@livekit/agents-plugin-assemblyai 1.4.9 → 1.5.0

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 CHANGED
@@ -22,7 +22,7 @@ class AssemblyAIPlugin extends import_agents.Plugin {
22
22
  constructor() {
23
23
  super({
24
24
  title: "assemblyai",
25
- version: "1.4.9",
25
+ version: "1.5.0",
26
26
  package: "@livekit/agents-plugin-assemblyai"
27
27
  });
28
28
  }
package/dist/index.js CHANGED
@@ -5,7 +5,7 @@ class AssemblyAIPlugin extends Plugin {
5
5
  constructor() {
6
6
  super({
7
7
  title: "assemblyai",
8
- version: "1.4.9",
8
+ version: "1.5.0",
9
9
  package: "@livekit/agents-plugin-assemblyai"
10
10
  });
11
11
  }
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/models.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2026 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\nexport type STTModels =\n | 'universal-streaming-english'\n | 'universal-streaming-multilingual'\n | 'u3-rt-pro'\n | 'u3-rt-pro-beta-1'\n | 'universal-3-5-pro'\n // Deprecated alias — AssemblyAI maps this to `u3-rt-pro` server-side, but the\n // Python plugin emits a warning and rewrites it. Kept here so TS users don't\n // break if they already pass it.\n | 'u3-pro';\n\nexport type STTEncoding = 'pcm_s16le' | 'pcm_mulaw';\n\nexport type VoiceFocus = 'near-field' | 'far-field';\n"],"mappings":";;;;;;;;;;;;;;AAAA;AAAA;","names":[]}
1
+ {"version":3,"sources":["../src/models.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2026 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\nexport type STTModels =\n | 'universal-streaming-english'\n | 'universal-streaming-multilingual'\n | 'u3-rt-pro'\n | 'u3-rt-pro-beta-1'\n | 'universal-3-5-pro'\n // Deprecated alias — AssemblyAI maps this to `universal-3-5-pro`, but the\n // Python plugin emits a warning and rewrites it. Kept here so TS users don't\n // break if they already pass it.\n | 'u3-pro';\n\nexport type STTEncoding = 'pcm_s16le' | 'pcm_mulaw';\n\nexport type VoiceFocus = 'near-field' | 'far-field';\n"],"mappings":";;;;;;;;;;;;;;AAAA;AAAA;","names":[]}
@@ -0,0 +1,122 @@
1
+ "use strict";
2
+ var import_agents = require("@livekit/agents");
3
+ var import_rtc_node = require("@livekit/rtc-node");
4
+ var import_node_events = require("node:events");
5
+ var import_vitest = require("vitest");
6
+ var import_ws = require("ws");
7
+ var import_stt = require("./stt.cjs");
8
+ function makeFrame(samplesPerChannel = 800, sampleRate = 16e3) {
9
+ const data = new Int16Array(samplesPerChannel);
10
+ data.fill(1);
11
+ return new import_rtc_node.AudioFrame(data, sampleRate, 1, samplesPerChannel);
12
+ }
13
+ async function startWebSocketServer() {
14
+ const wss = new import_ws.WebSocketServer({ host: "127.0.0.1", port: 0 });
15
+ await (0, import_node_events.once)(wss, "listening");
16
+ const address = wss.address();
17
+ return { wss, baseUrl: `ws://127.0.0.1:${address.port}` };
18
+ }
19
+ async function closeWebSocketServer(wss) {
20
+ await new Promise((resolve) => wss.close(() => resolve()));
21
+ }
22
+ async function waitUntil(predicate, timeoutMs = 1e3) {
23
+ const startedAt = Date.now();
24
+ while (Date.now() - startedAt < timeoutMs) {
25
+ if (predicate()) return;
26
+ await new Promise((resolve) => setTimeout(resolve, 10));
27
+ }
28
+ throw new Error("timed out waiting for condition");
29
+ }
30
+ async function collectUntilEnd(stream) {
31
+ const events = [];
32
+ for await (const event of stream) {
33
+ events.push(event);
34
+ if (event.type === import_agents.stt.SpeechEventType.END_OF_SPEECH) break;
35
+ }
36
+ return events;
37
+ }
38
+ (0, import_vitest.describe)("AssemblyAI STT metadata", () => {
39
+ (0, import_vitest.it)("maps turn confidence fields onto speech data metadata", async () => {
40
+ const { wss, baseUrl } = await startWebSocketServer();
41
+ let requestUrl = "";
42
+ wss.on("connection", (ws, req) => {
43
+ requestUrl = req.url ?? "";
44
+ ws.on("message", () => {
45
+ ws.send(
46
+ JSON.stringify({
47
+ type: "Turn",
48
+ transcript: "hola mundo",
49
+ utterance: "hola mundo",
50
+ end_of_turn: true,
51
+ language_code: "es",
52
+ language_confidence: 0.94,
53
+ words: [
54
+ { text: "hola", start: 0, end: 200, confidence: 0.96 },
55
+ { text: "mundo", start: 200, end: 500, confidence: 0.98 }
56
+ ]
57
+ })
58
+ );
59
+ });
60
+ });
61
+ try {
62
+ const assemblyai = new import_stt.STT({
63
+ apiKey: "test-key",
64
+ baseUrl,
65
+ speechModel: "u3-rt-pro"
66
+ });
67
+ const stream = assemblyai.stream({
68
+ connOptions: { maxRetry: 0, retryIntervalMs: 1, timeoutMs: 1e3 }
69
+ });
70
+ await waitUntil(() => requestUrl !== "");
71
+ stream.pushFrame(makeFrame());
72
+ stream.endInput();
73
+ const events = await collectUntilEnd(stream);
74
+ stream.close();
75
+ (0, import_vitest.expect)(new URL(`ws://127.0.0.1${requestUrl}`).pathname).toBe("/v3/ws");
76
+ (0, import_vitest.expect)(
77
+ events.filter((event) => {
78
+ var _a;
79
+ return (_a = event.alternatives) == null ? void 0 : _a[0];
80
+ }).map((event) => {
81
+ var _a, _b, _c, _d;
82
+ return {
83
+ type: event.type,
84
+ language: (_b = (_a = event.alternatives) == null ? void 0 : _a[0]) == null ? void 0 : _b.language,
85
+ metadata: (_d = (_c = event.alternatives) == null ? void 0 : _c[0]) == null ? void 0 : _d.metadata
86
+ };
87
+ })
88
+ ).toEqual([
89
+ {
90
+ type: import_agents.stt.SpeechEventType.INTERIM_TRANSCRIPT,
91
+ language: "es",
92
+ metadata: {
93
+ assemblyai: {
94
+ languageConfidence: 0.94
95
+ }
96
+ }
97
+ },
98
+ {
99
+ type: import_agents.stt.SpeechEventType.PREFLIGHT_TRANSCRIPT,
100
+ language: "es",
101
+ metadata: {
102
+ assemblyai: {
103
+ languageConfidence: 0.94
104
+ }
105
+ }
106
+ },
107
+ {
108
+ type: import_agents.stt.SpeechEventType.FINAL_TRANSCRIPT,
109
+ language: "es",
110
+ metadata: {
111
+ assemblyai: {
112
+ languageConfidence: 0.94
113
+ }
114
+ }
115
+ }
116
+ ]);
117
+ } finally {
118
+ await closeWebSocketServer(wss);
119
+ }
120
+ });
121
+ });
122
+ //# sourceMappingURL=stt-metadata.test.cjs.map
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["../src/stt-metadata.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2026 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { stt as sttLib } from '@livekit/agents';\nimport { AudioFrame } from '@livekit/rtc-node';\nimport { once } from 'node:events';\nimport type { AddressInfo } from 'node:net';\nimport { describe, expect, it } from 'vitest';\nimport { WebSocketServer } from 'ws';\nimport { STT } from './stt.js';\n\nfunction makeFrame(samplesPerChannel = 800, sampleRate = 16000): AudioFrame {\n const data = new Int16Array(samplesPerChannel);\n data.fill(1);\n return new AudioFrame(data, sampleRate, 1, samplesPerChannel);\n}\n\nasync function startWebSocketServer() {\n const wss = new WebSocketServer({ host: '127.0.0.1', port: 0 });\n await once(wss, 'listening');\n const address = wss.address() as AddressInfo;\n return { wss, baseUrl: `ws://127.0.0.1:${address.port}` };\n}\n\nasync function closeWebSocketServer(wss: WebSocketServer): Promise<void> {\n await new Promise<void>((resolve) => wss.close(() => resolve()));\n}\n\nasync function waitUntil(predicate: () => boolean, timeoutMs = 1000): Promise<void> {\n const startedAt = Date.now();\n while (Date.now() - startedAt < timeoutMs) {\n if (predicate()) return;\n await new Promise((resolve) => setTimeout(resolve, 10));\n }\n throw new Error('timed out waiting for condition');\n}\n\nasync function collectUntilEnd(stream: sttLib.SpeechStream): Promise<sttLib.SpeechEvent[]> {\n const events: sttLib.SpeechEvent[] = [];\n for await (const event of stream) {\n events.push(event);\n if (event.type === sttLib.SpeechEventType.END_OF_SPEECH) break;\n }\n return events;\n}\n\ndescribe('AssemblyAI STT metadata', () => {\n it('maps turn confidence fields onto speech data metadata', async () => {\n const { wss, baseUrl } = await startWebSocketServer();\n let requestUrl = '';\n\n wss.on('connection', (ws, req) => {\n requestUrl = req.url ?? '';\n ws.on('message', () => {\n ws.send(\n JSON.stringify({\n type: 'Turn',\n transcript: 'hola mundo',\n utterance: 'hola mundo',\n end_of_turn: true,\n language_code: 'es',\n language_confidence: 0.94,\n words: [\n { text: 'hola', start: 0, end: 200, confidence: 0.96 },\n { text: 'mundo', start: 200, end: 500, confidence: 0.98 },\n ],\n }),\n );\n });\n });\n\n try {\n const assemblyai = new STT({\n apiKey: 'test-key',\n baseUrl,\n speechModel: 'u3-rt-pro',\n });\n const stream = assemblyai.stream({\n connOptions: { maxRetry: 0, retryIntervalMs: 1, timeoutMs: 1000 },\n });\n\n await waitUntil(() => requestUrl !== '');\n\n stream.pushFrame(makeFrame());\n stream.endInput();\n\n const events = await collectUntilEnd(stream);\n stream.close();\n\n expect(new URL(`ws://127.0.0.1${requestUrl}`).pathname).toBe('/v3/ws');\n expect(\n events\n .filter((event) => event.alternatives?.[0])\n .map((event) => ({\n type: event.type,\n language: event.alternatives?.[0]?.language,\n metadata: event.alternatives?.[0]?.metadata,\n })),\n ).toEqual([\n {\n type: sttLib.SpeechEventType.INTERIM_TRANSCRIPT,\n language: 'es',\n metadata: {\n assemblyai: {\n languageConfidence: 0.94,\n },\n },\n },\n {\n type: sttLib.SpeechEventType.PREFLIGHT_TRANSCRIPT,\n language: 'es',\n metadata: {\n assemblyai: {\n languageConfidence: 0.94,\n },\n },\n },\n {\n type: sttLib.SpeechEventType.FINAL_TRANSCRIPT,\n language: 'es',\n metadata: {\n assemblyai: {\n languageConfidence: 0.94,\n },\n },\n },\n ]);\n } finally {\n await closeWebSocketServer(wss);\n }\n });\n});\n"],"mappings":";AAGA,oBAA8B;AAC9B,sBAA2B;AAC3B,yBAAqB;AAErB,oBAAqC;AACrC,gBAAgC;AAChC,iBAAoB;AAEpB,SAAS,UAAU,oBAAoB,KAAK,aAAa,MAAmB;AAC1E,QAAM,OAAO,IAAI,WAAW,iBAAiB;AAC7C,OAAK,KAAK,CAAC;AACX,SAAO,IAAI,2BAAW,MAAM,YAAY,GAAG,iBAAiB;AAC9D;AAEA,eAAe,uBAAuB;AACpC,QAAM,MAAM,IAAI,0BAAgB,EAAE,MAAM,aAAa,MAAM,EAAE,CAAC;AAC9D,YAAM,yBAAK,KAAK,WAAW;AAC3B,QAAM,UAAU,IAAI,QAAQ;AAC5B,SAAO,EAAE,KAAK,SAAS,kBAAkB,QAAQ,IAAI,GAAG;AAC1D;AAEA,eAAe,qBAAqB,KAAqC;AACvE,QAAM,IAAI,QAAc,CAAC,YAAY,IAAI,MAAM,MAAM,QAAQ,CAAC,CAAC;AACjE;AAEA,eAAe,UAAU,WAA0B,YAAY,KAAqB;AAClF,QAAM,YAAY,KAAK,IAAI;AAC3B,SAAO,KAAK,IAAI,IAAI,YAAY,WAAW;AACzC,QAAI,UAAU,EAAG;AACjB,UAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,EAAE,CAAC;AAAA,EACxD;AACA,QAAM,IAAI,MAAM,iCAAiC;AACnD;AAEA,eAAe,gBAAgB,QAA4D;AACzF,QAAM,SAA+B,CAAC;AACtC,mBAAiB,SAAS,QAAQ;AAChC,WAAO,KAAK,KAAK;AACjB,QAAI,MAAM,SAAS,cAAAA,IAAO,gBAAgB,cAAe;AAAA,EAC3D;AACA,SAAO;AACT;AAAA,IAEA,wBAAS,2BAA2B,MAAM;AACxC,wBAAG,yDAAyD,YAAY;AACtE,UAAM,EAAE,KAAK,QAAQ,IAAI,MAAM,qBAAqB;AACpD,QAAI,aAAa;AAEjB,QAAI,GAAG,cAAc,CAAC,IAAI,QAAQ;AAChC,mBAAa,IAAI,OAAO;AACxB,SAAG,GAAG,WAAW,MAAM;AACrB,WAAG;AAAA,UACD,KAAK,UAAU;AAAA,YACb,MAAM;AAAA,YACN,YAAY;AAAA,YACZ,WAAW;AAAA,YACX,aAAa;AAAA,YACb,eAAe;AAAA,YACf,qBAAqB;AAAA,YACrB,OAAO;AAAA,cACL,EAAE,MAAM,QAAQ,OAAO,GAAG,KAAK,KAAK,YAAY,KAAK;AAAA,cACrD,EAAE,MAAM,SAAS,OAAO,KAAK,KAAK,KAAK,YAAY,KAAK;AAAA,YAC1D;AAAA,UACF,CAAC;AAAA,QACH;AAAA,MACF,CAAC;AAAA,IACH,CAAC;AAED,QAAI;AACF,YAAM,aAAa,IAAI,eAAI;AAAA,QACzB,QAAQ;AAAA,QACR;AAAA,QACA,aAAa;AAAA,MACf,CAAC;AACD,YAAM,SAAS,WAAW,OAAO;AAAA,QAC/B,aAAa,EAAE,UAAU,GAAG,iBAAiB,GAAG,WAAW,IAAK;AAAA,MAClE,CAAC;AAED,YAAM,UAAU,MAAM,eAAe,EAAE;AAEvC,aAAO,UAAU,UAAU,CAAC;AAC5B,aAAO,SAAS;AAEhB,YAAM,SAAS,MAAM,gBAAgB,MAAM;AAC3C,aAAO,MAAM;AAEb,gCAAO,IAAI,IAAI,iBAAiB,UAAU,EAAE,EAAE,QAAQ,EAAE,KAAK,QAAQ;AACrE;AAAA,QACE,OACG,OAAO,CAAC,UAAO;AA5F1B;AA4F6B,6BAAM,iBAAN,mBAAqB;AAAA,SAAE,EACzC,IAAI,CAAC,UAAO;AA7FvB;AA6F2B;AAAA,YACf,MAAM,MAAM;AAAA,YACZ,WAAU,iBAAM,iBAAN,mBAAqB,OAArB,mBAAyB;AAAA,YACnC,WAAU,iBAAM,iBAAN,mBAAqB,OAArB,mBAAyB;AAAA,UACrC;AAAA,SAAE;AAAA,MACN,EAAE,QAAQ;AAAA,QACR;AAAA,UACE,MAAM,cAAAA,IAAO,gBAAgB;AAAA,UAC7B,UAAU;AAAA,UACV,UAAU;AAAA,YACR,YAAY;AAAA,cACV,oBAAoB;AAAA,YACtB;AAAA,UACF;AAAA,QACF;AAAA,QACA;AAAA,UACE,MAAM,cAAAA,IAAO,gBAAgB;AAAA,UAC7B,UAAU;AAAA,UACV,UAAU;AAAA,YACR,YAAY;AAAA,cACV,oBAAoB;AAAA,YACtB;AAAA,UACF;AAAA,QACF;AAAA,QACA;AAAA,UACE,MAAM,cAAAA,IAAO,gBAAgB;AAAA,UAC7B,UAAU;AAAA,UACV,UAAU;AAAA,YACR,YAAY;AAAA,cACV,oBAAoB;AAAA,YACtB;AAAA,UACF;AAAA,QACF;AAAA,MACF,CAAC;AAAA,IACH,UAAE;AACA,YAAM,qBAAqB,GAAG;AAAA,IAChC;AAAA,EACF,CAAC;AACH,CAAC;","names":["sttLib"]}
@@ -0,0 +1,2 @@
1
+ export {};
2
+ //# sourceMappingURL=stt-metadata.test.d.ts.map
@@ -0,0 +1,2 @@
1
+ export {};
2
+ //# sourceMappingURL=stt-metadata.test.d.ts.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"stt-metadata.test.d.ts","sourceRoot":"","sources":["../src/stt-metadata.test.ts"],"names":[],"mappings":""}
@@ -0,0 +1,121 @@
1
+ import { stt as sttLib } from "@livekit/agents";
2
+ import { AudioFrame } from "@livekit/rtc-node";
3
+ import { once } from "node:events";
4
+ import { describe, expect, it } from "vitest";
5
+ import { WebSocketServer } from "ws";
6
+ import { STT } from "./stt.js";
7
+ function makeFrame(samplesPerChannel = 800, sampleRate = 16e3) {
8
+ const data = new Int16Array(samplesPerChannel);
9
+ data.fill(1);
10
+ return new AudioFrame(data, sampleRate, 1, samplesPerChannel);
11
+ }
12
+ async function startWebSocketServer() {
13
+ const wss = new WebSocketServer({ host: "127.0.0.1", port: 0 });
14
+ await once(wss, "listening");
15
+ const address = wss.address();
16
+ return { wss, baseUrl: `ws://127.0.0.1:${address.port}` };
17
+ }
18
+ async function closeWebSocketServer(wss) {
19
+ await new Promise((resolve) => wss.close(() => resolve()));
20
+ }
21
+ async function waitUntil(predicate, timeoutMs = 1e3) {
22
+ const startedAt = Date.now();
23
+ while (Date.now() - startedAt < timeoutMs) {
24
+ if (predicate()) return;
25
+ await new Promise((resolve) => setTimeout(resolve, 10));
26
+ }
27
+ throw new Error("timed out waiting for condition");
28
+ }
29
+ async function collectUntilEnd(stream) {
30
+ const events = [];
31
+ for await (const event of stream) {
32
+ events.push(event);
33
+ if (event.type === sttLib.SpeechEventType.END_OF_SPEECH) break;
34
+ }
35
+ return events;
36
+ }
37
+ describe("AssemblyAI STT metadata", () => {
38
+ it("maps turn confidence fields onto speech data metadata", async () => {
39
+ const { wss, baseUrl } = await startWebSocketServer();
40
+ let requestUrl = "";
41
+ wss.on("connection", (ws, req) => {
42
+ requestUrl = req.url ?? "";
43
+ ws.on("message", () => {
44
+ ws.send(
45
+ JSON.stringify({
46
+ type: "Turn",
47
+ transcript: "hola mundo",
48
+ utterance: "hola mundo",
49
+ end_of_turn: true,
50
+ language_code: "es",
51
+ language_confidence: 0.94,
52
+ words: [
53
+ { text: "hola", start: 0, end: 200, confidence: 0.96 },
54
+ { text: "mundo", start: 200, end: 500, confidence: 0.98 }
55
+ ]
56
+ })
57
+ );
58
+ });
59
+ });
60
+ try {
61
+ const assemblyai = new STT({
62
+ apiKey: "test-key",
63
+ baseUrl,
64
+ speechModel: "u3-rt-pro"
65
+ });
66
+ const stream = assemblyai.stream({
67
+ connOptions: { maxRetry: 0, retryIntervalMs: 1, timeoutMs: 1e3 }
68
+ });
69
+ await waitUntil(() => requestUrl !== "");
70
+ stream.pushFrame(makeFrame());
71
+ stream.endInput();
72
+ const events = await collectUntilEnd(stream);
73
+ stream.close();
74
+ expect(new URL(`ws://127.0.0.1${requestUrl}`).pathname).toBe("/v3/ws");
75
+ expect(
76
+ events.filter((event) => {
77
+ var _a;
78
+ return (_a = event.alternatives) == null ? void 0 : _a[0];
79
+ }).map((event) => {
80
+ var _a, _b, _c, _d;
81
+ return {
82
+ type: event.type,
83
+ language: (_b = (_a = event.alternatives) == null ? void 0 : _a[0]) == null ? void 0 : _b.language,
84
+ metadata: (_d = (_c = event.alternatives) == null ? void 0 : _c[0]) == null ? void 0 : _d.metadata
85
+ };
86
+ })
87
+ ).toEqual([
88
+ {
89
+ type: sttLib.SpeechEventType.INTERIM_TRANSCRIPT,
90
+ language: "es",
91
+ metadata: {
92
+ assemblyai: {
93
+ languageConfidence: 0.94
94
+ }
95
+ }
96
+ },
97
+ {
98
+ type: sttLib.SpeechEventType.PREFLIGHT_TRANSCRIPT,
99
+ language: "es",
100
+ metadata: {
101
+ assemblyai: {
102
+ languageConfidence: 0.94
103
+ }
104
+ }
105
+ },
106
+ {
107
+ type: sttLib.SpeechEventType.FINAL_TRANSCRIPT,
108
+ language: "es",
109
+ metadata: {
110
+ assemblyai: {
111
+ languageConfidence: 0.94
112
+ }
113
+ }
114
+ }
115
+ ]);
116
+ } finally {
117
+ await closeWebSocketServer(wss);
118
+ }
119
+ });
120
+ });
121
+ //# sourceMappingURL=stt-metadata.test.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["../src/stt-metadata.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2026 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { stt as sttLib } from '@livekit/agents';\nimport { AudioFrame } from '@livekit/rtc-node';\nimport { once } from 'node:events';\nimport type { AddressInfo } from 'node:net';\nimport { describe, expect, it } from 'vitest';\nimport { WebSocketServer } from 'ws';\nimport { STT } from './stt.js';\n\nfunction makeFrame(samplesPerChannel = 800, sampleRate = 16000): AudioFrame {\n const data = new Int16Array(samplesPerChannel);\n data.fill(1);\n return new AudioFrame(data, sampleRate, 1, samplesPerChannel);\n}\n\nasync function startWebSocketServer() {\n const wss = new WebSocketServer({ host: '127.0.0.1', port: 0 });\n await once(wss, 'listening');\n const address = wss.address() as AddressInfo;\n return { wss, baseUrl: `ws://127.0.0.1:${address.port}` };\n}\n\nasync function closeWebSocketServer(wss: WebSocketServer): Promise<void> {\n await new Promise<void>((resolve) => wss.close(() => resolve()));\n}\n\nasync function waitUntil(predicate: () => boolean, timeoutMs = 1000): Promise<void> {\n const startedAt = Date.now();\n while (Date.now() - startedAt < timeoutMs) {\n if (predicate()) return;\n await new Promise((resolve) => setTimeout(resolve, 10));\n }\n throw new Error('timed out waiting for condition');\n}\n\nasync function collectUntilEnd(stream: sttLib.SpeechStream): Promise<sttLib.SpeechEvent[]> {\n const events: sttLib.SpeechEvent[] = [];\n for await (const event of stream) {\n events.push(event);\n if (event.type === sttLib.SpeechEventType.END_OF_SPEECH) break;\n }\n return events;\n}\n\ndescribe('AssemblyAI STT metadata', () => {\n it('maps turn confidence fields onto speech data metadata', async () => {\n const { wss, baseUrl } = await startWebSocketServer();\n let requestUrl = '';\n\n wss.on('connection', (ws, req) => {\n requestUrl = req.url ?? '';\n ws.on('message', () => {\n ws.send(\n JSON.stringify({\n type: 'Turn',\n transcript: 'hola mundo',\n utterance: 'hola mundo',\n end_of_turn: true,\n language_code: 'es',\n language_confidence: 0.94,\n words: [\n { text: 'hola', start: 0, end: 200, confidence: 0.96 },\n { text: 'mundo', start: 200, end: 500, confidence: 0.98 },\n ],\n }),\n );\n });\n });\n\n try {\n const assemblyai = new STT({\n apiKey: 'test-key',\n baseUrl,\n speechModel: 'u3-rt-pro',\n });\n const stream = assemblyai.stream({\n connOptions: { maxRetry: 0, retryIntervalMs: 1, timeoutMs: 1000 },\n });\n\n await waitUntil(() => requestUrl !== '');\n\n stream.pushFrame(makeFrame());\n stream.endInput();\n\n const events = await collectUntilEnd(stream);\n stream.close();\n\n expect(new URL(`ws://127.0.0.1${requestUrl}`).pathname).toBe('/v3/ws');\n expect(\n events\n .filter((event) => event.alternatives?.[0])\n .map((event) => ({\n type: event.type,\n language: event.alternatives?.[0]?.language,\n metadata: event.alternatives?.[0]?.metadata,\n })),\n ).toEqual([\n {\n type: sttLib.SpeechEventType.INTERIM_TRANSCRIPT,\n language: 'es',\n metadata: {\n assemblyai: {\n languageConfidence: 0.94,\n },\n },\n },\n {\n type: sttLib.SpeechEventType.PREFLIGHT_TRANSCRIPT,\n language: 'es',\n metadata: {\n assemblyai: {\n languageConfidence: 0.94,\n },\n },\n },\n {\n type: sttLib.SpeechEventType.FINAL_TRANSCRIPT,\n language: 'es',\n metadata: {\n assemblyai: {\n languageConfidence: 0.94,\n },\n },\n },\n ]);\n } finally {\n await closeWebSocketServer(wss);\n }\n });\n});\n"],"mappings":"AAGA,SAAS,OAAO,cAAc;AAC9B,SAAS,kBAAkB;AAC3B,SAAS,YAAY;AAErB,SAAS,UAAU,QAAQ,UAAU;AACrC,SAAS,uBAAuB;AAChC,SAAS,WAAW;AAEpB,SAAS,UAAU,oBAAoB,KAAK,aAAa,MAAmB;AAC1E,QAAM,OAAO,IAAI,WAAW,iBAAiB;AAC7C,OAAK,KAAK,CAAC;AACX,SAAO,IAAI,WAAW,MAAM,YAAY,GAAG,iBAAiB;AAC9D;AAEA,eAAe,uBAAuB;AACpC,QAAM,MAAM,IAAI,gBAAgB,EAAE,MAAM,aAAa,MAAM,EAAE,CAAC;AAC9D,QAAM,KAAK,KAAK,WAAW;AAC3B,QAAM,UAAU,IAAI,QAAQ;AAC5B,SAAO,EAAE,KAAK,SAAS,kBAAkB,QAAQ,IAAI,GAAG;AAC1D;AAEA,eAAe,qBAAqB,KAAqC;AACvE,QAAM,IAAI,QAAc,CAAC,YAAY,IAAI,MAAM,MAAM,QAAQ,CAAC,CAAC;AACjE;AAEA,eAAe,UAAU,WAA0B,YAAY,KAAqB;AAClF,QAAM,YAAY,KAAK,IAAI;AAC3B,SAAO,KAAK,IAAI,IAAI,YAAY,WAAW;AACzC,QAAI,UAAU,EAAG;AACjB,UAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,EAAE,CAAC;AAAA,EACxD;AACA,QAAM,IAAI,MAAM,iCAAiC;AACnD;AAEA,eAAe,gBAAgB,QAA4D;AACzF,QAAM,SAA+B,CAAC;AACtC,mBAAiB,SAAS,QAAQ;AAChC,WAAO,KAAK,KAAK;AACjB,QAAI,MAAM,SAAS,OAAO,gBAAgB,cAAe;AAAA,EAC3D;AACA,SAAO;AACT;AAEA,SAAS,2BAA2B,MAAM;AACxC,KAAG,yDAAyD,YAAY;AACtE,UAAM,EAAE,KAAK,QAAQ,IAAI,MAAM,qBAAqB;AACpD,QAAI,aAAa;AAEjB,QAAI,GAAG,cAAc,CAAC,IAAI,QAAQ;AAChC,mBAAa,IAAI,OAAO;AACxB,SAAG,GAAG,WAAW,MAAM;AACrB,WAAG;AAAA,UACD,KAAK,UAAU;AAAA,YACb,MAAM;AAAA,YACN,YAAY;AAAA,YACZ,WAAW;AAAA,YACX,aAAa;AAAA,YACb,eAAe;AAAA,YACf,qBAAqB;AAAA,YACrB,OAAO;AAAA,cACL,EAAE,MAAM,QAAQ,OAAO,GAAG,KAAK,KAAK,YAAY,KAAK;AAAA,cACrD,EAAE,MAAM,SAAS,OAAO,KAAK,KAAK,KAAK,YAAY,KAAK;AAAA,YAC1D;AAAA,UACF,CAAC;AAAA,QACH;AAAA,MACF,CAAC;AAAA,IACH,CAAC;AAED,QAAI;AACF,YAAM,aAAa,IAAI,IAAI;AAAA,QACzB,QAAQ;AAAA,QACR;AAAA,QACA,aAAa;AAAA,MACf,CAAC;AACD,YAAM,SAAS,WAAW,OAAO;AAAA,QAC/B,aAAa,EAAE,UAAU,GAAG,iBAAiB,GAAG,WAAW,IAAK;AAAA,MAClE,CAAC;AAED,YAAM,UAAU,MAAM,eAAe,EAAE;AAEvC,aAAO,UAAU,UAAU,CAAC;AAC5B,aAAO,SAAS;AAEhB,YAAM,SAAS,MAAM,gBAAgB,MAAM;AAC3C,aAAO,MAAM;AAEb,aAAO,IAAI,IAAI,iBAAiB,UAAU,EAAE,EAAE,QAAQ,EAAE,KAAK,QAAQ;AACrE;AAAA,QACE,OACG,OAAO,CAAC,UAAO;AA5F1B;AA4F6B,6BAAM,iBAAN,mBAAqB;AAAA,SAAE,EACzC,IAAI,CAAC,UAAO;AA7FvB;AA6F2B;AAAA,YACf,MAAM,MAAM;AAAA,YACZ,WAAU,iBAAM,iBAAN,mBAAqB,OAArB,mBAAyB;AAAA,YACnC,WAAU,iBAAM,iBAAN,mBAAqB,OAArB,mBAAyB;AAAA,UACrC;AAAA,SAAE;AAAA,MACN,EAAE,QAAQ;AAAA,QACR;AAAA,UACE,MAAM,OAAO,gBAAgB;AAAA,UAC7B,UAAU;AAAA,UACV,UAAU;AAAA,YACR,YAAY;AAAA,cACV,oBAAoB;AAAA,YACtB;AAAA,UACF;AAAA,QACF;AAAA,QACA;AAAA,UACE,MAAM,OAAO,gBAAgB;AAAA,UAC7B,UAAU;AAAA,UACV,UAAU;AAAA,YACR,YAAY;AAAA,cACV,oBAAoB;AAAA,YACtB;AAAA,UACF;AAAA,QACF;AAAA,QACA;AAAA,UACE,MAAM,OAAO,gBAAgB;AAAA,UAC7B,UAAU;AAAA,UACV,UAAU;AAAA,YACR,YAAY;AAAA,cACV,oBAAoB;AAAA,YACtB;AAAA,UACF;AAAA,QACF;AAAA,MACF,CAAC;AAAA,IACH,UAAE;AACA,YAAM,qBAAqB,GAAG;AAAA,IAChC;AAAA,EACF,CAAC;AACH,CAAC;","names":[]}
package/dist/stt.cjs CHANGED
@@ -28,6 +28,14 @@ const U3_PRO_MODELS = ["u3-rt-pro", "u3-rt-pro-beta-1", "universal-3-5-pro"];
28
28
  function isU3ProModel(model) {
29
29
  return U3_PRO_MODELS.includes(model);
30
30
  }
31
+ function speechDataMetadata(data) {
32
+ const assemblyai = {};
33
+ if (typeof data.language_confidence === "number") {
34
+ assemblyai.languageConfidence = data.language_confidence;
35
+ }
36
+ if (Object.keys(assemblyai).length === 0) return void 0;
37
+ return { assemblyai };
38
+ }
31
39
  const defaultSTTOptions = {
32
40
  apiKey: process.env.ASSEMBLYAI_API_KEY,
33
41
  sampleRate: 16e3,
@@ -53,12 +61,19 @@ class STT extends import_agents.stt.STT {
53
61
  alignedTranscript: "word"
54
62
  });
55
63
  if (opts.speechModel === "u3-pro") {
56
- (0, import_agents.log)().warn("'u3-pro' is deprecated, use 'u3-rt-pro' instead.");
57
- opts.speechModel = "u3-rt-pro";
64
+ (0, import_agents.log)().warn("'u3-pro' is deprecated, use 'universal-3-5-pro' instead.");
65
+ opts.speechModel = "universal-3-5-pro";
58
66
  }
59
67
  const speechModel = opts.speechModel ?? defaultSTTOptions.speechModel;
60
68
  if (!isU3ProModel(speechModel)) {
61
- for (const param of ["prompt", "voiceFocus", "voiceFocusThreshold", "mode"]) {
69
+ for (const param of [
70
+ "prompt",
71
+ "agentContext",
72
+ "previousContextNTurns",
73
+ "voiceFocus",
74
+ "voiceFocusThreshold",
75
+ "mode"
76
+ ]) {
62
77
  if (opts[param] !== void 0) {
63
78
  throw new Error(
64
79
  `The '${param}' parameter is only supported with the ${U3_PRO_MODELS.join(", ")} models.`
@@ -72,7 +87,7 @@ class STT extends import_agents.stt.STT {
72
87
  "AssemblyAI API key is required. Pass one in via the `apiKey` parameter, or set it as the `ASSEMBLYAI_API_KEY` environment variable"
73
88
  );
74
89
  }
75
- const minTurnSilence = opts.minTurnSilence ?? 100;
90
+ const minTurnSilence = opts.minTurnSilence ?? (opts.mode === void 0 ? 100 : void 0);
76
91
  this.#opts = {
77
92
  ...defaultSTTOptions,
78
93
  ...opts,
@@ -136,6 +151,7 @@ class SpeechStream extends import_agents.stt.SpeechStream {
136
151
  this.#opts = { ...this.#opts, ...opts };
137
152
  const configMsg = { type: "UpdateConfiguration" };
138
153
  if (opts.prompt !== void 0) configMsg.prompt = opts.prompt;
154
+ if (opts.agentContext !== void 0) configMsg.agent_context = opts.agentContext;
139
155
  if (opts.keytermsPrompt !== void 0) configMsg.keyterms_prompt = opts.keytermsPrompt;
140
156
  if (opts.maxTurnSilence !== void 0) configMsg.max_turn_silence = opts.maxTurnSilence;
141
157
  if (opts.minTurnSilence !== void 0) configMsg.min_turn_silence = opts.minTurnSilence;
@@ -188,7 +204,7 @@ class SpeechStream extends import_agents.stt.SpeechStream {
188
204
  let minSilence = this.#opts.minTurnSilence;
189
205
  let maxSilence = this.#opts.maxTurnSilence;
190
206
  if (isU3ProModel(this.#opts.speechModel)) {
191
- if (minSilence === void 0) minSilence = 100;
207
+ if (minSilence === void 0 && this.#opts.mode === void 0) minSilence = 100;
192
208
  if (maxSilence === void 0) maxSilence = minSilence;
193
209
  }
194
210
  const defaultLanguageDetection = this.#opts.speechModel.includes("multilingual") || isU3ProModel(this.#opts.speechModel);
@@ -203,7 +219,10 @@ class SpeechStream extends import_agents.stt.SpeechStream {
203
219
  max_turn_silence: maxSilence,
204
220
  keyterms_prompt: this.#opts.keytermsPrompt !== void 0 ? JSON.stringify(this.#opts.keytermsPrompt) : void 0,
205
221
  language_detection: languageDetection,
222
+ inactivity_timeout: this.#opts.inactivityTimeout,
206
223
  prompt: this.#opts.prompt,
224
+ agent_context: this.#opts.agentContext,
225
+ previous_context_n_turns: this.#opts.previousContextNTurns,
207
226
  vad_threshold: this.#opts.vadThreshold,
208
227
  speaker_labels: this.#opts.speakerLabels,
209
228
  max_speakers: this.#opts.maxSpeakers,
@@ -367,6 +386,7 @@ class SpeechStream extends import_agents.stt.SpeechStream {
367
386
  const utterance = data.utterance ?? "";
368
387
  const transcript = data.transcript ?? "";
369
388
  const language = (0, import_agents.normalizeLanguage)(data.language_code ?? "en");
389
+ const metadata = speechDataMetadata(data);
370
390
  const timedWords = words.map(
371
391
  (word) => (0, import_agents.createTimedString)({
372
392
  text: word.text ?? "",
@@ -393,7 +413,8 @@ class SpeechStream extends import_agents.stt.SpeechStream {
393
413
  startTime,
394
414
  endTime,
395
415
  confidence,
396
- words: timedWords
416
+ words: timedWords,
417
+ ...metadata ? { metadata } : {}
397
418
  }
398
419
  ]
399
420
  });
@@ -415,7 +436,8 @@ class SpeechStream extends import_agents.stt.SpeechStream {
415
436
  startTime: this.#lastPreflightStartTime,
416
437
  endTime,
417
438
  confidence: utteranceConfidence,
418
- words: utteranceWords
439
+ words: utteranceWords,
440
+ ...metadata ? { metadata } : {}
419
441
  }
420
442
  ]
421
443
  });
@@ -432,7 +454,8 @@ class SpeechStream extends import_agents.stt.SpeechStream {
432
454
  startTime,
433
455
  endTime,
434
456
  confidence,
435
- words: timedWords
457
+ words: timedWords,
458
+ ...metadata ? { metadata } : {}
436
459
  }
437
460
  ]
438
461
  });
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\n//\nimport {\n type APIConnectOptions,\n type AudioBuffer,\n AudioByteStream,\n Future,\n Task,\n createTimedString,\n delay,\n log,\n normalizeLanguage,\n stt,\n waitForAbort,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { RawData } from 'ws';\nimport { WebSocket } from 'ws';\nimport type { STTEncoding, STTModels, VoiceFocus } from './models.js';\n\nconst U3_PRO_MODELS = ['u3-rt-pro', 'u3-rt-pro-beta-1', 'universal-3-5-pro'] as const;\n\nfunction isU3ProModel(model: STTModels): boolean {\n return U3_PRO_MODELS.includes(model as (typeof U3_PRO_MODELS)[number]);\n}\n\n// AssemblyAI Universal-Streaming (v3) message envelope. All fields are optional\n// since we narrow on `type` before reading anything else.\ninterface StreamEventMessage {\n type?: 'Begin' | 'SpeechStarted' | 'Turn' | 'Termination' | string;\n // Begin\n id?: string;\n expires_at?: number;\n // Turn\n transcript?: string;\n utterance?: string;\n end_of_turn?: boolean;\n end_of_turn_confidence?: number;\n turn_is_formatted?: boolean;\n language_code?: string;\n speaker_label?: string;\n words?: Array<{\n text?: string;\n start?: number;\n end?: number;\n confidence?: number;\n speaker?: string;\n }>;\n // Termination\n audio_duration_seconds?: number;\n session_duration_seconds?: number;\n}\n\nexport interface STTOptions {\n apiKey?: string;\n sampleRate: number;\n /**\n * How large each chunk of audio is before being sent to AssemblyAI, in\n * milliseconds. Corresponds to Python's `buffer_size_seconds` (seconds there,\n * ms here per this repo's time-unit convention).\n */\n bufferSizeMs: number;\n encoding: STTEncoding;\n speechModel: STTModels;\n languageDetection?: boolean;\n endOfTurnConfidenceThreshold?: number;\n /** Minimum silence (ms) before a confident end-of-turn is finalized. */\n minTurnSilence?: number;\n /** Maximum silence (ms) before end-of-turn is forced regardless of confidence. */\n maxTurnSilence?: number;\n formatTurns?: boolean;\n keytermsPrompt?: string[];\n /** Only supported with the `u3-rt-pro` model. */\n prompt?: string;\n vadThreshold?: number;\n /**\n * Enable speaker diarization. Note: AssemblyAI will return per-word speaker\n * labels, but the JS framework's `stt.SpeechData` type does not yet expose\n * a `speakerId` field (unlike the Python framework), so the labels are not\n * currently surfaced on emitted events. Setting this to `true` still has\n * effect server-side. Once the base `SpeechData` interface gains speaker\n * support, `#processStreamEvent` should forward `data.words[].speaker` too.\n */\n speakerLabels?: boolean;\n maxSpeakers?: number;\n domain?: string;\n /** Isolate the primary voice and suppress background noise. Connect-time only. */\n voiceFocus?: VoiceFocus;\n /** Background audio suppression aggressiveness, from 0.0 to 1.0. Connect-time only. */\n voiceFocusThreshold?: number;\n /**\n * Accuracy/latency preset for u3-rt-pro: `min_latency`, `balanced`, or `max_accuracy`.\n * Explicit silence, partials, or VAD options still take precedence over mode defaults.\n */\n mode?: 'min_latency' | 'balanced' | 'max_accuracy';\n baseUrl: string;\n}\n\nconst defaultSTTOptions: STTOptions = {\n apiKey: process.env.ASSEMBLYAI_API_KEY,\n sampleRate: 16000,\n bufferSizeMs: 50,\n encoding: 'pcm_s16le',\n speechModel: 'universal-3-5-pro',\n baseUrl: 'wss://streaming.assemblyai.com',\n};\n\nexport class STT extends stt.STT {\n #opts: STTOptions;\n #streams = new Set<WeakRef<SpeechStream>>();\n label = 'assemblyai.STT';\n\n get model(): string {\n return this.#opts.speechModel;\n }\n\n get provider(): string {\n return 'AssemblyAI';\n }\n\n constructor(opts: Partial<STTOptions> = {}) {\n super({\n streaming: true,\n interimResults: true,\n alignedTranscript: 'word',\n });\n\n if (opts.speechModel === 'u3-pro') {\n log().warn(\"'u3-pro' is deprecated, use 'u3-rt-pro' instead.\");\n opts.speechModel = 'u3-rt-pro';\n }\n\n const speechModel = opts.speechModel ?? defaultSTTOptions.speechModel;\n if (!isU3ProModel(speechModel)) {\n for (const param of ['prompt', 'voiceFocus', 'voiceFocusThreshold', 'mode'] as const) {\n if (opts[param] !== undefined) {\n throw new Error(\n `The '${param}' parameter is only supported with the ${U3_PRO_MODELS.join(', ')} models.`,\n );\n }\n }\n }\n\n const apiKey = opts.apiKey ?? defaultSTTOptions.apiKey;\n if (!apiKey) {\n throw new Error(\n 'AssemblyAI API key is required. Pass one in via the `apiKey` parameter, or set it as the `ASSEMBLYAI_API_KEY` environment variable',\n );\n }\n\n // Minimize latency; matches LK's end-of-turn detector well.\n const minTurnSilence = opts.minTurnSilence ?? 100;\n\n this.#opts = {\n ...defaultSTTOptions,\n ...opts,\n apiKey,\n minTurnSilence,\n };\n }\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n async _recognize(_: AudioBuffer): Promise<stt.SpeechEvent> {\n throw new Error('Non-streaming recognize is not supported on AssemblyAI STT');\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n for (const ref of this.#streams) {\n const stream = ref.deref();\n if (stream) {\n stream.updateOptions(opts);\n } else {\n this.#streams.delete(ref);\n }\n }\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): SpeechStream {\n const stream = new SpeechStream(this, this.#opts, options?.connOptions);\n this.#streams.add(new WeakRef(stream));\n return stream;\n }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n static readonly CLOSE_MSG = JSON.stringify({ type: 'Terminate' });\n\n #opts: STTOptions;\n #logger = log();\n #speechDurationInS = 0;\n #lastPreflightStartTime = 0;\n #pendingConfigMessages: Record<string, unknown>[] = [];\n #configMessagePending = new Future();\n #sessionId: string | null = null;\n #expiresAt: number | null = null;\n label = 'assemblyai.SpeechStream';\n\n constructor(stt: STT, opts: STTOptions, connOptions?: APIConnectOptions) {\n super(stt, opts.sampleRate, connOptions);\n this.#opts = opts;\n this.closed = false;\n }\n\n /**\n * The AssemblyAI session ID. Set when the WebSocket connection is established\n * (before any speech events). Null until the connection completes.\n * Share this with the AssemblyAI team when reporting issues.\n */\n get sessionId(): string | null {\n return this.#sessionId;\n }\n\n /**\n * Unix timestamp when the AssemblyAI session expires. Set alongside\n * {@link sessionId} when the WebSocket connection is established.\n */\n get expiresAt(): number | null {\n return this.#expiresAt;\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n\n const configMsg: Record<string, unknown> = { type: 'UpdateConfiguration' };\n if (opts.prompt !== undefined) configMsg.prompt = opts.prompt;\n if (opts.keytermsPrompt !== undefined) configMsg.keyterms_prompt = opts.keytermsPrompt;\n if (opts.maxTurnSilence !== undefined) configMsg.max_turn_silence = opts.maxTurnSilence;\n if (opts.minTurnSilence !== undefined) configMsg.min_turn_silence = opts.minTurnSilence;\n if (opts.endOfTurnConfidenceThreshold !== undefined) {\n configMsg.end_of_turn_confidence_threshold = opts.endOfTurnConfidenceThreshold;\n }\n if (opts.vadThreshold !== undefined) configMsg.vad_threshold = opts.vadThreshold;\n\n // Only send if any actual fields (besides `type`) were specified.\n if (Object.keys(configMsg).length > 1) {\n this.#pendingConfigMessages.push(configMsg);\n if (!this.#configMessagePending.done) this.#configMessagePending.resolve();\n }\n }\n\n /**\n * Force-finalize the current turn immediately.\n */\n forceEndpoint() {\n this.#pendingConfigMessages.push({ type: 'ForceEndpoint' });\n if (!this.#configMessagePending.done) this.#configMessagePending.resolve();\n }\n\n // Deepgram-style reconnect loop around a single websocket lifetime.\n protected async run() {\n const maxRetry = 32;\n let retries = 0;\n\n while (!this.input.closed && !this.closed) {\n try {\n const ws = await this.#connectWS();\n await this.#runWS(ws);\n retries = 0;\n } catch (e) {\n if (!this.closed && !this.input.closed) {\n if (retries >= maxRetry) {\n throw new Error(`failed to connect to AssemblyAI after ${retries} attempts: ${e}`);\n }\n\n const retryDelaySeconds = Math.min(retries * 5, 10);\n retries++;\n\n this.#logger.warn(\n `failed to connect to AssemblyAI, retrying in ${retryDelaySeconds} seconds: ${e} (${retries}/${maxRetry})`,\n );\n await delay(retryDelaySeconds * 1000);\n } else {\n this.#logger.warn(\n `AssemblyAI disconnected, connection is closed: ${e} (inputClosed: ${this.input.closed}, isClosed: ${this.closed})`,\n );\n }\n }\n }\n\n this.closed = true;\n }\n\n async #connectWS(): Promise<WebSocket> {\n // u3-rt-pro family models default both min and max silence to 100ms when unset.\n let minSilence = this.#opts.minTurnSilence;\n let maxSilence = this.#opts.maxTurnSilence;\n if (isU3ProModel(this.#opts.speechModel)) {\n if (minSilence === undefined) minSilence = 100;\n if (maxSilence === undefined) maxSilence = minSilence;\n }\n\n // Default language_detection to true for multilingual / u3-rt-pro-family models, false otherwise.\n const defaultLanguageDetection =\n this.#opts.speechModel.includes('multilingual') || isU3ProModel(this.#opts.speechModel);\n const languageDetection = this.#opts.languageDetection ?? defaultLanguageDetection;\n\n const liveConfig: Record<string, unknown> = {\n sample_rate: this.#opts.sampleRate,\n encoding: this.#opts.encoding,\n speech_model: this.#opts.speechModel,\n format_turns: this.#opts.formatTurns,\n end_of_turn_confidence_threshold: this.#opts.endOfTurnConfidenceThreshold,\n min_turn_silence: minSilence,\n max_turn_silence: maxSilence,\n keyterms_prompt:\n this.#opts.keytermsPrompt !== undefined\n ? JSON.stringify(this.#opts.keytermsPrompt)\n : undefined,\n language_detection: languageDetection,\n prompt: this.#opts.prompt,\n vad_threshold: this.#opts.vadThreshold,\n speaker_labels: this.#opts.speakerLabels,\n max_speakers: this.#opts.maxSpeakers,\n domain: this.#opts.domain,\n voice_focus: this.#opts.voiceFocus,\n voice_focus_threshold: this.#opts.voiceFocusThreshold,\n mode: this.#opts.mode,\n };\n\n const url = new URL(`${this.#opts.baseUrl}/v3/ws`);\n // Python serializes booleans as the strings \"true\"/\"false\", so we mirror that.\n for (const [key, value] of Object.entries(liveConfig)) {\n if (value === undefined || value === null) continue;\n if (typeof value === 'boolean') {\n url.searchParams.append(key, value ? 'true' : 'false');\n } else {\n url.searchParams.append(key, String(value));\n }\n }\n\n const ws = new WebSocket(url, {\n headers: {\n Authorization: this.#opts.apiKey!,\n 'Content-Type': 'application/json',\n 'User-Agent': 'AssemblyAI/1.0 (integration=Livekit)',\n },\n });\n\n await new Promise<void>((resolve, reject) => {\n ws.on('open', () => resolve());\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(new Error(`WebSocket returned ${code}`)));\n });\n\n return ws;\n }\n\n async #runWS(ws: WebSocket) {\n let closing = false;\n const sessionController = new AbortController();\n\n // gets cancelled also when sendTask is complete\n const wsMonitor = Task.from(async (controller) => {\n const closed = new Promise<void>((_, reject) => {\n ws.once('close', (code, reason) => {\n if (!closing) {\n this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);\n reject(new Error('WebSocket closed'));\n }\n });\n });\n\n await Promise.race([closed, waitForAbort(controller.signal)]);\n });\n\n const sendTask = async () => {\n const samplesPerBuffer = Math.floor((this.#opts.sampleRate * this.#opts.bufferSizeMs) / 1000);\n const audioStream = new AudioByteStream(this.#opts.sampleRate, 1, samplesPerBuffer);\n\n const abortPromise = waitForAbort(this.abortSignal);\n const sessionAbort = waitForAbort(sessionController.signal);\n\n try {\n while (!this.closed) {\n const result = await Promise.race([this.input.next(), abortPromise, sessionAbort]);\n\n if (result === undefined) return; // aborted\n if (result.done) break;\n\n const data = result.value;\n\n let frames: AudioFrame[];\n if (data === SpeechStream.FLUSH_SENTINEL) {\n frames = audioStream.flush();\n } else if (data.sampleRate === this.#opts.sampleRate && data.channels === 1) {\n // AssemblyAI expects mono PCM. The base SpeechStream only resamples\n // sample rate, so reject any frame that is not already downmixed.\n frames = audioStream.write(data.data.buffer as ArrayBuffer);\n } else {\n throw new Error('sample rate or channel count of frame does not match');\n }\n\n for (const frame of frames) {\n this.#speechDurationInS += frame.samplesPerChannel / frame.sampleRate;\n ws.send(frame.data.buffer);\n }\n }\n } finally {\n closing = true;\n try {\n ws.send(SpeechStream.CLOSE_MSG);\n } catch {\n // ignore — socket may already be closing\n }\n wsMonitor.cancel();\n }\n };\n\n let messageHandler: ((msg: RawData, isBinary: boolean) => void) | null = null;\n const listenTask = Task.from(async (controller) => {\n const listenMessage = new Promise<void>((resolve, reject) => {\n messageHandler = (msg, isBinary) => {\n if (isBinary) {\n this.#logger.error('unexpected binary message from AssemblyAI');\n return;\n }\n try {\n const json = JSON.parse(msg.toString()) as StreamEventMessage;\n this.#processStreamEvent(json);\n if (this.closed || closing) {\n resolve();\n }\n } catch (err) {\n this.#logger.error(`AssemblyAI: error processing message: ${msg}`);\n reject(err);\n }\n };\n ws.on('message', messageHandler);\n });\n\n await Promise.race([listenMessage, waitForAbort(controller.signal)]);\n });\n\n const configTask = Task.from(async (controller) => {\n // Drain any messages queued while the socket was reconnecting.\n while (this.#pendingConfigMessages.length > 0) {\n const msg = this.#pendingConfigMessages.shift()!;\n ws.send(JSON.stringify(msg));\n }\n\n while (!controller.signal.aborted) {\n await Promise.race([this.#configMessagePending.await, waitForAbort(controller.signal)]);\n if (controller.signal.aborted) return;\n\n this.#configMessagePending = new Future();\n while (this.#pendingConfigMessages.length > 0) {\n const msg = this.#pendingConfigMessages.shift()!;\n ws.send(JSON.stringify(msg));\n }\n }\n });\n\n try {\n await Promise.all([sendTask(), listenTask.result, wsMonitor.result]);\n } finally {\n closing = true;\n sessionController.abort();\n listenTask.cancel();\n configTask.cancel();\n if (messageHandler) ws.off('message', messageHandler);\n try {\n ws.close();\n } catch {\n // ignore\n }\n }\n }\n\n #averageConfidence(words: Array<{ confidence?: number }>): number {\n if (words.length === 0) return 0;\n return words.reduce((sum, w) => sum + (w.confidence ?? 0), 0) / words.length;\n }\n\n #processStreamEvent(data: StreamEventMessage) {\n const messageType = data.type;\n\n if (messageType === 'Begin') {\n this.#sessionId = data.id ?? null;\n this.#expiresAt = data.expires_at ?? null;\n this.#logger.info(\n `AssemblyAI session started id=${this.#sessionId} expires_at=${this.#expiresAt}`,\n );\n return;\n }\n\n if (messageType === 'SpeechStarted') {\n this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n return;\n }\n\n if (messageType === 'Termination') {\n this.#logger.debug(\n `AssemblyAI session terminated audio_duration=${data.audio_duration_seconds}s session_duration=${data.session_duration_seconds}s`,\n );\n return;\n }\n\n if (messageType !== 'Turn') {\n return;\n }\n\n const words = data.words ?? [];\n const endOfTurn = Boolean(data.end_of_turn);\n const turnIsFormatted = Boolean(data.turn_is_formatted);\n const utterance = data.utterance ?? '';\n const transcript = data.transcript ?? '';\n const language = normalizeLanguage(data.language_code ?? 'en');\n\n // Word timestamps are in milliseconds:\n // https://www.assemblyai.com/docs/api-reference/streaming-api/streaming-api#receive.receiveTurn.words\n const timedWords = words.map((word) =>\n createTimedString({\n text: word.text ?? '',\n startTime: (word.start ?? 0) / 1000 + this.startTimeOffset,\n endTime: (word.end ?? 0) / 1000 + this.startTimeOffset,\n confidence: word.confidence ?? 0,\n startTimeOffset: this.startTimeOffset,\n }),\n );\n\n let startTime = 0;\n let endTime = 0;\n let confidence = 0;\n\n // `words` are cumulative for the turn — emit as an interim transcript.\n if (timedWords.length > 0) {\n const interimText = timedWords.map((w) => w.text).join(' ');\n startTime = timedWords[0]!.startTime ?? 0;\n endTime = timedWords[timedWords.length - 1]!.endTime ?? 0;\n confidence = this.#averageConfidence(timedWords);\n\n this.queue.put({\n type: stt.SpeechEventType.INTERIM_TRANSCRIPT,\n alternatives: [\n {\n language,\n text: interimText,\n startTime,\n endTime,\n confidence,\n words: timedWords,\n },\n ],\n });\n }\n\n // `utterance` is chunk-based (not cumulative) — emit as a preflight transcript\n // covering only the words since the last preflight.\n if (utterance) {\n if (this.#lastPreflightStartTime === 0) {\n this.#lastPreflightStartTime = startTime;\n }\n\n const utteranceWords = timedWords.filter(\n (w) => w.startTime !== undefined && w.startTime >= this.#lastPreflightStartTime,\n );\n const utteranceConfidence = this.#averageConfidence(utteranceWords);\n\n this.queue.put({\n type: stt.SpeechEventType.PREFLIGHT_TRANSCRIPT,\n alternatives: [\n {\n language,\n text: utterance,\n startTime: this.#lastPreflightStartTime,\n endTime,\n confidence: utteranceConfidence,\n words: utteranceWords,\n },\n ],\n });\n this.#lastPreflightStartTime = endTime;\n }\n\n // End-of-turn: emit FINAL_TRANSCRIPT + END_OF_SPEECH.\n // If the user asked for formatted turns, wait for a formatted final.\n const waitingForFormatted = this.#opts.formatTurns === true && !turnIsFormatted;\n if (endOfTurn && !waitingForFormatted) {\n this.queue.put({\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n alternatives: [\n {\n language,\n text: transcript,\n startTime,\n endTime,\n confidence,\n words: timedWords,\n },\n ],\n });\n\n this.queue.put({ type: stt.SpeechEventType.END_OF_SPEECH });\n if (this.#speechDurationInS > 0) {\n this.queue.put({\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n // Propagate the AssemblyAI session id as the request id so metrics\n // can be correlated back to a specific connection, mirroring how\n // Deepgram surfaces its `request_id`.\n requestId: this.#sessionId ?? undefined,\n recognitionUsage: {\n audioDuration: this.#speechDurationInS,\n },\n });\n this.#speechDurationInS = 0;\n this.#lastPreflightStartTime = 0;\n }\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,oBAYO;AAGP,gBAA0B;AAG1B,MAAM,gBAAgB,CAAC,aAAa,oBAAoB,mBAAmB;AAE3E,SAAS,aAAa,OAA2B;AAC/C,SAAO,cAAc,SAAS,KAAuC;AACvE;AA0EA,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,YAAY;AAAA,EACZ,cAAc;AAAA,EACd,UAAU;AAAA,EACV,aAAa;AAAA,EACb,SAAS;AACX;AAEO,MAAM,YAAY,kBAAI,IAAI;AAAA,EAC/B;AAAA,EACA,WAAW,oBAAI,IAA2B;AAAA,EAC1C,QAAQ;AAAA,EAER,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,YAAY,OAA4B,CAAC,GAAG;AAC1C,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB;AAAA,MAChB,mBAAmB;AAAA,IACrB,CAAC;AAED,QAAI,KAAK,gBAAgB,UAAU;AACjC,6BAAI,EAAE,KAAK,kDAAkD;AAC7D,WAAK,cAAc;AAAA,IACrB;AAEA,UAAM,cAAc,KAAK,eAAe,kBAAkB;AAC1D,QAAI,CAAC,aAAa,WAAW,GAAG;AAC9B,iBAAW,SAAS,CAAC,UAAU,cAAc,uBAAuB,MAAM,GAAY;AACpF,YAAI,KAAK,KAAK,MAAM,QAAW;AAC7B,gBAAM,IAAI;AAAA,YACR,QAAQ,KAAK,0CAA0C,cAAc,KAAK,IAAI,CAAC;AAAA,UACjF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,SAAS,KAAK,UAAU,kBAAkB;AAChD,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAGA,UAAM,iBAAiB,KAAK,kBAAkB;AAE9C,SAAK,QAAQ;AAAA,MACX,GAAG;AAAA,MACH,GAAG;AAAA,MACH;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,WAAW,GAA0C;AACzD,UAAM,IAAI,MAAM,4DAA4D;AAAA,EAC9E;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,eAAW,OAAO,KAAK,UAAU;AAC/B,YAAM,SAAS,IAAI,MAAM;AACzB,UAAI,QAAQ;AACV,eAAO,cAAc,IAAI;AAAA,MAC3B,OAAO;AACL,aAAK,SAAS,OAAO,GAAG;AAAA,MAC1B;AAAA,IACF;AAAA,EACF;AAAA,EAEA,OAAO,SAA6D;AAClE,UAAM,SAAS,IAAI,aAAa,MAAM,KAAK,OAAO,mCAAS,WAAW;AACtE,SAAK,SAAS,IAAI,IAAI,QAAQ,MAAM,CAAC;AACrC,WAAO;AAAA,EACT;AACF;AAEO,MAAM,qBAAqB,kBAAI,aAAa;AAAA,EACjD,OAAgB,YAAY,KAAK,UAAU,EAAE,MAAM,YAAY,CAAC;AAAA,EAEhE;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,qBAAqB;AAAA,EACrB,0BAA0B;AAAA,EAC1B,yBAAoD,CAAC;AAAA,EACrD,wBAAwB,IAAI,qBAAO;AAAA,EACnC,aAA4B;AAAA,EAC5B,aAA4B;AAAA,EAC5B,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB,aAAiC;AACvE,UAAMA,MAAK,KAAK,YAAY,WAAW;AACvC,SAAK,QAAQ;AACb,SAAK,SAAS;AAAA,EAChB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,IAAI,YAA2B;AAC7B,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,IAAI,YAA2B;AAC7B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAEtC,UAAM,YAAqC,EAAE,MAAM,sBAAsB;AACzE,QAAI,KAAK,WAAW,OAAW,WAAU,SAAS,KAAK;AACvD,QAAI,KAAK,mBAAmB,OAAW,WAAU,kBAAkB,KAAK;AACxE,QAAI,KAAK,mBAAmB,OAAW,WAAU,mBAAmB,KAAK;AACzE,QAAI,KAAK,mBAAmB,OAAW,WAAU,mBAAmB,KAAK;AACzE,QAAI,KAAK,iCAAiC,QAAW;AACnD,gBAAU,mCAAmC,KAAK;AAAA,IACpD;AACA,QAAI,KAAK,iBAAiB,OAAW,WAAU,gBAAgB,KAAK;AAGpE,QAAI,OAAO,KAAK,SAAS,EAAE,SAAS,GAAG;AACrC,WAAK,uBAAuB,KAAK,SAAS;AAC1C,UAAI,CAAC,KAAK,sBAAsB,KAAM,MAAK,sBAAsB,QAAQ;AAAA,IAC3E;AAAA,EACF;AAAA;AAAA;AAAA;AAAA,EAKA,gBAAgB;AACd,SAAK,uBAAuB,KAAK,EAAE,MAAM,gBAAgB,CAAC;AAC1D,QAAI,CAAC,KAAK,sBAAsB,KAAM,MAAK,sBAAsB,QAAQ;AAAA,EAC3E;AAAA;AAAA,EAGA,MAAgB,MAAM;AACpB,UAAM,WAAW;AACjB,QAAI,UAAU;AAEd,WAAO,CAAC,KAAK,MAAM,UAAU,CAAC,KAAK,QAAQ;AACzC,UAAI;AACF,cAAM,KAAK,MAAM,KAAK,WAAW;AACjC,cAAM,KAAK,OAAO,EAAE;AACpB,kBAAU;AAAA,MACZ,SAAS,GAAG;AACV,YAAI,CAAC,KAAK,UAAU,CAAC,KAAK,MAAM,QAAQ;AACtC,cAAI,WAAW,UAAU;AACvB,kBAAM,IAAI,MAAM,yCAAyC,OAAO,cAAc,CAAC,EAAE;AAAA,UACnF;AAEA,gBAAM,oBAAoB,KAAK,IAAI,UAAU,GAAG,EAAE;AAClD;AAEA,eAAK,QAAQ;AAAA,YACX,gDAAgD,iBAAiB,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ;AAAA,UACzG;AACA,oBAAM,qBAAM,oBAAoB,GAAI;AAAA,QACtC,OAAO;AACL,eAAK,QAAQ;AAAA,YACX,kDAAkD,CAAC,kBAAkB,KAAK,MAAM,MAAM,eAAe,KAAK,MAAM;AAAA,UAClH;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAM,aAAiC;AAErC,QAAI,aAAa,KAAK,MAAM;AAC5B,QAAI,aAAa,KAAK,MAAM;AAC5B,QAAI,aAAa,KAAK,MAAM,WAAW,GAAG;AACxC,UAAI,eAAe,OAAW,cAAa;AAC3C,UAAI,eAAe,OAAW,cAAa;AAAA,IAC7C;AAGA,UAAM,2BACJ,KAAK,MAAM,YAAY,SAAS,cAAc,KAAK,aAAa,KAAK,MAAM,WAAW;AACxF,UAAM,oBAAoB,KAAK,MAAM,qBAAqB;AAE1D,UAAM,aAAsC;AAAA,MAC1C,aAAa,KAAK,MAAM;AAAA,MACxB,UAAU,KAAK,MAAM;AAAA,MACrB,cAAc,KAAK,MAAM;AAAA,MACzB,cAAc,KAAK,MAAM;AAAA,MACzB,kCAAkC,KAAK,MAAM;AAAA,MAC7C,kBAAkB;AAAA,MAClB,kBAAkB;AAAA,MAClB,iBACE,KAAK,MAAM,mBAAmB,SAC1B,KAAK,UAAU,KAAK,MAAM,cAAc,IACxC;AAAA,MACN,oBAAoB;AAAA,MACpB,QAAQ,KAAK,MAAM;AAAA,MACnB,eAAe,KAAK,MAAM;AAAA,MAC1B,gBAAgB,KAAK,MAAM;AAAA,MAC3B,cAAc,KAAK,MAAM;AAAA,MACzB,QAAQ,KAAK,MAAM;AAAA,MACnB,aAAa,KAAK,MAAM;AAAA,MACxB,uBAAuB,KAAK,MAAM;AAAA,MAClC,MAAM,KAAK,MAAM;AAAA,IACnB;AAEA,UAAM,MAAM,IAAI,IAAI,GAAG,KAAK,MAAM,OAAO,QAAQ;AAEjD,eAAW,CAAC,KAAK,KAAK,KAAK,OAAO,QAAQ,UAAU,GAAG;AACrD,UAAI,UAAU,UAAa,UAAU,KAAM;AAC3C,UAAI,OAAO,UAAU,WAAW;AAC9B,YAAI,aAAa,OAAO,KAAK,QAAQ,SAAS,OAAO;AAAA,MACvD,OAAO;AACL,YAAI,aAAa,OAAO,KAAK,OAAO,KAAK,CAAC;AAAA,MAC5C;AAAA,IACF;AAEA,UAAM,KAAK,IAAI,oBAAU,KAAK;AAAA,MAC5B,SAAS;AAAA,QACP,eAAe,KAAK,MAAM;AAAA,QAC1B,gBAAgB;AAAA,QAChB,cAAc;AAAA,MAChB;AAAA,IACF,CAAC;AAED,UAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,SAAG,GAAG,QAAQ,MAAM,QAAQ,CAAC;AAC7B,SAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,SAAG,GAAG,SAAS,CAAC,SAAS,OAAO,IAAI,MAAM,sBAAsB,IAAI,EAAE,CAAC,CAAC;AAAA,IAC1E,CAAC;AAED,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,OAAO,IAAe;AAC1B,QAAI,UAAU;AACd,UAAM,oBAAoB,IAAI,gBAAgB;AAG9C,UAAM,YAAY,mBAAK,KAAK,OAAO,eAAe;AAChD,YAAM,SAAS,IAAI,QAAc,CAAC,GAAG,WAAW;AAC9C,WAAG,KAAK,SAAS,CAAC,MAAM,WAAW;AACjC,cAAI,CAAC,SAAS;AACZ,iBAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,MAAM,EAAE;AAClE,mBAAO,IAAI,MAAM,kBAAkB,CAAC;AAAA,UACtC;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,YAAQ,4BAAa,WAAW,MAAM,CAAC,CAAC;AAAA,IAC9D,CAAC;AAED,UAAM,WAAW,YAAY;AAC3B,YAAM,mBAAmB,KAAK,MAAO,KAAK,MAAM,aAAa,KAAK,MAAM,eAAgB,GAAI;AAC5F,YAAM,cAAc,IAAI,8BAAgB,KAAK,MAAM,YAAY,GAAG,gBAAgB;AAElF,YAAM,mBAAe,4BAAa,KAAK,WAAW;AAClD,YAAM,mBAAe,4BAAa,kBAAkB,MAAM;AAE1D,UAAI;AACF,eAAO,CAAC,KAAK,QAAQ;AACnB,gBAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,KAAK,MAAM,KAAK,GAAG,cAAc,YAAY,CAAC;AAEjF,cAAI,WAAW,OAAW;AAC1B,cAAI,OAAO,KAAM;AAEjB,gBAAM,OAAO,OAAO;AAEpB,cAAI;AACJ,cAAI,SAAS,aAAa,gBAAgB;AACxC,qBAAS,YAAY,MAAM;AAAA,UAC7B,WAAW,KAAK,eAAe,KAAK,MAAM,cAAc,KAAK,aAAa,GAAG;AAG3E,qBAAS,YAAY,MAAM,KAAK,KAAK,MAAqB;AAAA,UAC5D,OAAO;AACL,kBAAM,IAAI,MAAM,sDAAsD;AAAA,UACxE;AAEA,qBAAW,SAAS,QAAQ;AAC1B,iBAAK,sBAAsB,MAAM,oBAAoB,MAAM;AAC3D,eAAG,KAAK,MAAM,KAAK,MAAM;AAAA,UAC3B;AAAA,QACF;AAAA,MACF,UAAE;AACA,kBAAU;AACV,YAAI;AACF,aAAG,KAAK,aAAa,SAAS;AAAA,QAChC,QAAQ;AAAA,QAER;AACA,kBAAU,OAAO;AAAA,MACnB;AAAA,IACF;AAEA,QAAI,iBAAqE;AACzE,UAAM,aAAa,mBAAK,KAAK,OAAO,eAAe;AACjD,YAAM,gBAAgB,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3D,yBAAiB,CAAC,KAAK,aAAa;AAClC,cAAI,UAAU;AACZ,iBAAK,QAAQ,MAAM,2CAA2C;AAC9D;AAAA,UACF;AACA,cAAI;AACF,kBAAM,OAAO,KAAK,MAAM,IAAI,SAAS,CAAC;AACtC,iBAAK,oBAAoB,IAAI;AAC7B,gBAAI,KAAK,UAAU,SAAS;AAC1B,sBAAQ;AAAA,YACV;AAAA,UACF,SAAS,KAAK;AACZ,iBAAK,QAAQ,MAAM,yCAAyC,GAAG,EAAE;AACjE,mBAAO,GAAG;AAAA,UACZ;AAAA,QACF;AACA,WAAG,GAAG,WAAW,cAAc;AAAA,MACjC,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,mBAAe,4BAAa,WAAW,MAAM,CAAC,CAAC;AAAA,IACrE,CAAC;AAED,UAAM,aAAa,mBAAK,KAAK,OAAO,eAAe;AAEjD,aAAO,KAAK,uBAAuB,SAAS,GAAG;AAC7C,cAAM,MAAM,KAAK,uBAAuB,MAAM;AAC9C,WAAG,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,MAC7B;AAEA,aAAO,CAAC,WAAW,OAAO,SAAS;AACjC,cAAM,QAAQ,KAAK,CAAC,KAAK,sBAAsB,WAAO,4BAAa,WAAW,MAAM,CAAC,CAAC;AACtF,YAAI,WAAW,OAAO,QAAS;AAE/B,aAAK,wBAAwB,IAAI,qBAAO;AACxC,eAAO,KAAK,uBAAuB,SAAS,GAAG;AAC7C,gBAAM,MAAM,KAAK,uBAAuB,MAAM;AAC9C,aAAG,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,QAC7B;AAAA,MACF;AAAA,IACF,CAAC;AAED,QAAI;AACF,YAAM,QAAQ,IAAI,CAAC,SAAS,GAAG,WAAW,QAAQ,UAAU,MAAM,CAAC;AAAA,IACrE,UAAE;AACA,gBAAU;AACV,wBAAkB,MAAM;AACxB,iBAAW,OAAO;AAClB,iBAAW,OAAO;AAClB,UAAI,eAAgB,IAAG,IAAI,WAAW,cAAc;AACpD,UAAI;AACF,WAAG,MAAM;AAAA,MACX,QAAQ;AAAA,MAER;AAAA,IACF;AAAA,EACF;AAAA,EAEA,mBAAmB,OAA+C;AAChE,QAAI,MAAM,WAAW,EAAG,QAAO;AAC/B,WAAO,MAAM,OAAO,CAAC,KAAK,MAAM,OAAO,EAAE,cAAc,IAAI,CAAC,IAAI,MAAM;AAAA,EACxE;AAAA,EAEA,oBAAoB,MAA0B;AAC5C,UAAM,cAAc,KAAK;AAEzB,QAAI,gBAAgB,SAAS;AAC3B,WAAK,aAAa,KAAK,MAAM;AAC7B,WAAK,aAAa,KAAK,cAAc;AACrC,WAAK,QAAQ;AAAA,QACX,iCAAiC,KAAK,UAAU,eAAe,KAAK,UAAU;AAAA,MAChF;AACA;AAAA,IACF;AAEA,QAAI,gBAAgB,iBAAiB;AACnC,WAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,gBAAgB,CAAC;AAC5D;AAAA,IACF;AAEA,QAAI,gBAAgB,eAAe;AACjC,WAAK,QAAQ;AAAA,QACX,gDAAgD,KAAK,sBAAsB,sBAAsB,KAAK,wBAAwB;AAAA,MAChI;AACA;AAAA,IACF;AAEA,QAAI,gBAAgB,QAAQ;AAC1B;AAAA,IACF;AAEA,UAAM,QAAQ,KAAK,SAAS,CAAC;AAC7B,UAAM,YAAY,QAAQ,KAAK,WAAW;AAC1C,UAAM,kBAAkB,QAAQ,KAAK,iBAAiB;AACtD,UAAM,YAAY,KAAK,aAAa;AACpC,UAAM,aAAa,KAAK,cAAc;AACtC,UAAM,eAAW,iCAAkB,KAAK,iBAAiB,IAAI;AAI7D,UAAM,aAAa,MAAM;AAAA,MAAI,CAAC,aAC5B,iCAAkB;AAAA,QAChB,MAAM,KAAK,QAAQ;AAAA,QACnB,YAAY,KAAK,SAAS,KAAK,MAAO,KAAK;AAAA,QAC3C,UAAU,KAAK,OAAO,KAAK,MAAO,KAAK;AAAA,QACvC,YAAY,KAAK,cAAc;AAAA,QAC/B,iBAAiB,KAAK;AAAA,MACxB,CAAC;AAAA,IACH;AAEA,QAAI,YAAY;AAChB,QAAI,UAAU;AACd,QAAI,aAAa;AAGjB,QAAI,WAAW,SAAS,GAAG;AACzB,YAAM,cAAc,WAAW,IAAI,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,GAAG;AAC1D,kBAAY,WAAW,CAAC,EAAG,aAAa;AACxC,gBAAU,WAAW,WAAW,SAAS,CAAC,EAAG,WAAW;AACxD,mBAAa,KAAK,mBAAmB,UAAU;AAE/C,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,kBAAI,gBAAgB;AAAA,QAC1B,cAAc;AAAA,UACZ;AAAA,YACE;AAAA,YACA,MAAM;AAAA,YACN;AAAA,YACA;AAAA,YACA;AAAA,YACA,OAAO;AAAA,UACT;AAAA,QACF;AAAA,MACF,CAAC;AAAA,IACH;AAIA,QAAI,WAAW;AACb,UAAI,KAAK,4BAA4B,GAAG;AACtC,aAAK,0BAA0B;AAAA,MACjC;AAEA,YAAM,iBAAiB,WAAW;AAAA,QAChC,CAAC,MAAM,EAAE,cAAc,UAAa,EAAE,aAAa,KAAK;AAAA,MAC1D;AACA,YAAM,sBAAsB,KAAK,mBAAmB,cAAc;AAElE,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,kBAAI,gBAAgB;AAAA,QAC1B,cAAc;AAAA,UACZ;AAAA,YACE;AAAA,YACA,MAAM;AAAA,YACN,WAAW,KAAK;AAAA,YAChB;AAAA,YACA,YAAY;AAAA,YACZ,OAAO;AAAA,UACT;AAAA,QACF;AAAA,MACF,CAAC;AACD,WAAK,0BAA0B;AAAA,IACjC;AAIA,UAAM,sBAAsB,KAAK,MAAM,gBAAgB,QAAQ,CAAC;AAChE,QAAI,aAAa,CAAC,qBAAqB;AACrC,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,kBAAI,gBAAgB;AAAA,QAC1B,cAAc;AAAA,UACZ;AAAA,YACE;AAAA,YACA,MAAM;AAAA,YACN;AAAA,YACA;AAAA,YACA;AAAA,YACA,OAAO;AAAA,UACT;AAAA,QACF;AAAA,MACF,CAAC;AAED,WAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,cAAc,CAAC;AAC1D,UAAI,KAAK,qBAAqB,GAAG;AAC/B,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,kBAAI,gBAAgB;AAAA;AAAA;AAAA;AAAA,UAI1B,WAAW,KAAK,cAAc;AAAA,UAC9B,kBAAkB;AAAA,YAChB,eAAe,KAAK;AAAA,UACtB;AAAA,QACF,CAAC;AACD,aAAK,qBAAqB;AAC1B,aAAK,0BAA0B;AAAA,MACjC;AAAA,IACF;AAAA,EACF;AACF;","names":["stt"]}
1
+ {"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2026 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n//\nimport {\n type APIConnectOptions,\n type AudioBuffer,\n AudioByteStream,\n Future,\n Task,\n createTimedString,\n delay,\n log,\n normalizeLanguage,\n stt,\n waitForAbort,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { RawData } from 'ws';\nimport { WebSocket } from 'ws';\nimport type { STTEncoding, STTModels, VoiceFocus } from './models.js';\n\n// Speech models in the Universal-3 Pro family, which share the same parameter support.\nconst U3_PRO_MODELS = ['u3-rt-pro', 'u3-rt-pro-beta-1', 'universal-3-5-pro'] as const;\n\nfunction isU3ProModel(model: STTModels): boolean {\n return U3_PRO_MODELS.includes(model as (typeof U3_PRO_MODELS)[number]);\n}\n\n// AssemblyAI Universal-Streaming (v3) message envelope. All fields are optional\n// since we narrow on `type` before reading anything else.\ninterface StreamEventMessage {\n type?: 'Begin' | 'SpeechStarted' | 'Turn' | 'Termination' | string;\n // Begin\n id?: string;\n expires_at?: number;\n // Turn\n transcript?: string;\n utterance?: string;\n end_of_turn?: boolean;\n end_of_turn_confidence?: number;\n turn_is_formatted?: boolean;\n language_code?: string;\n language_confidence?: number;\n speaker_label?: string;\n words?: Array<{\n text?: string;\n start?: number;\n end?: number;\n confidence?: number;\n speaker?: string;\n }>;\n // Termination\n audio_duration_seconds?: number;\n session_duration_seconds?: number;\n}\n\nfunction speechDataMetadata(data: StreamEventMessage): stt.SpeechData['metadata'] | undefined {\n const assemblyai: Record<string, number> = {};\n\n if (typeof data.language_confidence === 'number') {\n assemblyai.languageConfidence = data.language_confidence;\n }\n\n if (Object.keys(assemblyai).length === 0) return undefined;\n\n return { assemblyai };\n}\n\nexport interface STTOptions {\n apiKey?: string;\n sampleRate: number;\n /**\n * How large each chunk of audio is before being sent to AssemblyAI, in\n * milliseconds. Corresponds to Python's `buffer_size_seconds` (seconds there,\n * ms here per this repo's time-unit convention).\n */\n bufferSizeMs: number;\n encoding: STTEncoding;\n speechModel: STTModels;\n languageDetection?: boolean;\n /**\n * Session inactivity timeout in seconds. AssemblyAI accepts integer values\n * from 5 to 3600; when unset, no inactivity timeout is applied.\n */\n inactivityTimeout?: number;\n endOfTurnConfidenceThreshold?: number;\n /** Minimum silence (ms) before a confident end-of-turn is finalized. */\n minTurnSilence?: number;\n /** Maximum silence (ms) before end-of-turn is forced regardless of confidence. */\n maxTurnSilence?: number;\n formatTurns?: boolean;\n keytermsPrompt?: string[];\n /** Only supported with the Universal-3 Pro model family. */\n prompt?: string;\n /** Only supported with the Universal-3 Pro model family. */\n agentContext?: string;\n /** Only supported with the Universal-3 Pro model family. Set at connection time only. */\n previousContextNTurns?: number;\n vadThreshold?: number;\n /**\n * Enable speaker diarization. Note: AssemblyAI will return per-word speaker\n * labels, but the JS framework's `stt.SpeechData` type does not yet expose\n * a `speakerId` field (unlike the Python framework), so the labels are not\n * currently surfaced on emitted events. Setting this to `true` still has\n * effect server-side. Once the base `SpeechData` interface gains speaker\n * support, `#processStreamEvent` should forward `data.words[].speaker` too.\n */\n speakerLabels?: boolean;\n maxSpeakers?: number;\n domain?: string;\n /** Isolate the primary voice and suppress background noise. Connect-time only. */\n voiceFocus?: VoiceFocus;\n /** Background audio suppression aggressiveness, from 0.0 to 1.0. Connect-time only. */\n voiceFocusThreshold?: number;\n /**\n * Accuracy/latency preset for the Universal-3 Pro model family: `min_latency`, `balanced`,\n * or `max_accuracy`. Explicit turn-silence values still take precedence over mode defaults.\n */\n mode?: 'min_latency' | 'balanced' | 'max_accuracy';\n baseUrl: string;\n}\n\nconst defaultSTTOptions: STTOptions = {\n apiKey: process.env.ASSEMBLYAI_API_KEY,\n sampleRate: 16000,\n bufferSizeMs: 50,\n encoding: 'pcm_s16le',\n speechModel: 'universal-3-5-pro',\n baseUrl: 'wss://streaming.assemblyai.com',\n};\n\nexport class STT extends stt.STT {\n #opts: STTOptions;\n #streams = new Set<WeakRef<SpeechStream>>();\n label = 'assemblyai.STT';\n\n get model(): string {\n return this.#opts.speechModel;\n }\n\n get provider(): string {\n return 'AssemblyAI';\n }\n\n constructor(opts: Partial<STTOptions> = {}) {\n super({\n streaming: true,\n interimResults: true,\n alignedTranscript: 'word',\n });\n\n if (opts.speechModel === 'u3-pro') {\n log().warn(\"'u3-pro' is deprecated, use 'universal-3-5-pro' instead.\");\n opts.speechModel = 'universal-3-5-pro';\n }\n\n const speechModel = opts.speechModel ?? defaultSTTOptions.speechModel;\n if (!isU3ProModel(speechModel)) {\n for (const param of [\n 'prompt',\n 'agentContext',\n 'previousContextNTurns',\n 'voiceFocus',\n 'voiceFocusThreshold',\n 'mode',\n ] as const) {\n if (opts[param] !== undefined) {\n throw new Error(\n `The '${param}' parameter is only supported with the ${U3_PRO_MODELS.join(', ')} models.`,\n );\n }\n }\n }\n\n const apiKey = opts.apiKey ?? defaultSTTOptions.apiKey;\n if (!apiKey) {\n throw new Error(\n 'AssemblyAI API key is required. Pass one in via the `apiKey` parameter, or set it as the `ASSEMBLYAI_API_KEY` environment variable',\n );\n }\n\n // Minimize latency by default, but let AssemblyAI's mode preset control silence tuning.\n const minTurnSilence = opts.minTurnSilence ?? (opts.mode === undefined ? 100 : undefined);\n\n this.#opts = {\n ...defaultSTTOptions,\n ...opts,\n apiKey,\n minTurnSilence,\n };\n }\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n async _recognize(_: AudioBuffer): Promise<stt.SpeechEvent> {\n throw new Error('Non-streaming recognize is not supported on AssemblyAI STT');\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n for (const ref of this.#streams) {\n const stream = ref.deref();\n if (stream) {\n stream.updateOptions(opts);\n } else {\n this.#streams.delete(ref);\n }\n }\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): SpeechStream {\n const stream = new SpeechStream(this, this.#opts, options?.connOptions);\n this.#streams.add(new WeakRef(stream));\n return stream;\n }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n static readonly CLOSE_MSG = JSON.stringify({ type: 'Terminate' });\n\n #opts: STTOptions;\n #logger = log();\n #speechDurationInS = 0;\n #lastPreflightStartTime = 0;\n #pendingConfigMessages: Record<string, unknown>[] = [];\n #configMessagePending = new Future();\n #sessionId: string | null = null;\n #expiresAt: number | null = null;\n label = 'assemblyai.SpeechStream';\n\n constructor(stt: STT, opts: STTOptions, connOptions?: APIConnectOptions) {\n super(stt, opts.sampleRate, connOptions);\n this.#opts = opts;\n this.closed = false;\n }\n\n /**\n * The AssemblyAI session ID. Set when the WebSocket connection is established\n * (before any speech events). Null until the connection completes.\n * Share this with the AssemblyAI team when reporting issues.\n */\n get sessionId(): string | null {\n return this.#sessionId;\n }\n\n /**\n * Unix timestamp when the AssemblyAI session expires. Set alongside\n * {@link sessionId} when the WebSocket connection is established.\n */\n get expiresAt(): number | null {\n return this.#expiresAt;\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n\n const configMsg: Record<string, unknown> = { type: 'UpdateConfiguration' };\n if (opts.prompt !== undefined) configMsg.prompt = opts.prompt;\n if (opts.agentContext !== undefined) configMsg.agent_context = opts.agentContext;\n if (opts.keytermsPrompt !== undefined) configMsg.keyterms_prompt = opts.keytermsPrompt;\n if (opts.maxTurnSilence !== undefined) configMsg.max_turn_silence = opts.maxTurnSilence;\n if (opts.minTurnSilence !== undefined) configMsg.min_turn_silence = opts.minTurnSilence;\n if (opts.endOfTurnConfidenceThreshold !== undefined) {\n configMsg.end_of_turn_confidence_threshold = opts.endOfTurnConfidenceThreshold;\n }\n if (opts.vadThreshold !== undefined) configMsg.vad_threshold = opts.vadThreshold;\n\n // Only send if any actual fields (besides `type`) were specified.\n if (Object.keys(configMsg).length > 1) {\n this.#pendingConfigMessages.push(configMsg);\n if (!this.#configMessagePending.done) this.#configMessagePending.resolve();\n }\n }\n\n /**\n * Force-finalize the current turn immediately.\n */\n forceEndpoint() {\n this.#pendingConfigMessages.push({ type: 'ForceEndpoint' });\n if (!this.#configMessagePending.done) this.#configMessagePending.resolve();\n }\n\n // Deepgram-style reconnect loop around a single websocket lifetime.\n protected async run() {\n const maxRetry = 32;\n let retries = 0;\n\n while (!this.input.closed && !this.closed) {\n try {\n const ws = await this.#connectWS();\n await this.#runWS(ws);\n retries = 0;\n } catch (e) {\n if (!this.closed && !this.input.closed) {\n if (retries >= maxRetry) {\n throw new Error(`failed to connect to AssemblyAI after ${retries} attempts: ${e}`);\n }\n\n const retryDelaySeconds = Math.min(retries * 5, 10);\n retries++;\n\n this.#logger.warn(\n `failed to connect to AssemblyAI, retrying in ${retryDelaySeconds} seconds: ${e} (${retries}/${maxRetry})`,\n );\n await delay(retryDelaySeconds * 1000);\n } else {\n this.#logger.warn(\n `AssemblyAI disconnected, connection is closed: ${e} (inputClosed: ${this.input.closed}, isClosed: ${this.closed})`,\n );\n }\n }\n }\n\n this.closed = true;\n }\n\n async #connectWS(): Promise<WebSocket> {\n // Universal-3 Pro family models default both min and max silence to 100ms when unset.\n // When a mode preset is selected, leave them unset unless explicitly provided so the\n // server's per-mode silence tuning is not overridden by the latency-optimized default.\n let minSilence = this.#opts.minTurnSilence;\n let maxSilence = this.#opts.maxTurnSilence;\n if (isU3ProModel(this.#opts.speechModel)) {\n if (minSilence === undefined && this.#opts.mode === undefined) minSilence = 100;\n if (maxSilence === undefined) maxSilence = minSilence;\n }\n\n // Default language_detection to true for multilingual / u3-rt-pro-family models, false otherwise.\n const defaultLanguageDetection =\n this.#opts.speechModel.includes('multilingual') || isU3ProModel(this.#opts.speechModel);\n const languageDetection = this.#opts.languageDetection ?? defaultLanguageDetection;\n\n const liveConfig: Record<string, unknown> = {\n sample_rate: this.#opts.sampleRate,\n encoding: this.#opts.encoding,\n speech_model: this.#opts.speechModel,\n format_turns: this.#opts.formatTurns,\n end_of_turn_confidence_threshold: this.#opts.endOfTurnConfidenceThreshold,\n min_turn_silence: minSilence,\n max_turn_silence: maxSilence,\n keyterms_prompt:\n this.#opts.keytermsPrompt !== undefined\n ? JSON.stringify(this.#opts.keytermsPrompt)\n : undefined,\n language_detection: languageDetection,\n inactivity_timeout: this.#opts.inactivityTimeout,\n prompt: this.#opts.prompt,\n agent_context: this.#opts.agentContext,\n previous_context_n_turns: this.#opts.previousContextNTurns,\n vad_threshold: this.#opts.vadThreshold,\n speaker_labels: this.#opts.speakerLabels,\n max_speakers: this.#opts.maxSpeakers,\n domain: this.#opts.domain,\n voice_focus: this.#opts.voiceFocus,\n voice_focus_threshold: this.#opts.voiceFocusThreshold,\n mode: this.#opts.mode,\n };\n\n const url = new URL(`${this.#opts.baseUrl}/v3/ws`);\n // Python serializes booleans as the strings \"true\"/\"false\", so we mirror that.\n for (const [key, value] of Object.entries(liveConfig)) {\n if (value === undefined || value === null) continue;\n if (typeof value === 'boolean') {\n url.searchParams.append(key, value ? 'true' : 'false');\n } else {\n url.searchParams.append(key, String(value));\n }\n }\n\n const ws = new WebSocket(url, {\n headers: {\n Authorization: this.#opts.apiKey!,\n 'Content-Type': 'application/json',\n 'User-Agent': 'AssemblyAI/1.0 (integration=Livekit)',\n },\n });\n\n await new Promise<void>((resolve, reject) => {\n ws.on('open', () => resolve());\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(new Error(`WebSocket returned ${code}`)));\n });\n\n return ws;\n }\n\n async #runWS(ws: WebSocket) {\n let closing = false;\n const sessionController = new AbortController();\n\n // gets cancelled also when sendTask is complete\n const wsMonitor = Task.from(async (controller) => {\n const closed = new Promise<void>((_, reject) => {\n ws.once('close', (code, reason) => {\n if (!closing) {\n this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);\n reject(new Error('WebSocket closed'));\n }\n });\n });\n\n await Promise.race([closed, waitForAbort(controller.signal)]);\n });\n\n const sendTask = async () => {\n const samplesPerBuffer = Math.floor((this.#opts.sampleRate * this.#opts.bufferSizeMs) / 1000);\n const audioStream = new AudioByteStream(this.#opts.sampleRate, 1, samplesPerBuffer);\n\n const abortPromise = waitForAbort(this.abortSignal);\n const sessionAbort = waitForAbort(sessionController.signal);\n\n try {\n while (!this.closed) {\n const result = await Promise.race([this.input.next(), abortPromise, sessionAbort]);\n\n if (result === undefined) return; // aborted\n if (result.done) break;\n\n const data = result.value;\n\n let frames: AudioFrame[];\n if (data === SpeechStream.FLUSH_SENTINEL) {\n frames = audioStream.flush();\n } else if (data.sampleRate === this.#opts.sampleRate && data.channels === 1) {\n // AssemblyAI expects mono PCM. The base SpeechStream only resamples\n // sample rate, so reject any frame that is not already downmixed.\n frames = audioStream.write(data.data.buffer as ArrayBuffer);\n } else {\n throw new Error('sample rate or channel count of frame does not match');\n }\n\n for (const frame of frames) {\n this.#speechDurationInS += frame.samplesPerChannel / frame.sampleRate;\n ws.send(frame.data.buffer);\n }\n }\n } finally {\n closing = true;\n try {\n ws.send(SpeechStream.CLOSE_MSG);\n } catch {\n // ignore — socket may already be closing\n }\n wsMonitor.cancel();\n }\n };\n\n let messageHandler: ((msg: RawData, isBinary: boolean) => void) | null = null;\n const listenTask = Task.from(async (controller) => {\n const listenMessage = new Promise<void>((resolve, reject) => {\n messageHandler = (msg, isBinary) => {\n if (isBinary) {\n this.#logger.error('unexpected binary message from AssemblyAI');\n return;\n }\n try {\n const json = JSON.parse(msg.toString()) as StreamEventMessage;\n this.#processStreamEvent(json);\n if (this.closed || closing) {\n resolve();\n }\n } catch (err) {\n this.#logger.error(`AssemblyAI: error processing message: ${msg}`);\n reject(err);\n }\n };\n ws.on('message', messageHandler);\n });\n\n await Promise.race([listenMessage, waitForAbort(controller.signal)]);\n });\n\n const configTask = Task.from(async (controller) => {\n // Drain any messages queued while the socket was reconnecting.\n while (this.#pendingConfigMessages.length > 0) {\n const msg = this.#pendingConfigMessages.shift()!;\n ws.send(JSON.stringify(msg));\n }\n\n while (!controller.signal.aborted) {\n await Promise.race([this.#configMessagePending.await, waitForAbort(controller.signal)]);\n if (controller.signal.aborted) return;\n\n this.#configMessagePending = new Future();\n while (this.#pendingConfigMessages.length > 0) {\n const msg = this.#pendingConfigMessages.shift()!;\n ws.send(JSON.stringify(msg));\n }\n }\n });\n\n try {\n await Promise.all([sendTask(), listenTask.result, wsMonitor.result]);\n } finally {\n closing = true;\n sessionController.abort();\n listenTask.cancel();\n configTask.cancel();\n if (messageHandler) ws.off('message', messageHandler);\n try {\n ws.close();\n } catch {\n // ignore\n }\n }\n }\n\n #averageConfidence(words: Array<{ confidence?: number }>): number {\n if (words.length === 0) return 0;\n return words.reduce((sum, w) => sum + (w.confidence ?? 0), 0) / words.length;\n }\n\n #processStreamEvent(data: StreamEventMessage) {\n const messageType = data.type;\n\n if (messageType === 'Begin') {\n this.#sessionId = data.id ?? null;\n this.#expiresAt = data.expires_at ?? null;\n this.#logger.info(\n `AssemblyAI session started id=${this.#sessionId} expires_at=${this.#expiresAt}`,\n );\n return;\n }\n\n if (messageType === 'SpeechStarted') {\n this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n return;\n }\n\n if (messageType === 'Termination') {\n this.#logger.debug(\n `AssemblyAI session terminated audio_duration=${data.audio_duration_seconds}s session_duration=${data.session_duration_seconds}s`,\n );\n return;\n }\n\n if (messageType !== 'Turn') {\n return;\n }\n\n const words = data.words ?? [];\n const endOfTurn = Boolean(data.end_of_turn);\n const turnIsFormatted = Boolean(data.turn_is_formatted);\n const utterance = data.utterance ?? '';\n const transcript = data.transcript ?? '';\n const language = normalizeLanguage(data.language_code ?? 'en');\n const metadata = speechDataMetadata(data);\n\n // Word timestamps are in milliseconds:\n // https://www.assemblyai.com/docs/api-reference/streaming-api/streaming-api#receive.receiveTurn.words\n const timedWords = words.map((word) =>\n createTimedString({\n text: word.text ?? '',\n startTime: (word.start ?? 0) / 1000 + this.startTimeOffset,\n endTime: (word.end ?? 0) / 1000 + this.startTimeOffset,\n confidence: word.confidence ?? 0,\n startTimeOffset: this.startTimeOffset,\n }),\n );\n\n let startTime = 0;\n let endTime = 0;\n let confidence = 0;\n\n // `words` are cumulative for the turn — emit as an interim transcript.\n if (timedWords.length > 0) {\n const interimText = timedWords.map((w) => w.text).join(' ');\n startTime = timedWords[0]!.startTime ?? 0;\n endTime = timedWords[timedWords.length - 1]!.endTime ?? 0;\n confidence = this.#averageConfidence(timedWords);\n\n this.queue.put({\n type: stt.SpeechEventType.INTERIM_TRANSCRIPT,\n alternatives: [\n {\n language,\n text: interimText,\n startTime,\n endTime,\n confidence,\n words: timedWords,\n ...(metadata ? { metadata } : {}),\n },\n ],\n });\n }\n\n // `utterance` is chunk-based (not cumulative) — emit as a preflight transcript\n // covering only the words since the last preflight.\n if (utterance) {\n if (this.#lastPreflightStartTime === 0) {\n this.#lastPreflightStartTime = startTime;\n }\n\n const utteranceWords = timedWords.filter(\n (w) => w.startTime !== undefined && w.startTime >= this.#lastPreflightStartTime,\n );\n const utteranceConfidence = this.#averageConfidence(utteranceWords);\n\n this.queue.put({\n type: stt.SpeechEventType.PREFLIGHT_TRANSCRIPT,\n alternatives: [\n {\n language,\n text: utterance,\n startTime: this.#lastPreflightStartTime,\n endTime,\n confidence: utteranceConfidence,\n words: utteranceWords,\n ...(metadata ? { metadata } : {}),\n },\n ],\n });\n this.#lastPreflightStartTime = endTime;\n }\n\n // End-of-turn: emit FINAL_TRANSCRIPT + END_OF_SPEECH.\n // If the user asked for formatted turns, wait for a formatted final.\n const waitingForFormatted = this.#opts.formatTurns === true && !turnIsFormatted;\n if (endOfTurn && !waitingForFormatted) {\n this.queue.put({\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n alternatives: [\n {\n language,\n text: transcript,\n startTime,\n endTime,\n confidence,\n words: timedWords,\n ...(metadata ? { metadata } : {}),\n },\n ],\n });\n\n this.queue.put({ type: stt.SpeechEventType.END_OF_SPEECH });\n if (this.#speechDurationInS > 0) {\n this.queue.put({\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n // Propagate the AssemblyAI session id as the request id so metrics\n // can be correlated back to a specific connection, mirroring how\n // Deepgram surfaces its `request_id`.\n requestId: this.#sessionId ?? undefined,\n recognitionUsage: {\n audioDuration: this.#speechDurationInS,\n },\n });\n this.#speechDurationInS = 0;\n this.#lastPreflightStartTime = 0;\n }\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,oBAYO;AAGP,gBAA0B;AAI1B,MAAM,gBAAgB,CAAC,aAAa,oBAAoB,mBAAmB;AAE3E,SAAS,aAAa,OAA2B;AAC/C,SAAO,cAAc,SAAS,KAAuC;AACvE;AA8BA,SAAS,mBAAmB,MAAkE;AAC5F,QAAM,aAAqC,CAAC;AAE5C,MAAI,OAAO,KAAK,wBAAwB,UAAU;AAChD,eAAW,qBAAqB,KAAK;AAAA,EACvC;AAEA,MAAI,OAAO,KAAK,UAAU,EAAE,WAAW,EAAG,QAAO;AAEjD,SAAO,EAAE,WAAW;AACtB;AAwDA,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,YAAY;AAAA,EACZ,cAAc;AAAA,EACd,UAAU;AAAA,EACV,aAAa;AAAA,EACb,SAAS;AACX;AAEO,MAAM,YAAY,kBAAI,IAAI;AAAA,EAC/B;AAAA,EACA,WAAW,oBAAI,IAA2B;AAAA,EAC1C,QAAQ;AAAA,EAER,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,YAAY,OAA4B,CAAC,GAAG;AAC1C,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB;AAAA,MAChB,mBAAmB;AAAA,IACrB,CAAC;AAED,QAAI,KAAK,gBAAgB,UAAU;AACjC,6BAAI,EAAE,KAAK,0DAA0D;AACrE,WAAK,cAAc;AAAA,IACrB;AAEA,UAAM,cAAc,KAAK,eAAe,kBAAkB;AAC1D,QAAI,CAAC,aAAa,WAAW,GAAG;AAC9B,iBAAW,SAAS;AAAA,QAClB;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MACF,GAAY;AACV,YAAI,KAAK,KAAK,MAAM,QAAW;AAC7B,gBAAM,IAAI;AAAA,YACR,QAAQ,KAAK,0CAA0C,cAAc,KAAK,IAAI,CAAC;AAAA,UACjF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,SAAS,KAAK,UAAU,kBAAkB;AAChD,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAGA,UAAM,iBAAiB,KAAK,mBAAmB,KAAK,SAAS,SAAY,MAAM;AAE/E,SAAK,QAAQ;AAAA,MACX,GAAG;AAAA,MACH,GAAG;AAAA,MACH;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,WAAW,GAA0C;AACzD,UAAM,IAAI,MAAM,4DAA4D;AAAA,EAC9E;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,eAAW,OAAO,KAAK,UAAU;AAC/B,YAAM,SAAS,IAAI,MAAM;AACzB,UAAI,QAAQ;AACV,eAAO,cAAc,IAAI;AAAA,MAC3B,OAAO;AACL,aAAK,SAAS,OAAO,GAAG;AAAA,MAC1B;AAAA,IACF;AAAA,EACF;AAAA,EAEA,OAAO,SAA6D;AAClE,UAAM,SAAS,IAAI,aAAa,MAAM,KAAK,OAAO,mCAAS,WAAW;AACtE,SAAK,SAAS,IAAI,IAAI,QAAQ,MAAM,CAAC;AACrC,WAAO;AAAA,EACT;AACF;AAEO,MAAM,qBAAqB,kBAAI,aAAa;AAAA,EACjD,OAAgB,YAAY,KAAK,UAAU,EAAE,MAAM,YAAY,CAAC;AAAA,EAEhE;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,qBAAqB;AAAA,EACrB,0BAA0B;AAAA,EAC1B,yBAAoD,CAAC;AAAA,EACrD,wBAAwB,IAAI,qBAAO;AAAA,EACnC,aAA4B;AAAA,EAC5B,aAA4B;AAAA,EAC5B,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB,aAAiC;AACvE,UAAMA,MAAK,KAAK,YAAY,WAAW;AACvC,SAAK,QAAQ;AACb,SAAK,SAAS;AAAA,EAChB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,IAAI,YAA2B;AAC7B,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,IAAI,YAA2B;AAC7B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAEtC,UAAM,YAAqC,EAAE,MAAM,sBAAsB;AACzE,QAAI,KAAK,WAAW,OAAW,WAAU,SAAS,KAAK;AACvD,QAAI,KAAK,iBAAiB,OAAW,WAAU,gBAAgB,KAAK;AACpE,QAAI,KAAK,mBAAmB,OAAW,WAAU,kBAAkB,KAAK;AACxE,QAAI,KAAK,mBAAmB,OAAW,WAAU,mBAAmB,KAAK;AACzE,QAAI,KAAK,mBAAmB,OAAW,WAAU,mBAAmB,KAAK;AACzE,QAAI,KAAK,iCAAiC,QAAW;AACnD,gBAAU,mCAAmC,KAAK;AAAA,IACpD;AACA,QAAI,KAAK,iBAAiB,OAAW,WAAU,gBAAgB,KAAK;AAGpE,QAAI,OAAO,KAAK,SAAS,EAAE,SAAS,GAAG;AACrC,WAAK,uBAAuB,KAAK,SAAS;AAC1C,UAAI,CAAC,KAAK,sBAAsB,KAAM,MAAK,sBAAsB,QAAQ;AAAA,IAC3E;AAAA,EACF;AAAA;AAAA;AAAA;AAAA,EAKA,gBAAgB;AACd,SAAK,uBAAuB,KAAK,EAAE,MAAM,gBAAgB,CAAC;AAC1D,QAAI,CAAC,KAAK,sBAAsB,KAAM,MAAK,sBAAsB,QAAQ;AAAA,EAC3E;AAAA;AAAA,EAGA,MAAgB,MAAM;AACpB,UAAM,WAAW;AACjB,QAAI,UAAU;AAEd,WAAO,CAAC,KAAK,MAAM,UAAU,CAAC,KAAK,QAAQ;AACzC,UAAI;AACF,cAAM,KAAK,MAAM,KAAK,WAAW;AACjC,cAAM,KAAK,OAAO,EAAE;AACpB,kBAAU;AAAA,MACZ,SAAS,GAAG;AACV,YAAI,CAAC,KAAK,UAAU,CAAC,KAAK,MAAM,QAAQ;AACtC,cAAI,WAAW,UAAU;AACvB,kBAAM,IAAI,MAAM,yCAAyC,OAAO,cAAc,CAAC,EAAE;AAAA,UACnF;AAEA,gBAAM,oBAAoB,KAAK,IAAI,UAAU,GAAG,EAAE;AAClD;AAEA,eAAK,QAAQ;AAAA,YACX,gDAAgD,iBAAiB,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ;AAAA,UACzG;AACA,oBAAM,qBAAM,oBAAoB,GAAI;AAAA,QACtC,OAAO;AACL,eAAK,QAAQ;AAAA,YACX,kDAAkD,CAAC,kBAAkB,KAAK,MAAM,MAAM,eAAe,KAAK,MAAM;AAAA,UAClH;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAM,aAAiC;AAIrC,QAAI,aAAa,KAAK,MAAM;AAC5B,QAAI,aAAa,KAAK,MAAM;AAC5B,QAAI,aAAa,KAAK,MAAM,WAAW,GAAG;AACxC,UAAI,eAAe,UAAa,KAAK,MAAM,SAAS,OAAW,cAAa;AAC5E,UAAI,eAAe,OAAW,cAAa;AAAA,IAC7C;AAGA,UAAM,2BACJ,KAAK,MAAM,YAAY,SAAS,cAAc,KAAK,aAAa,KAAK,MAAM,WAAW;AACxF,UAAM,oBAAoB,KAAK,MAAM,qBAAqB;AAE1D,UAAM,aAAsC;AAAA,MAC1C,aAAa,KAAK,MAAM;AAAA,MACxB,UAAU,KAAK,MAAM;AAAA,MACrB,cAAc,KAAK,MAAM;AAAA,MACzB,cAAc,KAAK,MAAM;AAAA,MACzB,kCAAkC,KAAK,MAAM;AAAA,MAC7C,kBAAkB;AAAA,MAClB,kBAAkB;AAAA,MAClB,iBACE,KAAK,MAAM,mBAAmB,SAC1B,KAAK,UAAU,KAAK,MAAM,cAAc,IACxC;AAAA,MACN,oBAAoB;AAAA,MACpB,oBAAoB,KAAK,MAAM;AAAA,MAC/B,QAAQ,KAAK,MAAM;AAAA,MACnB,eAAe,KAAK,MAAM;AAAA,MAC1B,0BAA0B,KAAK,MAAM;AAAA,MACrC,eAAe,KAAK,MAAM;AAAA,MAC1B,gBAAgB,KAAK,MAAM;AAAA,MAC3B,cAAc,KAAK,MAAM;AAAA,MACzB,QAAQ,KAAK,MAAM;AAAA,MACnB,aAAa,KAAK,MAAM;AAAA,MACxB,uBAAuB,KAAK,MAAM;AAAA,MAClC,MAAM,KAAK,MAAM;AAAA,IACnB;AAEA,UAAM,MAAM,IAAI,IAAI,GAAG,KAAK,MAAM,OAAO,QAAQ;AAEjD,eAAW,CAAC,KAAK,KAAK,KAAK,OAAO,QAAQ,UAAU,GAAG;AACrD,UAAI,UAAU,UAAa,UAAU,KAAM;AAC3C,UAAI,OAAO,UAAU,WAAW;AAC9B,YAAI,aAAa,OAAO,KAAK,QAAQ,SAAS,OAAO;AAAA,MACvD,OAAO;AACL,YAAI,aAAa,OAAO,KAAK,OAAO,KAAK,CAAC;AAAA,MAC5C;AAAA,IACF;AAEA,UAAM,KAAK,IAAI,oBAAU,KAAK;AAAA,MAC5B,SAAS;AAAA,QACP,eAAe,KAAK,MAAM;AAAA,QAC1B,gBAAgB;AAAA,QAChB,cAAc;AAAA,MAChB;AAAA,IACF,CAAC;AAED,UAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,SAAG,GAAG,QAAQ,MAAM,QAAQ,CAAC;AAC7B,SAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,SAAG,GAAG,SAAS,CAAC,SAAS,OAAO,IAAI,MAAM,sBAAsB,IAAI,EAAE,CAAC,CAAC;AAAA,IAC1E,CAAC;AAED,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,OAAO,IAAe;AAC1B,QAAI,UAAU;AACd,UAAM,oBAAoB,IAAI,gBAAgB;AAG9C,UAAM,YAAY,mBAAK,KAAK,OAAO,eAAe;AAChD,YAAM,SAAS,IAAI,QAAc,CAAC,GAAG,WAAW;AAC9C,WAAG,KAAK,SAAS,CAAC,MAAM,WAAW;AACjC,cAAI,CAAC,SAAS;AACZ,iBAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,MAAM,EAAE;AAClE,mBAAO,IAAI,MAAM,kBAAkB,CAAC;AAAA,UACtC;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,YAAQ,4BAAa,WAAW,MAAM,CAAC,CAAC;AAAA,IAC9D,CAAC;AAED,UAAM,WAAW,YAAY;AAC3B,YAAM,mBAAmB,KAAK,MAAO,KAAK,MAAM,aAAa,KAAK,MAAM,eAAgB,GAAI;AAC5F,YAAM,cAAc,IAAI,8BAAgB,KAAK,MAAM,YAAY,GAAG,gBAAgB;AAElF,YAAM,mBAAe,4BAAa,KAAK,WAAW;AAClD,YAAM,mBAAe,4BAAa,kBAAkB,MAAM;AAE1D,UAAI;AACF,eAAO,CAAC,KAAK,QAAQ;AACnB,gBAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,KAAK,MAAM,KAAK,GAAG,cAAc,YAAY,CAAC;AAEjF,cAAI,WAAW,OAAW;AAC1B,cAAI,OAAO,KAAM;AAEjB,gBAAM,OAAO,OAAO;AAEpB,cAAI;AACJ,cAAI,SAAS,aAAa,gBAAgB;AACxC,qBAAS,YAAY,MAAM;AAAA,UAC7B,WAAW,KAAK,eAAe,KAAK,MAAM,cAAc,KAAK,aAAa,GAAG;AAG3E,qBAAS,YAAY,MAAM,KAAK,KAAK,MAAqB;AAAA,UAC5D,OAAO;AACL,kBAAM,IAAI,MAAM,sDAAsD;AAAA,UACxE;AAEA,qBAAW,SAAS,QAAQ;AAC1B,iBAAK,sBAAsB,MAAM,oBAAoB,MAAM;AAC3D,eAAG,KAAK,MAAM,KAAK,MAAM;AAAA,UAC3B;AAAA,QACF;AAAA,MACF,UAAE;AACA,kBAAU;AACV,YAAI;AACF,aAAG,KAAK,aAAa,SAAS;AAAA,QAChC,QAAQ;AAAA,QAER;AACA,kBAAU,OAAO;AAAA,MACnB;AAAA,IACF;AAEA,QAAI,iBAAqE;AACzE,UAAM,aAAa,mBAAK,KAAK,OAAO,eAAe;AACjD,YAAM,gBAAgB,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3D,yBAAiB,CAAC,KAAK,aAAa;AAClC,cAAI,UAAU;AACZ,iBAAK,QAAQ,MAAM,2CAA2C;AAC9D;AAAA,UACF;AACA,cAAI;AACF,kBAAM,OAAO,KAAK,MAAM,IAAI,SAAS,CAAC;AACtC,iBAAK,oBAAoB,IAAI;AAC7B,gBAAI,KAAK,UAAU,SAAS;AAC1B,sBAAQ;AAAA,YACV;AAAA,UACF,SAAS,KAAK;AACZ,iBAAK,QAAQ,MAAM,yCAAyC,GAAG,EAAE;AACjE,mBAAO,GAAG;AAAA,UACZ;AAAA,QACF;AACA,WAAG,GAAG,WAAW,cAAc;AAAA,MACjC,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,mBAAe,4BAAa,WAAW,MAAM,CAAC,CAAC;AAAA,IACrE,CAAC;AAED,UAAM,aAAa,mBAAK,KAAK,OAAO,eAAe;AAEjD,aAAO,KAAK,uBAAuB,SAAS,GAAG;AAC7C,cAAM,MAAM,KAAK,uBAAuB,MAAM;AAC9C,WAAG,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,MAC7B;AAEA,aAAO,CAAC,WAAW,OAAO,SAAS;AACjC,cAAM,QAAQ,KAAK,CAAC,KAAK,sBAAsB,WAAO,4BAAa,WAAW,MAAM,CAAC,CAAC;AACtF,YAAI,WAAW,OAAO,QAAS;AAE/B,aAAK,wBAAwB,IAAI,qBAAO;AACxC,eAAO,KAAK,uBAAuB,SAAS,GAAG;AAC7C,gBAAM,MAAM,KAAK,uBAAuB,MAAM;AAC9C,aAAG,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,QAC7B;AAAA,MACF;AAAA,IACF,CAAC;AAED,QAAI;AACF,YAAM,QAAQ,IAAI,CAAC,SAAS,GAAG,WAAW,QAAQ,UAAU,MAAM,CAAC;AAAA,IACrE,UAAE;AACA,gBAAU;AACV,wBAAkB,MAAM;AACxB,iBAAW,OAAO;AAClB,iBAAW,OAAO;AAClB,UAAI,eAAgB,IAAG,IAAI,WAAW,cAAc;AACpD,UAAI;AACF,WAAG,MAAM;AAAA,MACX,QAAQ;AAAA,MAER;AAAA,IACF;AAAA,EACF;AAAA,EAEA,mBAAmB,OAA+C;AAChE,QAAI,MAAM,WAAW,EAAG,QAAO;AAC/B,WAAO,MAAM,OAAO,CAAC,KAAK,MAAM,OAAO,EAAE,cAAc,IAAI,CAAC,IAAI,MAAM;AAAA,EACxE;AAAA,EAEA,oBAAoB,MAA0B;AAC5C,UAAM,cAAc,KAAK;AAEzB,QAAI,gBAAgB,SAAS;AAC3B,WAAK,aAAa,KAAK,MAAM;AAC7B,WAAK,aAAa,KAAK,cAAc;AACrC,WAAK,QAAQ;AAAA,QACX,iCAAiC,KAAK,UAAU,eAAe,KAAK,UAAU;AAAA,MAChF;AACA;AAAA,IACF;AAEA,QAAI,gBAAgB,iBAAiB;AACnC,WAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,gBAAgB,CAAC;AAC5D;AAAA,IACF;AAEA,QAAI,gBAAgB,eAAe;AACjC,WAAK,QAAQ;AAAA,QACX,gDAAgD,KAAK,sBAAsB,sBAAsB,KAAK,wBAAwB;AAAA,MAChI;AACA;AAAA,IACF;AAEA,QAAI,gBAAgB,QAAQ;AAC1B;AAAA,IACF;AAEA,UAAM,QAAQ,KAAK,SAAS,CAAC;AAC7B,UAAM,YAAY,QAAQ,KAAK,WAAW;AAC1C,UAAM,kBAAkB,QAAQ,KAAK,iBAAiB;AACtD,UAAM,YAAY,KAAK,aAAa;AACpC,UAAM,aAAa,KAAK,cAAc;AACtC,UAAM,eAAW,iCAAkB,KAAK,iBAAiB,IAAI;AAC7D,UAAM,WAAW,mBAAmB,IAAI;AAIxC,UAAM,aAAa,MAAM;AAAA,MAAI,CAAC,aAC5B,iCAAkB;AAAA,QAChB,MAAM,KAAK,QAAQ;AAAA,QACnB,YAAY,KAAK,SAAS,KAAK,MAAO,KAAK;AAAA,QAC3C,UAAU,KAAK,OAAO,KAAK,MAAO,KAAK;AAAA,QACvC,YAAY,KAAK,cAAc;AAAA,QAC/B,iBAAiB,KAAK;AAAA,MACxB,CAAC;AAAA,IACH;AAEA,QAAI,YAAY;AAChB,QAAI,UAAU;AACd,QAAI,aAAa;AAGjB,QAAI,WAAW,SAAS,GAAG;AACzB,YAAM,cAAc,WAAW,IAAI,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,GAAG;AAC1D,kBAAY,WAAW,CAAC,EAAG,aAAa;AACxC,gBAAU,WAAW,WAAW,SAAS,CAAC,EAAG,WAAW;AACxD,mBAAa,KAAK,mBAAmB,UAAU;AAE/C,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,kBAAI,gBAAgB;AAAA,QAC1B,cAAc;AAAA,UACZ;AAAA,YACE;AAAA,YACA,MAAM;AAAA,YACN;AAAA,YACA;AAAA,YACA;AAAA,YACA,OAAO;AAAA,YACP,GAAI,WAAW,EAAE,SAAS,IAAI,CAAC;AAAA,UACjC;AAAA,QACF;AAAA,MACF,CAAC;AAAA,IACH;AAIA,QAAI,WAAW;AACb,UAAI,KAAK,4BAA4B,GAAG;AACtC,aAAK,0BAA0B;AAAA,MACjC;AAEA,YAAM,iBAAiB,WAAW;AAAA,QAChC,CAAC,MAAM,EAAE,cAAc,UAAa,EAAE,aAAa,KAAK;AAAA,MAC1D;AACA,YAAM,sBAAsB,KAAK,mBAAmB,cAAc;AAElE,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,kBAAI,gBAAgB;AAAA,QAC1B,cAAc;AAAA,UACZ;AAAA,YACE;AAAA,YACA,MAAM;AAAA,YACN,WAAW,KAAK;AAAA,YAChB;AAAA,YACA,YAAY;AAAA,YACZ,OAAO;AAAA,YACP,GAAI,WAAW,EAAE,SAAS,IAAI,CAAC;AAAA,UACjC;AAAA,QACF;AAAA,MACF,CAAC;AACD,WAAK,0BAA0B;AAAA,IACjC;AAIA,UAAM,sBAAsB,KAAK,MAAM,gBAAgB,QAAQ,CAAC;AAChE,QAAI,aAAa,CAAC,qBAAqB;AACrC,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,kBAAI,gBAAgB;AAAA,QAC1B,cAAc;AAAA,UACZ;AAAA,YACE;AAAA,YACA,MAAM;AAAA,YACN;AAAA,YACA;AAAA,YACA;AAAA,YACA,OAAO;AAAA,YACP,GAAI,WAAW,EAAE,SAAS,IAAI,CAAC;AAAA,UACjC;AAAA,QACF;AAAA,MACF,CAAC;AAED,WAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,cAAc,CAAC;AAC1D,UAAI,KAAK,qBAAqB,GAAG;AAC/B,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,kBAAI,gBAAgB;AAAA;AAAA;AAAA;AAAA,UAI1B,WAAW,KAAK,cAAc;AAAA,UAC9B,kBAAkB;AAAA,YAChB,eAAe,KAAK;AAAA,UACtB;AAAA,QACF,CAAC;AACD,aAAK,qBAAqB;AAC1B,aAAK,0BAA0B;AAAA,MACjC;AAAA,IACF;AAAA,EACF;AACF;","names":["stt"]}
package/dist/stt.d.cts CHANGED
@@ -12,6 +12,11 @@ export interface STTOptions {
12
12
  encoding: STTEncoding;
13
13
  speechModel: STTModels;
14
14
  languageDetection?: boolean;
15
+ /**
16
+ * Session inactivity timeout in seconds. AssemblyAI accepts integer values
17
+ * from 5 to 3600; when unset, no inactivity timeout is applied.
18
+ */
19
+ inactivityTimeout?: number;
15
20
  endOfTurnConfidenceThreshold?: number;
16
21
  /** Minimum silence (ms) before a confident end-of-turn is finalized. */
17
22
  minTurnSilence?: number;
@@ -19,8 +24,12 @@ export interface STTOptions {
19
24
  maxTurnSilence?: number;
20
25
  formatTurns?: boolean;
21
26
  keytermsPrompt?: string[];
22
- /** Only supported with the `u3-rt-pro` model. */
27
+ /** Only supported with the Universal-3 Pro model family. */
23
28
  prompt?: string;
29
+ /** Only supported with the Universal-3 Pro model family. */
30
+ agentContext?: string;
31
+ /** Only supported with the Universal-3 Pro model family. Set at connection time only. */
32
+ previousContextNTurns?: number;
24
33
  vadThreshold?: number;
25
34
  /**
26
35
  * Enable speaker diarization. Note: AssemblyAI will return per-word speaker
@@ -38,8 +47,8 @@ export interface STTOptions {
38
47
  /** Background audio suppression aggressiveness, from 0.0 to 1.0. Connect-time only. */
39
48
  voiceFocusThreshold?: number;
40
49
  /**
41
- * Accuracy/latency preset for u3-rt-pro: `min_latency`, `balanced`, or `max_accuracy`.
42
- * Explicit silence, partials, or VAD options still take precedence over mode defaults.
50
+ * Accuracy/latency preset for the Universal-3 Pro model family: `min_latency`, `balanced`,
51
+ * or `max_accuracy`. Explicit turn-silence values still take precedence over mode defaults.
43
52
  */
44
53
  mode?: 'min_latency' | 'balanced' | 'max_accuracy';
45
54
  baseUrl: string;