@livekit/agents-plugin-deepgram 0.5.7 → 1.0.0-next.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/_utils.cjs +56 -0
- package/dist/_utils.cjs.map +1 -0
- package/dist/_utils.d.cts +12 -0
- package/dist/_utils.d.ts +12 -0
- package/dist/_utils.d.ts.map +1 -0
- package/dist/_utils.js +32 -0
- package/dist/_utils.js.map +1 -0
- package/dist/index.cjs +14 -3
- package/dist/index.cjs.map +1 -1
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +11 -0
- package/dist/index.js.map +1 -1
- package/dist/models.cjs.map +1 -1
- package/dist/models.d.cts +2 -2
- package/dist/models.d.ts +2 -2
- package/dist/models.d.ts.map +1 -1
- package/dist/stt.cjs +28 -5
- 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 +28 -5
- package/dist/stt.js.map +1 -1
- package/package.json +7 -7
- package/src/_utils.ts +50 -0
- package/src/index.ts +14 -1
- package/src/models.ts +4 -1
- package/src/stt.ts +29 -6
package/dist/_utils.cjs
ADDED
|
@@ -0,0 +1,56 @@
|
|
|
1
|
+
"use strict";
|
|
2
|
+
var __defProp = Object.defineProperty;
|
|
3
|
+
var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
|
|
4
|
+
var __getOwnPropNames = Object.getOwnPropertyNames;
|
|
5
|
+
var __hasOwnProp = Object.prototype.hasOwnProperty;
|
|
6
|
+
var __export = (target, all) => {
|
|
7
|
+
for (var name in all)
|
|
8
|
+
__defProp(target, name, { get: all[name], enumerable: true });
|
|
9
|
+
};
|
|
10
|
+
var __copyProps = (to, from, except, desc) => {
|
|
11
|
+
if (from && typeof from === "object" || typeof from === "function") {
|
|
12
|
+
for (let key of __getOwnPropNames(from))
|
|
13
|
+
if (!__hasOwnProp.call(to, key) && key !== except)
|
|
14
|
+
__defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
|
|
15
|
+
}
|
|
16
|
+
return to;
|
|
17
|
+
};
|
|
18
|
+
var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
|
|
19
|
+
var utils_exports = {};
|
|
20
|
+
__export(utils_exports, {
|
|
21
|
+
PeriodicCollector: () => PeriodicCollector
|
|
22
|
+
});
|
|
23
|
+
module.exports = __toCommonJS(utils_exports);
|
|
24
|
+
class PeriodicCollector {
|
|
25
|
+
duration;
|
|
26
|
+
callback;
|
|
27
|
+
lastFlushTime;
|
|
28
|
+
total = null;
|
|
29
|
+
constructor(callback, options) {
|
|
30
|
+
this.duration = options.duration;
|
|
31
|
+
this.callback = callback;
|
|
32
|
+
this.lastFlushTime = performance.now() / 1e3;
|
|
33
|
+
}
|
|
34
|
+
push(value) {
|
|
35
|
+
if (this.total === null) {
|
|
36
|
+
this.total = value;
|
|
37
|
+
} else {
|
|
38
|
+
this.total = this.total + value;
|
|
39
|
+
}
|
|
40
|
+
if (performance.now() / 1e3 - this.lastFlushTime >= this.duration) {
|
|
41
|
+
this.flush();
|
|
42
|
+
}
|
|
43
|
+
}
|
|
44
|
+
flush() {
|
|
45
|
+
if (this.total !== null) {
|
|
46
|
+
this.callback(this.total);
|
|
47
|
+
this.total = null;
|
|
48
|
+
}
|
|
49
|
+
this.lastFlushTime = performance.now() / 1e3;
|
|
50
|
+
}
|
|
51
|
+
}
|
|
52
|
+
// Annotate the CommonJS export names for ESM import in node:
|
|
53
|
+
0 && (module.exports = {
|
|
54
|
+
PeriodicCollector
|
|
55
|
+
});
|
|
56
|
+
//# sourceMappingURL=_utils.cjs.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"sources":["../src/_utils.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\nexport class PeriodicCollector<T> {\n private duration: number;\n private callback: (value: T) => void;\n private lastFlushTime: number;\n private total: T | null = null;\n\n constructor(callback: (value: T) => void, options: { duration: number }) {\n /**\n * Create a new periodic collector that accumulates values and calls the callback\n * after the specified duration if there are values to report.\n *\n * @param callback Function to call with accumulated value when duration expires\n * @param options.duration Time in seconds between callback invocations\n */\n this.duration = options.duration;\n this.callback = callback;\n this.lastFlushTime = performance.now() / 1000; // Convert to seconds\n }\n\n push(value: T): void {\n /**\n * Add a value to the accumulator\n */\n if (this.total === null) {\n this.total = value;\n } else {\n // Type assertion needed for generic addition\n this.total = (this.total as any) + (value as any);\n }\n\n if (performance.now() / 1000 - this.lastFlushTime >= this.duration) {\n this.flush();\n }\n }\n\n flush(): void {\n /**\n * Force callback to be called with current total if non-zero\n */\n if (this.total !== null) {\n this.callback(this.total);\n this.total = null;\n }\n this.lastFlushTime = performance.now() / 1000;\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAIO,MAAM,kBAAqB;AAAA,EACxB;AAAA,EACA;AAAA,EACA;AAAA,EACA,QAAkB;AAAA,EAE1B,YAAY,UAA8B,SAA+B;AAQvE,SAAK,WAAW,QAAQ;AACxB,SAAK,WAAW;AAChB,SAAK,gBAAgB,YAAY,IAAI,IAAI;AAAA,EAC3C;AAAA,EAEA,KAAK,OAAgB;AAInB,QAAI,KAAK,UAAU,MAAM;AACvB,WAAK,QAAQ;AAAA,IACf,OAAO;AAEL,WAAK,QAAS,KAAK,QAAiB;AAAA,IACtC;AAEA,QAAI,YAAY,IAAI,IAAI,MAAO,KAAK,iBAAiB,KAAK,UAAU;AAClE,WAAK,MAAM;AAAA,IACb;AAAA,EACF;AAAA,EAEA,QAAc;AAIZ,QAAI,KAAK,UAAU,MAAM;AACvB,WAAK,SAAS,KAAK,KAAK;AACxB,WAAK,QAAQ;AAAA,IACf;AACA,SAAK,gBAAgB,YAAY,IAAI,IAAI;AAAA,EAC3C;AACF;","names":[]}
|
|
@@ -0,0 +1,12 @@
|
|
|
1
|
+
export declare class PeriodicCollector<T> {
|
|
2
|
+
private duration;
|
|
3
|
+
private callback;
|
|
4
|
+
private lastFlushTime;
|
|
5
|
+
private total;
|
|
6
|
+
constructor(callback: (value: T) => void, options: {
|
|
7
|
+
duration: number;
|
|
8
|
+
});
|
|
9
|
+
push(value: T): void;
|
|
10
|
+
flush(): void;
|
|
11
|
+
}
|
|
12
|
+
//# sourceMappingURL=_utils.d.ts.map
|
package/dist/_utils.d.ts
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
1
|
+
export declare class PeriodicCollector<T> {
|
|
2
|
+
private duration;
|
|
3
|
+
private callback;
|
|
4
|
+
private lastFlushTime;
|
|
5
|
+
private total;
|
|
6
|
+
constructor(callback: (value: T) => void, options: {
|
|
7
|
+
duration: number;
|
|
8
|
+
});
|
|
9
|
+
push(value: T): void;
|
|
10
|
+
flush(): void;
|
|
11
|
+
}
|
|
12
|
+
//# sourceMappingURL=_utils.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"_utils.d.ts","sourceRoot":"","sources":["../src/_utils.ts"],"names":[],"mappings":"AAIA,qBAAa,iBAAiB,CAAC,CAAC;IAC9B,OAAO,CAAC,QAAQ,CAAS;IACzB,OAAO,CAAC,QAAQ,CAAqB;IACrC,OAAO,CAAC,aAAa,CAAS;IAC9B,OAAO,CAAC,KAAK,CAAkB;gBAEnB,QAAQ,EAAE,CAAC,KAAK,EAAE,CAAC,KAAK,IAAI,EAAE,OAAO,EAAE;QAAE,QAAQ,EAAE,MAAM,CAAA;KAAE;IAavE,IAAI,CAAC,KAAK,EAAE,CAAC,GAAG,IAAI;IAgBpB,KAAK,IAAI,IAAI;CAUd"}
|
package/dist/_utils.js
ADDED
|
@@ -0,0 +1,32 @@
|
|
|
1
|
+
class PeriodicCollector {
|
|
2
|
+
duration;
|
|
3
|
+
callback;
|
|
4
|
+
lastFlushTime;
|
|
5
|
+
total = null;
|
|
6
|
+
constructor(callback, options) {
|
|
7
|
+
this.duration = options.duration;
|
|
8
|
+
this.callback = callback;
|
|
9
|
+
this.lastFlushTime = performance.now() / 1e3;
|
|
10
|
+
}
|
|
11
|
+
push(value) {
|
|
12
|
+
if (this.total === null) {
|
|
13
|
+
this.total = value;
|
|
14
|
+
} else {
|
|
15
|
+
this.total = this.total + value;
|
|
16
|
+
}
|
|
17
|
+
if (performance.now() / 1e3 - this.lastFlushTime >= this.duration) {
|
|
18
|
+
this.flush();
|
|
19
|
+
}
|
|
20
|
+
}
|
|
21
|
+
flush() {
|
|
22
|
+
if (this.total !== null) {
|
|
23
|
+
this.callback(this.total);
|
|
24
|
+
this.total = null;
|
|
25
|
+
}
|
|
26
|
+
this.lastFlushTime = performance.now() / 1e3;
|
|
27
|
+
}
|
|
28
|
+
}
|
|
29
|
+
export {
|
|
30
|
+
PeriodicCollector
|
|
31
|
+
};
|
|
32
|
+
//# sourceMappingURL=_utils.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"sources":["../src/_utils.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\nexport class PeriodicCollector<T> {\n private duration: number;\n private callback: (value: T) => void;\n private lastFlushTime: number;\n private total: T | null = null;\n\n constructor(callback: (value: T) => void, options: { duration: number }) {\n /**\n * Create a new periodic collector that accumulates values and calls the callback\n * after the specified duration if there are values to report.\n *\n * @param callback Function to call with accumulated value when duration expires\n * @param options.duration Time in seconds between callback invocations\n */\n this.duration = options.duration;\n this.callback = callback;\n this.lastFlushTime = performance.now() / 1000; // Convert to seconds\n }\n\n push(value: T): void {\n /**\n * Add a value to the accumulator\n */\n if (this.total === null) {\n this.total = value;\n } else {\n // Type assertion needed for generic addition\n this.total = (this.total as any) + (value as any);\n }\n\n if (performance.now() / 1000 - this.lastFlushTime >= this.duration) {\n this.flush();\n }\n }\n\n flush(): void {\n /**\n * Force callback to be called with current total if non-zero\n */\n if (this.total !== null) {\n this.callback(this.total);\n this.total = null;\n }\n this.lastFlushTime = performance.now() / 1000;\n }\n}\n"],"mappings":"AAIO,MAAM,kBAAqB;AAAA,EACxB;AAAA,EACA;AAAA,EACA;AAAA,EACA,QAAkB;AAAA,EAE1B,YAAY,UAA8B,SAA+B;AAQvE,SAAK,WAAW,QAAQ;AACxB,SAAK,WAAW;AAChB,SAAK,gBAAgB,YAAY,IAAI,IAAI;AAAA,EAC3C;AAAA,EAEA,KAAK,OAAgB;AAInB,QAAI,KAAK,UAAU,MAAM;AACvB,WAAK,QAAQ;AAAA,IACf,OAAO;AAEL,WAAK,QAAS,KAAK,QAAiB;AAAA,IACtC;AAEA,QAAI,YAAY,IAAI,IAAI,MAAO,KAAK,iBAAiB,KAAK,UAAU;AAClE,WAAK,MAAM;AAAA,IACb;AAAA,EACF;AAAA,EAEA,QAAc;AAIZ,QAAI,KAAK,UAAU,MAAM;AACvB,WAAK,SAAS,KAAK,KAAK;AACxB,WAAK,QAAQ;AAAA,IACf;AACA,SAAK,gBAAgB,YAAY,IAAI,IAAI;AAAA,EAC3C;AACF;","names":[]}
|
package/dist/index.cjs
CHANGED
|
@@ -13,9 +13,20 @@ var __copyProps = (to, from, except, desc) => {
|
|
|
13
13
|
};
|
|
14
14
|
var __reExport = (target, mod, secondTarget) => (__copyProps(target, mod, "default"), secondTarget && __copyProps(secondTarget, mod, "default"));
|
|
15
15
|
var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
|
|
16
|
-
var
|
|
17
|
-
module.exports = __toCommonJS(
|
|
18
|
-
|
|
16
|
+
var index_exports = {};
|
|
17
|
+
module.exports = __toCommonJS(index_exports);
|
|
18
|
+
var import_agents = require("@livekit/agents");
|
|
19
|
+
__reExport(index_exports, require("./stt.cjs"), module.exports);
|
|
20
|
+
class DeepgramPlugin extends import_agents.Plugin {
|
|
21
|
+
constructor() {
|
|
22
|
+
super({
|
|
23
|
+
title: "deepgram",
|
|
24
|
+
version: "0.5.6",
|
|
25
|
+
package: "@livekit/agents-plugin-deepgram"
|
|
26
|
+
});
|
|
27
|
+
}
|
|
28
|
+
}
|
|
29
|
+
import_agents.Plugin.registerPlugin(new DeepgramPlugin());
|
|
19
30
|
// Annotate the CommonJS export names for ESM import in node:
|
|
20
31
|
0 && (module.exports = {
|
|
21
32
|
...require("./stt.cjs")
|
package/dist/index.cjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText:
|
|
1
|
+
{"version":3,"sources":["../src/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { Plugin } from '@livekit/agents';\n\nexport * from './stt.js';\n\nclass DeepgramPlugin extends Plugin {\n constructor() {\n super({\n title: 'deepgram',\n version: '0.5.6',\n package: '@livekit/agents-plugin-deepgram',\n });\n }\n}\n\nPlugin.registerPlugin(new DeepgramPlugin());\n"],"mappings":";;;;;;;;;;;;;;;AAAA;AAAA;AAGA,oBAAuB;AAEvB,0BAAc,qBALd;AAOA,MAAM,uBAAuB,qBAAO;AAAA,EAClC,cAAc;AACZ,UAAM;AAAA,MACJ,OAAO;AAAA,MACP,SAAS;AAAA,MACT,SAAS;AAAA,IACX,CAAC;AAAA,EACH;AACF;AAEA,qBAAO,eAAe,IAAI,eAAe,CAAC;","names":[]}
|
package/dist/index.d.ts.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"
|
|
1
|
+
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAKA,cAAc,UAAU,CAAC"}
|
package/dist/index.js
CHANGED
|
@@ -1,2 +1,13 @@
|
|
|
1
|
+
import { Plugin } from "@livekit/agents";
|
|
1
2
|
export * from "./stt.js";
|
|
3
|
+
class DeepgramPlugin extends Plugin {
|
|
4
|
+
constructor() {
|
|
5
|
+
super({
|
|
6
|
+
title: "deepgram",
|
|
7
|
+
version: "0.5.6",
|
|
8
|
+
package: "@livekit/agents-plugin-deepgram"
|
|
9
|
+
});
|
|
10
|
+
}
|
|
11
|
+
}
|
|
12
|
+
Plugin.registerPlugin(new DeepgramPlugin());
|
|
2
13
|
//# sourceMappingURL=index.js.map
|
package/dist/index.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText:
|
|
1
|
+
{"version":3,"sources":["../src/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { Plugin } from '@livekit/agents';\n\nexport * from './stt.js';\n\nclass DeepgramPlugin extends Plugin {\n constructor() {\n super({\n title: 'deepgram',\n version: '0.5.6',\n package: '@livekit/agents-plugin-deepgram',\n });\n }\n}\n\nPlugin.registerPlugin(new DeepgramPlugin());\n"],"mappings":"AAGA,SAAS,cAAc;AAEvB,cAAc;AAEd,MAAM,uBAAuB,OAAO;AAAA,EAClC,cAAc;AACZ,UAAM;AAAA,MACJ,OAAO;AAAA,MACP,SAAS;AAAA,MACT,SAAS;AAAA,IACX,CAAC;AAAA,EACH;AACF;AAEA,OAAO,eAAe,IAAI,eAAe,CAAC;","names":[]}
|
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\nexport type STTModels =\n | 'nova-general'\n | 'nova-phonecall'\n | 'nova-meeting'\n | 'nova-2-general'\n | 'nova-2-meeting'\n | 'nova-2-phonecall'\n | 'nova-2-finance'\n | 'nova-2-conversationalai'\n | 'nova-2-voicemail'\n | 'nova-2-video'\n | 'nova-2-medical'\n | 'nova-2-drivethru'\n | 'nova-2-automotive'\n | 'nova-3-general'\n | 'enhanced-general'\n | 'enhanced-meeting'\n | 'enhanced-phonecall'\n | 'enhanced-finance'\n | 'base'\n | 'meeting'\n | 'phonecall'\n | 'finance'\n | 'conversationalai'\n | 'voicemail'\n | 'video'\n | 'whisper-tiny'\n | 'whisper-base'\n | 'whisper-small'\n | 'whisper-medium'\n | 'whisper-large';\n\nexport type STTLanguages =\n | 'da'\n | 'de'\n | 'en'\n | 'en-AU'\n | 'en-GB'\n | 'en-IN'\n | 'en-NZ'\n | 'en-US'\n | 'es'\n | 'es-419'\n | 'es-LATAM'\n | 'fr'\n | 'fr-CA'\n | 'hi'\n | 'hi-Latn'\n | 'id'\n | 'it'\n | 'ja'\n | 'ko'\n | 'nl'\n | 'no'\n | 'pl'\n | 'pt'\n | 'pt-BR'\n | 'ru'\n | 'sv'\n | 'ta'\n | 'taq'\n | 'th'\n | 'tr'\n | 'uk'\n | 'zh'\n | 'zh-CN'\n | 'zh-TW';\n"],"mappings":";;;;;;;;;;;;;;AAAA;AAAA;","names":[]}
|
|
1
|
+
{"version":3,"sources":["../src/models.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\nexport type STTModels =\n | 'nova-general'\n | 'nova-phonecall'\n | 'nova-meeting'\n | 'nova-2-general'\n | 'nova-2-meeting'\n | 'nova-2-phonecall'\n | 'nova-2-finance'\n | 'nova-2-conversationalai'\n | 'nova-2-voicemail'\n | 'nova-2-video'\n | 'nova-2-medical'\n | 'nova-2-drivethru'\n | 'nova-2-automotive'\n | 'nova-3'\n | 'nova-3-general'\n | 'nova-3-medical'\n | 'enhanced-general'\n | 'enhanced-meeting'\n | 'enhanced-phonecall'\n | 'enhanced-finance'\n | 'base'\n | 'meeting'\n | 'phonecall'\n | 'finance'\n | 'conversationalai'\n | 'voicemail'\n | 'video'\n | 'whisper-tiny'\n | 'whisper-base'\n | 'whisper-small'\n | 'whisper-medium'\n | 'whisper-large';\n\nexport type STTLanguages =\n | 'da'\n | 'de'\n | 'en'\n | 'en-AU'\n | 'en-GB'\n | 'en-IN'\n | 'en-NZ'\n | 'en-US'\n | 'es'\n | 'es-419'\n | 'es-LATAM'\n | 'fr'\n | 'fr-CA'\n | 'hi'\n | 'hi-Latn'\n | 'id'\n | 'it'\n | 'ja'\n | 'ko'\n | 'nl'\n | 'no'\n | 'pl'\n | 'pt'\n | 'pt-BR'\n | 'ru'\n | 'sv'\n | 'ta'\n | 'taq'\n | 'th'\n | 'tr'\n | 'uk'\n | 'zh'\n | 'zh-CN'\n | 'zh-TW'\n | 'multi';\n"],"mappings":";;;;;;;;;;;;;;AAAA;AAAA;","names":[]}
|
package/dist/models.d.cts
CHANGED
|
@@ -1,3 +1,3 @@
|
|
|
1
|
-
export type STTModels = 'nova-general' | 'nova-phonecall' | 'nova-meeting' | 'nova-2-general' | 'nova-2-meeting' | 'nova-2-phonecall' | 'nova-2-finance' | 'nova-2-conversationalai' | 'nova-2-voicemail' | 'nova-2-video' | 'nova-2-medical' | 'nova-2-drivethru' | 'nova-2-automotive' | 'nova-3-general' | 'enhanced-general' | 'enhanced-meeting' | 'enhanced-phonecall' | 'enhanced-finance' | 'base' | 'meeting' | 'phonecall' | 'finance' | 'conversationalai' | 'voicemail' | 'video' | 'whisper-tiny' | 'whisper-base' | 'whisper-small' | 'whisper-medium' | 'whisper-large';
|
|
2
|
-
export type STTLanguages = 'da' | 'de' | 'en' | 'en-AU' | 'en-GB' | 'en-IN' | 'en-NZ' | 'en-US' | 'es' | 'es-419' | 'es-LATAM' | 'fr' | 'fr-CA' | 'hi' | 'hi-Latn' | 'id' | 'it' | 'ja' | 'ko' | 'nl' | 'no' | 'pl' | 'pt' | 'pt-BR' | 'ru' | 'sv' | 'ta' | 'taq' | 'th' | 'tr' | 'uk' | 'zh' | 'zh-CN' | 'zh-TW';
|
|
1
|
+
export type STTModels = 'nova-general' | 'nova-phonecall' | 'nova-meeting' | 'nova-2-general' | 'nova-2-meeting' | 'nova-2-phonecall' | 'nova-2-finance' | 'nova-2-conversationalai' | 'nova-2-voicemail' | 'nova-2-video' | 'nova-2-medical' | 'nova-2-drivethru' | 'nova-2-automotive' | 'nova-3' | 'nova-3-general' | 'nova-3-medical' | 'enhanced-general' | 'enhanced-meeting' | 'enhanced-phonecall' | 'enhanced-finance' | 'base' | 'meeting' | 'phonecall' | 'finance' | 'conversationalai' | 'voicemail' | 'video' | 'whisper-tiny' | 'whisper-base' | 'whisper-small' | 'whisper-medium' | 'whisper-large';
|
|
2
|
+
export type STTLanguages = 'da' | 'de' | 'en' | 'en-AU' | 'en-GB' | 'en-IN' | 'en-NZ' | 'en-US' | 'es' | 'es-419' | 'es-LATAM' | 'fr' | 'fr-CA' | 'hi' | 'hi-Latn' | 'id' | 'it' | 'ja' | 'ko' | 'nl' | 'no' | 'pl' | 'pt' | 'pt-BR' | 'ru' | 'sv' | 'ta' | 'taq' | 'th' | 'tr' | 'uk' | 'zh' | 'zh-CN' | 'zh-TW' | 'multi';
|
|
3
3
|
//# sourceMappingURL=models.d.ts.map
|
package/dist/models.d.ts
CHANGED
|
@@ -1,3 +1,3 @@
|
|
|
1
|
-
export type STTModels = 'nova-general' | 'nova-phonecall' | 'nova-meeting' | 'nova-2-general' | 'nova-2-meeting' | 'nova-2-phonecall' | 'nova-2-finance' | 'nova-2-conversationalai' | 'nova-2-voicemail' | 'nova-2-video' | 'nova-2-medical' | 'nova-2-drivethru' | 'nova-2-automotive' | 'nova-3-general' | 'enhanced-general' | 'enhanced-meeting' | 'enhanced-phonecall' | 'enhanced-finance' | 'base' | 'meeting' | 'phonecall' | 'finance' | 'conversationalai' | 'voicemail' | 'video' | 'whisper-tiny' | 'whisper-base' | 'whisper-small' | 'whisper-medium' | 'whisper-large';
|
|
2
|
-
export type STTLanguages = 'da' | 'de' | 'en' | 'en-AU' | 'en-GB' | 'en-IN' | 'en-NZ' | 'en-US' | 'es' | 'es-419' | 'es-LATAM' | 'fr' | 'fr-CA' | 'hi' | 'hi-Latn' | 'id' | 'it' | 'ja' | 'ko' | 'nl' | 'no' | 'pl' | 'pt' | 'pt-BR' | 'ru' | 'sv' | 'ta' | 'taq' | 'th' | 'tr' | 'uk' | 'zh' | 'zh-CN' | 'zh-TW';
|
|
1
|
+
export type STTModels = 'nova-general' | 'nova-phonecall' | 'nova-meeting' | 'nova-2-general' | 'nova-2-meeting' | 'nova-2-phonecall' | 'nova-2-finance' | 'nova-2-conversationalai' | 'nova-2-voicemail' | 'nova-2-video' | 'nova-2-medical' | 'nova-2-drivethru' | 'nova-2-automotive' | 'nova-3' | 'nova-3-general' | 'nova-3-medical' | 'enhanced-general' | 'enhanced-meeting' | 'enhanced-phonecall' | 'enhanced-finance' | 'base' | 'meeting' | 'phonecall' | 'finance' | 'conversationalai' | 'voicemail' | 'video' | 'whisper-tiny' | 'whisper-base' | 'whisper-small' | 'whisper-medium' | 'whisper-large';
|
|
2
|
+
export type STTLanguages = 'da' | 'de' | 'en' | 'en-AU' | 'en-GB' | 'en-IN' | 'en-NZ' | 'en-US' | 'es' | 'es-419' | 'es-LATAM' | 'fr' | 'fr-CA' | 'hi' | 'hi-Latn' | 'id' | 'it' | 'ja' | 'ko' | 'nl' | 'no' | 'pl' | 'pt' | 'pt-BR' | 'ru' | 'sv' | 'ta' | 'taq' | 'th' | 'tr' | 'uk' | 'zh' | 'zh-CN' | 'zh-TW' | 'multi';
|
|
3
3
|
//# 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,MAAM,MAAM,SAAS,GACjB,cAAc,GACd,gBAAgB,GAChB,cAAc,GACd,gBAAgB,GAChB,gBAAgB,GAChB,kBAAkB,GAClB,gBAAgB,GAChB,yBAAyB,GACzB,kBAAkB,GAClB,cAAc,GACd,gBAAgB,GAChB,kBAAkB,GAClB,mBAAmB,GACnB,gBAAgB,GAChB,kBAAkB,GAClB,kBAAkB,GAClB,oBAAoB,GACpB,kBAAkB,GAClB,MAAM,GACN,SAAS,GACT,WAAW,GACX,SAAS,GACT,kBAAkB,GAClB,WAAW,GACX,OAAO,GACP,cAAc,GACd,cAAc,GACd,eAAe,GACf,gBAAgB,GAChB,eAAe,CAAC;AAEpB,MAAM,MAAM,YAAY,GACpB,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,OAAO,GACP,OAAO,GACP,OAAO,GACP,OAAO,GACP,OAAO,GACP,IAAI,GACJ,QAAQ,GACR,UAAU,GACV,IAAI,GACJ,OAAO,GACP,IAAI,GACJ,SAAS,GACT,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,OAAO,GACP,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,KAAK,GACL,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,OAAO,GACP,OAAO,CAAC"}
|
|
1
|
+
{"version":3,"file":"models.d.ts","sourceRoot":"","sources":["../src/models.ts"],"names":[],"mappings":"AAIA,MAAM,MAAM,SAAS,GACjB,cAAc,GACd,gBAAgB,GAChB,cAAc,GACd,gBAAgB,GAChB,gBAAgB,GAChB,kBAAkB,GAClB,gBAAgB,GAChB,yBAAyB,GACzB,kBAAkB,GAClB,cAAc,GACd,gBAAgB,GAChB,kBAAkB,GAClB,mBAAmB,GACnB,QAAQ,GACR,gBAAgB,GAChB,gBAAgB,GAChB,kBAAkB,GAClB,kBAAkB,GAClB,oBAAoB,GACpB,kBAAkB,GAClB,MAAM,GACN,SAAS,GACT,WAAW,GACX,SAAS,GACT,kBAAkB,GAClB,WAAW,GACX,OAAO,GACP,cAAc,GACd,cAAc,GACd,eAAe,GACf,gBAAgB,GAChB,eAAe,CAAC;AAEpB,MAAM,MAAM,YAAY,GACpB,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,OAAO,GACP,OAAO,GACP,OAAO,GACP,OAAO,GACP,OAAO,GACP,IAAI,GACJ,QAAQ,GACR,UAAU,GACV,IAAI,GACJ,OAAO,GACP,IAAI,GACJ,SAAS,GACT,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,OAAO,GACP,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,KAAK,GACL,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,IAAI,GACJ,OAAO,GACP,OAAO,GACP,OAAO,CAAC"}
|
package/dist/stt.cjs
CHANGED
|
@@ -24,6 +24,7 @@ __export(stt_exports, {
|
|
|
24
24
|
module.exports = __toCommonJS(stt_exports);
|
|
25
25
|
var import_agents = require("@livekit/agents");
|
|
26
26
|
var import_ws = require("ws");
|
|
27
|
+
var import_utils = require("./_utils.cjs");
|
|
27
28
|
const API_BASE_URL_V1 = "wss://api.deepgram.com/v1/listen";
|
|
28
29
|
const defaultSTTOptions = {
|
|
29
30
|
apiKey: process.env.DEEPGRAM_API_KEY,
|
|
@@ -31,7 +32,7 @@ const defaultSTTOptions = {
|
|
|
31
32
|
detectLanguage: false,
|
|
32
33
|
interimResults: true,
|
|
33
34
|
punctuate: true,
|
|
34
|
-
model: "nova-
|
|
35
|
+
model: "nova-3",
|
|
35
36
|
smartFormat: true,
|
|
36
37
|
noDelay: true,
|
|
37
38
|
endpointing: 25,
|
|
@@ -97,15 +98,21 @@ class SpeechStream extends import_agents.stt.SpeechStream {
|
|
|
97
98
|
#logger = (0, import_agents.log)();
|
|
98
99
|
#speaking = false;
|
|
99
100
|
#resetWS = new import_agents.Future();
|
|
101
|
+
#requestId = "";
|
|
102
|
+
#audioDurationCollector;
|
|
100
103
|
label = "deepgram.SpeechStream";
|
|
101
104
|
constructor(stt2, opts) {
|
|
102
|
-
super(stt2);
|
|
105
|
+
super(stt2, opts.sampleRate);
|
|
103
106
|
this.#opts = opts;
|
|
104
107
|
this.closed = false;
|
|
105
108
|
this.#audioEnergyFilter = new import_agents.AudioEnergyFilter();
|
|
106
|
-
this.#
|
|
109
|
+
this.#audioDurationCollector = new import_utils.PeriodicCollector(
|
|
110
|
+
(duration) => this.onAudioDurationReport(duration),
|
|
111
|
+
{ duration: 5 }
|
|
112
|
+
);
|
|
107
113
|
}
|
|
108
|
-
async
|
|
114
|
+
async run() {
|
|
115
|
+
const maxRetry = 32;
|
|
109
116
|
let retries = 0;
|
|
110
117
|
let ws;
|
|
111
118
|
while (!this.input.closed) {
|
|
@@ -189,6 +196,7 @@ class SpeechStream extends import_agents.stt.SpeechStream {
|
|
|
189
196
|
let frames;
|
|
190
197
|
if (data === SpeechStream.FLUSH_SENTINEL) {
|
|
191
198
|
frames = stream.flush();
|
|
199
|
+
this.#audioDurationCollector.flush();
|
|
192
200
|
} else if (data.sampleRate === this.#opts.sampleRate || data.channels === this.#opts.numChannels) {
|
|
193
201
|
frames = stream.write(data.data.buffer);
|
|
194
202
|
} else {
|
|
@@ -196,6 +204,8 @@ class SpeechStream extends import_agents.stt.SpeechStream {
|
|
|
196
204
|
}
|
|
197
205
|
for await (const frame of frames) {
|
|
198
206
|
if (this.#audioEnergyFilter.pushFrame(frame)) {
|
|
207
|
+
const frameDuration = frame.samplesPerChannel / frame.sampleRate;
|
|
208
|
+
this.#audioDurationCollector.push(frameDuration);
|
|
199
209
|
ws.send(frame.data.buffer);
|
|
200
210
|
}
|
|
201
211
|
}
|
|
@@ -207,7 +217,7 @@ class SpeechStream extends import_agents.stt.SpeechStream {
|
|
|
207
217
|
(_, reject) => ws.once("close", (code, reason) => {
|
|
208
218
|
if (!closing) {
|
|
209
219
|
this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);
|
|
210
|
-
reject();
|
|
220
|
+
reject(new Error("WebSocket closed"));
|
|
211
221
|
}
|
|
212
222
|
})
|
|
213
223
|
);
|
|
@@ -229,8 +239,11 @@ class SpeechStream extends import_agents.stt.SpeechStream {
|
|
|
229
239
|
// https://developers.deepgram.com/docs/understand-endpointing-interim-results#using-endpointing-speech_final
|
|
230
240
|
// for more information about the different types of events
|
|
231
241
|
case "Results": {
|
|
242
|
+
const metadata = json["metadata"];
|
|
243
|
+
const requestId = metadata["request_id"];
|
|
232
244
|
const isFinal = json["is_final"];
|
|
233
245
|
const isEndpoint = json["speech_final"];
|
|
246
|
+
this.#requestId = requestId;
|
|
234
247
|
const alternatives = liveTranscriptionToSpeechData(this.#opts.language, json);
|
|
235
248
|
if (alternatives[0] && alternatives[0].text) {
|
|
236
249
|
if (!this.#speaking) {
|
|
@@ -275,6 +288,16 @@ class SpeechStream extends import_agents.stt.SpeechStream {
|
|
|
275
288
|
ws.close();
|
|
276
289
|
clearInterval(keepalive);
|
|
277
290
|
}
|
|
291
|
+
onAudioDurationReport(duration) {
|
|
292
|
+
const usageEvent = {
|
|
293
|
+
type: import_agents.stt.SpeechEventType.RECOGNITION_USAGE,
|
|
294
|
+
requestId: this.#requestId,
|
|
295
|
+
recognitionUsage: {
|
|
296
|
+
audioDuration: duration
|
|
297
|
+
}
|
|
298
|
+
};
|
|
299
|
+
this.queue.put(usageEvent);
|
|
300
|
+
}
|
|
278
301
|
}
|
|
279
302
|
const liveTranscriptionToSpeechData = (language, data) => {
|
|
280
303
|
const alts = data["channel"]["alternatives"];
|
package/dist/stt.cjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type AudioBuffer,\n AudioByteStream,\n AudioEnergyFilter,\n Future,\n log,\n stt,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { type RawData, WebSocket } from 'ws';\nimport type { STTLanguages, STTModels } from './models.js';\n\nconst API_BASE_URL_V1 = 'wss://api.deepgram.com/v1/listen';\n\nexport interface STTOptions {\n apiKey?: string;\n language?: STTLanguages | string;\n detectLanguage: boolean;\n interimResults: boolean;\n punctuate: boolean;\n model: STTModels;\n smartFormat: boolean;\n noDelay: boolean;\n endpointing: number;\n fillerWords: boolean;\n sampleRate: number;\n numChannels: number;\n keywords: [string, number][];\n keyterm: string[];\n profanityFilter: boolean;\n dictation: boolean;\n diarize: boolean;\n numerals: boolean;\n}\n\nconst defaultSTTOptions: STTOptions = {\n apiKey: process.env.DEEPGRAM_API_KEY,\n language: 'en-US',\n detectLanguage: false,\n interimResults: true,\n punctuate: true,\n model: 'nova-2-general',\n smartFormat: true,\n noDelay: true,\n endpointing: 25,\n fillerWords: false,\n sampleRate: 16000,\n numChannels: 1,\n keywords: [],\n keyterm: [],\n profanityFilter: false,\n dictation: false,\n diarize: false,\n numerals: false,\n};\n\nexport class STT extends stt.STT {\n #opts: STTOptions;\n #logger = log();\n label = 'deepgram.STT';\n\n constructor(opts: Partial<STTOptions> = defaultSTTOptions) {\n super({\n streaming: true,\n interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,\n });\n if (opts.apiKey === undefined && defaultSTTOptions.apiKey === undefined) {\n throw new Error(\n 'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n );\n }\n\n this.#opts = { ...defaultSTTOptions, ...opts };\n\n if (this.#opts.detectLanguage) {\n this.#opts.language = undefined;\n } else if (\n this.#opts.language &&\n !['en-US', 'en'].includes(this.#opts.language) &&\n [\n 'nova-2-meeting',\n 'nova-2-phonecall',\n 'nova-2-finance',\n 'nova-2-conversationalai',\n 'nova-2-voicemail',\n 'nova-2-video',\n 'nova-2-medical',\n 'nova-2-drivethru',\n 'nova-2-automotive',\n 'nova-3-general',\n ].includes(this.#opts.model)\n ) {\n this.#logger.warn(\n `${this.#opts.model} does not support language ${this.#opts.language}, falling back to nova-2-general`,\n );\n this.#opts.model = 'nova-2-general';\n }\n }\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n async _recognize(_: AudioBuffer): Promise<stt.SpeechEvent> {\n throw new Error('Recognize is not supported on Deepgram STT');\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n }\n\n stream(): SpeechStream {\n return new SpeechStream(this, this.#opts);\n }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n #opts: STTOptions;\n #audioEnergyFilter: AudioEnergyFilter;\n #logger = log();\n #speaking = false;\n #resetWS = new Future();\n label = 'deepgram.SpeechStream';\n\n constructor(stt: STT, opts: STTOptions) {\n super(stt);\n this.#opts = opts;\n this.closed = false;\n this.#audioEnergyFilter = new AudioEnergyFilter();\n\n this.#run();\n }\n\n async #run(maxRetry = 32) {\n let retries = 0;\n let ws: WebSocket;\n while (!this.input.closed) {\n const streamURL = new URL(API_BASE_URL_V1);\n const params = {\n model: this.#opts.model,\n punctuate: this.#opts.punctuate,\n smart_format: this.#opts.smartFormat,\n dictation: this.#opts.dictation,\n diarize: this.#opts.diarize,\n numerals: this.#opts.numerals,\n no_delay: this.#opts.noDelay,\n interim_results: this.#opts.interimResults,\n encoding: 'linear16',\n vad_events: true,\n sample_rate: this.#opts.sampleRate,\n channels: this.#opts.numChannels,\n endpointing: this.#opts.endpointing || false,\n filler_words: this.#opts.fillerWords,\n keywords: this.#opts.keywords.map((x) => x.join(':')),\n keyterm: this.#opts.keyterm,\n profanity_filter: this.#opts.profanityFilter,\n language: this.#opts.language,\n };\n Object.entries(params).forEach(([k, v]) => {\n if (v !== undefined) {\n if (typeof v === 'string' || typeof v === 'number' || typeof v === 'boolean') {\n streamURL.searchParams.append(k, encodeURIComponent(v));\n } else {\n v.forEach((x) => streamURL.searchParams.append(k, encodeURIComponent(x)));\n }\n }\n });\n\n ws = new WebSocket(streamURL, {\n headers: { Authorization: `Token ${this.#opts.apiKey}` },\n });\n\n try {\n await new Promise((resolve, reject) => {\n ws.on('open', resolve);\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n });\n\n await this.#runWS(ws);\n } catch (e) {\n if (retries >= maxRetry) {\n throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);\n }\n\n const delay = Math.min(retries * 5, 10);\n retries++;\n\n this.#logger.warn(\n `failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`,\n );\n await new Promise((resolve) => setTimeout(resolve, delay * 1000));\n }\n }\n\n this.closed = true;\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n this.#resetWS.resolve();\n }\n\n async #runWS(ws: WebSocket) {\n this.#resetWS = new Future();\n let closing = false;\n\n const keepalive = setInterval(() => {\n try {\n ws.send(JSON.stringify({ type: 'KeepAlive' }));\n } catch {\n clearInterval(keepalive);\n return;\n }\n }, 5000);\n\n const sendTask = async () => {\n const samples100Ms = Math.floor(this.#opts.sampleRate / 10);\n const stream = new AudioByteStream(\n this.#opts.sampleRate,\n this.#opts.numChannels,\n samples100Ms,\n );\n\n for await (const data of this.input) {\n let frames: AudioFrame[];\n if (data === SpeechStream.FLUSH_SENTINEL) {\n frames = stream.flush();\n } else if (\n data.sampleRate === this.#opts.sampleRate ||\n data.channels === this.#opts.numChannels\n ) {\n frames = stream.write(data.data.buffer);\n } else {\n throw new Error(`sample rate or channel count of frame does not match`);\n }\n\n for await (const frame of frames) {\n if (this.#audioEnergyFilter.pushFrame(frame)) {\n ws.send(frame.data.buffer);\n }\n }\n }\n\n closing = true;\n ws.send(JSON.stringify({ type: 'CloseStream' }));\n };\n\n const wsMonitor = 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();\n }\n }),\n );\n\n const listenTask = async () => {\n while (!this.closed && !closing) {\n try {\n await new Promise<RawData>((resolve) => {\n ws.once('message', (data) => resolve(data));\n }).then((msg) => {\n const json = JSON.parse(msg.toString());\n switch (json['type']) {\n case 'SpeechStarted': {\n // This is a normal case. Deepgram's SpeechStarted events\n // are not correlated with speech_final or utterance end.\n // It's possible that we receive two in a row without an endpoint\n // It's also possible we receive a transcript without a SpeechStarted event.\n if (this.#speaking) return;\n this.#speaking = true;\n this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n break;\n }\n // see this page:\n // https://developers.deepgram.com/docs/understand-endpointing-interim-results#using-endpointing-speech_final\n // for more information about the different types of events\n case 'Results': {\n const isFinal = json['is_final'];\n const isEndpoint = json['speech_final'];\n\n const alternatives = liveTranscriptionToSpeechData(this.#opts.language!, json);\n\n // If, for some reason, we didn't get a SpeechStarted event but we got\n // a transcript with text, we should start speaking. It's rare but has\n // been observed.\n if (alternatives[0] && alternatives[0].text) {\n if (!this.#speaking) {\n this.#speaking = true;\n this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n }\n\n if (isFinal) {\n this.queue.put({\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n alternatives: [alternatives[0], ...alternatives.slice(1)],\n });\n } else {\n this.queue.put({\n type: stt.SpeechEventType.INTERIM_TRANSCRIPT,\n alternatives: [alternatives[0], ...alternatives.slice(1)],\n });\n }\n }\n\n // if we receive an endpoint, only end the speech if\n // we either had a SpeechStarted event or we have a seen\n // a non-empty transcript (deepgram doesn't have a SpeechEnded event)\n if (isEndpoint && this.#speaking) {\n this.#speaking = false;\n this.queue.put({ type: stt.SpeechEventType.END_OF_SPEECH });\n }\n\n break;\n }\n case 'Metadata': {\n break;\n }\n default: {\n this.#logger.child({ msg: json }).warn('received unexpected message from Deepgram');\n break;\n }\n }\n });\n } catch (error) {\n this.#logger.child({ error }).warn('unrecoverable error, exiting');\n break;\n }\n }\n };\n\n await Promise.race([this.#resetWS.await, Promise.all([sendTask(), listenTask(), wsMonitor])]);\n closing = true;\n ws.close();\n clearInterval(keepalive);\n }\n}\n\nconst liveTranscriptionToSpeechData = (\n language: STTLanguages | string,\n data: { [id: string]: any },\n): stt.SpeechData[] => {\n const alts: any[] = data['channel']['alternatives'];\n\n return alts.map((alt) => ({\n language,\n startTime: alt['words'].length ? alt['words'][0]['start'] : 0,\n endTime: alt['words'].length ? alt['words'][alt['words'].length - 1]['end'] : 0,\n confidence: alt['confidence'],\n text: alt['transcript'],\n }));\n};\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBAOO;AAEP,gBAAwC;AAGxC,MAAM,kBAAkB;AAuBxB,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,WAAW;AAAA,EACX,OAAO;AAAA,EACP,aAAa;AAAA,EACb,SAAS;AAAA,EACT,aAAa;AAAA,EACb,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,UAAU,CAAC;AAAA,EACX,SAAS,CAAC;AAAA,EACV,iBAAiB;AAAA,EACjB,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AACZ;AAEO,MAAM,YAAY,kBAAI,IAAI;AAAA,EAC/B;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,QAAQ;AAAA,EAER,YAAY,OAA4B,mBAAmB;AACzD,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB,KAAK,kBAAkB,kBAAkB;AAAA,IAC3D,CAAC;AACD,QAAI,KAAK,WAAW,UAAa,kBAAkB,WAAW,QAAW;AACvE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ,EAAE,GAAG,mBAAmB,GAAG,KAAK;AAE7C,QAAI,KAAK,MAAM,gBAAgB;AAC7B,WAAK,MAAM,WAAW;AAAA,IACxB,WACE,KAAK,MAAM,YACX,CAAC,CAAC,SAAS,IAAI,EAAE,SAAS,KAAK,MAAM,QAAQ,KAC7C;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,EAAE,SAAS,KAAK,MAAM,KAAK,GAC3B;AACA,WAAK,QAAQ;AAAA,QACX,GAAG,KAAK,MAAM,KAAK,8BAA8B,KAAK,MAAM,QAAQ;AAAA,MACtE;AACA,WAAK,MAAM,QAAQ;AAAA,IACrB;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,WAAW,GAA0C;AACzD,UAAM,IAAI,MAAM,4CAA4C;AAAA,EAC9D;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAAA,EACxC;AAAA,EAEA,SAAuB;AACrB,WAAO,IAAI,aAAa,MAAM,KAAK,KAAK;AAAA,EAC1C;AACF;AAEO,MAAM,qBAAqB,kBAAI,aAAa;AAAA,EACjD;AAAA,EACA;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,YAAY;AAAA,EACZ,WAAW,IAAI,qBAAO;AAAA,EACtB,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB;AACtC,UAAMA,IAAG;AACT,SAAK,QAAQ;AACb,SAAK,SAAS;AACd,SAAK,qBAAqB,IAAI,gCAAkB;AAEhD,SAAK,KAAK;AAAA,EACZ;AAAA,EAEA,MAAM,KAAK,WAAW,IAAI;AACxB,QAAI,UAAU;AACd,QAAI;AACJ,WAAO,CAAC,KAAK,MAAM,QAAQ;AACzB,YAAM,YAAY,IAAI,IAAI,eAAe;AACzC,YAAM,SAAS;AAAA,QACb,OAAO,KAAK,MAAM;AAAA,QAClB,WAAW,KAAK,MAAM;AAAA,QACtB,cAAc,KAAK,MAAM;AAAA,QACzB,WAAW,KAAK,MAAM;AAAA,QACtB,SAAS,KAAK,MAAM;AAAA,QACpB,UAAU,KAAK,MAAM;AAAA,QACrB,UAAU,KAAK,MAAM;AAAA,QACrB,iBAAiB,KAAK,MAAM;AAAA,QAC5B,UAAU;AAAA,QACV,YAAY;AAAA,QACZ,aAAa,KAAK,MAAM;AAAA,QACxB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM,eAAe;AAAA,QACvC,cAAc,KAAK,MAAM;AAAA,QACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,QACpD,SAAS,KAAK,MAAM;AAAA,QACpB,kBAAkB,KAAK,MAAM;AAAA,QAC7B,UAAU,KAAK,MAAM;AAAA,MACvB;AACA,aAAO,QAAQ,MAAM,EAAE,QAAQ,CAAC,CAAC,GAAG,CAAC,MAAM;AACzC,YAAI,MAAM,QAAW;AACnB,cAAI,OAAO,MAAM,YAAY,OAAO,MAAM,YAAY,OAAO,MAAM,WAAW;AAC5E,sBAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC;AAAA,UACxD,OAAO;AACL,cAAE,QAAQ,CAAC,MAAM,UAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC,CAAC;AAAA,UAC1E;AAAA,QACF;AAAA,MACF,CAAC;AAED,WAAK,IAAI,oBAAU,WAAW;AAAA,QAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,MACzD,CAAC;AAED,UAAI;AACF,cAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,aAAG,GAAG,QAAQ,OAAO;AACrB,aAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,aAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,QAC/D,CAAC;AAED,cAAM,KAAK,OAAO,EAAE;AAAA,MACtB,SAAS,GAAG;AACV,YAAI,WAAW,UAAU;AACvB,gBAAM,IAAI,MAAM,uCAAuC,OAAO,cAAc,CAAC,EAAE;AAAA,QACjF;AAEA,cAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,EAAE;AACtC;AAEA,aAAK,QAAQ;AAAA,UACX,8CAA8C,KAAK,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ;AAAA,QAC3F;AACA,cAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,QAAQ,GAAI,CAAC;AAAA,MAClE;AAAA,IACF;AAEA,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,SAAK,SAAS,QAAQ;AAAA,EACxB;AAAA,EAEA,MAAM,OAAO,IAAe;AAC1B,SAAK,WAAW,IAAI,qBAAO;AAC3B,QAAI,UAAU;AAEd,UAAM,YAAY,YAAY,MAAM;AAClC,UAAI;AACF,WAAG,KAAK,KAAK,UAAU,EAAE,MAAM,YAAY,CAAC,CAAC;AAAA,MAC/C,QAAQ;AACN,sBAAc,SAAS;AACvB;AAAA,MACF;AAAA,IACF,GAAG,GAAI;AAEP,UAAM,WAAW,YAAY;AAC3B,YAAM,eAAe,KAAK,MAAM,KAAK,MAAM,aAAa,EAAE;AAC1D,YAAM,SAAS,IAAI;AAAA,QACjB,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX;AAAA,MACF;AAEA,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI;AACJ,YAAI,SAAS,aAAa,gBAAgB;AACxC,mBAAS,OAAO,MAAM;AAAA,QACxB,WACE,KAAK,eAAe,KAAK,MAAM,cAC/B,KAAK,aAAa,KAAK,MAAM,aAC7B;AACA,mBAAS,OAAO,MAAM,KAAK,KAAK,MAAM;AAAA,QACxC,OAAO;AACL,gBAAM,IAAI,MAAM,sDAAsD;AAAA,QACxE;AAEA,yBAAiB,SAAS,QAAQ;AAChC,cAAI,KAAK,mBAAmB,UAAU,KAAK,GAAG;AAC5C,eAAG,KAAK,MAAM,KAAK,MAAM;AAAA,UAC3B;AAAA,QACF;AAAA,MACF;AAEA,gBAAU;AACV,SAAG,KAAK,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC,CAAC;AAAA,IACjD;AAEA,UAAM,YAAY,IAAI;AAAA,MAAc,CAAC,GAAG,WACtC,GAAG,KAAK,SAAS,CAAC,MAAM,WAAW;AACjC,YAAI,CAAC,SAAS;AACZ,eAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,MAAM,EAAE;AAClE,iBAAO;AAAA,QACT;AAAA,MACF,CAAC;AAAA,IACH;AAEA,UAAM,aAAa,YAAY;AAC7B,aAAO,CAAC,KAAK,UAAU,CAAC,SAAS;AAC/B,YAAI;AACF,gBAAM,IAAI,QAAiB,CAAC,YAAY;AACtC,eAAG,KAAK,WAAW,CAAC,SAAS,QAAQ,IAAI,CAAC;AAAA,UAC5C,CAAC,EAAE,KAAK,CAAC,QAAQ;AACf,kBAAM,OAAO,KAAK,MAAM,IAAI,SAAS,CAAC;AACtC,oBAAQ,KAAK,MAAM,GAAG;AAAA,cACpB,KAAK,iBAAiB;AAKpB,oBAAI,KAAK,UAAW;AACpB,qBAAK,YAAY;AACjB,qBAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,gBAAgB,CAAC;AAC5D;AAAA,cACF;AAAA;AAAA;AAAA;AAAA,cAIA,KAAK,WAAW;AACd,sBAAM,UAAU,KAAK,UAAU;AAC/B,sBAAM,aAAa,KAAK,cAAc;AAEtC,sBAAM,eAAe,8BAA8B,KAAK,MAAM,UAAW,IAAI;AAK7E,oBAAI,aAAa,CAAC,KAAK,aAAa,CAAC,EAAE,MAAM;AAC3C,sBAAI,CAAC,KAAK,WAAW;AACnB,yBAAK,YAAY;AACjB,yBAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,gBAAgB,CAAC;AAAA,kBAC9D;AAEA,sBAAI,SAAS;AACX,yBAAK,MAAM,IAAI;AAAA,sBACb,MAAM,kBAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH,OAAO;AACL,yBAAK,MAAM,IAAI;AAAA,sBACb,MAAM,kBAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH;AAAA,gBACF;AAKA,oBAAI,cAAc,KAAK,WAAW;AAChC,uBAAK,YAAY;AACjB,uBAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,cAAc,CAAC;AAAA,gBAC5D;AAEA;AAAA,cACF;AAAA,cACA,KAAK,YAAY;AACf;AAAA,cACF;AAAA,cACA,SAAS;AACP,qBAAK,QAAQ,MAAM,EAAE,KAAK,KAAK,CAAC,EAAE,KAAK,2CAA2C;AAClF;AAAA,cACF;AAAA,YACF;AAAA,UACF,CAAC;AAAA,QACH,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,EAAE,MAAM,CAAC,EAAE,KAAK,8BAA8B;AACjE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,QAAQ,KAAK,CAAC,KAAK,SAAS,OAAO,QAAQ,IAAI,CAAC,SAAS,GAAG,WAAW,GAAG,SAAS,CAAC,CAAC,CAAC;AAC5F,cAAU;AACV,OAAG,MAAM;AACT,kBAAc,SAAS;AAAA,EACzB;AACF;AAEA,MAAM,gCAAgC,CACpC,UACA,SACqB;AACrB,QAAM,OAAc,KAAK,SAAS,EAAE,cAAc;AAElD,SAAO,KAAK,IAAI,CAAC,SAAS;AAAA,IACxB;AAAA,IACA,WAAW,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,CAAC,EAAE,OAAO,IAAI;AAAA,IAC5D,SAAS,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,IAAI,OAAO,EAAE,SAAS,CAAC,EAAE,KAAK,IAAI;AAAA,IAC9E,YAAY,IAAI,YAAY;AAAA,IAC5B,MAAM,IAAI,YAAY;AAAA,EACxB,EAAE;AACJ;","names":["stt"]}
|
|
1
|
+
{"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type AudioBuffer,\n AudioByteStream,\n AudioEnergyFilter,\n Future,\n log,\n stt,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { type RawData, WebSocket } from 'ws';\nimport { PeriodicCollector } from './_utils.js';\nimport type { STTLanguages, STTModels } from './models.js';\n\nconst API_BASE_URL_V1 = 'wss://api.deepgram.com/v1/listen';\n\nexport interface STTOptions {\n apiKey?: string;\n language?: STTLanguages | string;\n detectLanguage: boolean;\n interimResults: boolean;\n punctuate: boolean;\n model: STTModels;\n smartFormat: boolean;\n noDelay: boolean;\n endpointing: number;\n fillerWords: boolean;\n sampleRate: number;\n numChannels: number;\n keywords: [string, number][];\n keyterm: string[];\n profanityFilter: boolean;\n dictation: boolean;\n diarize: boolean;\n numerals: boolean;\n}\n\nconst defaultSTTOptions: STTOptions = {\n apiKey: process.env.DEEPGRAM_API_KEY,\n language: 'en-US',\n detectLanguage: false,\n interimResults: true,\n punctuate: true,\n model: 'nova-3',\n smartFormat: true,\n noDelay: true,\n endpointing: 25,\n fillerWords: false,\n sampleRate: 16000,\n numChannels: 1,\n keywords: [],\n keyterm: [],\n profanityFilter: false,\n dictation: false,\n diarize: false,\n numerals: false,\n};\n\nexport class STT extends stt.STT {\n #opts: STTOptions;\n #logger = log();\n label = 'deepgram.STT';\n\n constructor(opts: Partial<STTOptions> = defaultSTTOptions) {\n super({\n streaming: true,\n interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,\n });\n if (opts.apiKey === undefined && defaultSTTOptions.apiKey === undefined) {\n throw new Error(\n 'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n );\n }\n\n this.#opts = { ...defaultSTTOptions, ...opts };\n\n if (this.#opts.detectLanguage) {\n this.#opts.language = undefined;\n } else if (\n this.#opts.language &&\n !['en-US', 'en'].includes(this.#opts.language) &&\n [\n 'nova-2-meeting',\n 'nova-2-phonecall',\n 'nova-2-finance',\n 'nova-2-conversationalai',\n 'nova-2-voicemail',\n 'nova-2-video',\n 'nova-2-medical',\n 'nova-2-drivethru',\n 'nova-2-automotive',\n 'nova-3-general',\n ].includes(this.#opts.model)\n ) {\n this.#logger.warn(\n `${this.#opts.model} does not support language ${this.#opts.language}, falling back to nova-2-general`,\n );\n this.#opts.model = 'nova-2-general';\n }\n }\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n async _recognize(_: AudioBuffer): Promise<stt.SpeechEvent> {\n throw new Error('Recognize is not supported on Deepgram STT');\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n }\n\n stream(): SpeechStream {\n return new SpeechStream(this, this.#opts);\n }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n #opts: STTOptions;\n #audioEnergyFilter: AudioEnergyFilter;\n #logger = log();\n #speaking = false;\n #resetWS = new Future();\n #requestId = '';\n #audioDurationCollector: PeriodicCollector<number>;\n label = 'deepgram.SpeechStream';\n\n constructor(stt: STT, opts: STTOptions) {\n super(stt, opts.sampleRate);\n this.#opts = opts;\n this.closed = false;\n this.#audioEnergyFilter = new AudioEnergyFilter();\n this.#audioDurationCollector = new PeriodicCollector(\n (duration) => this.onAudioDurationReport(duration),\n { duration: 5.0 },\n );\n }\n\n protected async run() {\n const maxRetry = 32;\n let retries = 0;\n let ws: WebSocket;\n while (!this.input.closed) {\n const streamURL = new URL(API_BASE_URL_V1);\n const params = {\n model: this.#opts.model,\n punctuate: this.#opts.punctuate,\n smart_format: this.#opts.smartFormat,\n dictation: this.#opts.dictation,\n diarize: this.#opts.diarize,\n numerals: this.#opts.numerals,\n no_delay: this.#opts.noDelay,\n interim_results: this.#opts.interimResults,\n encoding: 'linear16',\n vad_events: true,\n sample_rate: this.#opts.sampleRate,\n channels: this.#opts.numChannels,\n endpointing: this.#opts.endpointing || false,\n filler_words: this.#opts.fillerWords,\n keywords: this.#opts.keywords.map((x) => x.join(':')),\n keyterm: this.#opts.keyterm,\n profanity_filter: this.#opts.profanityFilter,\n language: this.#opts.language,\n };\n Object.entries(params).forEach(([k, v]) => {\n if (v !== undefined) {\n if (typeof v === 'string' || typeof v === 'number' || typeof v === 'boolean') {\n streamURL.searchParams.append(k, encodeURIComponent(v));\n } else {\n v.forEach((x) => streamURL.searchParams.append(k, encodeURIComponent(x)));\n }\n }\n });\n\n ws = new WebSocket(streamURL, {\n headers: { Authorization: `Token ${this.#opts.apiKey}` },\n });\n\n try {\n await new Promise((resolve, reject) => {\n ws.on('open', resolve);\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n });\n\n await this.#runWS(ws);\n } catch (e) {\n if (retries >= maxRetry) {\n throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);\n }\n\n const delay = Math.min(retries * 5, 10);\n retries++;\n\n this.#logger.warn(\n `failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`,\n );\n await new Promise((resolve) => setTimeout(resolve, delay * 1000));\n }\n }\n\n this.closed = true;\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n this.#resetWS.resolve();\n }\n\n async #runWS(ws: WebSocket) {\n this.#resetWS = new Future();\n let closing = false;\n\n const keepalive = setInterval(() => {\n try {\n ws.send(JSON.stringify({ type: 'KeepAlive' }));\n } catch {\n clearInterval(keepalive);\n return;\n }\n }, 5000);\n\n const sendTask = async () => {\n const samples100Ms = Math.floor(this.#opts.sampleRate / 10);\n const stream = new AudioByteStream(\n this.#opts.sampleRate,\n this.#opts.numChannels,\n samples100Ms,\n );\n\n for await (const data of this.input) {\n let frames: AudioFrame[];\n if (data === SpeechStream.FLUSH_SENTINEL) {\n frames = stream.flush();\n this.#audioDurationCollector.flush();\n } else if (\n data.sampleRate === this.#opts.sampleRate ||\n data.channels === this.#opts.numChannels\n ) {\n frames = stream.write(data.data.buffer);\n } else {\n throw new Error(`sample rate or channel count of frame does not match`);\n }\n\n for await (const frame of frames) {\n if (this.#audioEnergyFilter.pushFrame(frame)) {\n const frameDuration = frame.samplesPerChannel / frame.sampleRate;\n this.#audioDurationCollector.push(frameDuration);\n ws.send(frame.data.buffer);\n }\n }\n }\n\n closing = true;\n ws.send(JSON.stringify({ type: 'CloseStream' }));\n };\n\n const wsMonitor = 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 const listenTask = async () => {\n while (!this.closed && !closing) {\n try {\n await new Promise<RawData>((resolve) => {\n ws.once('message', (data) => resolve(data));\n }).then((msg) => {\n const json = JSON.parse(msg.toString());\n switch (json['type']) {\n case 'SpeechStarted': {\n // This is a normal case. Deepgram's SpeechStarted events\n // are not correlated with speech_final or utterance end.\n // It's possible that we receive two in a row without an endpoint\n // It's also possible we receive a transcript without a SpeechStarted event.\n if (this.#speaking) return;\n this.#speaking = true;\n this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n break;\n }\n // see this page:\n // https://developers.deepgram.com/docs/understand-endpointing-interim-results#using-endpointing-speech_final\n // for more information about the different types of events\n case 'Results': {\n const metadata = json['metadata'];\n const requestId = metadata['request_id'];\n const isFinal = json['is_final'];\n const isEndpoint = json['speech_final'];\n this.#requestId = requestId;\n\n const alternatives = liveTranscriptionToSpeechData(this.#opts.language!, json);\n\n // If, for some reason, we didn't get a SpeechStarted event but we got\n // a transcript with text, we should start speaking. It's rare but has\n // been observed.\n if (alternatives[0] && alternatives[0].text) {\n if (!this.#speaking) {\n this.#speaking = true;\n this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n }\n\n if (isFinal) {\n this.queue.put({\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n alternatives: [alternatives[0], ...alternatives.slice(1)],\n });\n } else {\n this.queue.put({\n type: stt.SpeechEventType.INTERIM_TRANSCRIPT,\n alternatives: [alternatives[0], ...alternatives.slice(1)],\n });\n }\n }\n\n // if we receive an endpoint, only end the speech if\n // we either had a SpeechStarted event or we have a seen\n // a non-empty transcript (deepgram doesn't have a SpeechEnded event)\n if (isEndpoint && this.#speaking) {\n this.#speaking = false;\n this.queue.put({ type: stt.SpeechEventType.END_OF_SPEECH });\n }\n\n break;\n }\n case 'Metadata': {\n break;\n }\n default: {\n this.#logger.child({ msg: json }).warn('received unexpected message from Deepgram');\n break;\n }\n }\n });\n } catch (error) {\n this.#logger.child({ error }).warn('unrecoverable error, exiting');\n break;\n }\n }\n };\n\n await Promise.race([this.#resetWS.await, Promise.all([sendTask(), listenTask(), wsMonitor])]);\n closing = true;\n ws.close();\n clearInterval(keepalive);\n }\n\n private onAudioDurationReport(duration: number) {\n const usageEvent: stt.SpeechEvent = {\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: {\n audioDuration: duration,\n },\n };\n this.queue.put(usageEvent);\n }\n}\n\nconst liveTranscriptionToSpeechData = (\n language: STTLanguages | string,\n data: { [id: string]: any },\n): stt.SpeechData[] => {\n const alts: any[] = data['channel']['alternatives'];\n\n return alts.map((alt) => ({\n language,\n startTime: alt['words'].length ? alt['words'][0]['start'] : 0,\n endTime: alt['words'].length ? alt['words'][alt['words'].length - 1]['end'] : 0,\n confidence: alt['confidence'],\n text: alt['transcript'],\n }));\n};\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBAOO;AAEP,gBAAwC;AACxC,mBAAkC;AAGlC,MAAM,kBAAkB;AAuBxB,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,WAAW;AAAA,EACX,OAAO;AAAA,EACP,aAAa;AAAA,EACb,SAAS;AAAA,EACT,aAAa;AAAA,EACb,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,UAAU,CAAC;AAAA,EACX,SAAS,CAAC;AAAA,EACV,iBAAiB;AAAA,EACjB,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AACZ;AAEO,MAAM,YAAY,kBAAI,IAAI;AAAA,EAC/B;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,QAAQ;AAAA,EAER,YAAY,OAA4B,mBAAmB;AACzD,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB,KAAK,kBAAkB,kBAAkB;AAAA,IAC3D,CAAC;AACD,QAAI,KAAK,WAAW,UAAa,kBAAkB,WAAW,QAAW;AACvE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ,EAAE,GAAG,mBAAmB,GAAG,KAAK;AAE7C,QAAI,KAAK,MAAM,gBAAgB;AAC7B,WAAK,MAAM,WAAW;AAAA,IACxB,WACE,KAAK,MAAM,YACX,CAAC,CAAC,SAAS,IAAI,EAAE,SAAS,KAAK,MAAM,QAAQ,KAC7C;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,EAAE,SAAS,KAAK,MAAM,KAAK,GAC3B;AACA,WAAK,QAAQ;AAAA,QACX,GAAG,KAAK,MAAM,KAAK,8BAA8B,KAAK,MAAM,QAAQ;AAAA,MACtE;AACA,WAAK,MAAM,QAAQ;AAAA,IACrB;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,WAAW,GAA0C;AACzD,UAAM,IAAI,MAAM,4CAA4C;AAAA,EAC9D;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAAA,EACxC;AAAA,EAEA,SAAuB;AACrB,WAAO,IAAI,aAAa,MAAM,KAAK,KAAK;AAAA,EAC1C;AACF;AAEO,MAAM,qBAAqB,kBAAI,aAAa;AAAA,EACjD;AAAA,EACA;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,YAAY;AAAA,EACZ,WAAW,IAAI,qBAAO;AAAA,EACtB,aAAa;AAAA,EACb;AAAA,EACA,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB;AACtC,UAAMA,MAAK,KAAK,UAAU;AAC1B,SAAK,QAAQ;AACb,SAAK,SAAS;AACd,SAAK,qBAAqB,IAAI,gCAAkB;AAChD,SAAK,0BAA0B,IAAI;AAAA,MACjC,CAAC,aAAa,KAAK,sBAAsB,QAAQ;AAAA,MACjD,EAAE,UAAU,EAAI;AAAA,IAClB;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,WAAW;AACjB,QAAI,UAAU;AACd,QAAI;AACJ,WAAO,CAAC,KAAK,MAAM,QAAQ;AACzB,YAAM,YAAY,IAAI,IAAI,eAAe;AACzC,YAAM,SAAS;AAAA,QACb,OAAO,KAAK,MAAM;AAAA,QAClB,WAAW,KAAK,MAAM;AAAA,QACtB,cAAc,KAAK,MAAM;AAAA,QACzB,WAAW,KAAK,MAAM;AAAA,QACtB,SAAS,KAAK,MAAM;AAAA,QACpB,UAAU,KAAK,MAAM;AAAA,QACrB,UAAU,KAAK,MAAM;AAAA,QACrB,iBAAiB,KAAK,MAAM;AAAA,QAC5B,UAAU;AAAA,QACV,YAAY;AAAA,QACZ,aAAa,KAAK,MAAM;AAAA,QACxB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM,eAAe;AAAA,QACvC,cAAc,KAAK,MAAM;AAAA,QACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,QACpD,SAAS,KAAK,MAAM;AAAA,QACpB,kBAAkB,KAAK,MAAM;AAAA,QAC7B,UAAU,KAAK,MAAM;AAAA,MACvB;AACA,aAAO,QAAQ,MAAM,EAAE,QAAQ,CAAC,CAAC,GAAG,CAAC,MAAM;AACzC,YAAI,MAAM,QAAW;AACnB,cAAI,OAAO,MAAM,YAAY,OAAO,MAAM,YAAY,OAAO,MAAM,WAAW;AAC5E,sBAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC;AAAA,UACxD,OAAO;AACL,cAAE,QAAQ,CAAC,MAAM,UAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC,CAAC;AAAA,UAC1E;AAAA,QACF;AAAA,MACF,CAAC;AAED,WAAK,IAAI,oBAAU,WAAW;AAAA,QAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,MACzD,CAAC;AAED,UAAI;AACF,cAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,aAAG,GAAG,QAAQ,OAAO;AACrB,aAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,aAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,QAC/D,CAAC;AAED,cAAM,KAAK,OAAO,EAAE;AAAA,MACtB,SAAS,GAAG;AACV,YAAI,WAAW,UAAU;AACvB,gBAAM,IAAI,MAAM,uCAAuC,OAAO,cAAc,CAAC,EAAE;AAAA,QACjF;AAEA,cAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,EAAE;AACtC;AAEA,aAAK,QAAQ;AAAA,UACX,8CAA8C,KAAK,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ;AAAA,QAC3F;AACA,cAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,QAAQ,GAAI,CAAC;AAAA,MAClE;AAAA,IACF;AAEA,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,SAAK,SAAS,QAAQ;AAAA,EACxB;AAAA,EAEA,MAAM,OAAO,IAAe;AAC1B,SAAK,WAAW,IAAI,qBAAO;AAC3B,QAAI,UAAU;AAEd,UAAM,YAAY,YAAY,MAAM;AAClC,UAAI;AACF,WAAG,KAAK,KAAK,UAAU,EAAE,MAAM,YAAY,CAAC,CAAC;AAAA,MAC/C,QAAQ;AACN,sBAAc,SAAS;AACvB;AAAA,MACF;AAAA,IACF,GAAG,GAAI;AAEP,UAAM,WAAW,YAAY;AAC3B,YAAM,eAAe,KAAK,MAAM,KAAK,MAAM,aAAa,EAAE;AAC1D,YAAM,SAAS,IAAI;AAAA,QACjB,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX;AAAA,MACF;AAEA,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI;AACJ,YAAI,SAAS,aAAa,gBAAgB;AACxC,mBAAS,OAAO,MAAM;AACtB,eAAK,wBAAwB,MAAM;AAAA,QACrC,WACE,KAAK,eAAe,KAAK,MAAM,cAC/B,KAAK,aAAa,KAAK,MAAM,aAC7B;AACA,mBAAS,OAAO,MAAM,KAAK,KAAK,MAAM;AAAA,QACxC,OAAO;AACL,gBAAM,IAAI,MAAM,sDAAsD;AAAA,QACxE;AAEA,yBAAiB,SAAS,QAAQ;AAChC,cAAI,KAAK,mBAAmB,UAAU,KAAK,GAAG;AAC5C,kBAAM,gBAAgB,MAAM,oBAAoB,MAAM;AACtD,iBAAK,wBAAwB,KAAK,aAAa;AAC/C,eAAG,KAAK,MAAM,KAAK,MAAM;AAAA,UAC3B;AAAA,QACF;AAAA,MACF;AAEA,gBAAU;AACV,SAAG,KAAK,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC,CAAC;AAAA,IACjD;AAEA,UAAM,YAAY,IAAI;AAAA,MAAc,CAAC,GAAG,WACtC,GAAG,KAAK,SAAS,CAAC,MAAM,WAAW;AACjC,YAAI,CAAC,SAAS;AACZ,eAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,MAAM,EAAE;AAClE,iBAAO,IAAI,MAAM,kBAAkB,CAAC;AAAA,QACtC;AAAA,MACF,CAAC;AAAA,IACH;AAEA,UAAM,aAAa,YAAY;AAC7B,aAAO,CAAC,KAAK,UAAU,CAAC,SAAS;AAC/B,YAAI;AACF,gBAAM,IAAI,QAAiB,CAAC,YAAY;AACtC,eAAG,KAAK,WAAW,CAAC,SAAS,QAAQ,IAAI,CAAC;AAAA,UAC5C,CAAC,EAAE,KAAK,CAAC,QAAQ;AACf,kBAAM,OAAO,KAAK,MAAM,IAAI,SAAS,CAAC;AACtC,oBAAQ,KAAK,MAAM,GAAG;AAAA,cACpB,KAAK,iBAAiB;AAKpB,oBAAI,KAAK,UAAW;AACpB,qBAAK,YAAY;AACjB,qBAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,gBAAgB,CAAC;AAC5D;AAAA,cACF;AAAA;AAAA;AAAA;AAAA,cAIA,KAAK,WAAW;AACd,sBAAM,WAAW,KAAK,UAAU;AAChC,sBAAM,YAAY,SAAS,YAAY;AACvC,sBAAM,UAAU,KAAK,UAAU;AAC/B,sBAAM,aAAa,KAAK,cAAc;AACtC,qBAAK,aAAa;AAElB,sBAAM,eAAe,8BAA8B,KAAK,MAAM,UAAW,IAAI;AAK7E,oBAAI,aAAa,CAAC,KAAK,aAAa,CAAC,EAAE,MAAM;AAC3C,sBAAI,CAAC,KAAK,WAAW;AACnB,yBAAK,YAAY;AACjB,yBAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,gBAAgB,CAAC;AAAA,kBAC9D;AAEA,sBAAI,SAAS;AACX,yBAAK,MAAM,IAAI;AAAA,sBACb,MAAM,kBAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH,OAAO;AACL,yBAAK,MAAM,IAAI;AAAA,sBACb,MAAM,kBAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH;AAAA,gBACF;AAKA,oBAAI,cAAc,KAAK,WAAW;AAChC,uBAAK,YAAY;AACjB,uBAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,cAAc,CAAC;AAAA,gBAC5D;AAEA;AAAA,cACF;AAAA,cACA,KAAK,YAAY;AACf;AAAA,cACF;AAAA,cACA,SAAS;AACP,qBAAK,QAAQ,MAAM,EAAE,KAAK,KAAK,CAAC,EAAE,KAAK,2CAA2C;AAClF;AAAA,cACF;AAAA,YACF;AAAA,UACF,CAAC;AAAA,QACH,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,EAAE,MAAM,CAAC,EAAE,KAAK,8BAA8B;AACjE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,QAAQ,KAAK,CAAC,KAAK,SAAS,OAAO,QAAQ,IAAI,CAAC,SAAS,GAAG,WAAW,GAAG,SAAS,CAAC,CAAC,CAAC;AAC5F,cAAU;AACV,OAAG,MAAM;AACT,kBAAc,SAAS;AAAA,EACzB;AAAA,EAEQ,sBAAsB,UAAkB;AAC9C,UAAM,aAA8B;AAAA,MAClC,MAAM,kBAAI,gBAAgB;AAAA,MAC1B,WAAW,KAAK;AAAA,MAChB,kBAAkB;AAAA,QAChB,eAAe;AAAA,MACjB;AAAA,IACF;AACA,SAAK,MAAM,IAAI,UAAU;AAAA,EAC3B;AACF;AAEA,MAAM,gCAAgC,CACpC,UACA,SACqB;AACrB,QAAM,OAAc,KAAK,SAAS,EAAE,cAAc;AAElD,SAAO,KAAK,IAAI,CAAC,SAAS;AAAA,IACxB;AAAA,IACA,WAAW,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,CAAC,EAAE,OAAO,IAAI;AAAA,IAC5D,SAAS,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,IAAI,OAAO,EAAE,SAAS,CAAC,EAAE,KAAK,IAAI;AAAA,IAC9E,YAAY,IAAI,YAAY;AAAA,IAC5B,MAAM,IAAI,YAAY;AAAA,EACxB,EAAE;AACJ;","names":["stt"]}
|
package/dist/stt.d.cts
CHANGED
|
@@ -32,6 +32,8 @@ export declare class SpeechStream extends stt.SpeechStream {
|
|
|
32
32
|
#private;
|
|
33
33
|
label: string;
|
|
34
34
|
constructor(stt: STT, opts: STTOptions);
|
|
35
|
+
protected run(): Promise<void>;
|
|
35
36
|
updateOptions(opts: Partial<STTOptions>): void;
|
|
37
|
+
private onAudioDurationReport;
|
|
36
38
|
}
|
|
37
39
|
//# sourceMappingURL=stt.d.ts.map
|
package/dist/stt.d.ts
CHANGED
|
@@ -32,6 +32,8 @@ export declare class SpeechStream extends stt.SpeechStream {
|
|
|
32
32
|
#private;
|
|
33
33
|
label: string;
|
|
34
34
|
constructor(stt: STT, opts: STTOptions);
|
|
35
|
+
protected run(): Promise<void>;
|
|
35
36
|
updateOptions(opts: Partial<STTOptions>): void;
|
|
37
|
+
private onAudioDurationReport;
|
|
36
38
|
}
|
|
37
39
|
//# sourceMappingURL=stt.d.ts.map
|
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,WAAW,EAKhB,GAAG,EACJ,MAAM,iBAAiB,CAAC;
|
|
1
|
+
{"version":3,"file":"stt.d.ts","sourceRoot":"","sources":["../src/stt.ts"],"names":[],"mappings":"AAGA,OAAO,EACL,KAAK,WAAW,EAKhB,GAAG,EACJ,MAAM,iBAAiB,CAAC;AAIzB,OAAO,KAAK,EAAE,YAAY,EAAE,SAAS,EAAE,MAAM,aAAa,CAAC;AAI3D,MAAM,WAAW,UAAU;IACzB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,QAAQ,CAAC,EAAE,YAAY,GAAG,MAAM,CAAC;IACjC,cAAc,EAAE,OAAO,CAAC;IACxB,cAAc,EAAE,OAAO,CAAC;IACxB,SAAS,EAAE,OAAO,CAAC;IACnB,KAAK,EAAE,SAAS,CAAC;IACjB,WAAW,EAAE,OAAO,CAAC;IACrB,OAAO,EAAE,OAAO,CAAC;IACjB,WAAW,EAAE,MAAM,CAAC;IACpB,WAAW,EAAE,OAAO,CAAC;IACrB,UAAU,EAAE,MAAM,CAAC;IACnB,WAAW,EAAE,MAAM,CAAC;IACpB,QAAQ,EAAE,CAAC,MAAM,EAAE,MAAM,CAAC,EAAE,CAAC;IAC7B,OAAO,EAAE,MAAM,EAAE,CAAC;IAClB,eAAe,EAAE,OAAO,CAAC;IACzB,SAAS,EAAE,OAAO,CAAC;IACnB,OAAO,EAAE,OAAO,CAAC;IACjB,QAAQ,EAAE,OAAO,CAAC;CACnB;AAuBD,qBAAa,GAAI,SAAQ,GAAG,CAAC,GAAG;;IAG9B,KAAK,SAAkB;gBAEX,IAAI,GAAE,OAAO,CAAC,UAAU,CAAqB;IAuCnD,UAAU,CAAC,CAAC,EAAE,WAAW,GAAG,OAAO,CAAC,GAAG,CAAC,WAAW,CAAC;IAI1D,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,UAAU,CAAC;IAIvC,MAAM,IAAI,YAAY;CAGvB;AAED,qBAAa,YAAa,SAAQ,GAAG,CAAC,YAAY;;IAQhD,KAAK,SAA2B;gBAEpB,GAAG,EAAE,GAAG,EAAE,IAAI,EAAE,UAAU;cAWtB,GAAG;IAkEnB,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,UAAU,CAAC;IAkJvC,OAAO,CAAC,qBAAqB;CAU9B"}
|
package/dist/stt.js
CHANGED
|
@@ -6,6 +6,7 @@ import {
|
|
|
6
6
|
stt
|
|
7
7
|
} from "@livekit/agents";
|
|
8
8
|
import { WebSocket } from "ws";
|
|
9
|
+
import { PeriodicCollector } from "./_utils.js";
|
|
9
10
|
const API_BASE_URL_V1 = "wss://api.deepgram.com/v1/listen";
|
|
10
11
|
const defaultSTTOptions = {
|
|
11
12
|
apiKey: process.env.DEEPGRAM_API_KEY,
|
|
@@ -13,7 +14,7 @@ const defaultSTTOptions = {
|
|
|
13
14
|
detectLanguage: false,
|
|
14
15
|
interimResults: true,
|
|
15
16
|
punctuate: true,
|
|
16
|
-
model: "nova-
|
|
17
|
+
model: "nova-3",
|
|
17
18
|
smartFormat: true,
|
|
18
19
|
noDelay: true,
|
|
19
20
|
endpointing: 25,
|
|
@@ -79,15 +80,21 @@ class SpeechStream extends stt.SpeechStream {
|
|
|
79
80
|
#logger = log();
|
|
80
81
|
#speaking = false;
|
|
81
82
|
#resetWS = new Future();
|
|
83
|
+
#requestId = "";
|
|
84
|
+
#audioDurationCollector;
|
|
82
85
|
label = "deepgram.SpeechStream";
|
|
83
86
|
constructor(stt2, opts) {
|
|
84
|
-
super(stt2);
|
|
87
|
+
super(stt2, opts.sampleRate);
|
|
85
88
|
this.#opts = opts;
|
|
86
89
|
this.closed = false;
|
|
87
90
|
this.#audioEnergyFilter = new AudioEnergyFilter();
|
|
88
|
-
this.#
|
|
91
|
+
this.#audioDurationCollector = new PeriodicCollector(
|
|
92
|
+
(duration) => this.onAudioDurationReport(duration),
|
|
93
|
+
{ duration: 5 }
|
|
94
|
+
);
|
|
89
95
|
}
|
|
90
|
-
async
|
|
96
|
+
async run() {
|
|
97
|
+
const maxRetry = 32;
|
|
91
98
|
let retries = 0;
|
|
92
99
|
let ws;
|
|
93
100
|
while (!this.input.closed) {
|
|
@@ -171,6 +178,7 @@ class SpeechStream extends stt.SpeechStream {
|
|
|
171
178
|
let frames;
|
|
172
179
|
if (data === SpeechStream.FLUSH_SENTINEL) {
|
|
173
180
|
frames = stream.flush();
|
|
181
|
+
this.#audioDurationCollector.flush();
|
|
174
182
|
} else if (data.sampleRate === this.#opts.sampleRate || data.channels === this.#opts.numChannels) {
|
|
175
183
|
frames = stream.write(data.data.buffer);
|
|
176
184
|
} else {
|
|
@@ -178,6 +186,8 @@ class SpeechStream extends stt.SpeechStream {
|
|
|
178
186
|
}
|
|
179
187
|
for await (const frame of frames) {
|
|
180
188
|
if (this.#audioEnergyFilter.pushFrame(frame)) {
|
|
189
|
+
const frameDuration = frame.samplesPerChannel / frame.sampleRate;
|
|
190
|
+
this.#audioDurationCollector.push(frameDuration);
|
|
181
191
|
ws.send(frame.data.buffer);
|
|
182
192
|
}
|
|
183
193
|
}
|
|
@@ -189,7 +199,7 @@ class SpeechStream extends stt.SpeechStream {
|
|
|
189
199
|
(_, reject) => ws.once("close", (code, reason) => {
|
|
190
200
|
if (!closing) {
|
|
191
201
|
this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);
|
|
192
|
-
reject();
|
|
202
|
+
reject(new Error("WebSocket closed"));
|
|
193
203
|
}
|
|
194
204
|
})
|
|
195
205
|
);
|
|
@@ -211,8 +221,11 @@ class SpeechStream extends stt.SpeechStream {
|
|
|
211
221
|
// https://developers.deepgram.com/docs/understand-endpointing-interim-results#using-endpointing-speech_final
|
|
212
222
|
// for more information about the different types of events
|
|
213
223
|
case "Results": {
|
|
224
|
+
const metadata = json["metadata"];
|
|
225
|
+
const requestId = metadata["request_id"];
|
|
214
226
|
const isFinal = json["is_final"];
|
|
215
227
|
const isEndpoint = json["speech_final"];
|
|
228
|
+
this.#requestId = requestId;
|
|
216
229
|
const alternatives = liveTranscriptionToSpeechData(this.#opts.language, json);
|
|
217
230
|
if (alternatives[0] && alternatives[0].text) {
|
|
218
231
|
if (!this.#speaking) {
|
|
@@ -257,6 +270,16 @@ class SpeechStream extends stt.SpeechStream {
|
|
|
257
270
|
ws.close();
|
|
258
271
|
clearInterval(keepalive);
|
|
259
272
|
}
|
|
273
|
+
onAudioDurationReport(duration) {
|
|
274
|
+
const usageEvent = {
|
|
275
|
+
type: stt.SpeechEventType.RECOGNITION_USAGE,
|
|
276
|
+
requestId: this.#requestId,
|
|
277
|
+
recognitionUsage: {
|
|
278
|
+
audioDuration: duration
|
|
279
|
+
}
|
|
280
|
+
};
|
|
281
|
+
this.queue.put(usageEvent);
|
|
282
|
+
}
|
|
260
283
|
}
|
|
261
284
|
const liveTranscriptionToSpeechData = (language, data) => {
|
|
262
285
|
const alts = data["channel"]["alternatives"];
|
package/dist/stt.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type AudioBuffer,\n AudioByteStream,\n AudioEnergyFilter,\n Future,\n log,\n stt,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { type RawData, WebSocket } from 'ws';\nimport type { STTLanguages, STTModels } from './models.js';\n\nconst API_BASE_URL_V1 = 'wss://api.deepgram.com/v1/listen';\n\nexport interface STTOptions {\n apiKey?: string;\n language?: STTLanguages | string;\n detectLanguage: boolean;\n interimResults: boolean;\n punctuate: boolean;\n model: STTModels;\n smartFormat: boolean;\n noDelay: boolean;\n endpointing: number;\n fillerWords: boolean;\n sampleRate: number;\n numChannels: number;\n keywords: [string, number][];\n keyterm: string[];\n profanityFilter: boolean;\n dictation: boolean;\n diarize: boolean;\n numerals: boolean;\n}\n\nconst defaultSTTOptions: STTOptions = {\n apiKey: process.env.DEEPGRAM_API_KEY,\n language: 'en-US',\n detectLanguage: false,\n interimResults: true,\n punctuate: true,\n model: 'nova-2-general',\n smartFormat: true,\n noDelay: true,\n endpointing: 25,\n fillerWords: false,\n sampleRate: 16000,\n numChannels: 1,\n keywords: [],\n keyterm: [],\n profanityFilter: false,\n dictation: false,\n diarize: false,\n numerals: false,\n};\n\nexport class STT extends stt.STT {\n #opts: STTOptions;\n #logger = log();\n label = 'deepgram.STT';\n\n constructor(opts: Partial<STTOptions> = defaultSTTOptions) {\n super({\n streaming: true,\n interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,\n });\n if (opts.apiKey === undefined && defaultSTTOptions.apiKey === undefined) {\n throw new Error(\n 'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n );\n }\n\n this.#opts = { ...defaultSTTOptions, ...opts };\n\n if (this.#opts.detectLanguage) {\n this.#opts.language = undefined;\n } else if (\n this.#opts.language &&\n !['en-US', 'en'].includes(this.#opts.language) &&\n [\n 'nova-2-meeting',\n 'nova-2-phonecall',\n 'nova-2-finance',\n 'nova-2-conversationalai',\n 'nova-2-voicemail',\n 'nova-2-video',\n 'nova-2-medical',\n 'nova-2-drivethru',\n 'nova-2-automotive',\n 'nova-3-general',\n ].includes(this.#opts.model)\n ) {\n this.#logger.warn(\n `${this.#opts.model} does not support language ${this.#opts.language}, falling back to nova-2-general`,\n );\n this.#opts.model = 'nova-2-general';\n }\n }\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n async _recognize(_: AudioBuffer): Promise<stt.SpeechEvent> {\n throw new Error('Recognize is not supported on Deepgram STT');\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n }\n\n stream(): SpeechStream {\n return new SpeechStream(this, this.#opts);\n }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n #opts: STTOptions;\n #audioEnergyFilter: AudioEnergyFilter;\n #logger = log();\n #speaking = false;\n #resetWS = new Future();\n label = 'deepgram.SpeechStream';\n\n constructor(stt: STT, opts: STTOptions) {\n super(stt);\n this.#opts = opts;\n this.closed = false;\n this.#audioEnergyFilter = new AudioEnergyFilter();\n\n this.#run();\n }\n\n async #run(maxRetry = 32) {\n let retries = 0;\n let ws: WebSocket;\n while (!this.input.closed) {\n const streamURL = new URL(API_BASE_URL_V1);\n const params = {\n model: this.#opts.model,\n punctuate: this.#opts.punctuate,\n smart_format: this.#opts.smartFormat,\n dictation: this.#opts.dictation,\n diarize: this.#opts.diarize,\n numerals: this.#opts.numerals,\n no_delay: this.#opts.noDelay,\n interim_results: this.#opts.interimResults,\n encoding: 'linear16',\n vad_events: true,\n sample_rate: this.#opts.sampleRate,\n channels: this.#opts.numChannels,\n endpointing: this.#opts.endpointing || false,\n filler_words: this.#opts.fillerWords,\n keywords: this.#opts.keywords.map((x) => x.join(':')),\n keyterm: this.#opts.keyterm,\n profanity_filter: this.#opts.profanityFilter,\n language: this.#opts.language,\n };\n Object.entries(params).forEach(([k, v]) => {\n if (v !== undefined) {\n if (typeof v === 'string' || typeof v === 'number' || typeof v === 'boolean') {\n streamURL.searchParams.append(k, encodeURIComponent(v));\n } else {\n v.forEach((x) => streamURL.searchParams.append(k, encodeURIComponent(x)));\n }\n }\n });\n\n ws = new WebSocket(streamURL, {\n headers: { Authorization: `Token ${this.#opts.apiKey}` },\n });\n\n try {\n await new Promise((resolve, reject) => {\n ws.on('open', resolve);\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n });\n\n await this.#runWS(ws);\n } catch (e) {\n if (retries >= maxRetry) {\n throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);\n }\n\n const delay = Math.min(retries * 5, 10);\n retries++;\n\n this.#logger.warn(\n `failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`,\n );\n await new Promise((resolve) => setTimeout(resolve, delay * 1000));\n }\n }\n\n this.closed = true;\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n this.#resetWS.resolve();\n }\n\n async #runWS(ws: WebSocket) {\n this.#resetWS = new Future();\n let closing = false;\n\n const keepalive = setInterval(() => {\n try {\n ws.send(JSON.stringify({ type: 'KeepAlive' }));\n } catch {\n clearInterval(keepalive);\n return;\n }\n }, 5000);\n\n const sendTask = async () => {\n const samples100Ms = Math.floor(this.#opts.sampleRate / 10);\n const stream = new AudioByteStream(\n this.#opts.sampleRate,\n this.#opts.numChannels,\n samples100Ms,\n );\n\n for await (const data of this.input) {\n let frames: AudioFrame[];\n if (data === SpeechStream.FLUSH_SENTINEL) {\n frames = stream.flush();\n } else if (\n data.sampleRate === this.#opts.sampleRate ||\n data.channels === this.#opts.numChannels\n ) {\n frames = stream.write(data.data.buffer);\n } else {\n throw new Error(`sample rate or channel count of frame does not match`);\n }\n\n for await (const frame of frames) {\n if (this.#audioEnergyFilter.pushFrame(frame)) {\n ws.send(frame.data.buffer);\n }\n }\n }\n\n closing = true;\n ws.send(JSON.stringify({ type: 'CloseStream' }));\n };\n\n const wsMonitor = 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();\n }\n }),\n );\n\n const listenTask = async () => {\n while (!this.closed && !closing) {\n try {\n await new Promise<RawData>((resolve) => {\n ws.once('message', (data) => resolve(data));\n }).then((msg) => {\n const json = JSON.parse(msg.toString());\n switch (json['type']) {\n case 'SpeechStarted': {\n // This is a normal case. Deepgram's SpeechStarted events\n // are not correlated with speech_final or utterance end.\n // It's possible that we receive two in a row without an endpoint\n // It's also possible we receive a transcript without a SpeechStarted event.\n if (this.#speaking) return;\n this.#speaking = true;\n this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n break;\n }\n // see this page:\n // https://developers.deepgram.com/docs/understand-endpointing-interim-results#using-endpointing-speech_final\n // for more information about the different types of events\n case 'Results': {\n const isFinal = json['is_final'];\n const isEndpoint = json['speech_final'];\n\n const alternatives = liveTranscriptionToSpeechData(this.#opts.language!, json);\n\n // If, for some reason, we didn't get a SpeechStarted event but we got\n // a transcript with text, we should start speaking. It's rare but has\n // been observed.\n if (alternatives[0] && alternatives[0].text) {\n if (!this.#speaking) {\n this.#speaking = true;\n this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n }\n\n if (isFinal) {\n this.queue.put({\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n alternatives: [alternatives[0], ...alternatives.slice(1)],\n });\n } else {\n this.queue.put({\n type: stt.SpeechEventType.INTERIM_TRANSCRIPT,\n alternatives: [alternatives[0], ...alternatives.slice(1)],\n });\n }\n }\n\n // if we receive an endpoint, only end the speech if\n // we either had a SpeechStarted event or we have a seen\n // a non-empty transcript (deepgram doesn't have a SpeechEnded event)\n if (isEndpoint && this.#speaking) {\n this.#speaking = false;\n this.queue.put({ type: stt.SpeechEventType.END_OF_SPEECH });\n }\n\n break;\n }\n case 'Metadata': {\n break;\n }\n default: {\n this.#logger.child({ msg: json }).warn('received unexpected message from Deepgram');\n break;\n }\n }\n });\n } catch (error) {\n this.#logger.child({ error }).warn('unrecoverable error, exiting');\n break;\n }\n }\n };\n\n await Promise.race([this.#resetWS.await, Promise.all([sendTask(), listenTask(), wsMonitor])]);\n closing = true;\n ws.close();\n clearInterval(keepalive);\n }\n}\n\nconst liveTranscriptionToSpeechData = (\n language: STTLanguages | string,\n data: { [id: string]: any },\n): stt.SpeechData[] => {\n const alts: any[] = data['channel']['alternatives'];\n\n return alts.map((alt) => ({\n language,\n startTime: alt['words'].length ? alt['words'][0]['start'] : 0,\n endTime: alt['words'].length ? alt['words'][alt['words'].length - 1]['end'] : 0,\n confidence: alt['confidence'],\n text: alt['transcript'],\n }));\n};\n"],"mappings":"AAGA;AAAA,EAEE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAuB,iBAAiB;AAGxC,MAAM,kBAAkB;AAuBxB,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,WAAW;AAAA,EACX,OAAO;AAAA,EACP,aAAa;AAAA,EACb,SAAS;AAAA,EACT,aAAa;AAAA,EACb,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,UAAU,CAAC;AAAA,EACX,SAAS,CAAC;AAAA,EACV,iBAAiB;AAAA,EACjB,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AACZ;AAEO,MAAM,YAAY,IAAI,IAAI;AAAA,EAC/B;AAAA,EACA,UAAU,IAAI;AAAA,EACd,QAAQ;AAAA,EAER,YAAY,OAA4B,mBAAmB;AACzD,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB,KAAK,kBAAkB,kBAAkB;AAAA,IAC3D,CAAC;AACD,QAAI,KAAK,WAAW,UAAa,kBAAkB,WAAW,QAAW;AACvE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ,EAAE,GAAG,mBAAmB,GAAG,KAAK;AAE7C,QAAI,KAAK,MAAM,gBAAgB;AAC7B,WAAK,MAAM,WAAW;AAAA,IACxB,WACE,KAAK,MAAM,YACX,CAAC,CAAC,SAAS,IAAI,EAAE,SAAS,KAAK,MAAM,QAAQ,KAC7C;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,EAAE,SAAS,KAAK,MAAM,KAAK,GAC3B;AACA,WAAK,QAAQ;AAAA,QACX,GAAG,KAAK,MAAM,KAAK,8BAA8B,KAAK,MAAM,QAAQ;AAAA,MACtE;AACA,WAAK,MAAM,QAAQ;AAAA,IACrB;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,WAAW,GAA0C;AACzD,UAAM,IAAI,MAAM,4CAA4C;AAAA,EAC9D;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAAA,EACxC;AAAA,EAEA,SAAuB;AACrB,WAAO,IAAI,aAAa,MAAM,KAAK,KAAK;AAAA,EAC1C;AACF;AAEO,MAAM,qBAAqB,IAAI,aAAa;AAAA,EACjD;AAAA,EACA;AAAA,EACA,UAAU,IAAI;AAAA,EACd,YAAY;AAAA,EACZ,WAAW,IAAI,OAAO;AAAA,EACtB,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB;AACtC,UAAMA,IAAG;AACT,SAAK,QAAQ;AACb,SAAK,SAAS;AACd,SAAK,qBAAqB,IAAI,kBAAkB;AAEhD,SAAK,KAAK;AAAA,EACZ;AAAA,EAEA,MAAM,KAAK,WAAW,IAAI;AACxB,QAAI,UAAU;AACd,QAAI;AACJ,WAAO,CAAC,KAAK,MAAM,QAAQ;AACzB,YAAM,YAAY,IAAI,IAAI,eAAe;AACzC,YAAM,SAAS;AAAA,QACb,OAAO,KAAK,MAAM;AAAA,QAClB,WAAW,KAAK,MAAM;AAAA,QACtB,cAAc,KAAK,MAAM;AAAA,QACzB,WAAW,KAAK,MAAM;AAAA,QACtB,SAAS,KAAK,MAAM;AAAA,QACpB,UAAU,KAAK,MAAM;AAAA,QACrB,UAAU,KAAK,MAAM;AAAA,QACrB,iBAAiB,KAAK,MAAM;AAAA,QAC5B,UAAU;AAAA,QACV,YAAY;AAAA,QACZ,aAAa,KAAK,MAAM;AAAA,QACxB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM,eAAe;AAAA,QACvC,cAAc,KAAK,MAAM;AAAA,QACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,QACpD,SAAS,KAAK,MAAM;AAAA,QACpB,kBAAkB,KAAK,MAAM;AAAA,QAC7B,UAAU,KAAK,MAAM;AAAA,MACvB;AACA,aAAO,QAAQ,MAAM,EAAE,QAAQ,CAAC,CAAC,GAAG,CAAC,MAAM;AACzC,YAAI,MAAM,QAAW;AACnB,cAAI,OAAO,MAAM,YAAY,OAAO,MAAM,YAAY,OAAO,MAAM,WAAW;AAC5E,sBAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC;AAAA,UACxD,OAAO;AACL,cAAE,QAAQ,CAAC,MAAM,UAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC,CAAC;AAAA,UAC1E;AAAA,QACF;AAAA,MACF,CAAC;AAED,WAAK,IAAI,UAAU,WAAW;AAAA,QAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,MACzD,CAAC;AAED,UAAI;AACF,cAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,aAAG,GAAG,QAAQ,OAAO;AACrB,aAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,aAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,QAC/D,CAAC;AAED,cAAM,KAAK,OAAO,EAAE;AAAA,MACtB,SAAS,GAAG;AACV,YAAI,WAAW,UAAU;AACvB,gBAAM,IAAI,MAAM,uCAAuC,OAAO,cAAc,CAAC,EAAE;AAAA,QACjF;AAEA,cAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,EAAE;AACtC;AAEA,aAAK,QAAQ;AAAA,UACX,8CAA8C,KAAK,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ;AAAA,QAC3F;AACA,cAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,QAAQ,GAAI,CAAC;AAAA,MAClE;AAAA,IACF;AAEA,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,SAAK,SAAS,QAAQ;AAAA,EACxB;AAAA,EAEA,MAAM,OAAO,IAAe;AAC1B,SAAK,WAAW,IAAI,OAAO;AAC3B,QAAI,UAAU;AAEd,UAAM,YAAY,YAAY,MAAM;AAClC,UAAI;AACF,WAAG,KAAK,KAAK,UAAU,EAAE,MAAM,YAAY,CAAC,CAAC;AAAA,MAC/C,QAAQ;AACN,sBAAc,SAAS;AACvB;AAAA,MACF;AAAA,IACF,GAAG,GAAI;AAEP,UAAM,WAAW,YAAY;AAC3B,YAAM,eAAe,KAAK,MAAM,KAAK,MAAM,aAAa,EAAE;AAC1D,YAAM,SAAS,IAAI;AAAA,QACjB,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX;AAAA,MACF;AAEA,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI;AACJ,YAAI,SAAS,aAAa,gBAAgB;AACxC,mBAAS,OAAO,MAAM;AAAA,QACxB,WACE,KAAK,eAAe,KAAK,MAAM,cAC/B,KAAK,aAAa,KAAK,MAAM,aAC7B;AACA,mBAAS,OAAO,MAAM,KAAK,KAAK,MAAM;AAAA,QACxC,OAAO;AACL,gBAAM,IAAI,MAAM,sDAAsD;AAAA,QACxE;AAEA,yBAAiB,SAAS,QAAQ;AAChC,cAAI,KAAK,mBAAmB,UAAU,KAAK,GAAG;AAC5C,eAAG,KAAK,MAAM,KAAK,MAAM;AAAA,UAC3B;AAAA,QACF;AAAA,MACF;AAEA,gBAAU;AACV,SAAG,KAAK,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC,CAAC;AAAA,IACjD;AAEA,UAAM,YAAY,IAAI;AAAA,MAAc,CAAC,GAAG,WACtC,GAAG,KAAK,SAAS,CAAC,MAAM,WAAW;AACjC,YAAI,CAAC,SAAS;AACZ,eAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,MAAM,EAAE;AAClE,iBAAO;AAAA,QACT;AAAA,MACF,CAAC;AAAA,IACH;AAEA,UAAM,aAAa,YAAY;AAC7B,aAAO,CAAC,KAAK,UAAU,CAAC,SAAS;AAC/B,YAAI;AACF,gBAAM,IAAI,QAAiB,CAAC,YAAY;AACtC,eAAG,KAAK,WAAW,CAAC,SAAS,QAAQ,IAAI,CAAC;AAAA,UAC5C,CAAC,EAAE,KAAK,CAAC,QAAQ;AACf,kBAAM,OAAO,KAAK,MAAM,IAAI,SAAS,CAAC;AACtC,oBAAQ,KAAK,MAAM,GAAG;AAAA,cACpB,KAAK,iBAAiB;AAKpB,oBAAI,KAAK,UAAW;AACpB,qBAAK,YAAY;AACjB,qBAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,gBAAgB,CAAC;AAC5D;AAAA,cACF;AAAA;AAAA;AAAA;AAAA,cAIA,KAAK,WAAW;AACd,sBAAM,UAAU,KAAK,UAAU;AAC/B,sBAAM,aAAa,KAAK,cAAc;AAEtC,sBAAM,eAAe,8BAA8B,KAAK,MAAM,UAAW,IAAI;AAK7E,oBAAI,aAAa,CAAC,KAAK,aAAa,CAAC,EAAE,MAAM;AAC3C,sBAAI,CAAC,KAAK,WAAW;AACnB,yBAAK,YAAY;AACjB,yBAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,gBAAgB,CAAC;AAAA,kBAC9D;AAEA,sBAAI,SAAS;AACX,yBAAK,MAAM,IAAI;AAAA,sBACb,MAAM,IAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH,OAAO;AACL,yBAAK,MAAM,IAAI;AAAA,sBACb,MAAM,IAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH;AAAA,gBACF;AAKA,oBAAI,cAAc,KAAK,WAAW;AAChC,uBAAK,YAAY;AACjB,uBAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,cAAc,CAAC;AAAA,gBAC5D;AAEA;AAAA,cACF;AAAA,cACA,KAAK,YAAY;AACf;AAAA,cACF;AAAA,cACA,SAAS;AACP,qBAAK,QAAQ,MAAM,EAAE,KAAK,KAAK,CAAC,EAAE,KAAK,2CAA2C;AAClF;AAAA,cACF;AAAA,YACF;AAAA,UACF,CAAC;AAAA,QACH,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,EAAE,MAAM,CAAC,EAAE,KAAK,8BAA8B;AACjE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,QAAQ,KAAK,CAAC,KAAK,SAAS,OAAO,QAAQ,IAAI,CAAC,SAAS,GAAG,WAAW,GAAG,SAAS,CAAC,CAAC,CAAC;AAC5F,cAAU;AACV,OAAG,MAAM;AACT,kBAAc,SAAS;AAAA,EACzB;AACF;AAEA,MAAM,gCAAgC,CACpC,UACA,SACqB;AACrB,QAAM,OAAc,KAAK,SAAS,EAAE,cAAc;AAElD,SAAO,KAAK,IAAI,CAAC,SAAS;AAAA,IACxB;AAAA,IACA,WAAW,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,CAAC,EAAE,OAAO,IAAI;AAAA,IAC5D,SAAS,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,IAAI,OAAO,EAAE,SAAS,CAAC,EAAE,KAAK,IAAI;AAAA,IAC9E,YAAY,IAAI,YAAY;AAAA,IAC5B,MAAM,IAAI,YAAY;AAAA,EACxB,EAAE;AACJ;","names":["stt"]}
|
|
1
|
+
{"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type AudioBuffer,\n AudioByteStream,\n AudioEnergyFilter,\n Future,\n log,\n stt,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { type RawData, WebSocket } from 'ws';\nimport { PeriodicCollector } from './_utils.js';\nimport type { STTLanguages, STTModels } from './models.js';\n\nconst API_BASE_URL_V1 = 'wss://api.deepgram.com/v1/listen';\n\nexport interface STTOptions {\n apiKey?: string;\n language?: STTLanguages | string;\n detectLanguage: boolean;\n interimResults: boolean;\n punctuate: boolean;\n model: STTModels;\n smartFormat: boolean;\n noDelay: boolean;\n endpointing: number;\n fillerWords: boolean;\n sampleRate: number;\n numChannels: number;\n keywords: [string, number][];\n keyterm: string[];\n profanityFilter: boolean;\n dictation: boolean;\n diarize: boolean;\n numerals: boolean;\n}\n\nconst defaultSTTOptions: STTOptions = {\n apiKey: process.env.DEEPGRAM_API_KEY,\n language: 'en-US',\n detectLanguage: false,\n interimResults: true,\n punctuate: true,\n model: 'nova-3',\n smartFormat: true,\n noDelay: true,\n endpointing: 25,\n fillerWords: false,\n sampleRate: 16000,\n numChannels: 1,\n keywords: [],\n keyterm: [],\n profanityFilter: false,\n dictation: false,\n diarize: false,\n numerals: false,\n};\n\nexport class STT extends stt.STT {\n #opts: STTOptions;\n #logger = log();\n label = 'deepgram.STT';\n\n constructor(opts: Partial<STTOptions> = defaultSTTOptions) {\n super({\n streaming: true,\n interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,\n });\n if (opts.apiKey === undefined && defaultSTTOptions.apiKey === undefined) {\n throw new Error(\n 'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n );\n }\n\n this.#opts = { ...defaultSTTOptions, ...opts };\n\n if (this.#opts.detectLanguage) {\n this.#opts.language = undefined;\n } else if (\n this.#opts.language &&\n !['en-US', 'en'].includes(this.#opts.language) &&\n [\n 'nova-2-meeting',\n 'nova-2-phonecall',\n 'nova-2-finance',\n 'nova-2-conversationalai',\n 'nova-2-voicemail',\n 'nova-2-video',\n 'nova-2-medical',\n 'nova-2-drivethru',\n 'nova-2-automotive',\n 'nova-3-general',\n ].includes(this.#opts.model)\n ) {\n this.#logger.warn(\n `${this.#opts.model} does not support language ${this.#opts.language}, falling back to nova-2-general`,\n );\n this.#opts.model = 'nova-2-general';\n }\n }\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n async _recognize(_: AudioBuffer): Promise<stt.SpeechEvent> {\n throw new Error('Recognize is not supported on Deepgram STT');\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n }\n\n stream(): SpeechStream {\n return new SpeechStream(this, this.#opts);\n }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n #opts: STTOptions;\n #audioEnergyFilter: AudioEnergyFilter;\n #logger = log();\n #speaking = false;\n #resetWS = new Future();\n #requestId = '';\n #audioDurationCollector: PeriodicCollector<number>;\n label = 'deepgram.SpeechStream';\n\n constructor(stt: STT, opts: STTOptions) {\n super(stt, opts.sampleRate);\n this.#opts = opts;\n this.closed = false;\n this.#audioEnergyFilter = new AudioEnergyFilter();\n this.#audioDurationCollector = new PeriodicCollector(\n (duration) => this.onAudioDurationReport(duration),\n { duration: 5.0 },\n );\n }\n\n protected async run() {\n const maxRetry = 32;\n let retries = 0;\n let ws: WebSocket;\n while (!this.input.closed) {\n const streamURL = new URL(API_BASE_URL_V1);\n const params = {\n model: this.#opts.model,\n punctuate: this.#opts.punctuate,\n smart_format: this.#opts.smartFormat,\n dictation: this.#opts.dictation,\n diarize: this.#opts.diarize,\n numerals: this.#opts.numerals,\n no_delay: this.#opts.noDelay,\n interim_results: this.#opts.interimResults,\n encoding: 'linear16',\n vad_events: true,\n sample_rate: this.#opts.sampleRate,\n channels: this.#opts.numChannels,\n endpointing: this.#opts.endpointing || false,\n filler_words: this.#opts.fillerWords,\n keywords: this.#opts.keywords.map((x) => x.join(':')),\n keyterm: this.#opts.keyterm,\n profanity_filter: this.#opts.profanityFilter,\n language: this.#opts.language,\n };\n Object.entries(params).forEach(([k, v]) => {\n if (v !== undefined) {\n if (typeof v === 'string' || typeof v === 'number' || typeof v === 'boolean') {\n streamURL.searchParams.append(k, encodeURIComponent(v));\n } else {\n v.forEach((x) => streamURL.searchParams.append(k, encodeURIComponent(x)));\n }\n }\n });\n\n ws = new WebSocket(streamURL, {\n headers: { Authorization: `Token ${this.#opts.apiKey}` },\n });\n\n try {\n await new Promise((resolve, reject) => {\n ws.on('open', resolve);\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n });\n\n await this.#runWS(ws);\n } catch (e) {\n if (retries >= maxRetry) {\n throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);\n }\n\n const delay = Math.min(retries * 5, 10);\n retries++;\n\n this.#logger.warn(\n `failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`,\n );\n await new Promise((resolve) => setTimeout(resolve, delay * 1000));\n }\n }\n\n this.closed = true;\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n this.#resetWS.resolve();\n }\n\n async #runWS(ws: WebSocket) {\n this.#resetWS = new Future();\n let closing = false;\n\n const keepalive = setInterval(() => {\n try {\n ws.send(JSON.stringify({ type: 'KeepAlive' }));\n } catch {\n clearInterval(keepalive);\n return;\n }\n }, 5000);\n\n const sendTask = async () => {\n const samples100Ms = Math.floor(this.#opts.sampleRate / 10);\n const stream = new AudioByteStream(\n this.#opts.sampleRate,\n this.#opts.numChannels,\n samples100Ms,\n );\n\n for await (const data of this.input) {\n let frames: AudioFrame[];\n if (data === SpeechStream.FLUSH_SENTINEL) {\n frames = stream.flush();\n this.#audioDurationCollector.flush();\n } else if (\n data.sampleRate === this.#opts.sampleRate ||\n data.channels === this.#opts.numChannels\n ) {\n frames = stream.write(data.data.buffer);\n } else {\n throw new Error(`sample rate or channel count of frame does not match`);\n }\n\n for await (const frame of frames) {\n if (this.#audioEnergyFilter.pushFrame(frame)) {\n const frameDuration = frame.samplesPerChannel / frame.sampleRate;\n this.#audioDurationCollector.push(frameDuration);\n ws.send(frame.data.buffer);\n }\n }\n }\n\n closing = true;\n ws.send(JSON.stringify({ type: 'CloseStream' }));\n };\n\n const wsMonitor = 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 const listenTask = async () => {\n while (!this.closed && !closing) {\n try {\n await new Promise<RawData>((resolve) => {\n ws.once('message', (data) => resolve(data));\n }).then((msg) => {\n const json = JSON.parse(msg.toString());\n switch (json['type']) {\n case 'SpeechStarted': {\n // This is a normal case. Deepgram's SpeechStarted events\n // are not correlated with speech_final or utterance end.\n // It's possible that we receive two in a row without an endpoint\n // It's also possible we receive a transcript without a SpeechStarted event.\n if (this.#speaking) return;\n this.#speaking = true;\n this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n break;\n }\n // see this page:\n // https://developers.deepgram.com/docs/understand-endpointing-interim-results#using-endpointing-speech_final\n // for more information about the different types of events\n case 'Results': {\n const metadata = json['metadata'];\n const requestId = metadata['request_id'];\n const isFinal = json['is_final'];\n const isEndpoint = json['speech_final'];\n this.#requestId = requestId;\n\n const alternatives = liveTranscriptionToSpeechData(this.#opts.language!, json);\n\n // If, for some reason, we didn't get a SpeechStarted event but we got\n // a transcript with text, we should start speaking. It's rare but has\n // been observed.\n if (alternatives[0] && alternatives[0].text) {\n if (!this.#speaking) {\n this.#speaking = true;\n this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n }\n\n if (isFinal) {\n this.queue.put({\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n alternatives: [alternatives[0], ...alternatives.slice(1)],\n });\n } else {\n this.queue.put({\n type: stt.SpeechEventType.INTERIM_TRANSCRIPT,\n alternatives: [alternatives[0], ...alternatives.slice(1)],\n });\n }\n }\n\n // if we receive an endpoint, only end the speech if\n // we either had a SpeechStarted event or we have a seen\n // a non-empty transcript (deepgram doesn't have a SpeechEnded event)\n if (isEndpoint && this.#speaking) {\n this.#speaking = false;\n this.queue.put({ type: stt.SpeechEventType.END_OF_SPEECH });\n }\n\n break;\n }\n case 'Metadata': {\n break;\n }\n default: {\n this.#logger.child({ msg: json }).warn('received unexpected message from Deepgram');\n break;\n }\n }\n });\n } catch (error) {\n this.#logger.child({ error }).warn('unrecoverable error, exiting');\n break;\n }\n }\n };\n\n await Promise.race([this.#resetWS.await, Promise.all([sendTask(), listenTask(), wsMonitor])]);\n closing = true;\n ws.close();\n clearInterval(keepalive);\n }\n\n private onAudioDurationReport(duration: number) {\n const usageEvent: stt.SpeechEvent = {\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: {\n audioDuration: duration,\n },\n };\n this.queue.put(usageEvent);\n }\n}\n\nconst liveTranscriptionToSpeechData = (\n language: STTLanguages | string,\n data: { [id: string]: any },\n): stt.SpeechData[] => {\n const alts: any[] = data['channel']['alternatives'];\n\n return alts.map((alt) => ({\n language,\n startTime: alt['words'].length ? alt['words'][0]['start'] : 0,\n endTime: alt['words'].length ? alt['words'][alt['words'].length - 1]['end'] : 0,\n confidence: alt['confidence'],\n text: alt['transcript'],\n }));\n};\n"],"mappings":"AAGA;AAAA,EAEE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAuB,iBAAiB;AACxC,SAAS,yBAAyB;AAGlC,MAAM,kBAAkB;AAuBxB,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,WAAW;AAAA,EACX,OAAO;AAAA,EACP,aAAa;AAAA,EACb,SAAS;AAAA,EACT,aAAa;AAAA,EACb,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,UAAU,CAAC;AAAA,EACX,SAAS,CAAC;AAAA,EACV,iBAAiB;AAAA,EACjB,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AACZ;AAEO,MAAM,YAAY,IAAI,IAAI;AAAA,EAC/B;AAAA,EACA,UAAU,IAAI;AAAA,EACd,QAAQ;AAAA,EAER,YAAY,OAA4B,mBAAmB;AACzD,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB,KAAK,kBAAkB,kBAAkB;AAAA,IAC3D,CAAC;AACD,QAAI,KAAK,WAAW,UAAa,kBAAkB,WAAW,QAAW;AACvE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ,EAAE,GAAG,mBAAmB,GAAG,KAAK;AAE7C,QAAI,KAAK,MAAM,gBAAgB;AAC7B,WAAK,MAAM,WAAW;AAAA,IACxB,WACE,KAAK,MAAM,YACX,CAAC,CAAC,SAAS,IAAI,EAAE,SAAS,KAAK,MAAM,QAAQ,KAC7C;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,EAAE,SAAS,KAAK,MAAM,KAAK,GAC3B;AACA,WAAK,QAAQ;AAAA,QACX,GAAG,KAAK,MAAM,KAAK,8BAA8B,KAAK,MAAM,QAAQ;AAAA,MACtE;AACA,WAAK,MAAM,QAAQ;AAAA,IACrB;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,WAAW,GAA0C;AACzD,UAAM,IAAI,MAAM,4CAA4C;AAAA,EAC9D;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAAA,EACxC;AAAA,EAEA,SAAuB;AACrB,WAAO,IAAI,aAAa,MAAM,KAAK,KAAK;AAAA,EAC1C;AACF;AAEO,MAAM,qBAAqB,IAAI,aAAa;AAAA,EACjD;AAAA,EACA;AAAA,EACA,UAAU,IAAI;AAAA,EACd,YAAY;AAAA,EACZ,WAAW,IAAI,OAAO;AAAA,EACtB,aAAa;AAAA,EACb;AAAA,EACA,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB;AACtC,UAAMA,MAAK,KAAK,UAAU;AAC1B,SAAK,QAAQ;AACb,SAAK,SAAS;AACd,SAAK,qBAAqB,IAAI,kBAAkB;AAChD,SAAK,0BAA0B,IAAI;AAAA,MACjC,CAAC,aAAa,KAAK,sBAAsB,QAAQ;AAAA,MACjD,EAAE,UAAU,EAAI;AAAA,IAClB;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,WAAW;AACjB,QAAI,UAAU;AACd,QAAI;AACJ,WAAO,CAAC,KAAK,MAAM,QAAQ;AACzB,YAAM,YAAY,IAAI,IAAI,eAAe;AACzC,YAAM,SAAS;AAAA,QACb,OAAO,KAAK,MAAM;AAAA,QAClB,WAAW,KAAK,MAAM;AAAA,QACtB,cAAc,KAAK,MAAM;AAAA,QACzB,WAAW,KAAK,MAAM;AAAA,QACtB,SAAS,KAAK,MAAM;AAAA,QACpB,UAAU,KAAK,MAAM;AAAA,QACrB,UAAU,KAAK,MAAM;AAAA,QACrB,iBAAiB,KAAK,MAAM;AAAA,QAC5B,UAAU;AAAA,QACV,YAAY;AAAA,QACZ,aAAa,KAAK,MAAM;AAAA,QACxB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM,eAAe;AAAA,QACvC,cAAc,KAAK,MAAM;AAAA,QACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,QACpD,SAAS,KAAK,MAAM;AAAA,QACpB,kBAAkB,KAAK,MAAM;AAAA,QAC7B,UAAU,KAAK,MAAM;AAAA,MACvB;AACA,aAAO,QAAQ,MAAM,EAAE,QAAQ,CAAC,CAAC,GAAG,CAAC,MAAM;AACzC,YAAI,MAAM,QAAW;AACnB,cAAI,OAAO,MAAM,YAAY,OAAO,MAAM,YAAY,OAAO,MAAM,WAAW;AAC5E,sBAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC;AAAA,UACxD,OAAO;AACL,cAAE,QAAQ,CAAC,MAAM,UAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC,CAAC;AAAA,UAC1E;AAAA,QACF;AAAA,MACF,CAAC;AAED,WAAK,IAAI,UAAU,WAAW;AAAA,QAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,MACzD,CAAC;AAED,UAAI;AACF,cAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,aAAG,GAAG,QAAQ,OAAO;AACrB,aAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,aAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,QAC/D,CAAC;AAED,cAAM,KAAK,OAAO,EAAE;AAAA,MACtB,SAAS,GAAG;AACV,YAAI,WAAW,UAAU;AACvB,gBAAM,IAAI,MAAM,uCAAuC,OAAO,cAAc,CAAC,EAAE;AAAA,QACjF;AAEA,cAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,EAAE;AACtC;AAEA,aAAK,QAAQ;AAAA,UACX,8CAA8C,KAAK,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ;AAAA,QAC3F;AACA,cAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,QAAQ,GAAI,CAAC;AAAA,MAClE;AAAA,IACF;AAEA,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,SAAK,SAAS,QAAQ;AAAA,EACxB;AAAA,EAEA,MAAM,OAAO,IAAe;AAC1B,SAAK,WAAW,IAAI,OAAO;AAC3B,QAAI,UAAU;AAEd,UAAM,YAAY,YAAY,MAAM;AAClC,UAAI;AACF,WAAG,KAAK,KAAK,UAAU,EAAE,MAAM,YAAY,CAAC,CAAC;AAAA,MAC/C,QAAQ;AACN,sBAAc,SAAS;AACvB;AAAA,MACF;AAAA,IACF,GAAG,GAAI;AAEP,UAAM,WAAW,YAAY;AAC3B,YAAM,eAAe,KAAK,MAAM,KAAK,MAAM,aAAa,EAAE;AAC1D,YAAM,SAAS,IAAI;AAAA,QACjB,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX;AAAA,MACF;AAEA,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI;AACJ,YAAI,SAAS,aAAa,gBAAgB;AACxC,mBAAS,OAAO,MAAM;AACtB,eAAK,wBAAwB,MAAM;AAAA,QACrC,WACE,KAAK,eAAe,KAAK,MAAM,cAC/B,KAAK,aAAa,KAAK,MAAM,aAC7B;AACA,mBAAS,OAAO,MAAM,KAAK,KAAK,MAAM;AAAA,QACxC,OAAO;AACL,gBAAM,IAAI,MAAM,sDAAsD;AAAA,QACxE;AAEA,yBAAiB,SAAS,QAAQ;AAChC,cAAI,KAAK,mBAAmB,UAAU,KAAK,GAAG;AAC5C,kBAAM,gBAAgB,MAAM,oBAAoB,MAAM;AACtD,iBAAK,wBAAwB,KAAK,aAAa;AAC/C,eAAG,KAAK,MAAM,KAAK,MAAM;AAAA,UAC3B;AAAA,QACF;AAAA,MACF;AAEA,gBAAU;AACV,SAAG,KAAK,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC,CAAC;AAAA,IACjD;AAEA,UAAM,YAAY,IAAI;AAAA,MAAc,CAAC,GAAG,WACtC,GAAG,KAAK,SAAS,CAAC,MAAM,WAAW;AACjC,YAAI,CAAC,SAAS;AACZ,eAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,MAAM,EAAE;AAClE,iBAAO,IAAI,MAAM,kBAAkB,CAAC;AAAA,QACtC;AAAA,MACF,CAAC;AAAA,IACH;AAEA,UAAM,aAAa,YAAY;AAC7B,aAAO,CAAC,KAAK,UAAU,CAAC,SAAS;AAC/B,YAAI;AACF,gBAAM,IAAI,QAAiB,CAAC,YAAY;AACtC,eAAG,KAAK,WAAW,CAAC,SAAS,QAAQ,IAAI,CAAC;AAAA,UAC5C,CAAC,EAAE,KAAK,CAAC,QAAQ;AACf,kBAAM,OAAO,KAAK,MAAM,IAAI,SAAS,CAAC;AACtC,oBAAQ,KAAK,MAAM,GAAG;AAAA,cACpB,KAAK,iBAAiB;AAKpB,oBAAI,KAAK,UAAW;AACpB,qBAAK,YAAY;AACjB,qBAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,gBAAgB,CAAC;AAC5D;AAAA,cACF;AAAA;AAAA;AAAA;AAAA,cAIA,KAAK,WAAW;AACd,sBAAM,WAAW,KAAK,UAAU;AAChC,sBAAM,YAAY,SAAS,YAAY;AACvC,sBAAM,UAAU,KAAK,UAAU;AAC/B,sBAAM,aAAa,KAAK,cAAc;AACtC,qBAAK,aAAa;AAElB,sBAAM,eAAe,8BAA8B,KAAK,MAAM,UAAW,IAAI;AAK7E,oBAAI,aAAa,CAAC,KAAK,aAAa,CAAC,EAAE,MAAM;AAC3C,sBAAI,CAAC,KAAK,WAAW;AACnB,yBAAK,YAAY;AACjB,yBAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,gBAAgB,CAAC;AAAA,kBAC9D;AAEA,sBAAI,SAAS;AACX,yBAAK,MAAM,IAAI;AAAA,sBACb,MAAM,IAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH,OAAO;AACL,yBAAK,MAAM,IAAI;AAAA,sBACb,MAAM,IAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH;AAAA,gBACF;AAKA,oBAAI,cAAc,KAAK,WAAW;AAChC,uBAAK,YAAY;AACjB,uBAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,cAAc,CAAC;AAAA,gBAC5D;AAEA;AAAA,cACF;AAAA,cACA,KAAK,YAAY;AACf;AAAA,cACF;AAAA,cACA,SAAS;AACP,qBAAK,QAAQ,MAAM,EAAE,KAAK,KAAK,CAAC,EAAE,KAAK,2CAA2C;AAClF;AAAA,cACF;AAAA,YACF;AAAA,UACF,CAAC;AAAA,QACH,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,EAAE,MAAM,CAAC,EAAE,KAAK,8BAA8B;AACjE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,QAAQ,KAAK,CAAC,KAAK,SAAS,OAAO,QAAQ,IAAI,CAAC,SAAS,GAAG,WAAW,GAAG,SAAS,CAAC,CAAC,CAAC;AAC5F,cAAU;AACV,OAAG,MAAM;AACT,kBAAc,SAAS;AAAA,EACzB;AAAA,EAEQ,sBAAsB,UAAkB;AAC9C,UAAM,aAA8B;AAAA,MAClC,MAAM,IAAI,gBAAgB;AAAA,MAC1B,WAAW,KAAK;AAAA,MAChB,kBAAkB;AAAA,QAChB,eAAe;AAAA,MACjB;AAAA,IACF;AACA,SAAK,MAAM,IAAI,UAAU;AAAA,EAC3B;AACF;AAEA,MAAM,gCAAgC,CACpC,UACA,SACqB;AACrB,QAAM,OAAc,KAAK,SAAS,EAAE,cAAc;AAElD,SAAO,KAAK,IAAI,CAAC,SAAS;AAAA,IACxB;AAAA,IACA,WAAW,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,CAAC,EAAE,OAAO,IAAI;AAAA,IAC5D,SAAS,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,IAAI,OAAO,EAAE,SAAS,CAAC,EAAE,KAAK,IAAI;AAAA,IAC9E,YAAY,IAAI,YAAY;AAAA,IAC5B,MAAM,IAAI,YAAY;AAAA,EACxB,EAAE;AACJ;","names":["stt"]}
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@livekit/agents-plugin-deepgram",
|
|
3
|
-
"version": "0.
|
|
3
|
+
"version": "1.0.0-next.0",
|
|
4
4
|
"description": "Deepgram plugin for LiveKit Agents for Node.js",
|
|
5
5
|
"main": "dist/index.js",
|
|
6
6
|
"require": "dist/index.cjs",
|
|
@@ -25,10 +25,10 @@
|
|
|
25
25
|
"README.md"
|
|
26
26
|
],
|
|
27
27
|
"devDependencies": {
|
|
28
|
-
"@livekit/agents": "^
|
|
29
|
-
"@livekit/agents-plugin-silero": "^
|
|
30
|
-
"@livekit/agents-plugins-test": "^
|
|
31
|
-
"@livekit/rtc-node": "^0.13.
|
|
28
|
+
"@livekit/agents": "^1.0.0-next.0",
|
|
29
|
+
"@livekit/agents-plugin-silero": "^1.0.0-next.0",
|
|
30
|
+
"@livekit/agents-plugins-test": "^1.0.0-next.0",
|
|
31
|
+
"@livekit/rtc-node": "^0.13.12",
|
|
32
32
|
"@microsoft/api-extractor": "^7.35.0",
|
|
33
33
|
"@types/ws": "^8.5.10",
|
|
34
34
|
"tsup": "^8.3.5",
|
|
@@ -38,8 +38,8 @@
|
|
|
38
38
|
"ws": "^8.16.0"
|
|
39
39
|
},
|
|
40
40
|
"peerDependencies": {
|
|
41
|
-
"@livekit/rtc-node": "^0.13.
|
|
42
|
-
"@livekit/agents": "^0.
|
|
41
|
+
"@livekit/rtc-node": "^0.13.12",
|
|
42
|
+
"@livekit/agents": "^1.0.0-next.01.0.0-next.0"
|
|
43
43
|
},
|
|
44
44
|
"scripts": {
|
|
45
45
|
"build": "tsup --onSuccess \"pnpm build:types\"",
|
package/src/_utils.ts
ADDED
|
@@ -0,0 +1,50 @@
|
|
|
1
|
+
// SPDX-FileCopyrightText: 2025 LiveKit, Inc.
|
|
2
|
+
//
|
|
3
|
+
// SPDX-License-Identifier: Apache-2.0
|
|
4
|
+
|
|
5
|
+
export class PeriodicCollector<T> {
|
|
6
|
+
private duration: number;
|
|
7
|
+
private callback: (value: T) => void;
|
|
8
|
+
private lastFlushTime: number;
|
|
9
|
+
private total: T | null = null;
|
|
10
|
+
|
|
11
|
+
constructor(callback: (value: T) => void, options: { duration: number }) {
|
|
12
|
+
/**
|
|
13
|
+
* Create a new periodic collector that accumulates values and calls the callback
|
|
14
|
+
* after the specified duration if there are values to report.
|
|
15
|
+
*
|
|
16
|
+
* @param callback Function to call with accumulated value when duration expires
|
|
17
|
+
* @param options.duration Time in seconds between callback invocations
|
|
18
|
+
*/
|
|
19
|
+
this.duration = options.duration;
|
|
20
|
+
this.callback = callback;
|
|
21
|
+
this.lastFlushTime = performance.now() / 1000; // Convert to seconds
|
|
22
|
+
}
|
|
23
|
+
|
|
24
|
+
push(value: T): void {
|
|
25
|
+
/**
|
|
26
|
+
* Add a value to the accumulator
|
|
27
|
+
*/
|
|
28
|
+
if (this.total === null) {
|
|
29
|
+
this.total = value;
|
|
30
|
+
} else {
|
|
31
|
+
// Type assertion needed for generic addition
|
|
32
|
+
this.total = (this.total as any) + (value as any);
|
|
33
|
+
}
|
|
34
|
+
|
|
35
|
+
if (performance.now() / 1000 - this.lastFlushTime >= this.duration) {
|
|
36
|
+
this.flush();
|
|
37
|
+
}
|
|
38
|
+
}
|
|
39
|
+
|
|
40
|
+
flush(): void {
|
|
41
|
+
/**
|
|
42
|
+
* Force callback to be called with current total if non-zero
|
|
43
|
+
*/
|
|
44
|
+
if (this.total !== null) {
|
|
45
|
+
this.callback(this.total);
|
|
46
|
+
this.total = null;
|
|
47
|
+
}
|
|
48
|
+
this.lastFlushTime = performance.now() / 1000;
|
|
49
|
+
}
|
|
50
|
+
}
|
package/src/index.ts
CHANGED
|
@@ -1,5 +1,18 @@
|
|
|
1
|
-
// SPDX-FileCopyrightText:
|
|
1
|
+
// SPDX-FileCopyrightText: 2025 LiveKit, Inc.
|
|
2
2
|
//
|
|
3
3
|
// SPDX-License-Identifier: Apache-2.0
|
|
4
|
+
import { Plugin } from '@livekit/agents';
|
|
4
5
|
|
|
5
6
|
export * from './stt.js';
|
|
7
|
+
|
|
8
|
+
class DeepgramPlugin extends Plugin {
|
|
9
|
+
constructor() {
|
|
10
|
+
super({
|
|
11
|
+
title: 'deepgram',
|
|
12
|
+
version: '0.5.6',
|
|
13
|
+
package: '@livekit/agents-plugin-deepgram',
|
|
14
|
+
});
|
|
15
|
+
}
|
|
16
|
+
}
|
|
17
|
+
|
|
18
|
+
Plugin.registerPlugin(new DeepgramPlugin());
|
package/src/models.ts
CHANGED
|
@@ -16,7 +16,9 @@ export type STTModels =
|
|
|
16
16
|
| 'nova-2-medical'
|
|
17
17
|
| 'nova-2-drivethru'
|
|
18
18
|
| 'nova-2-automotive'
|
|
19
|
+
| 'nova-3'
|
|
19
20
|
| 'nova-3-general'
|
|
21
|
+
| 'nova-3-medical'
|
|
20
22
|
| 'enhanced-general'
|
|
21
23
|
| 'enhanced-meeting'
|
|
22
24
|
| 'enhanced-phonecall'
|
|
@@ -68,4 +70,5 @@ export type STTLanguages =
|
|
|
68
70
|
| 'uk'
|
|
69
71
|
| 'zh'
|
|
70
72
|
| 'zh-CN'
|
|
71
|
-
| 'zh-TW'
|
|
73
|
+
| 'zh-TW'
|
|
74
|
+
| 'multi';
|
package/src/stt.ts
CHANGED
|
@@ -11,6 +11,7 @@ import {
|
|
|
11
11
|
} from '@livekit/agents';
|
|
12
12
|
import type { AudioFrame } from '@livekit/rtc-node';
|
|
13
13
|
import { type RawData, WebSocket } from 'ws';
|
|
14
|
+
import { PeriodicCollector } from './_utils.js';
|
|
14
15
|
import type { STTLanguages, STTModels } from './models.js';
|
|
15
16
|
|
|
16
17
|
const API_BASE_URL_V1 = 'wss://api.deepgram.com/v1/listen';
|
|
@@ -42,7 +43,7 @@ const defaultSTTOptions: STTOptions = {
|
|
|
42
43
|
detectLanguage: false,
|
|
43
44
|
interimResults: true,
|
|
44
45
|
punctuate: true,
|
|
45
|
-
model: 'nova-
|
|
46
|
+
model: 'nova-3',
|
|
46
47
|
smartFormat: true,
|
|
47
48
|
noDelay: true,
|
|
48
49
|
endpointing: 25,
|
|
@@ -120,18 +121,23 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
120
121
|
#logger = log();
|
|
121
122
|
#speaking = false;
|
|
122
123
|
#resetWS = new Future();
|
|
124
|
+
#requestId = '';
|
|
125
|
+
#audioDurationCollector: PeriodicCollector<number>;
|
|
123
126
|
label = 'deepgram.SpeechStream';
|
|
124
127
|
|
|
125
128
|
constructor(stt: STT, opts: STTOptions) {
|
|
126
|
-
super(stt);
|
|
129
|
+
super(stt, opts.sampleRate);
|
|
127
130
|
this.#opts = opts;
|
|
128
131
|
this.closed = false;
|
|
129
132
|
this.#audioEnergyFilter = new AudioEnergyFilter();
|
|
130
|
-
|
|
131
|
-
|
|
133
|
+
this.#audioDurationCollector = new PeriodicCollector(
|
|
134
|
+
(duration) => this.onAudioDurationReport(duration),
|
|
135
|
+
{ duration: 5.0 },
|
|
136
|
+
);
|
|
132
137
|
}
|
|
133
138
|
|
|
134
|
-
async
|
|
139
|
+
protected async run() {
|
|
140
|
+
const maxRetry = 32;
|
|
135
141
|
let retries = 0;
|
|
136
142
|
let ws: WebSocket;
|
|
137
143
|
while (!this.input.closed) {
|
|
@@ -226,6 +232,7 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
226
232
|
let frames: AudioFrame[];
|
|
227
233
|
if (data === SpeechStream.FLUSH_SENTINEL) {
|
|
228
234
|
frames = stream.flush();
|
|
235
|
+
this.#audioDurationCollector.flush();
|
|
229
236
|
} else if (
|
|
230
237
|
data.sampleRate === this.#opts.sampleRate ||
|
|
231
238
|
data.channels === this.#opts.numChannels
|
|
@@ -237,6 +244,8 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
237
244
|
|
|
238
245
|
for await (const frame of frames) {
|
|
239
246
|
if (this.#audioEnergyFilter.pushFrame(frame)) {
|
|
247
|
+
const frameDuration = frame.samplesPerChannel / frame.sampleRate;
|
|
248
|
+
this.#audioDurationCollector.push(frameDuration);
|
|
240
249
|
ws.send(frame.data.buffer);
|
|
241
250
|
}
|
|
242
251
|
}
|
|
@@ -250,7 +259,7 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
250
259
|
ws.once('close', (code, reason) => {
|
|
251
260
|
if (!closing) {
|
|
252
261
|
this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);
|
|
253
|
-
reject();
|
|
262
|
+
reject(new Error('WebSocket closed'));
|
|
254
263
|
}
|
|
255
264
|
}),
|
|
256
265
|
);
|
|
@@ -277,8 +286,11 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
277
286
|
// https://developers.deepgram.com/docs/understand-endpointing-interim-results#using-endpointing-speech_final
|
|
278
287
|
// for more information about the different types of events
|
|
279
288
|
case 'Results': {
|
|
289
|
+
const metadata = json['metadata'];
|
|
290
|
+
const requestId = metadata['request_id'];
|
|
280
291
|
const isFinal = json['is_final'];
|
|
281
292
|
const isEndpoint = json['speech_final'];
|
|
293
|
+
this.#requestId = requestId;
|
|
282
294
|
|
|
283
295
|
const alternatives = liveTranscriptionToSpeechData(this.#opts.language!, json);
|
|
284
296
|
|
|
@@ -335,6 +347,17 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
335
347
|
ws.close();
|
|
336
348
|
clearInterval(keepalive);
|
|
337
349
|
}
|
|
350
|
+
|
|
351
|
+
private onAudioDurationReport(duration: number) {
|
|
352
|
+
const usageEvent: stt.SpeechEvent = {
|
|
353
|
+
type: stt.SpeechEventType.RECOGNITION_USAGE,
|
|
354
|
+
requestId: this.#requestId,
|
|
355
|
+
recognitionUsage: {
|
|
356
|
+
audioDuration: duration,
|
|
357
|
+
},
|
|
358
|
+
};
|
|
359
|
+
this.queue.put(usageEvent);
|
|
360
|
+
}
|
|
338
361
|
}
|
|
339
362
|
|
|
340
363
|
const liveTranscriptionToSpeechData = (
|