {@link stt.SpeechEventType.FINAL_TRANSCRIPT}.\n */\ntype STTTurnEndEvent = {\n type: 'turn.end';\n /**\n * Cumulative text for the current turn, i.e. the full text transcribed so far in this turn, not a delta.\n */\n transcript: string;\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\ntype STTErrorEvent = {\n type: 'error';\n error_code?: string;\n status_code?: number;\n title?: string;\n message?: string;\n doc_url?: string;\n request_id?: string;\n};\n\n/**\n * Server-sent event on the `/stt/turns/websocket` endpoint.\n *\n * See https://docs.cartesia.ai/api-reference/stt/turns/websocket.\n */\ntype STTEventMessage =\n | STTConnectedEvent\n | STTTurnStartEvent\n | STTTurnUpdateEvent\n | STTTurnEagerEndEvent\n | STTTurnResumeEvent\n | STTTurnEndEvent\n | STTErrorEvent;\n\nexport type STTOptions = {\n apiKey: string;\n // eslint-disable-next-line @typescript-eslint/ban-types\n model: STTModel | (string & {});\n sampleRate: number;\n baseUrl: string;\n audioChunkDurationMS: number;\n};\n\nconst defaultSTTOptions = {\n model: 'ink-2' satisfies STTModel,\n /** recommended default */\n sampleRate: 16_000,\n /** recommended default */\n audioChunkDurationMS: 160,\n baseUrl: 'https://api.cartesia.ai',\n};\n\nfunction mergeSTTOptions(base: STTOptions, override: Partial<STTOptions>): STTOptions {\n return {\n apiKey: override.apiKey ?? base.apiKey,\n baseUrl: override.baseUrl ?? base.baseUrl,\n model: override.model ?? base.model,\n sampleRate: override.sampleRate ?? base.sampleRate,\n audioChunkDurationMS: override.audioChunkDurationMS ?? base.audioChunkDurationMS,\n };\n}\n\n/**\n * Cartesia speech to text.\n *\n * Supports:\n * - Streaming\n * - Turn detection\n * - Interim results\n *\n * See https://docs.cartesia.ai/build-with-cartesia/stt-models/latest\n *\n * @example\n * ```typescript\n * import { voice } from '@livekit/agents';\n * import { STT, TTS } from '@livekit/agents-plugin-cartesia';\n *\n * const session = new voice.AgentSession({\n * stt: new STT(),\n * llm: new LLM(), // choose your favorite LLM\n * tts: new TTS(),\n * turnHandling: { turnDetection: 'stt' },\n * });\n * ```\n */\nexport class STT extends stt.STT {\n #opts: STTOptions;\n\n constructor(opts: Partial<STTOptions> = {}) {\n super({\n streaming: true,\n interimResults: true,\n alignedTranscript: false,\n diarization: false,\n });\n\n const apiKey = opts.apiKey ?? process.env.CARTESIA_API_KEY;\n if (!apiKey) {\n throw new Error(\n 'Cartesia API key is required, whether as an argument or as $CARTESIA_API_KEY',\n );\n }\n\n this.#opts = mergeSTTOptions({ ...defaultSTTOptions, apiKey }, opts);\n }\n\n override get label(): string {\n return 'cartesia.STT';\n }\n\n override get model(): string {\n return this.#opts.model;\n }\n\n override get provider(): string {\n return 'Cartesia';\n }\n\n protected override async _recognize(): Promise<stt.SpeechEvent> {\n throw new Error('Cartesia STT does not support batch recognition, use stream() instead');\n }\n\n override stream(options?: { connOptions?: APIConnectOptions }): SpeechStream {\n return new SpeechStream(this, this.#opts, options?.connOptions);\n }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n #opts: STTOptions;\n #logger = log();\n #ws: WebSocket | null = null;\n #requestId: string | undefined;\n #speaking = false;\n #currentTranscript = '';\n #speechDuration = 0;\n #closingWs = false;\n\n constructor(sttInstance: STT, opts: STTOptions, connOptions?: APIConnectOptions) {\n super(sttInstance, opts.sampleRate, connOptions);\n this.#opts = { ...opts };\n }\n\n override get label(): string {\n return 'cartesia.SpeechStream';\n }\n\n protected override async run() {\n if (this.closed) {\n this.close();\n return;\n }\n\n this.#speaking = false;\n this.#currentTranscript = '';\n this.#speechDuration = 0;\n this.#closingWs = false;\n\n let keepaliveInterval: NodeJS.Timeout | undefined;\n const abortController = new AbortController();\n\n try {\n const url = this.#getCartesiaUrl();\n this.#logger.debug(`Connecting to Cartesia STT: ${url}`);\n\n const ws = new WebSocket(url, {\n headers: {\n [AUTHORIZATION_HEADER]: this.#opts.apiKey,\n [VERSION_HEADER]: API_VERSION,\n },\n });\n this.#ws = ws;\n\n // Cartesia returns the request id on the WS upgrade response, before any\n // turn events are sent. Capture it so logs and metrics emitted on the\n // connection (and any pre-`turn.start` errors) are attributable.\n ws.once('upgrade', (response: IncomingMessage) => {\n const headerValue = response.headers[REQUEST_ID_HEADER];\n const requestId = Array.isArray(headerValue) ? headerValue[0] : headerValue;\n if (requestId) {\n this.#requestId = requestId;\n this.#logger.debug({ cartesiaRequestId: requestId }, 'Cartesia STT WebSocket connected');\n }\n });\n\n await new Promise<void>((resolve, reject) => {\n const onOpen = () => {\n ws.off('error', onError);\n resolve();\n };\n const onError = (err: Error) => {\n ws.off('open', onOpen);\n reject(new APIConnectionError({ message: err.message, options: { retryable: true } }));\n };\n\n ws.once('open', onOpen);\n ws.once('error', onError);\n });\n\n // If one task fails, abort its peer and close the WS so the peer can\n // exit before we re-throw. Otherwise a dangling task would survive\n // into the next retry attempt (sendTask blocked on input.next() would\n // steal frames; recvTask's close handler could still push events to\n // the queue and surface an unhandled rejection).\n let firstError: unknown;\n const stopPeer = (err: unknown) => {\n if (firstError === undefined) firstError = err;\n abortController.abort();\n if (ws.readyState === WebSocket.OPEN) ws.close();\n };\n const sendPromise = this.#sendTask(ws, abortController.signal).catch(stopPeer);\n const recvPromise = this.#recvTask(ws).catch(stopPeer);\n\n keepaliveInterval = setInterval(() => {\n try {\n if (ws.readyState === WebSocket.OPEN) {\n ws.ping();\n }\n } catch {\n // ignore\n }\n }, KEEPALIVE_INTERVAL_MS);\n await Promise.all([sendPromise, recvPromise]);\n if (firstError !== undefined) throw firstError;\n } catch (error) {\n this.#logger.error('Cartesia STT stream error', { error });\n throw error;\n } finally {\n abortController.abort();\n if (keepaliveInterval) clearInterval(keepaliveInterval);\n if (this.#ws?.readyState === WebSocket.OPEN) {\n this.#ws.close();\n }\n }\n\n this.close();\n }\n\n async #sendTask(ws: WebSocket, abortSignal: AbortSignal) {\n const samplesPerChunk = Math.floor(\n (this.#opts.sampleRate * this.#opts.audioChunkDurationMS) / 1000,\n );\n const audioBstream = new AudioByteStream(this.#opts.sampleRate, 1, samplesPerChunk);\n\n let hasEnded = false;\n const iterator = this.input[Symbol.asyncIterator]();\n const abortPromise = waitForAbort(abortSignal);\n\n while (true) {\n const result = await Promise.race([iterator.next(), abortPromise]);\n\n if (result === undefined) return; // aborted\n\n if (result.done) {\n hasEnded = true;\n } else {\n const data = result.value;\n const frames: AudioFrame[] = [];\n\n if (data === SpeechStream.FLUSH_SENTINEL) {\n frames.push(...audioBstream.flush());\n } else {\n // Pass the typed array directly so AudioByteStream uses byteOffset/\n // byteLength — `.buffer` would include foreign bytes when rtc-node\n // hands us a view over a pooled allocator.\n frames.push(...audioBstream.write(data.data));\n }\n\n for (const frame of frames) {\n this.#speechDuration += calculateAudioDurationSeconds(frame);\n if (ws.readyState === WebSocket.OPEN) {\n ws.send(frame.data);\n }\n }\n }\n\n if (hasEnded) break;\n }\n\n if (ws.readyState === WebSocket.OPEN) {\n this.#closingWs = true;\n this.#logger.debug('Sending close message to Cartesia STT');\n ws.send(JSON.stringify({ type: 'close' }));\n\n // After `close`, the server should flush any pending turn events and\n // close the socket. If it hangs, force the issue rather than letting\n // the stream wait forever.\n const drainTimer = setTimeout(() => {\n this.#logger.warn(\n `Cartesia STT did not close within ${DRAIN_TIMEOUT_MS}ms after done; forcing close`,\n );\n if (ws.readyState === WebSocket.OPEN || ws.readyState === WebSocket.CLOSING) {\n ws.terminate();\n }\n }, DRAIN_TIMEOUT_MS);\n ws.once('close', () => clearTimeout(drainTimer));\n }\n }\n\n async #recvTask(ws: WebSocket) {\n let pendingError: Error | undefined = undefined;\n\n return new Promise<void>((resolve, reject) => {\n const settle = (err?: Error) => {\n if (err) reject(err);\n else resolve();\n };\n\n ws.on('message', (data: Buffer, isBinary: boolean) => {\n if (isBinary) {\n this.#logger.warn('Received unexpected binary message from Cartesia STT');\n return;\n }\n let msg: STTEventMessage;\n try {\n msg = JSON.parse(data.toString());\n } catch (error) {\n this.#logger.error('Failed to parse Cartesia STT message', { error });\n return;\n }\n try {\n if (msg.type === 'error') {\n this.#logger.error('Cartesia sent an error', msg);\n\n // do not close the websocket on bad requests since that may be caused by invalid messages\n if (msg.status_code === undefined || msg.status_code >= 500) {\n // Defer until the WS is fully closed so we don't race the `close`\n // handler with a rejected `recvTask`. Close the socket explicitly\n // in case the server doesn't, otherwise recv would hang.\n pendingError = new APIConnectionError({\n message: msg.message || msg.title,\n options: {\n retryable: true,\n },\n });\n ws.close();\n return;\n }\n } else {\n this.#processStreamEvent(msg);\n }\n } catch (error) {\n this.#logger.error('Failed to process Cartesia STT message', { error });\n }\n });\n\n ws.on('close', (code, reason) => {\n // If a turn was in progress, close it out so consumers see a balanced\n // START_OF_SPEECH/END_OF_SPEECH pair and don't carry stale turn state\n // into the next reconnect.\n if (this.#speaking) {\n if (this.#currentTranscript) {\n this.#sendTranscriptEvent(\n stt.SpeechEventType.FINAL_TRANSCRIPT,\n this.#currentTranscript,\n );\n }\n if (!this.queue.closed)\n this.queue.put({\n type: stt.SpeechEventType.END_OF_SPEECH,\n requestId: this.#requestId,\n });\n this.#logger.debug('Cartesia STT END_OF_SPEECH');\n if (this.#speechDuration > 0 && !this.queue.closed) {\n this.queue.put({\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: { audioDuration: this.#speechDuration },\n });\n this.#speechDuration = 0;\n }\n this.#speaking = false;\n this.#currentTranscript = '';\n }\n\n if (pendingError) {\n settle(pendingError);\n return;\n }\n if (this.#closingWs || this.closed) {\n settle();\n return;\n }\n this.#logger.warn(`Cartesia STT WebSocket closed: ${code} ${reason.toString()}`);\n settle(\n new APIConnectionError({\n message: `Cartesia STT connection closed unexpectedly (code=${code})`,\n options: { retryable: true },\n }),\n );\n });\n\n ws.on('error', (err) => {\n if (this.closed) {\n settle();\n return;\n }\n settle(new APIConnectionError({ message: err.message, options: { retryable: true } }));\n });\n });\n }\n\n #processStreamEvent(data: Exclude<STTEventMessage, { type: 'error' }>) {\n if (data.request_id) {\n this.#requestId = data.request_id;\n }\n\n switch (data.type) {\n case 'connected':\n return;\n\n case 'turn.start': {\n if (this.#speaking) return;\n this.#speaking = true;\n this.#currentTranscript = '';\n if (!this.queue.closed)\n this.queue.put({\n type: stt.SpeechEventType.START_OF_SPEECH,\n requestId: this.#requestId,\n });\n this.#logger.debug('Cartesia STT START_OF_SPEECH');\n return;\n }\n\n case 'turn.update': {\n if (!this.#speaking) return;\n const transcript = data.transcript;\n if (!transcript) return;\n // Only emit interim updates when the cumulative transcript actually\n // changed; this avoids canceling preflight generation needlessly.\n if (this.#currentTranscript === transcript) return;\n this.#currentTranscript = transcript;\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, transcript);\n return;\n }\n\n case 'turn.eager_end': {\n if (!this.#speaking) return;\n const transcript = data.transcript;\n if (!transcript) return;\n this.#currentTranscript = transcript;\n this.#sendTranscriptEvent(stt.SpeechEventType.PREFLIGHT_TRANSCRIPT, transcript);\n return;\n }\n\n case 'turn.resume':\n // turn.resume has no transcript; re-emit the latest cumulative transcript\n // as an interim event so the pipeline cancels the pending preflight.\n if (!this.#speaking || !this.#currentTranscript) return;\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, this.#currentTranscript);\n return;\n\n case 'turn.end': {\n if (!this.#speaking) return;\n const transcript = data.transcript;\n\n this.#sendTranscriptEvent(stt.SpeechEventType.FINAL_TRANSCRIPT, transcript);\n\n this.#speaking = false;\n if (!this.queue.closed)\n this.queue.put({\n type: stt.SpeechEventType.END_OF_SPEECH,\n requestId: this.#requestId,\n });\n this.#logger.debug('Cartesia STT END_OF_SPEECH');\n this.#currentTranscript = '';\n\n if (this.#speechDuration > 0 && !this.queue.closed) {\n this.queue.put({\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: { audioDuration: this.#speechDuration },\n });\n this.#speechDuration = 0;\n }\n\n return;\n }\n\n default:\n this.#logger.warn('received unexpected message from Cartesia STT', { data });\n }\n }\n\n #sendTranscriptEvent(eventType: stt.SpeechEventType, transcript: string) {\n if (this.queue.closed) return;\n this.queue.put({\n type: eventType,\n requestId: this.#requestId,\n alternatives: [\n {\n // Cartesia STT only supports English at this time.\n language: asLanguageCode('en'),\n text: transcript,\n startTime: 0,\n endTime: 0,\n confidence: 0,\n },\n ],\n });\n }\n\n #getCartesiaUrl(): string {\n const params = new URLSearchParams({\n model: this.#opts.model,\n sample_rate: this.#opts.sampleRate.toString(),\n encoding: AUDIO_ENCODING,\n });\n\n const wsBase = this.#opts.baseUrl.replace(/^http/, 'ws');\n return `${wsBase}/stt/turns/websocket?${params.toString()}`;\n }\n\n override close() {\n super.close();\n this.#ws?.close();\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBASO;AAGP,gBAA0B;AAC1B,oBAA8C;AAE9C,MAAM,uBAAuB;AAC7B,MAAM,iBAAiB;AACvB,MAAM,oBAAoB;AAC1B,MAAM,cAAc;AACpB,MAAM,mBAAmB;AACzB,MAAM,wBAAwB;AAyH9B,MAAM,oBAAoB;AAAA,EACxB,OAAO;AAAA;AAAA,EAEP,YAAY;AAAA;AAAA,EAEZ,sBAAsB;AAAA,EACtB,SAAS;AACX;AAEA,SAAS,gBAAgB,MAAkB,UAA2C;AACpF,SAAO;AAAA,IACL,QAAQ,SAAS,UAAU,KAAK;AAAA,IAChC,SAAS,SAAS,WAAW,KAAK;AAAA,IAClC,OAAO,SAAS,SAAS,KAAK;AAAA,IAC9B,YAAY,SAAS,cAAc,KAAK;AAAA,IACxC,sBAAsB,SAAS,wBAAwB,KAAK;AAAA,EAC9D;AACF;AAyBO,MAAM,YAAY,kBAAI,IAAI;AAAA,EAC/B;AAAA,EAEA,YAAY,OAA4B,CAAC,GAAG;AAC1C,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB;AAAA,MAChB,mBAAmB;AAAA,MACnB,aAAa;AAAA,IACf,CAAC;AAED,UAAM,SAAS,KAAK,UAAU,QAAQ,IAAI;AAC1C,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ,gBAAgB,EAAE,GAAG,mBAAmB,OAAO,GAAG,IAAI;AAAA,EACrE;AAAA,EAEA,IAAa,QAAgB;AAC3B,WAAO;AAAA,EACT;AAAA,EAEA,IAAa,QAAgB;AAC3B,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAa,WAAmB;AAC9B,WAAO;AAAA,EACT;AAAA,EAEA,MAAyB,aAAuC;AAC9D,UAAM,IAAI,MAAM,uEAAuE;AAAA,EACzF;AAAA,EAES,OAAO,SAA6D;AAC3E,WAAO,IAAI,aAAa,MAAM,KAAK,OAAO,mCAAS,WAAW;AAAA,EAChE;AACF;AAEO,MAAM,qBAAqB,kBAAI,aAAa;AAAA,EACjD;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,MAAwB;AAAA,EACxB;AAAA,EACA,YAAY;AAAA,EACZ,qBAAqB;AAAA,EACrB,kBAAkB;AAAA,EAClB,aAAa;AAAA,EAEb,YAAY,aAAkB,MAAkB,aAAiC;AAC/E,UAAM,aAAa,KAAK,YAAY,WAAW;AAC/C,SAAK,QAAQ,EAAE,GAAG,KAAK;AAAA,EACzB;AAAA,EAEA,IAAa,QAAgB;AAC3B,WAAO;AAAA,EACT;AAAA,EAEA,MAAyB,MAAM;AAvPjC;AAwPI,QAAI,KAAK,QAAQ;AACf,WAAK,MAAM;AACX;AAAA,IACF;AAEA,SAAK,YAAY;AACjB,SAAK,qBAAqB;AAC1B,SAAK,kBAAkB;AACvB,SAAK,aAAa;AAElB,QAAI;AACJ,UAAM,kBAAkB,IAAI,gBAAgB;AAE5C,QAAI;AACF,YAAM,MAAM,KAAK,gBAAgB;AACjC,WAAK,QAAQ,MAAM,+BAA+B,GAAG,EAAE;AAEvD,YAAM,KAAK,IAAI,oBAAU,KAAK;AAAA,QAC5B,SAAS;AAAA,UACP,CAAC,oBAAoB,GAAG,KAAK,MAAM;AAAA,UACnC,CAAC,cAAc,GAAG;AAAA,QACpB;AAAA,MACF,CAAC;AACD,WAAK,MAAM;AAKX,SAAG,KAAK,WAAW,CAAC,aAA8B;AAChD,cAAM,cAAc,SAAS,QAAQ,iBAAiB;AACtD,cAAM,YAAY,MAAM,QAAQ,WAAW,IAAI,YAAY,CAAC,IAAI;AAChE,YAAI,WAAW;AACb,eAAK,aAAa;AAClB,eAAK,QAAQ,MAAM,EAAE,mBAAmB,UAAU,GAAG,kCAAkC;AAAA,QACzF;AAAA,MACF,CAAC;AAED,YAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,cAAM,SAAS,MAAM;AACnB,aAAG,IAAI,SAAS,OAAO;AACvB,kBAAQ;AAAA,QACV;AACA,cAAM,UAAU,CAAC,QAAe;AAC9B,aAAG,IAAI,QAAQ,MAAM;AACrB,iBAAO,IAAI,iCAAmB,EAAE,SAAS,IAAI,SAAS,SAAS,EAAE,WAAW,KAAK,EAAE,CAAC,CAAC;AAAA,QACvF;AAEA,WAAG,KAAK,QAAQ,MAAM;AACtB,WAAG,KAAK,SAAS,OAAO;AAAA,MAC1B,CAAC;AAOD,UAAI;AACJ,YAAM,WAAW,CAAC,QAAiB;AACjC,YAAI,eAAe,OAAW,cAAa;AAC3C,wBAAgB,MAAM;AACtB,YAAI,GAAG,eAAe,oBAAU,KAAM,IAAG,MAAM;AAAA,MACjD;AACA,YAAM,cAAc,KAAK,UAAU,IAAI,gBAAgB,MAAM,EAAE,MAAM,QAAQ;AAC7E,YAAM,cAAc,KAAK,UAAU,EAAE,EAAE,MAAM,QAAQ;AAErD,0BAAoB,YAAY,MAAM;AACpC,YAAI;AACF,cAAI,GAAG,eAAe,oBAAU,MAAM;AACpC,eAAG,KAAK;AAAA,UACV;AAAA,QACF,QAAQ;AAAA,QAER;AAAA,MACF,GAAG,qBAAqB;AACxB,YAAM,QAAQ,IAAI,CAAC,aAAa,WAAW,CAAC;AAC5C,UAAI,eAAe,OAAW,OAAM;AAAA,IACtC,SAAS,OAAO;AACd,WAAK,QAAQ,MAAM,6BAA6B,EAAE,MAAM,CAAC;AACzD,YAAM;AAAA,IACR,UAAE;AACA,sBAAgB,MAAM;AACtB,UAAI,kBAAmB,eAAc,iBAAiB;AACtD,YAAI,UAAK,QAAL,mBAAU,gBAAe,oBAAU,MAAM;AAC3C,aAAK,IAAI,MAAM;AAAA,MACjB;AAAA,IACF;AAEA,SAAK,MAAM;AAAA,EACb;AAAA,EAEA,MAAM,UAAU,IAAe,aAA0B;AACvD,UAAM,kBAAkB,KAAK;AAAA,MAC1B,KAAK,MAAM,aAAa,KAAK,MAAM,uBAAwB;AAAA,IAC9D;AACA,UAAM,eAAe,IAAI,8BAAgB,KAAK,MAAM,YAAY,GAAG,eAAe;AAElF,QAAI,WAAW;AACf,UAAM,WAAW,KAAK,MAAM,OAAO,aAAa,EAAE;AAClD,UAAM,mBAAe,4BAAa,WAAW;AAE7C,WAAO,MAAM;AACX,YAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,SAAS,KAAK,GAAG,YAAY,CAAC;AAEjE,UAAI,WAAW,OAAW;AAE1B,UAAI,OAAO,MAAM;AACf,mBAAW;AAAA,MACb,OAAO;AACL,cAAM,OAAO,OAAO;AACpB,cAAM,SAAuB,CAAC;AAE9B,YAAI,SAAS,aAAa,gBAAgB;AACxC,iBAAO,KAAK,GAAG,aAAa,MAAM,CAAC;AAAA,QACrC,OAAO;AAIL,iBAAO,KAAK,GAAG,aAAa,MAAM,KAAK,IAAI,CAAC;AAAA,QAC9C;AAEA,mBAAW,SAAS,QAAQ;AAC1B,eAAK,uBAAmB,6CAA8B,KAAK;AAC3D,cAAI,GAAG,eAAe,oBAAU,MAAM;AACpC,eAAG,KAAK,MAAM,IAAI;AAAA,UACpB;AAAA,QACF;AAAA,MACF;AAEA,UAAI,SAAU;AAAA,IAChB;AAEA,QAAI,GAAG,eAAe,oBAAU,MAAM;AACpC,WAAK,aAAa;AAClB,WAAK,QAAQ,MAAM,uCAAuC;AAC1D,SAAG,KAAK,KAAK,UAAU,EAAE,MAAM,QAAQ,CAAC,CAAC;AAKzC,YAAM,aAAa,WAAW,MAAM;AAClC,aAAK,QAAQ;AAAA,UACX,qCAAqC,gBAAgB;AAAA,QACvD;AACA,YAAI,GAAG,eAAe,oBAAU,QAAQ,GAAG,eAAe,oBAAU,SAAS;AAC3E,aAAG,UAAU;AAAA,QACf;AAAA,MACF,GAAG,gBAAgB;AACnB,SAAG,KAAK,SAAS,MAAM,aAAa,UAAU,CAAC;AAAA,IACjD;AAAA,EACF;AAAA,EAEA,MAAM,UAAU,IAAe;AAC7B,QAAI,eAAkC;AAEtC,WAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,YAAM,SAAS,CAAC,QAAgB;AAC9B,YAAI,IAAK,QAAO,GAAG;AAAA,YACd,SAAQ;AAAA,MACf;AAEA,SAAG,GAAG,WAAW,CAAC,MAAc,aAAsB;AACpD,YAAI,UAAU;AACZ,eAAK,QAAQ,KAAK,sDAAsD;AACxE;AAAA,QACF;AACA,YAAI;AACJ,YAAI;AACF,gBAAM,KAAK,MAAM,KAAK,SAAS,CAAC;AAAA,QAClC,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,wCAAwC,EAAE,MAAM,CAAC;AACpE;AAAA,QACF;AACA,YAAI;AACF,cAAI,IAAI,SAAS,SAAS;AACxB,iBAAK,QAAQ,MAAM,0BAA0B,GAAG;AAGhD,gBAAI,IAAI,gBAAgB,UAAa,IAAI,eAAe,KAAK;AAI3D,6BAAe,IAAI,iCAAmB;AAAA,gBACpC,SAAS,IAAI,WAAW,IAAI;AAAA,gBAC5B,SAAS;AAAA,kBACP,WAAW;AAAA,gBACb;AAAA,cACF,CAAC;AACD,iBAAG,MAAM;AACT;AAAA,YACF;AAAA,UACF,OAAO;AACL,iBAAK,oBAAoB,GAAG;AAAA,UAC9B;AAAA,QACF,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,0CAA0C,EAAE,MAAM,CAAC;AAAA,QACxE;AAAA,MACF,CAAC;AAED,SAAG,GAAG,SAAS,CAAC,MAAM,WAAW;AAI/B,YAAI,KAAK,WAAW;AAClB,cAAI,KAAK,oBAAoB;AAC3B,iBAAK;AAAA,cACH,kBAAI,gBAAgB;AAAA,cACpB,KAAK;AAAA,YACP;AAAA,UACF;AACA,cAAI,CAAC,KAAK,MAAM;AACd,iBAAK,MAAM,IAAI;AAAA,cACb,MAAM,kBAAI,gBAAgB;AAAA,cAC1B,WAAW,KAAK;AAAA,YAClB,CAAC;AACH,eAAK,QAAQ,MAAM,4BAA4B;AAC/C,cAAI,KAAK,kBAAkB,KAAK,CAAC,KAAK,MAAM,QAAQ;AAClD,iBAAK,MAAM,IAAI;AAAA,cACb,MAAM,kBAAI,gBAAgB;AAAA,cAC1B,WAAW,KAAK;AAAA,cAChB,kBAAkB,EAAE,eAAe,KAAK,gBAAgB;AAAA,YAC1D,CAAC;AACD,iBAAK,kBAAkB;AAAA,UACzB;AACA,eAAK,YAAY;AACjB,eAAK,qBAAqB;AAAA,QAC5B;AAEA,YAAI,cAAc;AAChB,iBAAO,YAAY;AACnB;AAAA,QACF;AACA,YAAI,KAAK,cAAc,KAAK,QAAQ;AAClC,iBAAO;AACP;AAAA,QACF;AACA,aAAK,QAAQ,KAAK,kCAAkC,IAAI,IAAI,OAAO,SAAS,CAAC,EAAE;AAC/E;AAAA,UACE,IAAI,iCAAmB;AAAA,YACrB,SAAS,qDAAqD,IAAI;AAAA,YAClE,SAAS,EAAE,WAAW,KAAK;AAAA,UAC7B,CAAC;AAAA,QACH;AAAA,MACF,CAAC;AAED,SAAG,GAAG,SAAS,CAAC,QAAQ;AACtB,YAAI,KAAK,QAAQ;AACf,iBAAO;AACP;AAAA,QACF;AACA,eAAO,IAAI,iCAAmB,EAAE,SAAS,IAAI,SAAS,SAAS,EAAE,WAAW,KAAK,EAAE,CAAC,CAAC;AAAA,MACvF,CAAC;AAAA,IACH,CAAC;AAAA,EACH;AAAA,EAEA,oBAAoB,MAAmD;AACrE,QAAI,KAAK,YAAY;AACnB,WAAK,aAAa,KAAK;AAAA,IACzB;AAEA,YAAQ,KAAK,MAAM;AAAA,MACjB,KAAK;AACH;AAAA,MAEF,KAAK,cAAc;AACjB,YAAI,KAAK,UAAW;AACpB,aAAK,YAAY;AACjB,aAAK,qBAAqB;AAC1B,YAAI,CAAC,KAAK,MAAM;AACd,eAAK,MAAM,IAAI;AAAA,YACb,MAAM,kBAAI,gBAAgB;AAAA,YAC1B,WAAW,KAAK;AAAA,UAClB,CAAC;AACH,aAAK,QAAQ,MAAM,8BAA8B;AACjD;AAAA,MACF;AAAA,MAEA,KAAK,eAAe;AAClB,YAAI,CAAC,KAAK,UAAW;AACrB,cAAM,aAAa,KAAK;AACxB,YAAI,CAAC,WAAY;AAGjB,YAAI,KAAK,uBAAuB,WAAY;AAC5C,aAAK,qBAAqB;AAC1B,aAAK,qBAAqB,kBAAI,gBAAgB,oBAAoB,UAAU;AAC5E;AAAA,MACF;AAAA,MAEA,KAAK,kBAAkB;AACrB,YAAI,CAAC,KAAK,UAAW;AACrB,cAAM,aAAa,KAAK;AACxB,YAAI,CAAC,WAAY;AACjB,aAAK,qBAAqB;AAC1B,aAAK,qBAAqB,kBAAI,gBAAgB,sBAAsB,UAAU;AAC9E;AAAA,MACF;AAAA,MAEA,KAAK;AAGH,YAAI,CAAC,KAAK,aAAa,CAAC,KAAK,mBAAoB;AACjD,aAAK,qBAAqB,kBAAI,gBAAgB,oBAAoB,KAAK,kBAAkB;AACzF;AAAA,MAEF,KAAK,YAAY;AACf,YAAI,CAAC,KAAK,UAAW;AACrB,cAAM,aAAa,KAAK;AAExB,aAAK,qBAAqB,kBAAI,gBAAgB,kBAAkB,UAAU;AAE1E,aAAK,YAAY;AACjB,YAAI,CAAC,KAAK,MAAM;AACd,eAAK,MAAM,IAAI;AAAA,YACb,MAAM,kBAAI,gBAAgB;AAAA,YAC1B,WAAW,KAAK;AAAA,UAClB,CAAC;AACH,aAAK,QAAQ,MAAM,4BAA4B;AAC/C,aAAK,qBAAqB;AAE1B,YAAI,KAAK,kBAAkB,KAAK,CAAC,KAAK,MAAM,QAAQ;AAClD,eAAK,MAAM,IAAI;AAAA,YACb,MAAM,kBAAI,gBAAgB;AAAA,YAC1B,WAAW,KAAK;AAAA,YAChB,kBAAkB,EAAE,eAAe,KAAK,gBAAgB;AAAA,UAC1D,CAAC;AACD,eAAK,kBAAkB;AAAA,QACzB;AAEA;AAAA,MACF;AAAA,MAEA;AACE,aAAK,QAAQ,KAAK,iDAAiD,EAAE,KAAK,CAAC;AAAA,IAC/E;AAAA,EACF;AAAA,EAEA,qBAAqB,WAAgC,YAAoB;AACvE,QAAI,KAAK,MAAM,OAAQ;AACvB,SAAK,MAAM,IAAI;AAAA,MACb,MAAM;AAAA,MACN,WAAW,KAAK;AAAA,MAChB,cAAc;AAAA,QACZ;AAAA;AAAA,UAEE,cAAU,8BAAe,IAAI;AAAA,UAC7B,MAAM;AAAA,UACN,WAAW;AAAA,UACX,SAAS;AAAA,UACT,YAAY;AAAA,QACd;AAAA,MACF;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,kBAA0B;AACxB,UAAM,SAAS,IAAI,gBAAgB;AAAA,MACjC,OAAO,KAAK,MAAM;AAAA,MAClB,aAAa,KAAK,MAAM,WAAW,SAAS;AAAA,MAC5C,UAAU;AAAA,IACZ,CAAC;AAED,UAAM,SAAS,KAAK,MAAM,QAAQ,QAAQ,SAAS,IAAI;AACvD,WAAO,GAAG,MAAM,wBAAwB,OAAO,SAAS,CAAC;AAAA,EAC3D;AAAA,EAES,QAAQ;AArmBnB;AAsmBI,UAAM,MAAM;AACZ,eAAK,QAAL,mBAAU;AAAA,EACZ;AACF;","names":[]}
|
|
1
|
+
{"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2026 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n APIConnectionError,\n AudioByteStream,\n asLanguageCode,\n calculateAudioDurationSeconds,\n getBaseLanguage,\n log,\n normalizeLanguage,\n stt,\n waitForAbort,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { IncomingMessage } from 'node:http';\nimport { WebSocket } from 'ws';\nimport { AUDIO_ENCODING, type STTModel } from './models.js';\n\nconst AUTHORIZATION_HEADER = 'X-API-Key';\nconst VERSION_HEADER = 'Cartesia-Version';\nconst REQUEST_ID_HEADER = 'cartesia-request-id';\nconst API_VERSION = '2026-03-01';\nconst DRAIN_TIMEOUT_MS = 5000;\nconst KEEPALIVE_INTERVAL_MS = 30000;\n\n/**\n * Fires once when the WebSocket connection is established.\n *\n * You do not need to wait for this event before sending audio.\n */\ntype STTConnectedEvent = {\n type: 'connected';\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\n/**\n * Model predicts {@link stt.SpeechEventType.START_OF_SPEECH}.\n */\ntype STTTurnStartEvent = {\n type: 'turn.start';\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\n/**\n * Fires repeatedly as the model transcribes the current user turn.\n *\n * Can be used for {@link stt.SpeechEventType.INTERIM_TRANSCRIPT}.\n */\ntype STTTurnUpdateEvent = {\n type: 'turn.update';\n /**\n * Cumulative text for the current turn, i.e. the full text transcribed so far in this turn, not a delta.\n */\n transcript: string;\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\n/**\n * Fires when the model predicts that the user might be done speaking.\n *\n * Can be used for {@link stt.SpeechEventType.PREFLIGHT_TRANSCRIPT}.\n */\ntype STTTurnEagerEndEvent = {\n type: 'turn.eager_end';\n /**\n * Cumulative text for the current turn, i.e. the full text transcribed so far in this turn, not a delta.\n */\n transcript: string;\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\n/**\n * Fires after {@link STTTurnEagerEndEvent} if the user turn has not actually ended.\n */\ntype STTTurnResumeEvent = {\n type: 'turn.resume';\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\n/**\n * Marks the end of a user turn.\n *\n * This is used for {@link stt.SpeechEventType.END_OF_SPEECH} and {@link stt.SpeechEventType.FINAL_TRANSCRIPT}.\n */\ntype STTTurnEndEvent = {\n type: 'turn.end';\n /**\n * Cumulative text for the current turn, i.e. the full text transcribed so far in this turn, not a delta.\n */\n transcript: string;\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\ntype STTErrorEvent = {\n type: 'error';\n error_code?: string;\n status_code?: number;\n title?: string;\n message?: string;\n doc_url?: string;\n request_id?: string;\n};\n\n/**\n * Server-sent event on the `/stt/turns/websocket` endpoint.\n *\n * See https://docs.cartesia.ai/api-reference/stt/turns/websocket.\n */\ntype STTEventMessage =\n | STTConnectedEvent\n | STTTurnStartEvent\n | STTTurnUpdateEvent\n | STTTurnEagerEndEvent\n | STTTurnResumeEvent\n | STTTurnEndEvent\n | STTErrorEvent;\n\nexport type STTOptions = {\n apiKey: string;\n // eslint-disable-next-line @typescript-eslint/ban-types\n model: STTModel | (string & {});\n sampleRate: number;\n baseUrl: string;\n audioChunkDurationMS: number;\n language: string;\n};\n\nconst defaultSTTOptions = {\n model: 'ink-2' satisfies STTModel,\n /** recommended default */\n sampleRate: 16_000,\n /** recommended default */\n audioChunkDurationMS: 160,\n baseUrl: 'https://api.cartesia.ai',\n language: 'en',\n};\n\nfunction mergeSTTOptions(base: STTOptions, override: Partial<STTOptions>): STTOptions {\n return {\n apiKey: override.apiKey ?? base.apiKey,\n baseUrl: override.baseUrl ?? base.baseUrl,\n model: override.model ?? base.model,\n sampleRate: override.sampleRate ?? base.sampleRate,\n audioChunkDurationMS: override.audioChunkDurationMS ?? base.audioChunkDurationMS,\n language:\n override.language !== undefined ? normalizeLanguage(override.language) : base.language,\n };\n}\n\n/**\n * Resolve the STT model from the configured language when the caller did not\n * pass one explicitly. Mirrors the Python plugin: `ink-2` only supports\n * English, so non-English languages route to the multilingual `ink-whisper`.\n */\nfunction resolveSTTModel(language: string): STTModel {\n return getBaseLanguage(language) === 'en' ? 'ink-2' : 'ink-whisper';\n}\n\n/**\n * Cartesia speech to text.\n *\n * Supports:\n * - Streaming\n * - Turn detection\n * - Interim results\n *\n * See https://docs.cartesia.ai/build-with-cartesia/stt-models/latest\n *\n * @example\n * ```typescript\n * import { voice } from '@livekit/agents';\n * import { STT, TTS } from '@livekit/agents-plugin-cartesia';\n *\n * const session = new voice.AgentSession({\n * stt: new STT(),\n * llm: new LLM(), // choose your favorite LLM\n * tts: new TTS(),\n * turnHandling: { turnDetection: 'stt' },\n * });\n * ```\n */\nexport class STT extends stt.STT {\n #opts: STTOptions;\n\n constructor(opts: Partial<STTOptions> = {}) {\n super({\n streaming: true,\n interimResults: true,\n alignedTranscript: false,\n diarization: false,\n });\n\n const apiKey = opts.apiKey ?? process.env.CARTESIA_API_KEY;\n if (!apiKey) {\n throw new Error(\n 'Cartesia API key is required, whether as an argument or as $CARTESIA_API_KEY',\n );\n }\n\n this.#opts = mergeSTTOptions({ ...defaultSTTOptions, apiKey }, opts);\n\n // Route to the multilingual model for non-English languages unless the\n // caller pinned a model. Matches the Python plugin's language→model logic.\n if (opts.model === undefined) {\n this.#opts.model = resolveSTTModel(this.#opts.language);\n }\n }\n\n override get label(): string {\n return 'cartesia.STT';\n }\n\n override get model(): string {\n return this.#opts.model;\n }\n\n override get provider(): string {\n return 'Cartesia';\n }\n\n protected override async _recognize(): Promise<stt.SpeechEvent> {\n throw new Error('Cartesia STT does not support batch recognition, use stream() instead');\n }\n\n override stream(options?: { connOptions?: APIConnectOptions }): SpeechStream {\n return new SpeechStream(this, this.#opts, options?.connOptions);\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = mergeSTTOptions(this.#opts, opts);\n\n // Keep the model in sync with a newly set language (e.g. switching to a\n // non-English language must move off the English-only ink-2), unless the\n // caller pinned a model in the same call. Mirrors the constructor.\n if (opts.language !== undefined && opts.model === undefined) {\n this.#opts.model = resolveSTTModel(this.#opts.language);\n }\n }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n #opts: STTOptions;\n #logger = log();\n #ws: WebSocket | null = null;\n #requestId: string | undefined;\n #speaking = false;\n #currentTranscript = '';\n #speechDuration = 0;\n #closingWs = false;\n\n constructor(sttInstance: STT, opts: STTOptions, connOptions?: APIConnectOptions) {\n super(sttInstance, opts.sampleRate, connOptions);\n this.#opts = { ...opts };\n }\n\n override get label(): string {\n return 'cartesia.SpeechStream';\n }\n\n protected override async run() {\n if (this.closed) {\n this.close();\n return;\n }\n\n this.#speaking = false;\n this.#currentTranscript = '';\n this.#speechDuration = 0;\n this.#closingWs = false;\n\n let keepaliveInterval: NodeJS.Timeout | undefined;\n const abortController = new AbortController();\n\n try {\n const url = this.#getCartesiaUrl();\n this.#logger.debug(`Connecting to Cartesia STT: ${url}`);\n\n const ws = new WebSocket(url, {\n headers: {\n [AUTHORIZATION_HEADER]: this.#opts.apiKey,\n [VERSION_HEADER]: API_VERSION,\n },\n });\n this.#ws = ws;\n\n // Cartesia returns the request id on the WS upgrade response, before any\n // turn events are sent. Capture it so logs and metrics emitted on the\n // connection (and any pre-`turn.start` errors) are attributable.\n ws.once('upgrade', (response: IncomingMessage) => {\n const headerValue = response.headers[REQUEST_ID_HEADER];\n const requestId = Array.isArray(headerValue) ? headerValue[0] : headerValue;\n if (requestId) {\n this.#requestId = requestId;\n this.#logger.debug({ cartesiaRequestId: requestId }, 'Cartesia STT WebSocket connected');\n }\n });\n\n await new Promise<void>((resolve, reject) => {\n const onOpen = () => {\n ws.off('error', onError);\n resolve();\n };\n const onError = (err: Error) => {\n ws.off('open', onOpen);\n reject(new APIConnectionError({ message: err.message, options: { retryable: true } }));\n };\n\n ws.once('open', onOpen);\n ws.once('error', onError);\n });\n\n // If one task fails, abort its peer and close the WS so the peer can\n // exit before we re-throw. Otherwise a dangling task would survive\n // into the next retry attempt (sendTask blocked on input.next() would\n // steal frames; recvTask's close handler could still push events to\n // the queue and surface an unhandled rejection).\n let firstError: unknown;\n const stopPeer = (err: unknown) => {\n if (firstError === undefined) firstError = err;\n abortController.abort();\n if (ws.readyState === WebSocket.OPEN) ws.close();\n };\n const sendPromise = this.#sendTask(ws, abortController.signal).catch(stopPeer);\n const recvPromise = this.#recvTask(ws).catch(stopPeer);\n\n keepaliveInterval = setInterval(() => {\n try {\n if (ws.readyState === WebSocket.OPEN) {\n ws.ping();\n }\n } catch {\n // ignore\n }\n }, KEEPALIVE_INTERVAL_MS);\n await Promise.all([sendPromise, recvPromise]);\n if (firstError !== undefined) throw firstError;\n } catch (error) {\n this.#logger.error('Cartesia STT stream error', { error });\n throw error;\n } finally {\n abortController.abort();\n if (keepaliveInterval) clearInterval(keepaliveInterval);\n if (this.#ws?.readyState === WebSocket.OPEN) {\n this.#ws.close();\n }\n }\n\n this.close();\n }\n\n async #sendTask(ws: WebSocket, abortSignal: AbortSignal) {\n const samplesPerChunk = Math.floor(\n (this.#opts.sampleRate * this.#opts.audioChunkDurationMS) / 1000,\n );\n const audioBstream = new AudioByteStream(this.#opts.sampleRate, 1, samplesPerChunk);\n\n let hasEnded = false;\n const iterator = this.input[Symbol.asyncIterator]();\n const abortPromise = waitForAbort(abortSignal);\n\n while (true) {\n const result = await Promise.race([iterator.next(), abortPromise]);\n\n if (result === undefined) return; // aborted\n\n if (result.done) {\n hasEnded = true;\n } else {\n const data = result.value;\n const frames: AudioFrame[] = [];\n\n if (data === SpeechStream.FLUSH_SENTINEL) {\n frames.push(...audioBstream.flush());\n } else {\n // Pass the typed array directly so AudioByteStream uses byteOffset/\n // byteLength — `.buffer` would include foreign bytes when rtc-node\n // hands us a view over a pooled allocator.\n frames.push(...audioBstream.write(data.data));\n }\n\n for (const frame of frames) {\n this.#speechDuration += calculateAudioDurationSeconds(frame);\n if (ws.readyState === WebSocket.OPEN) {\n ws.send(frame.data);\n }\n }\n }\n\n if (hasEnded) break;\n }\n\n if (ws.readyState === WebSocket.OPEN) {\n this.#closingWs = true;\n this.#logger.debug('Sending close message to Cartesia STT');\n ws.send(JSON.stringify({ type: 'close' }));\n\n // After `close`, the server should flush any pending turn events and\n // close the socket. If it hangs, force the issue rather than letting\n // the stream wait forever.\n const drainTimer = setTimeout(() => {\n this.#logger.warn(\n `Cartesia STT did not close within ${DRAIN_TIMEOUT_MS}ms after done; forcing close`,\n );\n if (ws.readyState === WebSocket.OPEN || ws.readyState === WebSocket.CLOSING) {\n ws.terminate();\n }\n }, DRAIN_TIMEOUT_MS);\n ws.once('close', () => clearTimeout(drainTimer));\n }\n }\n\n async #recvTask(ws: WebSocket) {\n let pendingError: Error | undefined = undefined;\n\n return new Promise<void>((resolve, reject) => {\n const settle = (err?: Error) => {\n if (err) reject(err);\n else resolve();\n };\n\n ws.on('message', (data: Buffer, isBinary: boolean) => {\n if (isBinary) {\n this.#logger.warn('Received unexpected binary message from Cartesia STT');\n return;\n }\n let msg: STTEventMessage;\n try {\n msg = JSON.parse(data.toString());\n } catch (error) {\n this.#logger.error('Failed to parse Cartesia STT message', { error });\n return;\n }\n try {\n if (msg.type === 'error') {\n this.#logger.error('Cartesia sent an error', msg);\n\n // do not close the websocket on bad requests since that may be caused by invalid messages\n if (msg.status_code === undefined || msg.status_code >= 500) {\n // Defer until the WS is fully closed so we don't race the `close`\n // handler with a rejected `recvTask`. Close the socket explicitly\n // in case the server doesn't, otherwise recv would hang.\n pendingError = new APIConnectionError({\n message: msg.message || msg.title,\n options: {\n retryable: true,\n },\n });\n ws.close();\n return;\n }\n } else {\n this.#processStreamEvent(msg);\n }\n } catch (error) {\n this.#logger.error('Failed to process Cartesia STT message', { error });\n }\n });\n\n ws.on('close', (code, reason) => {\n // If a turn was in progress, close it out so consumers see a balanced\n // START_OF_SPEECH/END_OF_SPEECH pair and don't carry stale turn state\n // into the next reconnect.\n if (this.#speaking) {\n if (this.#currentTranscript) {\n this.#sendTranscriptEvent(\n stt.SpeechEventType.FINAL_TRANSCRIPT,\n this.#currentTranscript,\n );\n }\n if (!this.queue.closed)\n this.queue.put({\n type: stt.SpeechEventType.END_OF_SPEECH,\n requestId: this.#requestId,\n });\n this.#logger.debug('Cartesia STT END_OF_SPEECH');\n if (this.#speechDuration > 0 && !this.queue.closed) {\n this.queue.put({\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: { audioDuration: this.#speechDuration },\n });\n this.#speechDuration = 0;\n }\n this.#speaking = false;\n this.#currentTranscript = '';\n }\n\n if (pendingError) {\n settle(pendingError);\n return;\n }\n if (this.#closingWs || this.closed) {\n settle();\n return;\n }\n this.#logger.warn(`Cartesia STT WebSocket closed: ${code} ${reason.toString()}`);\n settle(\n new APIConnectionError({\n message: `Cartesia STT connection closed unexpectedly (code=${code})`,\n options: { retryable: true },\n }),\n );\n });\n\n ws.on('error', (err) => {\n if (this.closed) {\n settle();\n return;\n }\n settle(new APIConnectionError({ message: err.message, options: { retryable: true } }));\n });\n });\n }\n\n #processStreamEvent(data: Exclude<STTEventMessage, { type: 'error' }>) {\n if (data.request_id) {\n this.#requestId = data.request_id;\n }\n\n switch (data.type) {\n case 'connected':\n return;\n\n case 'turn.start': {\n if (this.#speaking) return;\n this.#speaking = true;\n this.#currentTranscript = '';\n if (!this.queue.closed)\n this.queue.put({\n type: stt.SpeechEventType.START_OF_SPEECH,\n requestId: this.#requestId,\n });\n this.#logger.debug('Cartesia STT START_OF_SPEECH');\n return;\n }\n\n case 'turn.update': {\n if (!this.#speaking) return;\n const transcript = data.transcript;\n if (!transcript) return;\n // Only emit interim updates when the cumulative transcript actually\n // changed; this avoids canceling preflight generation needlessly.\n if (this.#currentTranscript === transcript) return;\n this.#currentTranscript = transcript;\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, transcript);\n return;\n }\n\n case 'turn.eager_end': {\n if (!this.#speaking) return;\n const transcript = data.transcript || this.#currentTranscript;\n if (!transcript) return;\n this.#currentTranscript = transcript;\n this.#sendTranscriptEvent(stt.SpeechEventType.PREFLIGHT_TRANSCRIPT, transcript);\n return;\n }\n\n case 'turn.resume':\n // turn.resume has no transcript; re-emit the latest cumulative transcript\n // as an interim event so the pipeline cancels the pending preflight.\n if (!this.#speaking || !this.#currentTranscript) return;\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, this.#currentTranscript);\n return;\n\n case 'turn.end': {\n if (!this.#speaking) return;\n const transcript = data.transcript || this.#currentTranscript;\n\n this.#sendTranscriptEvent(stt.SpeechEventType.FINAL_TRANSCRIPT, transcript);\n\n this.#speaking = false;\n if (!this.queue.closed)\n this.queue.put({\n type: stt.SpeechEventType.END_OF_SPEECH,\n requestId: this.#requestId,\n });\n this.#logger.debug('Cartesia STT END_OF_SPEECH');\n this.#currentTranscript = '';\n\n if (this.#speechDuration > 0 && !this.queue.closed) {\n this.queue.put({\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: { audioDuration: this.#speechDuration },\n });\n this.#speechDuration = 0;\n }\n\n return;\n }\n\n default:\n this.#logger.warn('received unexpected message from Cartesia STT', { data });\n }\n }\n\n #sendTranscriptEvent(eventType: stt.SpeechEventType, transcript: string) {\n if (this.queue.closed) return;\n this.queue.put({\n type: eventType,\n requestId: this.#requestId,\n alternatives: [\n {\n language: asLanguageCode(this.#opts.language),\n text: transcript,\n startTime: 0,\n endTime: 0,\n confidence: 0,\n },\n ],\n });\n }\n\n #getCartesiaUrl(): string {\n // The Cartesia /stt/turns/websocket endpoint only accepts model, sample_rate\n // and encoding — there is no `language` query param. Language selection is\n // expressed through the model (ink-2 for English, ink-whisper otherwise),\n // so #opts.language is used only to tag emitted transcripts.\n const params = new URLSearchParams({\n model: this.#opts.model,\n sample_rate: this.#opts.sampleRate.toString(),\n encoding: AUDIO_ENCODING,\n });\n\n const wsBase = this.#opts.baseUrl.replace(/^http/, 'ws');\n return `${wsBase}/stt/turns/websocket?${params.toString()}`;\n }\n\n override close() {\n super.close();\n this.#ws?.close();\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBAWO;AAGP,gBAA0B;AAC1B,oBAA8C;AAE9C,MAAM,uBAAuB;AAC7B,MAAM,iBAAiB;AACvB,MAAM,oBAAoB;AAC1B,MAAM,cAAc;AACpB,MAAM,mBAAmB;AACzB,MAAM,wBAAwB;AA0H9B,MAAM,oBAAoB;AAAA,EACxB,OAAO;AAAA;AAAA,EAEP,YAAY;AAAA;AAAA,EAEZ,sBAAsB;AAAA,EACtB,SAAS;AAAA,EACT,UAAU;AACZ;AAEA,SAAS,gBAAgB,MAAkB,UAA2C;AACpF,SAAO;AAAA,IACL,QAAQ,SAAS,UAAU,KAAK;AAAA,IAChC,SAAS,SAAS,WAAW,KAAK;AAAA,IAClC,OAAO,SAAS,SAAS,KAAK;AAAA,IAC9B,YAAY,SAAS,cAAc,KAAK;AAAA,IACxC,sBAAsB,SAAS,wBAAwB,KAAK;AAAA,IAC5D,UACE,SAAS,aAAa,aAAY,iCAAkB,SAAS,QAAQ,IAAI,KAAK;AAAA,EAClF;AACF;AAOA,SAAS,gBAAgB,UAA4B;AACnD,aAAO,+BAAgB,QAAQ,MAAM,OAAO,UAAU;AACxD;AAyBO,MAAM,YAAY,kBAAI,IAAI;AAAA,EAC/B;AAAA,EAEA,YAAY,OAA4B,CAAC,GAAG;AAC1C,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB;AAAA,MAChB,mBAAmB;AAAA,MACnB,aAAa;AAAA,IACf,CAAC;AAED,UAAM,SAAS,KAAK,UAAU,QAAQ,IAAI;AAC1C,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ,gBAAgB,EAAE,GAAG,mBAAmB,OAAO,GAAG,IAAI;AAInE,QAAI,KAAK,UAAU,QAAW;AAC5B,WAAK,MAAM,QAAQ,gBAAgB,KAAK,MAAM,QAAQ;AAAA,IACxD;AAAA,EACF;AAAA,EAEA,IAAa,QAAgB;AAC3B,WAAO;AAAA,EACT;AAAA,EAEA,IAAa,QAAgB;AAC3B,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAa,WAAmB;AAC9B,WAAO;AAAA,EACT;AAAA,EAEA,MAAyB,aAAuC;AAC9D,UAAM,IAAI,MAAM,uEAAuE;AAAA,EACzF;AAAA,EAES,OAAO,SAA6D;AAC3E,WAAO,IAAI,aAAa,MAAM,KAAK,OAAO,mCAAS,WAAW;AAAA,EAChE;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,gBAAgB,KAAK,OAAO,IAAI;AAK7C,QAAI,KAAK,aAAa,UAAa,KAAK,UAAU,QAAW;AAC3D,WAAK,MAAM,QAAQ,gBAAgB,KAAK,MAAM,QAAQ;AAAA,IACxD;AAAA,EACF;AACF;AAEO,MAAM,qBAAqB,kBAAI,aAAa;AAAA,EACjD;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,MAAwB;AAAA,EACxB;AAAA,EACA,YAAY;AAAA,EACZ,qBAAqB;AAAA,EACrB,kBAAkB;AAAA,EAClB,aAAa;AAAA,EAEb,YAAY,aAAkB,MAAkB,aAAiC;AAC/E,UAAM,aAAa,KAAK,YAAY,WAAW;AAC/C,SAAK,QAAQ,EAAE,GAAG,KAAK;AAAA,EACzB;AAAA,EAEA,IAAa,QAAgB;AAC3B,WAAO;AAAA,EACT;AAAA,EAEA,MAAyB,MAAM;AAvRjC;AAwRI,QAAI,KAAK,QAAQ;AACf,WAAK,MAAM;AACX;AAAA,IACF;AAEA,SAAK,YAAY;AACjB,SAAK,qBAAqB;AAC1B,SAAK,kBAAkB;AACvB,SAAK,aAAa;AAElB,QAAI;AACJ,UAAM,kBAAkB,IAAI,gBAAgB;AAE5C,QAAI;AACF,YAAM,MAAM,KAAK,gBAAgB;AACjC,WAAK,QAAQ,MAAM,+BAA+B,GAAG,EAAE;AAEvD,YAAM,KAAK,IAAI,oBAAU,KAAK;AAAA,QAC5B,SAAS;AAAA,UACP,CAAC,oBAAoB,GAAG,KAAK,MAAM;AAAA,UACnC,CAAC,cAAc,GAAG;AAAA,QACpB;AAAA,MACF,CAAC;AACD,WAAK,MAAM;AAKX,SAAG,KAAK,WAAW,CAAC,aAA8B;AAChD,cAAM,cAAc,SAAS,QAAQ,iBAAiB;AACtD,cAAM,YAAY,MAAM,QAAQ,WAAW,IAAI,YAAY,CAAC,IAAI;AAChE,YAAI,WAAW;AACb,eAAK,aAAa;AAClB,eAAK,QAAQ,MAAM,EAAE,mBAAmB,UAAU,GAAG,kCAAkC;AAAA,QACzF;AAAA,MACF,CAAC;AAED,YAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,cAAM,SAAS,MAAM;AACnB,aAAG,IAAI,SAAS,OAAO;AACvB,kBAAQ;AAAA,QACV;AACA,cAAM,UAAU,CAAC,QAAe;AAC9B,aAAG,IAAI,QAAQ,MAAM;AACrB,iBAAO,IAAI,iCAAmB,EAAE,SAAS,IAAI,SAAS,SAAS,EAAE,WAAW,KAAK,EAAE,CAAC,CAAC;AAAA,QACvF;AAEA,WAAG,KAAK,QAAQ,MAAM;AACtB,WAAG,KAAK,SAAS,OAAO;AAAA,MAC1B,CAAC;AAOD,UAAI;AACJ,YAAM,WAAW,CAAC,QAAiB;AACjC,YAAI,eAAe,OAAW,cAAa;AAC3C,wBAAgB,MAAM;AACtB,YAAI,GAAG,eAAe,oBAAU,KAAM,IAAG,MAAM;AAAA,MACjD;AACA,YAAM,cAAc,KAAK,UAAU,IAAI,gBAAgB,MAAM,EAAE,MAAM,QAAQ;AAC7E,YAAM,cAAc,KAAK,UAAU,EAAE,EAAE,MAAM,QAAQ;AAErD,0BAAoB,YAAY,MAAM;AACpC,YAAI;AACF,cAAI,GAAG,eAAe,oBAAU,MAAM;AACpC,eAAG,KAAK;AAAA,UACV;AAAA,QACF,QAAQ;AAAA,QAER;AAAA,MACF,GAAG,qBAAqB;AACxB,YAAM,QAAQ,IAAI,CAAC,aAAa,WAAW,CAAC;AAC5C,UAAI,eAAe,OAAW,OAAM;AAAA,IACtC,SAAS,OAAO;AACd,WAAK,QAAQ,MAAM,6BAA6B,EAAE,MAAM,CAAC;AACzD,YAAM;AAAA,IACR,UAAE;AACA,sBAAgB,MAAM;AACtB,UAAI,kBAAmB,eAAc,iBAAiB;AACtD,YAAI,UAAK,QAAL,mBAAU,gBAAe,oBAAU,MAAM;AAC3C,aAAK,IAAI,MAAM;AAAA,MACjB;AAAA,IACF;AAEA,SAAK,MAAM;AAAA,EACb;AAAA,EAEA,MAAM,UAAU,IAAe,aAA0B;AACvD,UAAM,kBAAkB,KAAK;AAAA,MAC1B,KAAK,MAAM,aAAa,KAAK,MAAM,uBAAwB;AAAA,IAC9D;AACA,UAAM,eAAe,IAAI,8BAAgB,KAAK,MAAM,YAAY,GAAG,eAAe;AAElF,QAAI,WAAW;AACf,UAAM,WAAW,KAAK,MAAM,OAAO,aAAa,EAAE;AAClD,UAAM,mBAAe,4BAAa,WAAW;AAE7C,WAAO,MAAM;AACX,YAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,SAAS,KAAK,GAAG,YAAY,CAAC;AAEjE,UAAI,WAAW,OAAW;AAE1B,UAAI,OAAO,MAAM;AACf,mBAAW;AAAA,MACb,OAAO;AACL,cAAM,OAAO,OAAO;AACpB,cAAM,SAAuB,CAAC;AAE9B,YAAI,SAAS,aAAa,gBAAgB;AACxC,iBAAO,KAAK,GAAG,aAAa,MAAM,CAAC;AAAA,QACrC,OAAO;AAIL,iBAAO,KAAK,GAAG,aAAa,MAAM,KAAK,IAAI,CAAC;AAAA,QAC9C;AAEA,mBAAW,SAAS,QAAQ;AAC1B,eAAK,uBAAmB,6CAA8B,KAAK;AAC3D,cAAI,GAAG,eAAe,oBAAU,MAAM;AACpC,eAAG,KAAK,MAAM,IAAI;AAAA,UACpB;AAAA,QACF;AAAA,MACF;AAEA,UAAI,SAAU;AAAA,IAChB;AAEA,QAAI,GAAG,eAAe,oBAAU,MAAM;AACpC,WAAK,aAAa;AAClB,WAAK,QAAQ,MAAM,uCAAuC;AAC1D,SAAG,KAAK,KAAK,UAAU,EAAE,MAAM,QAAQ,CAAC,CAAC;AAKzC,YAAM,aAAa,WAAW,MAAM;AAClC,aAAK,QAAQ;AAAA,UACX,qCAAqC,gBAAgB;AAAA,QACvD;AACA,YAAI,GAAG,eAAe,oBAAU,QAAQ,GAAG,eAAe,oBAAU,SAAS;AAC3E,aAAG,UAAU;AAAA,QACf;AAAA,MACF,GAAG,gBAAgB;AACnB,SAAG,KAAK,SAAS,MAAM,aAAa,UAAU,CAAC;AAAA,IACjD;AAAA,EACF;AAAA,EAEA,MAAM,UAAU,IAAe;AAC7B,QAAI,eAAkC;AAEtC,WAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,YAAM,SAAS,CAAC,QAAgB;AAC9B,YAAI,IAAK,QAAO,GAAG;AAAA,YACd,SAAQ;AAAA,MACf;AAEA,SAAG,GAAG,WAAW,CAAC,MAAc,aAAsB;AACpD,YAAI,UAAU;AACZ,eAAK,QAAQ,KAAK,sDAAsD;AACxE;AAAA,QACF;AACA,YAAI;AACJ,YAAI;AACF,gBAAM,KAAK,MAAM,KAAK,SAAS,CAAC;AAAA,QAClC,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,wCAAwC,EAAE,MAAM,CAAC;AACpE;AAAA,QACF;AACA,YAAI;AACF,cAAI,IAAI,SAAS,SAAS;AACxB,iBAAK,QAAQ,MAAM,0BAA0B,GAAG;AAGhD,gBAAI,IAAI,gBAAgB,UAAa,IAAI,eAAe,KAAK;AAI3D,6BAAe,IAAI,iCAAmB;AAAA,gBACpC,SAAS,IAAI,WAAW,IAAI;AAAA,gBAC5B,SAAS;AAAA,kBACP,WAAW;AAAA,gBACb;AAAA,cACF,CAAC;AACD,iBAAG,MAAM;AACT;AAAA,YACF;AAAA,UACF,OAAO;AACL,iBAAK,oBAAoB,GAAG;AAAA,UAC9B;AAAA,QACF,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,0CAA0C,EAAE,MAAM,CAAC;AAAA,QACxE;AAAA,MACF,CAAC;AAED,SAAG,GAAG,SAAS,CAAC,MAAM,WAAW;AAI/B,YAAI,KAAK,WAAW;AAClB,cAAI,KAAK,oBAAoB;AAC3B,iBAAK;AAAA,cACH,kBAAI,gBAAgB;AAAA,cACpB,KAAK;AAAA,YACP;AAAA,UACF;AACA,cAAI,CAAC,KAAK,MAAM;AACd,iBAAK,MAAM,IAAI;AAAA,cACb,MAAM,kBAAI,gBAAgB;AAAA,cAC1B,WAAW,KAAK;AAAA,YAClB,CAAC;AACH,eAAK,QAAQ,MAAM,4BAA4B;AAC/C,cAAI,KAAK,kBAAkB,KAAK,CAAC,KAAK,MAAM,QAAQ;AAClD,iBAAK,MAAM,IAAI;AAAA,cACb,MAAM,kBAAI,gBAAgB;AAAA,cAC1B,WAAW,KAAK;AAAA,cAChB,kBAAkB,EAAE,eAAe,KAAK,gBAAgB;AAAA,YAC1D,CAAC;AACD,iBAAK,kBAAkB;AAAA,UACzB;AACA,eAAK,YAAY;AACjB,eAAK,qBAAqB;AAAA,QAC5B;AAEA,YAAI,cAAc;AAChB,iBAAO,YAAY;AACnB;AAAA,QACF;AACA,YAAI,KAAK,cAAc,KAAK,QAAQ;AAClC,iBAAO;AACP;AAAA,QACF;AACA,aAAK,QAAQ,KAAK,kCAAkC,IAAI,IAAI,OAAO,SAAS,CAAC,EAAE;AAC/E;AAAA,UACE,IAAI,iCAAmB;AAAA,YACrB,SAAS,qDAAqD,IAAI;AAAA,YAClE,SAAS,EAAE,WAAW,KAAK;AAAA,UAC7B,CAAC;AAAA,QACH;AAAA,MACF,CAAC;AAED,SAAG,GAAG,SAAS,CAAC,QAAQ;AACtB,YAAI,KAAK,QAAQ;AACf,iBAAO;AACP;AAAA,QACF;AACA,eAAO,IAAI,iCAAmB,EAAE,SAAS,IAAI,SAAS,SAAS,EAAE,WAAW,KAAK,EAAE,CAAC,CAAC;AAAA,MACvF,CAAC;AAAA,IACH,CAAC;AAAA,EACH;AAAA,EAEA,oBAAoB,MAAmD;AACrE,QAAI,KAAK,YAAY;AACnB,WAAK,aAAa,KAAK;AAAA,IACzB;AAEA,YAAQ,KAAK,MAAM;AAAA,MACjB,KAAK;AACH;AAAA,MAEF,KAAK,cAAc;AACjB,YAAI,KAAK,UAAW;AACpB,aAAK,YAAY;AACjB,aAAK,qBAAqB;AAC1B,YAAI,CAAC,KAAK,MAAM;AACd,eAAK,MAAM,IAAI;AAAA,YACb,MAAM,kBAAI,gBAAgB;AAAA,YAC1B,WAAW,KAAK;AAAA,UAClB,CAAC;AACH,aAAK,QAAQ,MAAM,8BAA8B;AACjD;AAAA,MACF;AAAA,MAEA,KAAK,eAAe;AAClB,YAAI,CAAC,KAAK,UAAW;AACrB,cAAM,aAAa,KAAK;AACxB,YAAI,CAAC,WAAY;AAGjB,YAAI,KAAK,uBAAuB,WAAY;AAC5C,aAAK,qBAAqB;AAC1B,aAAK,qBAAqB,kBAAI,gBAAgB,oBAAoB,UAAU;AAC5E;AAAA,MACF;AAAA,MAEA,KAAK,kBAAkB;AACrB,YAAI,CAAC,KAAK,UAAW;AACrB,cAAM,aAAa,KAAK,cAAc,KAAK;AAC3C,YAAI,CAAC,WAAY;AACjB,aAAK,qBAAqB;AAC1B,aAAK,qBAAqB,kBAAI,gBAAgB,sBAAsB,UAAU;AAC9E;AAAA,MACF;AAAA,MAEA,KAAK;AAGH,YAAI,CAAC,KAAK,aAAa,CAAC,KAAK,mBAAoB;AACjD,aAAK,qBAAqB,kBAAI,gBAAgB,oBAAoB,KAAK,kBAAkB;AACzF;AAAA,MAEF,KAAK,YAAY;AACf,YAAI,CAAC,KAAK,UAAW;AACrB,cAAM,aAAa,KAAK,cAAc,KAAK;AAE3C,aAAK,qBAAqB,kBAAI,gBAAgB,kBAAkB,UAAU;AAE1E,aAAK,YAAY;AACjB,YAAI,CAAC,KAAK,MAAM;AACd,eAAK,MAAM,IAAI;AAAA,YACb,MAAM,kBAAI,gBAAgB;AAAA,YAC1B,WAAW,KAAK;AAAA,UAClB,CAAC;AACH,aAAK,QAAQ,MAAM,4BAA4B;AAC/C,aAAK,qBAAqB;AAE1B,YAAI,KAAK,kBAAkB,KAAK,CAAC,KAAK,MAAM,QAAQ;AAClD,eAAK,MAAM,IAAI;AAAA,YACb,MAAM,kBAAI,gBAAgB;AAAA,YAC1B,WAAW,KAAK;AAAA,YAChB,kBAAkB,EAAE,eAAe,KAAK,gBAAgB;AAAA,UAC1D,CAAC;AACD,eAAK,kBAAkB;AAAA,QACzB;AAEA;AAAA,MACF;AAAA,MAEA;AACE,aAAK,QAAQ,KAAK,iDAAiD,EAAE,KAAK,CAAC;AAAA,IAC/E;AAAA,EACF;AAAA,EAEA,qBAAqB,WAAgC,YAAoB;AACvE,QAAI,KAAK,MAAM,OAAQ;AACvB,SAAK,MAAM,IAAI;AAAA,MACb,MAAM;AAAA,MACN,WAAW,KAAK;AAAA,MAChB,cAAc;AAAA,QACZ;AAAA,UACE,cAAU,8BAAe,KAAK,MAAM,QAAQ;AAAA,UAC5C,MAAM;AAAA,UACN,WAAW;AAAA,UACX,SAAS;AAAA,UACT,YAAY;AAAA,QACd;AAAA,MACF;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,kBAA0B;AAKxB,UAAM,SAAS,IAAI,gBAAgB;AAAA,MACjC,OAAO,KAAK,MAAM;AAAA,MAClB,aAAa,KAAK,MAAM,WAAW,SAAS;AAAA,MAC5C,UAAU;AAAA,IACZ,CAAC;AAED,UAAM,SAAS,KAAK,MAAM,QAAQ,QAAQ,SAAS,IAAI;AACvD,WAAO,GAAG,MAAM,wBAAwB,OAAO,SAAS,CAAC;AAAA,EAC3D;AAAA,EAES,QAAQ;AAxoBnB;AAyoBI,UAAM,MAAM;AACZ,eAAK,QAAL,mBAAU;AAAA,EACZ;AACF;","names":[]}
|
package/dist/stt.d.cts
CHANGED
|
@@ -6,6 +6,7 @@ export type STTOptions = {
|
|
|
6
6
|
sampleRate: number;
|
|
7
7
|
baseUrl: string;
|
|
8
8
|
audioChunkDurationMS: number;
|
|
9
|
+
language: string;
|
|
9
10
|
};
|
|
10
11
|
/**
|
|
11
12
|
* Cartesia speech to text.
|
|
@@ -40,6 +41,7 @@ export declare class STT extends stt.STT {
|
|
|
40
41
|
stream(options?: {
|
|
41
42
|
connOptions?: APIConnectOptions;
|
|
42
43
|
}): SpeechStream;
|
|
44
|
+
updateOptions(opts: Partial<STTOptions>): void;
|
|
43
45
|
}
|
|
44
46
|
export declare class SpeechStream extends stt.SpeechStream {
|
|
45
47
|
#private;
|
package/dist/stt.d.ts
CHANGED
|
@@ -6,6 +6,7 @@ export type STTOptions = {
|
|
|
6
6
|
sampleRate: number;
|
|
7
7
|
baseUrl: string;
|
|
8
8
|
audioChunkDurationMS: number;
|
|
9
|
+
language: string;
|
|
9
10
|
};
|
|
10
11
|
/**
|
|
11
12
|
* Cartesia speech to text.
|
|
@@ -40,6 +41,7 @@ export declare class STT extends stt.STT {
|
|
|
40
41
|
stream(options?: {
|
|
41
42
|
connOptions?: APIConnectOptions;
|
|
42
43
|
}): SpeechStream;
|
|
44
|
+
updateOptions(opts: Partial<STTOptions>): void;
|
|
43
45
|
}
|
|
44
46
|
export declare class SpeechStream extends stt.SpeechStream {
|
|
45
47
|
#private;
|
package/dist/stt.d.ts.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"stt.d.ts","sourceRoot":"","sources":["../src/stt.ts"],"names":[],"mappings":"AAGA,OAAO,EACL,KAAK,iBAAiB,
|
|
1
|
+
{"version":3,"file":"stt.d.ts","sourceRoot":"","sources":["../src/stt.ts"],"names":[],"mappings":"AAGA,OAAO,EACL,KAAK,iBAAiB,EAQtB,GAAG,EAEJ,MAAM,iBAAiB,CAAC;AAIzB,OAAO,EAAkB,KAAK,QAAQ,EAAE,MAAM,aAAa,CAAC;AAuH5D,MAAM,MAAM,UAAU,GAAG;IACvB,MAAM,EAAE,MAAM,CAAC;IAEf,KAAK,EAAE,QAAQ,GAAG,CAAC,MAAM,GAAG,EAAE,CAAC,CAAC;IAChC,UAAU,EAAE,MAAM,CAAC;IACnB,OAAO,EAAE,MAAM,CAAC;IAChB,oBAAoB,EAAE,MAAM,CAAC;IAC7B,QAAQ,EAAE,MAAM,CAAC;CAClB,CAAC;AAiCF;;;;;;;;;;;;;;;;;;;;;;GAsBG;AACH,qBAAa,GAAI,SAAQ,GAAG,CAAC,GAAG;;gBAGlB,IAAI,GAAE,OAAO,CAAC,UAAU,CAAM;IAwB1C,IAAa,KAAK,IAAI,MAAM,CAE3B;IAED,IAAa,KAAK,IAAI,MAAM,CAE3B;IAED,IAAa,QAAQ,IAAI,MAAM,CAE9B;cAEwB,UAAU,IAAI,OAAO,CAAC,GAAG,CAAC,WAAW,CAAC;IAItD,MAAM,CAAC,OAAO,CAAC,EAAE;QAAE,WAAW,CAAC,EAAE,iBAAiB,CAAA;KAAE,GAAG,YAAY;IAI5E,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,UAAU,CAAC;CAUxC;AAED,qBAAa,YAAa,SAAQ,GAAG,CAAC,YAAY;;gBAUpC,WAAW,EAAE,GAAG,EAAE,IAAI,EAAE,UAAU,EAAE,WAAW,CAAC,EAAE,iBAAiB;IAK/E,IAAa,KAAK,IAAI,MAAM,CAE3B;cAEwB,GAAG;IAiXnB,KAAK;CAIf"}
|
package/dist/stt.js
CHANGED
|
@@ -3,7 +3,9 @@ import {
|
|
|
3
3
|
AudioByteStream,
|
|
4
4
|
asLanguageCode,
|
|
5
5
|
calculateAudioDurationSeconds,
|
|
6
|
+
getBaseLanguage,
|
|
6
7
|
log,
|
|
8
|
+
normalizeLanguage,
|
|
7
9
|
stt,
|
|
8
10
|
waitForAbort
|
|
9
11
|
} from "@livekit/agents";
|
|
@@ -21,7 +23,8 @@ const defaultSTTOptions = {
|
|
|
21
23
|
sampleRate: 16e3,
|
|
22
24
|
/** recommended default */
|
|
23
25
|
audioChunkDurationMS: 160,
|
|
24
|
-
baseUrl: "https://api.cartesia.ai"
|
|
26
|
+
baseUrl: "https://api.cartesia.ai",
|
|
27
|
+
language: "en"
|
|
25
28
|
};
|
|
26
29
|
function mergeSTTOptions(base, override) {
|
|
27
30
|
return {
|
|
@@ -29,9 +32,13 @@ function mergeSTTOptions(base, override) {
|
|
|
29
32
|
baseUrl: override.baseUrl ?? base.baseUrl,
|
|
30
33
|
model: override.model ?? base.model,
|
|
31
34
|
sampleRate: override.sampleRate ?? base.sampleRate,
|
|
32
|
-
audioChunkDurationMS: override.audioChunkDurationMS ?? base.audioChunkDurationMS
|
|
35
|
+
audioChunkDurationMS: override.audioChunkDurationMS ?? base.audioChunkDurationMS,
|
|
36
|
+
language: override.language !== void 0 ? normalizeLanguage(override.language) : base.language
|
|
33
37
|
};
|
|
34
38
|
}
|
|
39
|
+
function resolveSTTModel(language) {
|
|
40
|
+
return getBaseLanguage(language) === "en" ? "ink-2" : "ink-whisper";
|
|
41
|
+
}
|
|
35
42
|
class STT extends stt.STT {
|
|
36
43
|
#opts;
|
|
37
44
|
constructor(opts = {}) {
|
|
@@ -48,6 +55,9 @@ class STT extends stt.STT {
|
|
|
48
55
|
);
|
|
49
56
|
}
|
|
50
57
|
this.#opts = mergeSTTOptions({ ...defaultSTTOptions, apiKey }, opts);
|
|
58
|
+
if (opts.model === void 0) {
|
|
59
|
+
this.#opts.model = resolveSTTModel(this.#opts.language);
|
|
60
|
+
}
|
|
51
61
|
}
|
|
52
62
|
get label() {
|
|
53
63
|
return "cartesia.STT";
|
|
@@ -64,6 +74,12 @@ class STT extends stt.STT {
|
|
|
64
74
|
stream(options) {
|
|
65
75
|
return new SpeechStream(this, this.#opts, options == null ? void 0 : options.connOptions);
|
|
66
76
|
}
|
|
77
|
+
updateOptions(opts) {
|
|
78
|
+
this.#opts = mergeSTTOptions(this.#opts, opts);
|
|
79
|
+
if (opts.language !== void 0 && opts.model === void 0) {
|
|
80
|
+
this.#opts.model = resolveSTTModel(this.#opts.language);
|
|
81
|
+
}
|
|
82
|
+
}
|
|
67
83
|
}
|
|
68
84
|
class SpeechStream extends stt.SpeechStream {
|
|
69
85
|
#opts;
|
|
@@ -317,7 +333,7 @@ class SpeechStream extends stt.SpeechStream {
|
|
|
317
333
|
}
|
|
318
334
|
case "turn.eager_end": {
|
|
319
335
|
if (!this.#speaking) return;
|
|
320
|
-
const transcript = data.transcript;
|
|
336
|
+
const transcript = data.transcript || this.#currentTranscript;
|
|
321
337
|
if (!transcript) return;
|
|
322
338
|
this.#currentTranscript = transcript;
|
|
323
339
|
this.#sendTranscriptEvent(stt.SpeechEventType.PREFLIGHT_TRANSCRIPT, transcript);
|
|
@@ -329,7 +345,7 @@ class SpeechStream extends stt.SpeechStream {
|
|
|
329
345
|
return;
|
|
330
346
|
case "turn.end": {
|
|
331
347
|
if (!this.#speaking) return;
|
|
332
|
-
const transcript = data.transcript;
|
|
348
|
+
const transcript = data.transcript || this.#currentTranscript;
|
|
333
349
|
this.#sendTranscriptEvent(stt.SpeechEventType.FINAL_TRANSCRIPT, transcript);
|
|
334
350
|
this.#speaking = false;
|
|
335
351
|
if (!this.queue.closed)
|
|
@@ -360,8 +376,7 @@ class SpeechStream extends stt.SpeechStream {
|
|
|
360
376
|
requestId: this.#requestId,
|
|
361
377
|
alternatives: [
|
|
362
378
|
{
|
|
363
|
-
|
|
364
|
-
language: asLanguageCode("en"),
|
|
379
|
+
language: asLanguageCode(this.#opts.language),
|
|
365
380
|
text: transcript,
|
|
366
381
|
startTime: 0,
|
|
367
382
|
endTime: 0,
|
package/dist/stt.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2026 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n APIConnectionError,\n AudioByteStream,\n asLanguageCode,\n calculateAudioDurationSeconds,\n log,\n stt,\n waitForAbort,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { IncomingMessage } from 'node:http';\nimport { WebSocket } from 'ws';\nimport { AUDIO_ENCODING, type STTModel } from './models.js';\n\nconst AUTHORIZATION_HEADER = 'X-API-Key';\nconst VERSION_HEADER = 'Cartesia-Version';\nconst REQUEST_ID_HEADER = 'cartesia-request-id';\nconst API_VERSION = '2026-03-01';\nconst DRAIN_TIMEOUT_MS = 5000;\nconst KEEPALIVE_INTERVAL_MS = 30000;\n\n/**\n * Fires once when the WebSocket connection is established.\n *\n * You do not need to wait for this event before sending audio.\n */\ntype STTConnectedEvent = {\n type: 'connected';\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\n/**\n * Model predicts {@link stt.SpeechEventType.START_OF_SPEECH}.\n */\ntype STTTurnStartEvent = {\n type: 'turn.start';\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\n/**\n * Fires repeatedly as the model transcribes the current user turn.\n *\n * Can be used for {@link stt.SpeechEventType.INTERIM_TRANSCRIPT}.\n */\ntype STTTurnUpdateEvent = {\n type: 'turn.update';\n /**\n * Cumulative text for the current turn, i.e. the full text transcribed so far in this turn, not a delta.\n */\n transcript: string;\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\n/**\n * Fires when the model predicts that the user might be done speaking.\n *\n * Can be used for {@link stt.SpeechEventType.PREFLIGHT_TRANSCRIPT}.\n */\ntype STTTurnEagerEndEvent = {\n type: 'turn.eager_end';\n /**\n * Cumulative text for the current turn, i.e. the full text transcribed so far in this turn, not a delta.\n */\n transcript: string;\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\n/**\n * Fires after {@link STTTurnEagerEndEvent} if the user turn has not actually ended.\n */\ntype STTTurnResumeEvent = {\n type: 'turn.resume';\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\n/**\n * Marks the end of a user turn.\n *\n * This is used for {@link stt.SpeechEventType.END_OF_SPEECH} and {@link stt.SpeechEventType.FINAL_TRANSCRIPT}.\n */\ntype STTTurnEndEvent = {\n type: 'turn.end';\n /**\n * Cumulative text for the current turn, i.e. the full text transcribed so far in this turn, not a delta.\n */\n transcript: string;\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\ntype STTErrorEvent = {\n type: 'error';\n error_code?: string;\n status_code?: number;\n title?: string;\n message?: string;\n doc_url?: string;\n request_id?: string;\n};\n\n/**\n * Server-sent event on the `/stt/turns/websocket` endpoint.\n *\n * See https://docs.cartesia.ai/api-reference/stt/turns/websocket.\n */\ntype STTEventMessage =\n | STTConnectedEvent\n | STTTurnStartEvent\n | STTTurnUpdateEvent\n | STTTurnEagerEndEvent\n | STTTurnResumeEvent\n | STTTurnEndEvent\n | STTErrorEvent;\n\nexport type STTOptions = {\n apiKey: string;\n // eslint-disable-next-line @typescript-eslint/ban-types\n model: STTModel | (string & {});\n sampleRate: number;\n baseUrl: string;\n audioChunkDurationMS: number;\n};\n\nconst defaultSTTOptions = {\n model: 'ink-2' satisfies STTModel,\n /** recommended default */\n sampleRate: 16_000,\n /** recommended default */\n audioChunkDurationMS: 160,\n baseUrl: 'https://api.cartesia.ai',\n};\n\nfunction mergeSTTOptions(base: STTOptions, override: Partial<STTOptions>): STTOptions {\n return {\n apiKey: override.apiKey ?? base.apiKey,\n baseUrl: override.baseUrl ?? base.baseUrl,\n model: override.model ?? base.model,\n sampleRate: override.sampleRate ?? base.sampleRate,\n audioChunkDurationMS: override.audioChunkDurationMS ?? base.audioChunkDurationMS,\n };\n}\n\n/**\n * Cartesia speech to text.\n *\n * Supports:\n * - Streaming\n * - Turn detection\n * - Interim results\n *\n * See https://docs.cartesia.ai/build-with-cartesia/stt-models/latest\n *\n * @example\n * ```typescript\n * import { voice } from '@livekit/agents';\n * import { STT, TTS } from '@livekit/agents-plugin-cartesia';\n *\n * const session = new voice.AgentSession({\n * stt: new STT(),\n * llm: new LLM(), // choose your favorite LLM\n * tts: new TTS(),\n * turnHandling: { turnDetection: 'stt' },\n * });\n * ```\n */\nexport class STT extends stt.STT {\n #opts: STTOptions;\n\n constructor(opts: Partial<STTOptions> = {}) {\n super({\n streaming: true,\n interimResults: true,\n alignedTranscript: false,\n diarization: false,\n });\n\n const apiKey = opts.apiKey ?? process.env.CARTESIA_API_KEY;\n if (!apiKey) {\n throw new Error(\n 'Cartesia API key is required, whether as an argument or as $CARTESIA_API_KEY',\n );\n }\n\n this.#opts = mergeSTTOptions({ ...defaultSTTOptions, apiKey }, opts);\n }\n\n override get label(): string {\n return 'cartesia.STT';\n }\n\n override get model(): string {\n return this.#opts.model;\n }\n\n override get provider(): string {\n return 'Cartesia';\n }\n\n protected override async _recognize(): Promise<stt.SpeechEvent> {\n throw new Error('Cartesia STT does not support batch recognition, use stream() instead');\n }\n\n override stream(options?: { connOptions?: APIConnectOptions }): SpeechStream {\n return new SpeechStream(this, this.#opts, options?.connOptions);\n }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n #opts: STTOptions;\n #logger = log();\n #ws: WebSocket | null = null;\n #requestId: string | undefined;\n #speaking = false;\n #currentTranscript = '';\n #speechDuration = 0;\n #closingWs = false;\n\n constructor(sttInstance: STT, opts: STTOptions, connOptions?: APIConnectOptions) {\n super(sttInstance, opts.sampleRate, connOptions);\n this.#opts = { ...opts };\n }\n\n override get label(): string {\n return 'cartesia.SpeechStream';\n }\n\n protected override async run() {\n if (this.closed) {\n this.close();\n return;\n }\n\n this.#speaking = false;\n this.#currentTranscript = '';\n this.#speechDuration = 0;\n this.#closingWs = false;\n\n let keepaliveInterval: NodeJS.Timeout | undefined;\n const abortController = new AbortController();\n\n try {\n const url = this.#getCartesiaUrl();\n this.#logger.debug(`Connecting to Cartesia STT: ${url}`);\n\n const ws = new WebSocket(url, {\n headers: {\n [AUTHORIZATION_HEADER]: this.#opts.apiKey,\n [VERSION_HEADER]: API_VERSION,\n },\n });\n this.#ws = ws;\n\n // Cartesia returns the request id on the WS upgrade response, before any\n // turn events are sent. Capture it so logs and metrics emitted on the\n // connection (and any pre-`turn.start` errors) are attributable.\n ws.once('upgrade', (response: IncomingMessage) => {\n const headerValue = response.headers[REQUEST_ID_HEADER];\n const requestId = Array.isArray(headerValue) ? headerValue[0] : headerValue;\n if (requestId) {\n this.#requestId = requestId;\n this.#logger.debug({ cartesiaRequestId: requestId }, 'Cartesia STT WebSocket connected');\n }\n });\n\n await new Promise<void>((resolve, reject) => {\n const onOpen = () => {\n ws.off('error', onError);\n resolve();\n };\n const onError = (err: Error) => {\n ws.off('open', onOpen);\n reject(new APIConnectionError({ message: err.message, options: { retryable: true } }));\n };\n\n ws.once('open', onOpen);\n ws.once('error', onError);\n });\n\n // If one task fails, abort its peer and close the WS so the peer can\n // exit before we re-throw. Otherwise a dangling task would survive\n // into the next retry attempt (sendTask blocked on input.next() would\n // steal frames; recvTask's close handler could still push events to\n // the queue and surface an unhandled rejection).\n let firstError: unknown;\n const stopPeer = (err: unknown) => {\n if (firstError === undefined) firstError = err;\n abortController.abort();\n if (ws.readyState === WebSocket.OPEN) ws.close();\n };\n const sendPromise = this.#sendTask(ws, abortController.signal).catch(stopPeer);\n const recvPromise = this.#recvTask(ws).catch(stopPeer);\n\n keepaliveInterval = setInterval(() => {\n try {\n if (ws.readyState === WebSocket.OPEN) {\n ws.ping();\n }\n } catch {\n // ignore\n }\n }, KEEPALIVE_INTERVAL_MS);\n await Promise.all([sendPromise, recvPromise]);\n if (firstError !== undefined) throw firstError;\n } catch (error) {\n this.#logger.error('Cartesia STT stream error', { error });\n throw error;\n } finally {\n abortController.abort();\n if (keepaliveInterval) clearInterval(keepaliveInterval);\n if (this.#ws?.readyState === WebSocket.OPEN) {\n this.#ws.close();\n }\n }\n\n this.close();\n }\n\n async #sendTask(ws: WebSocket, abortSignal: AbortSignal) {\n const samplesPerChunk = Math.floor(\n (this.#opts.sampleRate * this.#opts.audioChunkDurationMS) / 1000,\n );\n const audioBstream = new AudioByteStream(this.#opts.sampleRate, 1, samplesPerChunk);\n\n let hasEnded = false;\n const iterator = this.input[Symbol.asyncIterator]();\n const abortPromise = waitForAbort(abortSignal);\n\n while (true) {\n const result = await Promise.race([iterator.next(), abortPromise]);\n\n if (result === undefined) return; // aborted\n\n if (result.done) {\n hasEnded = true;\n } else {\n const data = result.value;\n const frames: AudioFrame[] = [];\n\n if (data === SpeechStream.FLUSH_SENTINEL) {\n frames.push(...audioBstream.flush());\n } else {\n // Pass the typed array directly so AudioByteStream uses byteOffset/\n // byteLength — `.buffer` would include foreign bytes when rtc-node\n // hands us a view over a pooled allocator.\n frames.push(...audioBstream.write(data.data));\n }\n\n for (const frame of frames) {\n this.#speechDuration += calculateAudioDurationSeconds(frame);\n if (ws.readyState === WebSocket.OPEN) {\n ws.send(frame.data);\n }\n }\n }\n\n if (hasEnded) break;\n }\n\n if (ws.readyState === WebSocket.OPEN) {\n this.#closingWs = true;\n this.#logger.debug('Sending close message to Cartesia STT');\n ws.send(JSON.stringify({ type: 'close' }));\n\n // After `close`, the server should flush any pending turn events and\n // close the socket. If it hangs, force the issue rather than letting\n // the stream wait forever.\n const drainTimer = setTimeout(() => {\n this.#logger.warn(\n `Cartesia STT did not close within ${DRAIN_TIMEOUT_MS}ms after done; forcing close`,\n );\n if (ws.readyState === WebSocket.OPEN || ws.readyState === WebSocket.CLOSING) {\n ws.terminate();\n }\n }, DRAIN_TIMEOUT_MS);\n ws.once('close', () => clearTimeout(drainTimer));\n }\n }\n\n async #recvTask(ws: WebSocket) {\n let pendingError: Error | undefined = undefined;\n\n return new Promise<void>((resolve, reject) => {\n const settle = (err?: Error) => {\n if (err) reject(err);\n else resolve();\n };\n\n ws.on('message', (data: Buffer, isBinary: boolean) => {\n if (isBinary) {\n this.#logger.warn('Received unexpected binary message from Cartesia STT');\n return;\n }\n let msg: STTEventMessage;\n try {\n msg = JSON.parse(data.toString());\n } catch (error) {\n this.#logger.error('Failed to parse Cartesia STT message', { error });\n return;\n }\n try {\n if (msg.type === 'error') {\n this.#logger.error('Cartesia sent an error', msg);\n\n // do not close the websocket on bad requests since that may be caused by invalid messages\n if (msg.status_code === undefined || msg.status_code >= 500) {\n // Defer until the WS is fully closed so we don't race the `close`\n // handler with a rejected `recvTask`. Close the socket explicitly\n // in case the server doesn't, otherwise recv would hang.\n pendingError = new APIConnectionError({\n message: msg.message || msg.title,\n options: {\n retryable: true,\n },\n });\n ws.close();\n return;\n }\n } else {\n this.#processStreamEvent(msg);\n }\n } catch (error) {\n this.#logger.error('Failed to process Cartesia STT message', { error });\n }\n });\n\n ws.on('close', (code, reason) => {\n // If a turn was in progress, close it out so consumers see a balanced\n // START_OF_SPEECH/END_OF_SPEECH pair and don't carry stale turn state\n // into the next reconnect.\n if (this.#speaking) {\n if (this.#currentTranscript) {\n this.#sendTranscriptEvent(\n stt.SpeechEventType.FINAL_TRANSCRIPT,\n this.#currentTranscript,\n );\n }\n if (!this.queue.closed)\n this.queue.put({\n type: stt.SpeechEventType.END_OF_SPEECH,\n requestId: this.#requestId,\n });\n this.#logger.debug('Cartesia STT END_OF_SPEECH');\n if (this.#speechDuration > 0 && !this.queue.closed) {\n this.queue.put({\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: { audioDuration: this.#speechDuration },\n });\n this.#speechDuration = 0;\n }\n this.#speaking = false;\n this.#currentTranscript = '';\n }\n\n if (pendingError) {\n settle(pendingError);\n return;\n }\n if (this.#closingWs || this.closed) {\n settle();\n return;\n }\n this.#logger.warn(`Cartesia STT WebSocket closed: ${code} ${reason.toString()}`);\n settle(\n new APIConnectionError({\n message: `Cartesia STT connection closed unexpectedly (code=${code})`,\n options: { retryable: true },\n }),\n );\n });\n\n ws.on('error', (err) => {\n if (this.closed) {\n settle();\n return;\n }\n settle(new APIConnectionError({ message: err.message, options: { retryable: true } }));\n });\n });\n }\n\n #processStreamEvent(data: Exclude<STTEventMessage, { type: 'error' }>) {\n if (data.request_id) {\n this.#requestId = data.request_id;\n }\n\n switch (data.type) {\n case 'connected':\n return;\n\n case 'turn.start': {\n if (this.#speaking) return;\n this.#speaking = true;\n this.#currentTranscript = '';\n if (!this.queue.closed)\n this.queue.put({\n type: stt.SpeechEventType.START_OF_SPEECH,\n requestId: this.#requestId,\n });\n this.#logger.debug('Cartesia STT START_OF_SPEECH');\n return;\n }\n\n case 'turn.update': {\n if (!this.#speaking) return;\n const transcript = data.transcript;\n if (!transcript) return;\n // Only emit interim updates when the cumulative transcript actually\n // changed; this avoids canceling preflight generation needlessly.\n if (this.#currentTranscript === transcript) return;\n this.#currentTranscript = transcript;\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, transcript);\n return;\n }\n\n case 'turn.eager_end': {\n if (!this.#speaking) return;\n const transcript = data.transcript;\n if (!transcript) return;\n this.#currentTranscript = transcript;\n this.#sendTranscriptEvent(stt.SpeechEventType.PREFLIGHT_TRANSCRIPT, transcript);\n return;\n }\n\n case 'turn.resume':\n // turn.resume has no transcript; re-emit the latest cumulative transcript\n // as an interim event so the pipeline cancels the pending preflight.\n if (!this.#speaking || !this.#currentTranscript) return;\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, this.#currentTranscript);\n return;\n\n case 'turn.end': {\n if (!this.#speaking) return;\n const transcript = data.transcript;\n\n this.#sendTranscriptEvent(stt.SpeechEventType.FINAL_TRANSCRIPT, transcript);\n\n this.#speaking = false;\n if (!this.queue.closed)\n this.queue.put({\n type: stt.SpeechEventType.END_OF_SPEECH,\n requestId: this.#requestId,\n });\n this.#logger.debug('Cartesia STT END_OF_SPEECH');\n this.#currentTranscript = '';\n\n if (this.#speechDuration > 0 && !this.queue.closed) {\n this.queue.put({\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: { audioDuration: this.#speechDuration },\n });\n this.#speechDuration = 0;\n }\n\n return;\n }\n\n default:\n this.#logger.warn('received unexpected message from Cartesia STT', { data });\n }\n }\n\n #sendTranscriptEvent(eventType: stt.SpeechEventType, transcript: string) {\n if (this.queue.closed) return;\n this.queue.put({\n type: eventType,\n requestId: this.#requestId,\n alternatives: [\n {\n // Cartesia STT only supports English at this time.\n language: asLanguageCode('en'),\n text: transcript,\n startTime: 0,\n endTime: 0,\n confidence: 0,\n },\n ],\n });\n }\n\n #getCartesiaUrl(): string {\n const params = new URLSearchParams({\n model: this.#opts.model,\n sample_rate: this.#opts.sampleRate.toString(),\n encoding: AUDIO_ENCODING,\n });\n\n const wsBase = this.#opts.baseUrl.replace(/^http/, 'ws');\n return `${wsBase}/stt/turns/websocket?${params.toString()}`;\n }\n\n override close() {\n super.close();\n this.#ws?.close();\n }\n}\n"],"mappings":"AAGA;AAAA,EAEE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAGP,SAAS,iBAAiB;AAC1B,SAAS,sBAAqC;AAE9C,MAAM,uBAAuB;AAC7B,MAAM,iBAAiB;AACvB,MAAM,oBAAoB;AAC1B,MAAM,cAAc;AACpB,MAAM,mBAAmB;AACzB,MAAM,wBAAwB;AAyH9B,MAAM,oBAAoB;AAAA,EACxB,OAAO;AAAA;AAAA,EAEP,YAAY;AAAA;AAAA,EAEZ,sBAAsB;AAAA,EACtB,SAAS;AACX;AAEA,SAAS,gBAAgB,MAAkB,UAA2C;AACpF,SAAO;AAAA,IACL,QAAQ,SAAS,UAAU,KAAK;AAAA,IAChC,SAAS,SAAS,WAAW,KAAK;AAAA,IAClC,OAAO,SAAS,SAAS,KAAK;AAAA,IAC9B,YAAY,SAAS,cAAc,KAAK;AAAA,IACxC,sBAAsB,SAAS,wBAAwB,KAAK;AAAA,EAC9D;AACF;AAyBO,MAAM,YAAY,IAAI,IAAI;AAAA,EAC/B;AAAA,EAEA,YAAY,OAA4B,CAAC,GAAG;AAC1C,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB;AAAA,MAChB,mBAAmB;AAAA,MACnB,aAAa;AAAA,IACf,CAAC;AAED,UAAM,SAAS,KAAK,UAAU,QAAQ,IAAI;AAC1C,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ,gBAAgB,EAAE,GAAG,mBAAmB,OAAO,GAAG,IAAI;AAAA,EACrE;AAAA,EAEA,IAAa,QAAgB;AAC3B,WAAO;AAAA,EACT;AAAA,EAEA,IAAa,QAAgB;AAC3B,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAa,WAAmB;AAC9B,WAAO;AAAA,EACT;AAAA,EAEA,MAAyB,aAAuC;AAC9D,UAAM,IAAI,MAAM,uEAAuE;AAAA,EACzF;AAAA,EAES,OAAO,SAA6D;AAC3E,WAAO,IAAI,aAAa,MAAM,KAAK,OAAO,mCAAS,WAAW;AAAA,EAChE;AACF;AAEO,MAAM,qBAAqB,IAAI,aAAa;AAAA,EACjD;AAAA,EACA,UAAU,IAAI;AAAA,EACd,MAAwB;AAAA,EACxB;AAAA,EACA,YAAY;AAAA,EACZ,qBAAqB;AAAA,EACrB,kBAAkB;AAAA,EAClB,aAAa;AAAA,EAEb,YAAY,aAAkB,MAAkB,aAAiC;AAC/E,UAAM,aAAa,KAAK,YAAY,WAAW;AAC/C,SAAK,QAAQ,EAAE,GAAG,KAAK;AAAA,EACzB;AAAA,EAEA,IAAa,QAAgB;AAC3B,WAAO;AAAA,EACT;AAAA,EAEA,MAAyB,MAAM;AAvPjC;AAwPI,QAAI,KAAK,QAAQ;AACf,WAAK,MAAM;AACX;AAAA,IACF;AAEA,SAAK,YAAY;AACjB,SAAK,qBAAqB;AAC1B,SAAK,kBAAkB;AACvB,SAAK,aAAa;AAElB,QAAI;AACJ,UAAM,kBAAkB,IAAI,gBAAgB;AAE5C,QAAI;AACF,YAAM,MAAM,KAAK,gBAAgB;AACjC,WAAK,QAAQ,MAAM,+BAA+B,GAAG,EAAE;AAEvD,YAAM,KAAK,IAAI,UAAU,KAAK;AAAA,QAC5B,SAAS;AAAA,UACP,CAAC,oBAAoB,GAAG,KAAK,MAAM;AAAA,UACnC,CAAC,cAAc,GAAG;AAAA,QACpB;AAAA,MACF,CAAC;AACD,WAAK,MAAM;AAKX,SAAG,KAAK,WAAW,CAAC,aAA8B;AAChD,cAAM,cAAc,SAAS,QAAQ,iBAAiB;AACtD,cAAM,YAAY,MAAM,QAAQ,WAAW,IAAI,YAAY,CAAC,IAAI;AAChE,YAAI,WAAW;AACb,eAAK,aAAa;AAClB,eAAK,QAAQ,MAAM,EAAE,mBAAmB,UAAU,GAAG,kCAAkC;AAAA,QACzF;AAAA,MACF,CAAC;AAED,YAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,cAAM,SAAS,MAAM;AACnB,aAAG,IAAI,SAAS,OAAO;AACvB,kBAAQ;AAAA,QACV;AACA,cAAM,UAAU,CAAC,QAAe;AAC9B,aAAG,IAAI,QAAQ,MAAM;AACrB,iBAAO,IAAI,mBAAmB,EAAE,SAAS,IAAI,SAAS,SAAS,EAAE,WAAW,KAAK,EAAE,CAAC,CAAC;AAAA,QACvF;AAEA,WAAG,KAAK,QAAQ,MAAM;AACtB,WAAG,KAAK,SAAS,OAAO;AAAA,MAC1B,CAAC;AAOD,UAAI;AACJ,YAAM,WAAW,CAAC,QAAiB;AACjC,YAAI,eAAe,OAAW,cAAa;AAC3C,wBAAgB,MAAM;AACtB,YAAI,GAAG,eAAe,UAAU,KAAM,IAAG,MAAM;AAAA,MACjD;AACA,YAAM,cAAc,KAAK,UAAU,IAAI,gBAAgB,MAAM,EAAE,MAAM,QAAQ;AAC7E,YAAM,cAAc,KAAK,UAAU,EAAE,EAAE,MAAM,QAAQ;AAErD,0BAAoB,YAAY,MAAM;AACpC,YAAI;AACF,cAAI,GAAG,eAAe,UAAU,MAAM;AACpC,eAAG,KAAK;AAAA,UACV;AAAA,QACF,QAAQ;AAAA,QAER;AAAA,MACF,GAAG,qBAAqB;AACxB,YAAM,QAAQ,IAAI,CAAC,aAAa,WAAW,CAAC;AAC5C,UAAI,eAAe,OAAW,OAAM;AAAA,IACtC,SAAS,OAAO;AACd,WAAK,QAAQ,MAAM,6BAA6B,EAAE,MAAM,CAAC;AACzD,YAAM;AAAA,IACR,UAAE;AACA,sBAAgB,MAAM;AACtB,UAAI,kBAAmB,eAAc,iBAAiB;AACtD,YAAI,UAAK,QAAL,mBAAU,gBAAe,UAAU,MAAM;AAC3C,aAAK,IAAI,MAAM;AAAA,MACjB;AAAA,IACF;AAEA,SAAK,MAAM;AAAA,EACb;AAAA,EAEA,MAAM,UAAU,IAAe,aAA0B;AACvD,UAAM,kBAAkB,KAAK;AAAA,MAC1B,KAAK,MAAM,aAAa,KAAK,MAAM,uBAAwB;AAAA,IAC9D;AACA,UAAM,eAAe,IAAI,gBAAgB,KAAK,MAAM,YAAY,GAAG,eAAe;AAElF,QAAI,WAAW;AACf,UAAM,WAAW,KAAK,MAAM,OAAO,aAAa,EAAE;AAClD,UAAM,eAAe,aAAa,WAAW;AAE7C,WAAO,MAAM;AACX,YAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,SAAS,KAAK,GAAG,YAAY,CAAC;AAEjE,UAAI,WAAW,OAAW;AAE1B,UAAI,OAAO,MAAM;AACf,mBAAW;AAAA,MACb,OAAO;AACL,cAAM,OAAO,OAAO;AACpB,cAAM,SAAuB,CAAC;AAE9B,YAAI,SAAS,aAAa,gBAAgB;AACxC,iBAAO,KAAK,GAAG,aAAa,MAAM,CAAC;AAAA,QACrC,OAAO;AAIL,iBAAO,KAAK,GAAG,aAAa,MAAM,KAAK,IAAI,CAAC;AAAA,QAC9C;AAEA,mBAAW,SAAS,QAAQ;AAC1B,eAAK,mBAAmB,8BAA8B,KAAK;AAC3D,cAAI,GAAG,eAAe,UAAU,MAAM;AACpC,eAAG,KAAK,MAAM,IAAI;AAAA,UACpB;AAAA,QACF;AAAA,MACF;AAEA,UAAI,SAAU;AAAA,IAChB;AAEA,QAAI,GAAG,eAAe,UAAU,MAAM;AACpC,WAAK,aAAa;AAClB,WAAK,QAAQ,MAAM,uCAAuC;AAC1D,SAAG,KAAK,KAAK,UAAU,EAAE,MAAM,QAAQ,CAAC,CAAC;AAKzC,YAAM,aAAa,WAAW,MAAM;AAClC,aAAK,QAAQ;AAAA,UACX,qCAAqC,gBAAgB;AAAA,QACvD;AACA,YAAI,GAAG,eAAe,UAAU,QAAQ,GAAG,eAAe,UAAU,SAAS;AAC3E,aAAG,UAAU;AAAA,QACf;AAAA,MACF,GAAG,gBAAgB;AACnB,SAAG,KAAK,SAAS,MAAM,aAAa,UAAU,CAAC;AAAA,IACjD;AAAA,EACF;AAAA,EAEA,MAAM,UAAU,IAAe;AAC7B,QAAI,eAAkC;AAEtC,WAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,YAAM,SAAS,CAAC,QAAgB;AAC9B,YAAI,IAAK,QAAO,GAAG;AAAA,YACd,SAAQ;AAAA,MACf;AAEA,SAAG,GAAG,WAAW,CAAC,MAAc,aAAsB;AACpD,YAAI,UAAU;AACZ,eAAK,QAAQ,KAAK,sDAAsD;AACxE;AAAA,QACF;AACA,YAAI;AACJ,YAAI;AACF,gBAAM,KAAK,MAAM,KAAK,SAAS,CAAC;AAAA,QAClC,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,wCAAwC,EAAE,MAAM,CAAC;AACpE;AAAA,QACF;AACA,YAAI;AACF,cAAI,IAAI,SAAS,SAAS;AACxB,iBAAK,QAAQ,MAAM,0BAA0B,GAAG;AAGhD,gBAAI,IAAI,gBAAgB,UAAa,IAAI,eAAe,KAAK;AAI3D,6BAAe,IAAI,mBAAmB;AAAA,gBACpC,SAAS,IAAI,WAAW,IAAI;AAAA,gBAC5B,SAAS;AAAA,kBACP,WAAW;AAAA,gBACb;AAAA,cACF,CAAC;AACD,iBAAG,MAAM;AACT;AAAA,YACF;AAAA,UACF,OAAO;AACL,iBAAK,oBAAoB,GAAG;AAAA,UAC9B;AAAA,QACF,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,0CAA0C,EAAE,MAAM,CAAC;AAAA,QACxE;AAAA,MACF,CAAC;AAED,SAAG,GAAG,SAAS,CAAC,MAAM,WAAW;AAI/B,YAAI,KAAK,WAAW;AAClB,cAAI,KAAK,oBAAoB;AAC3B,iBAAK;AAAA,cACH,IAAI,gBAAgB;AAAA,cACpB,KAAK;AAAA,YACP;AAAA,UACF;AACA,cAAI,CAAC,KAAK,MAAM;AACd,iBAAK,MAAM,IAAI;AAAA,cACb,MAAM,IAAI,gBAAgB;AAAA,cAC1B,WAAW,KAAK;AAAA,YAClB,CAAC;AACH,eAAK,QAAQ,MAAM,4BAA4B;AAC/C,cAAI,KAAK,kBAAkB,KAAK,CAAC,KAAK,MAAM,QAAQ;AAClD,iBAAK,MAAM,IAAI;AAAA,cACb,MAAM,IAAI,gBAAgB;AAAA,cAC1B,WAAW,KAAK;AAAA,cAChB,kBAAkB,EAAE,eAAe,KAAK,gBAAgB;AAAA,YAC1D,CAAC;AACD,iBAAK,kBAAkB;AAAA,UACzB;AACA,eAAK,YAAY;AACjB,eAAK,qBAAqB;AAAA,QAC5B;AAEA,YAAI,cAAc;AAChB,iBAAO,YAAY;AACnB;AAAA,QACF;AACA,YAAI,KAAK,cAAc,KAAK,QAAQ;AAClC,iBAAO;AACP;AAAA,QACF;AACA,aAAK,QAAQ,KAAK,kCAAkC,IAAI,IAAI,OAAO,SAAS,CAAC,EAAE;AAC/E;AAAA,UACE,IAAI,mBAAmB;AAAA,YACrB,SAAS,qDAAqD,IAAI;AAAA,YAClE,SAAS,EAAE,WAAW,KAAK;AAAA,UAC7B,CAAC;AAAA,QACH;AAAA,MACF,CAAC;AAED,SAAG,GAAG,SAAS,CAAC,QAAQ;AACtB,YAAI,KAAK,QAAQ;AACf,iBAAO;AACP;AAAA,QACF;AACA,eAAO,IAAI,mBAAmB,EAAE,SAAS,IAAI,SAAS,SAAS,EAAE,WAAW,KAAK,EAAE,CAAC,CAAC;AAAA,MACvF,CAAC;AAAA,IACH,CAAC;AAAA,EACH;AAAA,EAEA,oBAAoB,MAAmD;AACrE,QAAI,KAAK,YAAY;AACnB,WAAK,aAAa,KAAK;AAAA,IACzB;AAEA,YAAQ,KAAK,MAAM;AAAA,MACjB,KAAK;AACH;AAAA,MAEF,KAAK,cAAc;AACjB,YAAI,KAAK,UAAW;AACpB,aAAK,YAAY;AACjB,aAAK,qBAAqB;AAC1B,YAAI,CAAC,KAAK,MAAM;AACd,eAAK,MAAM,IAAI;AAAA,YACb,MAAM,IAAI,gBAAgB;AAAA,YAC1B,WAAW,KAAK;AAAA,UAClB,CAAC;AACH,aAAK,QAAQ,MAAM,8BAA8B;AACjD;AAAA,MACF;AAAA,MAEA,KAAK,eAAe;AAClB,YAAI,CAAC,KAAK,UAAW;AACrB,cAAM,aAAa,KAAK;AACxB,YAAI,CAAC,WAAY;AAGjB,YAAI,KAAK,uBAAuB,WAAY;AAC5C,aAAK,qBAAqB;AAC1B,aAAK,qBAAqB,IAAI,gBAAgB,oBAAoB,UAAU;AAC5E;AAAA,MACF;AAAA,MAEA,KAAK,kBAAkB;AACrB,YAAI,CAAC,KAAK,UAAW;AACrB,cAAM,aAAa,KAAK;AACxB,YAAI,CAAC,WAAY;AACjB,aAAK,qBAAqB;AAC1B,aAAK,qBAAqB,IAAI,gBAAgB,sBAAsB,UAAU;AAC9E;AAAA,MACF;AAAA,MAEA,KAAK;AAGH,YAAI,CAAC,KAAK,aAAa,CAAC,KAAK,mBAAoB;AACjD,aAAK,qBAAqB,IAAI,gBAAgB,oBAAoB,KAAK,kBAAkB;AACzF;AAAA,MAEF,KAAK,YAAY;AACf,YAAI,CAAC,KAAK,UAAW;AACrB,cAAM,aAAa,KAAK;AAExB,aAAK,qBAAqB,IAAI,gBAAgB,kBAAkB,UAAU;AAE1E,aAAK,YAAY;AACjB,YAAI,CAAC,KAAK,MAAM;AACd,eAAK,MAAM,IAAI;AAAA,YACb,MAAM,IAAI,gBAAgB;AAAA,YAC1B,WAAW,KAAK;AAAA,UAClB,CAAC;AACH,aAAK,QAAQ,MAAM,4BAA4B;AAC/C,aAAK,qBAAqB;AAE1B,YAAI,KAAK,kBAAkB,KAAK,CAAC,KAAK,MAAM,QAAQ;AAClD,eAAK,MAAM,IAAI;AAAA,YACb,MAAM,IAAI,gBAAgB;AAAA,YAC1B,WAAW,KAAK;AAAA,YAChB,kBAAkB,EAAE,eAAe,KAAK,gBAAgB;AAAA,UAC1D,CAAC;AACD,eAAK,kBAAkB;AAAA,QACzB;AAEA;AAAA,MACF;AAAA,MAEA;AACE,aAAK,QAAQ,KAAK,iDAAiD,EAAE,KAAK,CAAC;AAAA,IAC/E;AAAA,EACF;AAAA,EAEA,qBAAqB,WAAgC,YAAoB;AACvE,QAAI,KAAK,MAAM,OAAQ;AACvB,SAAK,MAAM,IAAI;AAAA,MACb,MAAM;AAAA,MACN,WAAW,KAAK;AAAA,MAChB,cAAc;AAAA,QACZ;AAAA;AAAA,UAEE,UAAU,eAAe,IAAI;AAAA,UAC7B,MAAM;AAAA,UACN,WAAW;AAAA,UACX,SAAS;AAAA,UACT,YAAY;AAAA,QACd;AAAA,MACF;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,kBAA0B;AACxB,UAAM,SAAS,IAAI,gBAAgB;AAAA,MACjC,OAAO,KAAK,MAAM;AAAA,MAClB,aAAa,KAAK,MAAM,WAAW,SAAS;AAAA,MAC5C,UAAU;AAAA,IACZ,CAAC;AAED,UAAM,SAAS,KAAK,MAAM,QAAQ,QAAQ,SAAS,IAAI;AACvD,WAAO,GAAG,MAAM,wBAAwB,OAAO,SAAS,CAAC;AAAA,EAC3D;AAAA,EAES,QAAQ;AArmBnB;AAsmBI,UAAM,MAAM;AACZ,eAAK,QAAL,mBAAU;AAAA,EACZ;AACF;","names":[]}
|
|
1
|
+
{"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2026 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n APIConnectionError,\n AudioByteStream,\n asLanguageCode,\n calculateAudioDurationSeconds,\n getBaseLanguage,\n log,\n normalizeLanguage,\n stt,\n waitForAbort,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { IncomingMessage } from 'node:http';\nimport { WebSocket } from 'ws';\nimport { AUDIO_ENCODING, type STTModel } from './models.js';\n\nconst AUTHORIZATION_HEADER = 'X-API-Key';\nconst VERSION_HEADER = 'Cartesia-Version';\nconst REQUEST_ID_HEADER = 'cartesia-request-id';\nconst API_VERSION = '2026-03-01';\nconst DRAIN_TIMEOUT_MS = 5000;\nconst KEEPALIVE_INTERVAL_MS = 30000;\n\n/**\n * Fires once when the WebSocket connection is established.\n *\n * You do not need to wait for this event before sending audio.\n */\ntype STTConnectedEvent = {\n type: 'connected';\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\n/**\n * Model predicts {@link stt.SpeechEventType.START_OF_SPEECH}.\n */\ntype STTTurnStartEvent = {\n type: 'turn.start';\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\n/**\n * Fires repeatedly as the model transcribes the current user turn.\n *\n * Can be used for {@link stt.SpeechEventType.INTERIM_TRANSCRIPT}.\n */\ntype STTTurnUpdateEvent = {\n type: 'turn.update';\n /**\n * Cumulative text for the current turn, i.e. the full text transcribed so far in this turn, not a delta.\n */\n transcript: string;\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\n/**\n * Fires when the model predicts that the user might be done speaking.\n *\n * Can be used for {@link stt.SpeechEventType.PREFLIGHT_TRANSCRIPT}.\n */\ntype STTTurnEagerEndEvent = {\n type: 'turn.eager_end';\n /**\n * Cumulative text for the current turn, i.e. the full text transcribed so far in this turn, not a delta.\n */\n transcript: string;\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\n/**\n * Fires after {@link STTTurnEagerEndEvent} if the user turn has not actually ended.\n */\ntype STTTurnResumeEvent = {\n type: 'turn.resume';\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\n/**\n * Marks the end of a user turn.\n *\n * This is used for {@link stt.SpeechEventType.END_OF_SPEECH} and {@link stt.SpeechEventType.FINAL_TRANSCRIPT}.\n */\ntype STTTurnEndEvent = {\n type: 'turn.end';\n /**\n * Cumulative text for the current turn, i.e. the full text transcribed so far in this turn, not a delta.\n */\n transcript: string;\n /**\n * Unique identifier for this connection. Does not change between turns.\n */\n request_id: string;\n};\n\ntype STTErrorEvent = {\n type: 'error';\n error_code?: string;\n status_code?: number;\n title?: string;\n message?: string;\n doc_url?: string;\n request_id?: string;\n};\n\n/**\n * Server-sent event on the `/stt/turns/websocket` endpoint.\n *\n * See https://docs.cartesia.ai/api-reference/stt/turns/websocket.\n */\ntype STTEventMessage =\n | STTConnectedEvent\n | STTTurnStartEvent\n | STTTurnUpdateEvent\n | STTTurnEagerEndEvent\n | STTTurnResumeEvent\n | STTTurnEndEvent\n | STTErrorEvent;\n\nexport type STTOptions = {\n apiKey: string;\n // eslint-disable-next-line @typescript-eslint/ban-types\n model: STTModel | (string & {});\n sampleRate: number;\n baseUrl: string;\n audioChunkDurationMS: number;\n language: string;\n};\n\nconst defaultSTTOptions = {\n model: 'ink-2' satisfies STTModel,\n /** recommended default */\n sampleRate: 16_000,\n /** recommended default */\n audioChunkDurationMS: 160,\n baseUrl: 'https://api.cartesia.ai',\n language: 'en',\n};\n\nfunction mergeSTTOptions(base: STTOptions, override: Partial<STTOptions>): STTOptions {\n return {\n apiKey: override.apiKey ?? base.apiKey,\n baseUrl: override.baseUrl ?? base.baseUrl,\n model: override.model ?? base.model,\n sampleRate: override.sampleRate ?? base.sampleRate,\n audioChunkDurationMS: override.audioChunkDurationMS ?? base.audioChunkDurationMS,\n language:\n override.language !== undefined ? normalizeLanguage(override.language) : base.language,\n };\n}\n\n/**\n * Resolve the STT model from the configured language when the caller did not\n * pass one explicitly. Mirrors the Python plugin: `ink-2` only supports\n * English, so non-English languages route to the multilingual `ink-whisper`.\n */\nfunction resolveSTTModel(language: string): STTModel {\n return getBaseLanguage(language) === 'en' ? 'ink-2' : 'ink-whisper';\n}\n\n/**\n * Cartesia speech to text.\n *\n * Supports:\n * - Streaming\n * - Turn detection\n * - Interim results\n *\n * See https://docs.cartesia.ai/build-with-cartesia/stt-models/latest\n *\n * @example\n * ```typescript\n * import { voice } from '@livekit/agents';\n * import { STT, TTS } from '@livekit/agents-plugin-cartesia';\n *\n * const session = new voice.AgentSession({\n * stt: new STT(),\n * llm: new LLM(), // choose your favorite LLM\n * tts: new TTS(),\n * turnHandling: { turnDetection: 'stt' },\n * });\n * ```\n */\nexport class STT extends stt.STT {\n #opts: STTOptions;\n\n constructor(opts: Partial<STTOptions> = {}) {\n super({\n streaming: true,\n interimResults: true,\n alignedTranscript: false,\n diarization: false,\n });\n\n const apiKey = opts.apiKey ?? process.env.CARTESIA_API_KEY;\n if (!apiKey) {\n throw new Error(\n 'Cartesia API key is required, whether as an argument or as $CARTESIA_API_KEY',\n );\n }\n\n this.#opts = mergeSTTOptions({ ...defaultSTTOptions, apiKey }, opts);\n\n // Route to the multilingual model for non-English languages unless the\n // caller pinned a model. Matches the Python plugin's language→model logic.\n if (opts.model === undefined) {\n this.#opts.model = resolveSTTModel(this.#opts.language);\n }\n }\n\n override get label(): string {\n return 'cartesia.STT';\n }\n\n override get model(): string {\n return this.#opts.model;\n }\n\n override get provider(): string {\n return 'Cartesia';\n }\n\n protected override async _recognize(): Promise<stt.SpeechEvent> {\n throw new Error('Cartesia STT does not support batch recognition, use stream() instead');\n }\n\n override stream(options?: { connOptions?: APIConnectOptions }): SpeechStream {\n return new SpeechStream(this, this.#opts, options?.connOptions);\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = mergeSTTOptions(this.#opts, opts);\n\n // Keep the model in sync with a newly set language (e.g. switching to a\n // non-English language must move off the English-only ink-2), unless the\n // caller pinned a model in the same call. Mirrors the constructor.\n if (opts.language !== undefined && opts.model === undefined) {\n this.#opts.model = resolveSTTModel(this.#opts.language);\n }\n }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n #opts: STTOptions;\n #logger = log();\n #ws: WebSocket | null = null;\n #requestId: string | undefined;\n #speaking = false;\n #currentTranscript = '';\n #speechDuration = 0;\n #closingWs = false;\n\n constructor(sttInstance: STT, opts: STTOptions, connOptions?: APIConnectOptions) {\n super(sttInstance, opts.sampleRate, connOptions);\n this.#opts = { ...opts };\n }\n\n override get label(): string {\n return 'cartesia.SpeechStream';\n }\n\n protected override async run() {\n if (this.closed) {\n this.close();\n return;\n }\n\n this.#speaking = false;\n this.#currentTranscript = '';\n this.#speechDuration = 0;\n this.#closingWs = false;\n\n let keepaliveInterval: NodeJS.Timeout | undefined;\n const abortController = new AbortController();\n\n try {\n const url = this.#getCartesiaUrl();\n this.#logger.debug(`Connecting to Cartesia STT: ${url}`);\n\n const ws = new WebSocket(url, {\n headers: {\n [AUTHORIZATION_HEADER]: this.#opts.apiKey,\n [VERSION_HEADER]: API_VERSION,\n },\n });\n this.#ws = ws;\n\n // Cartesia returns the request id on the WS upgrade response, before any\n // turn events are sent. Capture it so logs and metrics emitted on the\n // connection (and any pre-`turn.start` errors) are attributable.\n ws.once('upgrade', (response: IncomingMessage) => {\n const headerValue = response.headers[REQUEST_ID_HEADER];\n const requestId = Array.isArray(headerValue) ? headerValue[0] : headerValue;\n if (requestId) {\n this.#requestId = requestId;\n this.#logger.debug({ cartesiaRequestId: requestId }, 'Cartesia STT WebSocket connected');\n }\n });\n\n await new Promise<void>((resolve, reject) => {\n const onOpen = () => {\n ws.off('error', onError);\n resolve();\n };\n const onError = (err: Error) => {\n ws.off('open', onOpen);\n reject(new APIConnectionError({ message: err.message, options: { retryable: true } }));\n };\n\n ws.once('open', onOpen);\n ws.once('error', onError);\n });\n\n // If one task fails, abort its peer and close the WS so the peer can\n // exit before we re-throw. Otherwise a dangling task would survive\n // into the next retry attempt (sendTask blocked on input.next() would\n // steal frames; recvTask's close handler could still push events to\n // the queue and surface an unhandled rejection).\n let firstError: unknown;\n const stopPeer = (err: unknown) => {\n if (firstError === undefined) firstError = err;\n abortController.abort();\n if (ws.readyState === WebSocket.OPEN) ws.close();\n };\n const sendPromise = this.#sendTask(ws, abortController.signal).catch(stopPeer);\n const recvPromise = this.#recvTask(ws).catch(stopPeer);\n\n keepaliveInterval = setInterval(() => {\n try {\n if (ws.readyState === WebSocket.OPEN) {\n ws.ping();\n }\n } catch {\n // ignore\n }\n }, KEEPALIVE_INTERVAL_MS);\n await Promise.all([sendPromise, recvPromise]);\n if (firstError !== undefined) throw firstError;\n } catch (error) {\n this.#logger.error('Cartesia STT stream error', { error });\n throw error;\n } finally {\n abortController.abort();\n if (keepaliveInterval) clearInterval(keepaliveInterval);\n if (this.#ws?.readyState === WebSocket.OPEN) {\n this.#ws.close();\n }\n }\n\n this.close();\n }\n\n async #sendTask(ws: WebSocket, abortSignal: AbortSignal) {\n const samplesPerChunk = Math.floor(\n (this.#opts.sampleRate * this.#opts.audioChunkDurationMS) / 1000,\n );\n const audioBstream = new AudioByteStream(this.#opts.sampleRate, 1, samplesPerChunk);\n\n let hasEnded = false;\n const iterator = this.input[Symbol.asyncIterator]();\n const abortPromise = waitForAbort(abortSignal);\n\n while (true) {\n const result = await Promise.race([iterator.next(), abortPromise]);\n\n if (result === undefined) return; // aborted\n\n if (result.done) {\n hasEnded = true;\n } else {\n const data = result.value;\n const frames: AudioFrame[] = [];\n\n if (data === SpeechStream.FLUSH_SENTINEL) {\n frames.push(...audioBstream.flush());\n } else {\n // Pass the typed array directly so AudioByteStream uses byteOffset/\n // byteLength — `.buffer` would include foreign bytes when rtc-node\n // hands us a view over a pooled allocator.\n frames.push(...audioBstream.write(data.data));\n }\n\n for (const frame of frames) {\n this.#speechDuration += calculateAudioDurationSeconds(frame);\n if (ws.readyState === WebSocket.OPEN) {\n ws.send(frame.data);\n }\n }\n }\n\n if (hasEnded) break;\n }\n\n if (ws.readyState === WebSocket.OPEN) {\n this.#closingWs = true;\n this.#logger.debug('Sending close message to Cartesia STT');\n ws.send(JSON.stringify({ type: 'close' }));\n\n // After `close`, the server should flush any pending turn events and\n // close the socket. If it hangs, force the issue rather than letting\n // the stream wait forever.\n const drainTimer = setTimeout(() => {\n this.#logger.warn(\n `Cartesia STT did not close within ${DRAIN_TIMEOUT_MS}ms after done; forcing close`,\n );\n if (ws.readyState === WebSocket.OPEN || ws.readyState === WebSocket.CLOSING) {\n ws.terminate();\n }\n }, DRAIN_TIMEOUT_MS);\n ws.once('close', () => clearTimeout(drainTimer));\n }\n }\n\n async #recvTask(ws: WebSocket) {\n let pendingError: Error | undefined = undefined;\n\n return new Promise<void>((resolve, reject) => {\n const settle = (err?: Error) => {\n if (err) reject(err);\n else resolve();\n };\n\n ws.on('message', (data: Buffer, isBinary: boolean) => {\n if (isBinary) {\n this.#logger.warn('Received unexpected binary message from Cartesia STT');\n return;\n }\n let msg: STTEventMessage;\n try {\n msg = JSON.parse(data.toString());\n } catch (error) {\n this.#logger.error('Failed to parse Cartesia STT message', { error });\n return;\n }\n try {\n if (msg.type === 'error') {\n this.#logger.error('Cartesia sent an error', msg);\n\n // do not close the websocket on bad requests since that may be caused by invalid messages\n if (msg.status_code === undefined || msg.status_code >= 500) {\n // Defer until the WS is fully closed so we don't race the `close`\n // handler with a rejected `recvTask`. Close the socket explicitly\n // in case the server doesn't, otherwise recv would hang.\n pendingError = new APIConnectionError({\n message: msg.message || msg.title,\n options: {\n retryable: true,\n },\n });\n ws.close();\n return;\n }\n } else {\n this.#processStreamEvent(msg);\n }\n } catch (error) {\n this.#logger.error('Failed to process Cartesia STT message', { error });\n }\n });\n\n ws.on('close', (code, reason) => {\n // If a turn was in progress, close it out so consumers see a balanced\n // START_OF_SPEECH/END_OF_SPEECH pair and don't carry stale turn state\n // into the next reconnect.\n if (this.#speaking) {\n if (this.#currentTranscript) {\n this.#sendTranscriptEvent(\n stt.SpeechEventType.FINAL_TRANSCRIPT,\n this.#currentTranscript,\n );\n }\n if (!this.queue.closed)\n this.queue.put({\n type: stt.SpeechEventType.END_OF_SPEECH,\n requestId: this.#requestId,\n });\n this.#logger.debug('Cartesia STT END_OF_SPEECH');\n if (this.#speechDuration > 0 && !this.queue.closed) {\n this.queue.put({\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: { audioDuration: this.#speechDuration },\n });\n this.#speechDuration = 0;\n }\n this.#speaking = false;\n this.#currentTranscript = '';\n }\n\n if (pendingError) {\n settle(pendingError);\n return;\n }\n if (this.#closingWs || this.closed) {\n settle();\n return;\n }\n this.#logger.warn(`Cartesia STT WebSocket closed: ${code} ${reason.toString()}`);\n settle(\n new APIConnectionError({\n message: `Cartesia STT connection closed unexpectedly (code=${code})`,\n options: { retryable: true },\n }),\n );\n });\n\n ws.on('error', (err) => {\n if (this.closed) {\n settle();\n return;\n }\n settle(new APIConnectionError({ message: err.message, options: { retryable: true } }));\n });\n });\n }\n\n #processStreamEvent(data: Exclude<STTEventMessage, { type: 'error' }>) {\n if (data.request_id) {\n this.#requestId = data.request_id;\n }\n\n switch (data.type) {\n case 'connected':\n return;\n\n case 'turn.start': {\n if (this.#speaking) return;\n this.#speaking = true;\n this.#currentTranscript = '';\n if (!this.queue.closed)\n this.queue.put({\n type: stt.SpeechEventType.START_OF_SPEECH,\n requestId: this.#requestId,\n });\n this.#logger.debug('Cartesia STT START_OF_SPEECH');\n return;\n }\n\n case 'turn.update': {\n if (!this.#speaking) return;\n const transcript = data.transcript;\n if (!transcript) return;\n // Only emit interim updates when the cumulative transcript actually\n // changed; this avoids canceling preflight generation needlessly.\n if (this.#currentTranscript === transcript) return;\n this.#currentTranscript = transcript;\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, transcript);\n return;\n }\n\n case 'turn.eager_end': {\n if (!this.#speaking) return;\n const transcript = data.transcript || this.#currentTranscript;\n if (!transcript) return;\n this.#currentTranscript = transcript;\n this.#sendTranscriptEvent(stt.SpeechEventType.PREFLIGHT_TRANSCRIPT, transcript);\n return;\n }\n\n case 'turn.resume':\n // turn.resume has no transcript; re-emit the latest cumulative transcript\n // as an interim event so the pipeline cancels the pending preflight.\n if (!this.#speaking || !this.#currentTranscript) return;\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, this.#currentTranscript);\n return;\n\n case 'turn.end': {\n if (!this.#speaking) return;\n const transcript = data.transcript || this.#currentTranscript;\n\n this.#sendTranscriptEvent(stt.SpeechEventType.FINAL_TRANSCRIPT, transcript);\n\n this.#speaking = false;\n if (!this.queue.closed)\n this.queue.put({\n type: stt.SpeechEventType.END_OF_SPEECH,\n requestId: this.#requestId,\n });\n this.#logger.debug('Cartesia STT END_OF_SPEECH');\n this.#currentTranscript = '';\n\n if (this.#speechDuration > 0 && !this.queue.closed) {\n this.queue.put({\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: { audioDuration: this.#speechDuration },\n });\n this.#speechDuration = 0;\n }\n\n return;\n }\n\n default:\n this.#logger.warn('received unexpected message from Cartesia STT', { data });\n }\n }\n\n #sendTranscriptEvent(eventType: stt.SpeechEventType, transcript: string) {\n if (this.queue.closed) return;\n this.queue.put({\n type: eventType,\n requestId: this.#requestId,\n alternatives: [\n {\n language: asLanguageCode(this.#opts.language),\n text: transcript,\n startTime: 0,\n endTime: 0,\n confidence: 0,\n },\n ],\n });\n }\n\n #getCartesiaUrl(): string {\n // The Cartesia /stt/turns/websocket endpoint only accepts model, sample_rate\n // and encoding — there is no `language` query param. Language selection is\n // expressed through the model (ink-2 for English, ink-whisper otherwise),\n // so #opts.language is used only to tag emitted transcripts.\n const params = new URLSearchParams({\n model: this.#opts.model,\n sample_rate: this.#opts.sampleRate.toString(),\n encoding: AUDIO_ENCODING,\n });\n\n const wsBase = this.#opts.baseUrl.replace(/^http/, 'ws');\n return `${wsBase}/stt/turns/websocket?${params.toString()}`;\n }\n\n override close() {\n super.close();\n this.#ws?.close();\n }\n}\n"],"mappings":"AAGA;AAAA,EAEE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAGP,SAAS,iBAAiB;AAC1B,SAAS,sBAAqC;AAE9C,MAAM,uBAAuB;AAC7B,MAAM,iBAAiB;AACvB,MAAM,oBAAoB;AAC1B,MAAM,cAAc;AACpB,MAAM,mBAAmB;AACzB,MAAM,wBAAwB;AA0H9B,MAAM,oBAAoB;AAAA,EACxB,OAAO;AAAA;AAAA,EAEP,YAAY;AAAA;AAAA,EAEZ,sBAAsB;AAAA,EACtB,SAAS;AAAA,EACT,UAAU;AACZ;AAEA,SAAS,gBAAgB,MAAkB,UAA2C;AACpF,SAAO;AAAA,IACL,QAAQ,SAAS,UAAU,KAAK;AAAA,IAChC,SAAS,SAAS,WAAW,KAAK;AAAA,IAClC,OAAO,SAAS,SAAS,KAAK;AAAA,IAC9B,YAAY,SAAS,cAAc,KAAK;AAAA,IACxC,sBAAsB,SAAS,wBAAwB,KAAK;AAAA,IAC5D,UACE,SAAS,aAAa,SAAY,kBAAkB,SAAS,QAAQ,IAAI,KAAK;AAAA,EAClF;AACF;AAOA,SAAS,gBAAgB,UAA4B;AACnD,SAAO,gBAAgB,QAAQ,MAAM,OAAO,UAAU;AACxD;AAyBO,MAAM,YAAY,IAAI,IAAI;AAAA,EAC/B;AAAA,EAEA,YAAY,OAA4B,CAAC,GAAG;AAC1C,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB;AAAA,MAChB,mBAAmB;AAAA,MACnB,aAAa;AAAA,IACf,CAAC;AAED,UAAM,SAAS,KAAK,UAAU,QAAQ,IAAI;AAC1C,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ,gBAAgB,EAAE,GAAG,mBAAmB,OAAO,GAAG,IAAI;AAInE,QAAI,KAAK,UAAU,QAAW;AAC5B,WAAK,MAAM,QAAQ,gBAAgB,KAAK,MAAM,QAAQ;AAAA,IACxD;AAAA,EACF;AAAA,EAEA,IAAa,QAAgB;AAC3B,WAAO;AAAA,EACT;AAAA,EAEA,IAAa,QAAgB;AAC3B,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAa,WAAmB;AAC9B,WAAO;AAAA,EACT;AAAA,EAEA,MAAyB,aAAuC;AAC9D,UAAM,IAAI,MAAM,uEAAuE;AAAA,EACzF;AAAA,EAES,OAAO,SAA6D;AAC3E,WAAO,IAAI,aAAa,MAAM,KAAK,OAAO,mCAAS,WAAW;AAAA,EAChE;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,gBAAgB,KAAK,OAAO,IAAI;AAK7C,QAAI,KAAK,aAAa,UAAa,KAAK,UAAU,QAAW;AAC3D,WAAK,MAAM,QAAQ,gBAAgB,KAAK,MAAM,QAAQ;AAAA,IACxD;AAAA,EACF;AACF;AAEO,MAAM,qBAAqB,IAAI,aAAa;AAAA,EACjD;AAAA,EACA,UAAU,IAAI;AAAA,EACd,MAAwB;AAAA,EACxB;AAAA,EACA,YAAY;AAAA,EACZ,qBAAqB;AAAA,EACrB,kBAAkB;AAAA,EAClB,aAAa;AAAA,EAEb,YAAY,aAAkB,MAAkB,aAAiC;AAC/E,UAAM,aAAa,KAAK,YAAY,WAAW;AAC/C,SAAK,QAAQ,EAAE,GAAG,KAAK;AAAA,EACzB;AAAA,EAEA,IAAa,QAAgB;AAC3B,WAAO;AAAA,EACT;AAAA,EAEA,MAAyB,MAAM;AAvRjC;AAwRI,QAAI,KAAK,QAAQ;AACf,WAAK,MAAM;AACX;AAAA,IACF;AAEA,SAAK,YAAY;AACjB,SAAK,qBAAqB;AAC1B,SAAK,kBAAkB;AACvB,SAAK,aAAa;AAElB,QAAI;AACJ,UAAM,kBAAkB,IAAI,gBAAgB;AAE5C,QAAI;AACF,YAAM,MAAM,KAAK,gBAAgB;AACjC,WAAK,QAAQ,MAAM,+BAA+B,GAAG,EAAE;AAEvD,YAAM,KAAK,IAAI,UAAU,KAAK;AAAA,QAC5B,SAAS;AAAA,UACP,CAAC,oBAAoB,GAAG,KAAK,MAAM;AAAA,UACnC,CAAC,cAAc,GAAG;AAAA,QACpB;AAAA,MACF,CAAC;AACD,WAAK,MAAM;AAKX,SAAG,KAAK,WAAW,CAAC,aAA8B;AAChD,cAAM,cAAc,SAAS,QAAQ,iBAAiB;AACtD,cAAM,YAAY,MAAM,QAAQ,WAAW,IAAI,YAAY,CAAC,IAAI;AAChE,YAAI,WAAW;AACb,eAAK,aAAa;AAClB,eAAK,QAAQ,MAAM,EAAE,mBAAmB,UAAU,GAAG,kCAAkC;AAAA,QACzF;AAAA,MACF,CAAC;AAED,YAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,cAAM,SAAS,MAAM;AACnB,aAAG,IAAI,SAAS,OAAO;AACvB,kBAAQ;AAAA,QACV;AACA,cAAM,UAAU,CAAC,QAAe;AAC9B,aAAG,IAAI,QAAQ,MAAM;AACrB,iBAAO,IAAI,mBAAmB,EAAE,SAAS,IAAI,SAAS,SAAS,EAAE,WAAW,KAAK,EAAE,CAAC,CAAC;AAAA,QACvF;AAEA,WAAG,KAAK,QAAQ,MAAM;AACtB,WAAG,KAAK,SAAS,OAAO;AAAA,MAC1B,CAAC;AAOD,UAAI;AACJ,YAAM,WAAW,CAAC,QAAiB;AACjC,YAAI,eAAe,OAAW,cAAa;AAC3C,wBAAgB,MAAM;AACtB,YAAI,GAAG,eAAe,UAAU,KAAM,IAAG,MAAM;AAAA,MACjD;AACA,YAAM,cAAc,KAAK,UAAU,IAAI,gBAAgB,MAAM,EAAE,MAAM,QAAQ;AAC7E,YAAM,cAAc,KAAK,UAAU,EAAE,EAAE,MAAM,QAAQ;AAErD,0BAAoB,YAAY,MAAM;AACpC,YAAI;AACF,cAAI,GAAG,eAAe,UAAU,MAAM;AACpC,eAAG,KAAK;AAAA,UACV;AAAA,QACF,QAAQ;AAAA,QAER;AAAA,MACF,GAAG,qBAAqB;AACxB,YAAM,QAAQ,IAAI,CAAC,aAAa,WAAW,CAAC;AAC5C,UAAI,eAAe,OAAW,OAAM;AAAA,IACtC,SAAS,OAAO;AACd,WAAK,QAAQ,MAAM,6BAA6B,EAAE,MAAM,CAAC;AACzD,YAAM;AAAA,IACR,UAAE;AACA,sBAAgB,MAAM;AACtB,UAAI,kBAAmB,eAAc,iBAAiB;AACtD,YAAI,UAAK,QAAL,mBAAU,gBAAe,UAAU,MAAM;AAC3C,aAAK,IAAI,MAAM;AAAA,MACjB;AAAA,IACF;AAEA,SAAK,MAAM;AAAA,EACb;AAAA,EAEA,MAAM,UAAU,IAAe,aAA0B;AACvD,UAAM,kBAAkB,KAAK;AAAA,MAC1B,KAAK,MAAM,aAAa,KAAK,MAAM,uBAAwB;AAAA,IAC9D;AACA,UAAM,eAAe,IAAI,gBAAgB,KAAK,MAAM,YAAY,GAAG,eAAe;AAElF,QAAI,WAAW;AACf,UAAM,WAAW,KAAK,MAAM,OAAO,aAAa,EAAE;AAClD,UAAM,eAAe,aAAa,WAAW;AAE7C,WAAO,MAAM;AACX,YAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,SAAS,KAAK,GAAG,YAAY,CAAC;AAEjE,UAAI,WAAW,OAAW;AAE1B,UAAI,OAAO,MAAM;AACf,mBAAW;AAAA,MACb,OAAO;AACL,cAAM,OAAO,OAAO;AACpB,cAAM,SAAuB,CAAC;AAE9B,YAAI,SAAS,aAAa,gBAAgB;AACxC,iBAAO,KAAK,GAAG,aAAa,MAAM,CAAC;AAAA,QACrC,OAAO;AAIL,iBAAO,KAAK,GAAG,aAAa,MAAM,KAAK,IAAI,CAAC;AAAA,QAC9C;AAEA,mBAAW,SAAS,QAAQ;AAC1B,eAAK,mBAAmB,8BAA8B,KAAK;AAC3D,cAAI,GAAG,eAAe,UAAU,MAAM;AACpC,eAAG,KAAK,MAAM,IAAI;AAAA,UACpB;AAAA,QACF;AAAA,MACF;AAEA,UAAI,SAAU;AAAA,IAChB;AAEA,QAAI,GAAG,eAAe,UAAU,MAAM;AACpC,WAAK,aAAa;AAClB,WAAK,QAAQ,MAAM,uCAAuC;AAC1D,SAAG,KAAK,KAAK,UAAU,EAAE,MAAM,QAAQ,CAAC,CAAC;AAKzC,YAAM,aAAa,WAAW,MAAM;AAClC,aAAK,QAAQ;AAAA,UACX,qCAAqC,gBAAgB;AAAA,QACvD;AACA,YAAI,GAAG,eAAe,UAAU,QAAQ,GAAG,eAAe,UAAU,SAAS;AAC3E,aAAG,UAAU;AAAA,QACf;AAAA,MACF,GAAG,gBAAgB;AACnB,SAAG,KAAK,SAAS,MAAM,aAAa,UAAU,CAAC;AAAA,IACjD;AAAA,EACF;AAAA,EAEA,MAAM,UAAU,IAAe;AAC7B,QAAI,eAAkC;AAEtC,WAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,YAAM,SAAS,CAAC,QAAgB;AAC9B,YAAI,IAAK,QAAO,GAAG;AAAA,YACd,SAAQ;AAAA,MACf;AAEA,SAAG,GAAG,WAAW,CAAC,MAAc,aAAsB;AACpD,YAAI,UAAU;AACZ,eAAK,QAAQ,KAAK,sDAAsD;AACxE;AAAA,QACF;AACA,YAAI;AACJ,YAAI;AACF,gBAAM,KAAK,MAAM,KAAK,SAAS,CAAC;AAAA,QAClC,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,wCAAwC,EAAE,MAAM,CAAC;AACpE;AAAA,QACF;AACA,YAAI;AACF,cAAI,IAAI,SAAS,SAAS;AACxB,iBAAK,QAAQ,MAAM,0BAA0B,GAAG;AAGhD,gBAAI,IAAI,gBAAgB,UAAa,IAAI,eAAe,KAAK;AAI3D,6BAAe,IAAI,mBAAmB;AAAA,gBACpC,SAAS,IAAI,WAAW,IAAI;AAAA,gBAC5B,SAAS;AAAA,kBACP,WAAW;AAAA,gBACb;AAAA,cACF,CAAC;AACD,iBAAG,MAAM;AACT;AAAA,YACF;AAAA,UACF,OAAO;AACL,iBAAK,oBAAoB,GAAG;AAAA,UAC9B;AAAA,QACF,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,0CAA0C,EAAE,MAAM,CAAC;AAAA,QACxE;AAAA,MACF,CAAC;AAED,SAAG,GAAG,SAAS,CAAC,MAAM,WAAW;AAI/B,YAAI,KAAK,WAAW;AAClB,cAAI,KAAK,oBAAoB;AAC3B,iBAAK;AAAA,cACH,IAAI,gBAAgB;AAAA,cACpB,KAAK;AAAA,YACP;AAAA,UACF;AACA,cAAI,CAAC,KAAK,MAAM;AACd,iBAAK,MAAM,IAAI;AAAA,cACb,MAAM,IAAI,gBAAgB;AAAA,cAC1B,WAAW,KAAK;AAAA,YAClB,CAAC;AACH,eAAK,QAAQ,MAAM,4BAA4B;AAC/C,cAAI,KAAK,kBAAkB,KAAK,CAAC,KAAK,MAAM,QAAQ;AAClD,iBAAK,MAAM,IAAI;AAAA,cACb,MAAM,IAAI,gBAAgB;AAAA,cAC1B,WAAW,KAAK;AAAA,cAChB,kBAAkB,EAAE,eAAe,KAAK,gBAAgB;AAAA,YAC1D,CAAC;AACD,iBAAK,kBAAkB;AAAA,UACzB;AACA,eAAK,YAAY;AACjB,eAAK,qBAAqB;AAAA,QAC5B;AAEA,YAAI,cAAc;AAChB,iBAAO,YAAY;AACnB;AAAA,QACF;AACA,YAAI,KAAK,cAAc,KAAK,QAAQ;AAClC,iBAAO;AACP;AAAA,QACF;AACA,aAAK,QAAQ,KAAK,kCAAkC,IAAI,IAAI,OAAO,SAAS,CAAC,EAAE;AAC/E;AAAA,UACE,IAAI,mBAAmB;AAAA,YACrB,SAAS,qDAAqD,IAAI;AAAA,YAClE,SAAS,EAAE,WAAW,KAAK;AAAA,UAC7B,CAAC;AAAA,QACH;AAAA,MACF,CAAC;AAED,SAAG,GAAG,SAAS,CAAC,QAAQ;AACtB,YAAI,KAAK,QAAQ;AACf,iBAAO;AACP;AAAA,QACF;AACA,eAAO,IAAI,mBAAmB,EAAE,SAAS,IAAI,SAAS,SAAS,EAAE,WAAW,KAAK,EAAE,CAAC,CAAC;AAAA,MACvF,CAAC;AAAA,IACH,CAAC;AAAA,EACH;AAAA,EAEA,oBAAoB,MAAmD;AACrE,QAAI,KAAK,YAAY;AACnB,WAAK,aAAa,KAAK;AAAA,IACzB;AAEA,YAAQ,KAAK,MAAM;AAAA,MACjB,KAAK;AACH;AAAA,MAEF,KAAK,cAAc;AACjB,YAAI,KAAK,UAAW;AACpB,aAAK,YAAY;AACjB,aAAK,qBAAqB;AAC1B,YAAI,CAAC,KAAK,MAAM;AACd,eAAK,MAAM,IAAI;AAAA,YACb,MAAM,IAAI,gBAAgB;AAAA,YAC1B,WAAW,KAAK;AAAA,UAClB,CAAC;AACH,aAAK,QAAQ,MAAM,8BAA8B;AACjD;AAAA,MACF;AAAA,MAEA,KAAK,eAAe;AAClB,YAAI,CAAC,KAAK,UAAW;AACrB,cAAM,aAAa,KAAK;AACxB,YAAI,CAAC,WAAY;AAGjB,YAAI,KAAK,uBAAuB,WAAY;AAC5C,aAAK,qBAAqB;AAC1B,aAAK,qBAAqB,IAAI,gBAAgB,oBAAoB,UAAU;AAC5E;AAAA,MACF;AAAA,MAEA,KAAK,kBAAkB;AACrB,YAAI,CAAC,KAAK,UAAW;AACrB,cAAM,aAAa,KAAK,cAAc,KAAK;AAC3C,YAAI,CAAC,WAAY;AACjB,aAAK,qBAAqB;AAC1B,aAAK,qBAAqB,IAAI,gBAAgB,sBAAsB,UAAU;AAC9E;AAAA,MACF;AAAA,MAEA,KAAK;AAGH,YAAI,CAAC,KAAK,aAAa,CAAC,KAAK,mBAAoB;AACjD,aAAK,qBAAqB,IAAI,gBAAgB,oBAAoB,KAAK,kBAAkB;AACzF;AAAA,MAEF,KAAK,YAAY;AACf,YAAI,CAAC,KAAK,UAAW;AACrB,cAAM,aAAa,KAAK,cAAc,KAAK;AAE3C,aAAK,qBAAqB,IAAI,gBAAgB,kBAAkB,UAAU;AAE1E,aAAK,YAAY;AACjB,YAAI,CAAC,KAAK,MAAM;AACd,eAAK,MAAM,IAAI;AAAA,YACb,MAAM,IAAI,gBAAgB;AAAA,YAC1B,WAAW,KAAK;AAAA,UAClB,CAAC;AACH,aAAK,QAAQ,MAAM,4BAA4B;AAC/C,aAAK,qBAAqB;AAE1B,YAAI,KAAK,kBAAkB,KAAK,CAAC,KAAK,MAAM,QAAQ;AAClD,eAAK,MAAM,IAAI;AAAA,YACb,MAAM,IAAI,gBAAgB;AAAA,YAC1B,WAAW,KAAK;AAAA,YAChB,kBAAkB,EAAE,eAAe,KAAK,gBAAgB;AAAA,UAC1D,CAAC;AACD,eAAK,kBAAkB;AAAA,QACzB;AAEA;AAAA,MACF;AAAA,MAEA;AACE,aAAK,QAAQ,KAAK,iDAAiD,EAAE,KAAK,CAAC;AAAA,IAC/E;AAAA,EACF;AAAA,EAEA,qBAAqB,WAAgC,YAAoB;AACvE,QAAI,KAAK,MAAM,OAAQ;AACvB,SAAK,MAAM,IAAI;AAAA,MACb,MAAM;AAAA,MACN,WAAW,KAAK;AAAA,MAChB,cAAc;AAAA,QACZ;AAAA,UACE,UAAU,eAAe,KAAK,MAAM,QAAQ;AAAA,UAC5C,MAAM;AAAA,UACN,WAAW;AAAA,UACX,SAAS;AAAA,UACT,YAAY;AAAA,QACd;AAAA,MACF;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,kBAA0B;AAKxB,UAAM,SAAS,IAAI,gBAAgB;AAAA,MACjC,OAAO,KAAK,MAAM;AAAA,MAClB,aAAa,KAAK,MAAM,WAAW,SAAS;AAAA,MAC5C,UAAU;AAAA,IACZ,CAAC;AAED,UAAM,SAAS,KAAK,MAAM,QAAQ,QAAQ,SAAS,IAAI;AACvD,WAAO,GAAG,MAAM,wBAAwB,OAAO,SAAS,CAAC;AAAA,EAC3D;AAAA,EAES,QAAQ;AAxoBnB;AAyoBI,UAAM,MAAM;AACZ,eAAK,QAAL,mBAAU;AAAA,EACZ;AACF;","names":[]}
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@livekit/agents-plugin-cartesia",
|
|
3
|
-
"version": "1.4.
|
|
3
|
+
"version": "1.4.11",
|
|
4
4
|
"description": "Cartesia plugin for LiveKit Node Agents",
|
|
5
5
|
"main": "dist/index.js",
|
|
6
6
|
"require": "dist/index.cjs",
|
|
@@ -30,10 +30,10 @@
|
|
|
30
30
|
"@types/ws": "^8.5.10",
|
|
31
31
|
"tsup": "^8.3.5",
|
|
32
32
|
"typescript": "^5.0.0",
|
|
33
|
-
"@livekit/agents": "1.4.
|
|
34
|
-
"@livekit/agents-plugin-openai": "1.4.
|
|
35
|
-
"@livekit/agents-plugin-silero": "1.4.
|
|
36
|
-
"@livekit/agents-plugins-test": "1.4.
|
|
33
|
+
"@livekit/agents": "1.4.11",
|
|
34
|
+
"@livekit/agents-plugin-openai": "1.4.11",
|
|
35
|
+
"@livekit/agents-plugin-silero": "1.4.11",
|
|
36
|
+
"@livekit/agents-plugins-test": "1.4.11"
|
|
37
37
|
},
|
|
38
38
|
"dependencies": {
|
|
39
39
|
"ws": "^8.18.0"
|
|
@@ -41,7 +41,7 @@
|
|
|
41
41
|
"peerDependencies": {
|
|
42
42
|
"@livekit/rtc-node": "^0.13.29",
|
|
43
43
|
"zod": "^3.25.76 || ^4.1.8",
|
|
44
|
-
"@livekit/agents": "1.4.
|
|
44
|
+
"@livekit/agents": "1.4.11"
|
|
45
45
|
},
|
|
46
46
|
"scripts": {
|
|
47
47
|
"build": "tsup --onSuccess \"pnpm build:types\"",
|
package/src/models.ts
CHANGED
|
@@ -71,5 +71,7 @@ export type TTSEncoding = 'pcm_s16le';
|
|
|
71
71
|
|
|
72
72
|
/**
|
|
73
73
|
* See [the docs](https://docs.cartesia.ai/build-with-cartesia/stt-models/latest) for all options.
|
|
74
|
+
*
|
|
75
|
+
* `ink-2` currently supports English only; `ink-whisper` is multilingual.
|
|
74
76
|
*/
|
|
75
|
-
export type STTModel = 'ink-2';
|
|
77
|
+
export type STTModel = 'ink-2' | 'ink-whisper';
|
package/src/stt.ts
CHANGED
|
@@ -7,7 +7,9 @@ import {
|
|
|
7
7
|
AudioByteStream,
|
|
8
8
|
asLanguageCode,
|
|
9
9
|
calculateAudioDurationSeconds,
|
|
10
|
+
getBaseLanguage,
|
|
10
11
|
log,
|
|
12
|
+
normalizeLanguage,
|
|
11
13
|
stt,
|
|
12
14
|
waitForAbort,
|
|
13
15
|
} from '@livekit/agents';
|
|
@@ -140,6 +142,7 @@ export type STTOptions = {
|
|
|
140
142
|
sampleRate: number;
|
|
141
143
|
baseUrl: string;
|
|
142
144
|
audioChunkDurationMS: number;
|
|
145
|
+
language: string;
|
|
143
146
|
};
|
|
144
147
|
|
|
145
148
|
const defaultSTTOptions = {
|
|
@@ -149,6 +152,7 @@ const defaultSTTOptions = {
|
|
|
149
152
|
/** recommended default */
|
|
150
153
|
audioChunkDurationMS: 160,
|
|
151
154
|
baseUrl: 'https://api.cartesia.ai',
|
|
155
|
+
language: 'en',
|
|
152
156
|
};
|
|
153
157
|
|
|
154
158
|
function mergeSTTOptions(base: STTOptions, override: Partial<STTOptions>): STTOptions {
|
|
@@ -158,9 +162,20 @@ function mergeSTTOptions(base: STTOptions, override: Partial<STTOptions>): STTOp
|
|
|
158
162
|
model: override.model ?? base.model,
|
|
159
163
|
sampleRate: override.sampleRate ?? base.sampleRate,
|
|
160
164
|
audioChunkDurationMS: override.audioChunkDurationMS ?? base.audioChunkDurationMS,
|
|
165
|
+
language:
|
|
166
|
+
override.language !== undefined ? normalizeLanguage(override.language) : base.language,
|
|
161
167
|
};
|
|
162
168
|
}
|
|
163
169
|
|
|
170
|
+
/**
|
|
171
|
+
* Resolve the STT model from the configured language when the caller did not
|
|
172
|
+
* pass one explicitly. Mirrors the Python plugin: `ink-2` only supports
|
|
173
|
+
* English, so non-English languages route to the multilingual `ink-whisper`.
|
|
174
|
+
*/
|
|
175
|
+
function resolveSTTModel(language: string): STTModel {
|
|
176
|
+
return getBaseLanguage(language) === 'en' ? 'ink-2' : 'ink-whisper';
|
|
177
|
+
}
|
|
178
|
+
|
|
164
179
|
/**
|
|
165
180
|
* Cartesia speech to text.
|
|
166
181
|
*
|
|
@@ -203,6 +218,12 @@ export class STT extends stt.STT {
|
|
|
203
218
|
}
|
|
204
219
|
|
|
205
220
|
this.#opts = mergeSTTOptions({ ...defaultSTTOptions, apiKey }, opts);
|
|
221
|
+
|
|
222
|
+
// Route to the multilingual model for non-English languages unless the
|
|
223
|
+
// caller pinned a model. Matches the Python plugin's language→model logic.
|
|
224
|
+
if (opts.model === undefined) {
|
|
225
|
+
this.#opts.model = resolveSTTModel(this.#opts.language);
|
|
226
|
+
}
|
|
206
227
|
}
|
|
207
228
|
|
|
208
229
|
override get label(): string {
|
|
@@ -224,6 +245,17 @@ export class STT extends stt.STT {
|
|
|
224
245
|
override stream(options?: { connOptions?: APIConnectOptions }): SpeechStream {
|
|
225
246
|
return new SpeechStream(this, this.#opts, options?.connOptions);
|
|
226
247
|
}
|
|
248
|
+
|
|
249
|
+
updateOptions(opts: Partial<STTOptions>) {
|
|
250
|
+
this.#opts = mergeSTTOptions(this.#opts, opts);
|
|
251
|
+
|
|
252
|
+
// Keep the model in sync with a newly set language (e.g. switching to a
|
|
253
|
+
// non-English language must move off the English-only ink-2), unless the
|
|
254
|
+
// caller pinned a model in the same call. Mirrors the constructor.
|
|
255
|
+
if (opts.language !== undefined && opts.model === undefined) {
|
|
256
|
+
this.#opts.model = resolveSTTModel(this.#opts.language);
|
|
257
|
+
}
|
|
258
|
+
}
|
|
227
259
|
}
|
|
228
260
|
|
|
229
261
|
export class SpeechStream extends stt.SpeechStream {
|
|
@@ -536,7 +568,7 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
536
568
|
|
|
537
569
|
case 'turn.eager_end': {
|
|
538
570
|
if (!this.#speaking) return;
|
|
539
|
-
const transcript = data.transcript;
|
|
571
|
+
const transcript = data.transcript || this.#currentTranscript;
|
|
540
572
|
if (!transcript) return;
|
|
541
573
|
this.#currentTranscript = transcript;
|
|
542
574
|
this.#sendTranscriptEvent(stt.SpeechEventType.PREFLIGHT_TRANSCRIPT, transcript);
|
|
@@ -552,7 +584,7 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
552
584
|
|
|
553
585
|
case 'turn.end': {
|
|
554
586
|
if (!this.#speaking) return;
|
|
555
|
-
const transcript = data.transcript;
|
|
587
|
+
const transcript = data.transcript || this.#currentTranscript;
|
|
556
588
|
|
|
557
589
|
this.#sendTranscriptEvent(stt.SpeechEventType.FINAL_TRANSCRIPT, transcript);
|
|
558
590
|
|
|
@@ -589,8 +621,7 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
589
621
|
requestId: this.#requestId,
|
|
590
622
|
alternatives: [
|
|
591
623
|
{
|
|
592
|
-
|
|
593
|
-
language: asLanguageCode('en'),
|
|
624
|
+
language: asLanguageCode(this.#opts.language),
|
|
594
625
|
text: transcript,
|
|
595
626
|
startTime: 0,
|
|
596
627
|
endTime: 0,
|
|
@@ -601,6 +632,10 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
601
632
|
}
|
|
602
633
|
|
|
603
634
|
#getCartesiaUrl(): string {
|
|
635
|
+
// The Cartesia /stt/turns/websocket endpoint only accepts model, sample_rate
|
|
636
|
+
// and encoding — there is no `language` query param. Language selection is
|
|
637
|
+
// expressed through the model (ink-2 for English, ink-whisper otherwise),
|
|
638
|
+
// so #opts.language is used only to tag emitted transcripts.
|
|
604
639
|
const params = new URLSearchParams({
|
|
605
640
|
model: this.#opts.model,
|
|
606
641
|
sample_rate: this.#opts.sampleRate.toString(),
|