@livekit/agents-plugin-openai 1.0.25 → 1.0.27
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/tts.cjs +7 -5
- package/dist/tts.cjs.map +1 -1
- package/dist/tts.d.cts +3 -3
- package/dist/tts.d.ts +3 -3
- package/dist/tts.d.ts.map +1 -1
- package/dist/tts.js +7 -5
- package/dist/tts.js.map +1 -1
- package/package.json +5 -5
- package/src/tts.ts +17 -5
package/dist/tts.cjs
CHANGED
|
@@ -58,7 +58,7 @@ class TTS extends import_agents.tts.TTS {
|
|
|
58
58
|
updateOptions(opts) {
|
|
59
59
|
this.#opts = { ...this.#opts, ...opts };
|
|
60
60
|
}
|
|
61
|
-
synthesize(text) {
|
|
61
|
+
synthesize(text, connOptions, abortSignal) {
|
|
62
62
|
return new ChunkedStream(
|
|
63
63
|
this,
|
|
64
64
|
text,
|
|
@@ -71,8 +71,10 @@ class TTS extends import_agents.tts.TTS {
|
|
|
71
71
|
response_format: "pcm",
|
|
72
72
|
speed: this.#opts.speed
|
|
73
73
|
},
|
|
74
|
-
{ signal:
|
|
75
|
-
)
|
|
74
|
+
{ signal: abortSignal }
|
|
75
|
+
),
|
|
76
|
+
connOptions,
|
|
77
|
+
abortSignal
|
|
76
78
|
);
|
|
77
79
|
}
|
|
78
80
|
stream() {
|
|
@@ -86,8 +88,8 @@ class ChunkedStream extends import_agents.tts.ChunkedStream {
|
|
|
86
88
|
label = "openai.ChunkedStream";
|
|
87
89
|
stream;
|
|
88
90
|
// set Promise<T> to any because OpenAI returns an annoying Response type
|
|
89
|
-
constructor(tts2, text, stream) {
|
|
90
|
-
super(text, tts2);
|
|
91
|
+
constructor(tts2, text, stream, connOptions, abortSignal) {
|
|
92
|
+
super(text, tts2, connOptions, abortSignal);
|
|
91
93
|
this.stream = stream;
|
|
92
94
|
}
|
|
93
95
|
async run() {
|
package/dist/tts.cjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { AudioByteStream, shortuuid, tts } from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { OpenAI } from 'openai';\nimport type { TTSModels, TTSVoices } from './models.js';\n\nconst OPENAI_TTS_SAMPLE_RATE = 24000;\nconst OPENAI_TTS_CHANNELS = 1;\n\nexport interface TTSOptions {\n model: TTSModels | string;\n voice: TTSVoices;\n speed: number;\n instructions?: string;\n baseURL?: string;\n client?: OpenAI;\n apiKey?: string;\n}\n\nconst defaultTTSOptions: TTSOptions = {\n apiKey: process.env.OPENAI_API_KEY,\n model: 'tts-1',\n voice: 'alloy',\n speed: 1,\n};\n\nexport class TTS extends tts.TTS {\n #opts: TTSOptions;\n #client: OpenAI;\n label = 'openai.TTS';\n private abortController = new AbortController();\n\n /**\n * Create a new instance of OpenAI TTS.\n *\n * @remarks\n * `apiKey` must be set to your OpenAI API key, either using the argument or by setting the\n * `OPENAI_API_KEY` environment variable.\n */\n constructor(opts: Partial<TTSOptions> = defaultTTSOptions) {\n super(OPENAI_TTS_SAMPLE_RATE, OPENAI_TTS_CHANNELS, { streaming: false });\n\n this.#opts = { ...defaultTTSOptions, ...opts };\n if (this.#opts.apiKey === undefined) {\n throw new Error('OpenAI API key is required, whether as an argument or as $OPENAI_API_KEY');\n }\n\n this.#client =\n this.#opts.client ||\n new OpenAI({\n baseURL: opts.baseURL,\n apiKey: opts.apiKey,\n });\n }\n\n updateOptions(opts: { model?: TTSModels | string; voice?: TTSVoices; speed?: number }) {\n this.#opts = { ...this.#opts, ...opts };\n }\n\n synthesize(text: string): ChunkedStream {\n return new ChunkedStream(\n this,\n text,\n this.#client.audio.speech.create(\n {\n input: text,\n model: this.#opts.model,\n voice: this.#opts.voice,\n instructions: this.#opts.instructions,\n response_format: 'pcm',\n speed: this.#opts.speed,\n },\n { signal:
|
|
1
|
+
{"version":3,"sources":["../src/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { type APIConnectOptions, AudioByteStream, shortuuid, tts } from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { OpenAI } from 'openai';\nimport type { TTSModels, TTSVoices } from './models.js';\n\nconst OPENAI_TTS_SAMPLE_RATE = 24000;\nconst OPENAI_TTS_CHANNELS = 1;\n\nexport interface TTSOptions {\n model: TTSModels | string;\n voice: TTSVoices;\n speed: number;\n instructions?: string;\n baseURL?: string;\n client?: OpenAI;\n apiKey?: string;\n}\n\nconst defaultTTSOptions: TTSOptions = {\n apiKey: process.env.OPENAI_API_KEY,\n model: 'tts-1',\n voice: 'alloy',\n speed: 1,\n};\n\nexport class TTS extends tts.TTS {\n #opts: TTSOptions;\n #client: OpenAI;\n label = 'openai.TTS';\n private abortController = new AbortController();\n\n /**\n * Create a new instance of OpenAI TTS.\n *\n * @remarks\n * `apiKey` must be set to your OpenAI API key, either using the argument or by setting the\n * `OPENAI_API_KEY` environment variable.\n */\n constructor(opts: Partial<TTSOptions> = defaultTTSOptions) {\n super(OPENAI_TTS_SAMPLE_RATE, OPENAI_TTS_CHANNELS, { streaming: false });\n\n this.#opts = { ...defaultTTSOptions, ...opts };\n if (this.#opts.apiKey === undefined) {\n throw new Error('OpenAI API key is required, whether as an argument or as $OPENAI_API_KEY');\n }\n\n this.#client =\n this.#opts.client ||\n new OpenAI({\n baseURL: opts.baseURL,\n apiKey: opts.apiKey,\n });\n }\n\n updateOptions(opts: { model?: TTSModels | string; voice?: TTSVoices; speed?: number }) {\n this.#opts = { ...this.#opts, ...opts };\n }\n\n synthesize(\n text: string,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ): ChunkedStream {\n return new ChunkedStream(\n this,\n text,\n this.#client.audio.speech.create(\n {\n input: text,\n model: this.#opts.model,\n voice: this.#opts.voice,\n instructions: this.#opts.instructions,\n response_format: 'pcm',\n speed: this.#opts.speed,\n },\n { signal: abortSignal },\n ),\n connOptions,\n abortSignal,\n );\n }\n\n stream(): tts.SynthesizeStream {\n throw new Error('Streaming is not supported on OpenAI TTS');\n }\n\n async close(): Promise<void> {\n this.abortController.abort();\n }\n}\n\nexport class ChunkedStream extends tts.ChunkedStream {\n label = 'openai.ChunkedStream';\n private stream: Promise<any>;\n\n // set Promise<T> to any because OpenAI returns an annoying Response type\n constructor(\n tts: TTS,\n text: string,\n stream: Promise<any>,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ) {\n super(text, tts, connOptions, abortSignal);\n this.stream = stream;\n }\n\n protected async run() {\n try {\n const buffer = await this.stream.then((r) => r.arrayBuffer());\n const requestId = shortuuid();\n const audioByteStream = new AudioByteStream(OPENAI_TTS_SAMPLE_RATE, OPENAI_TTS_CHANNELS);\n const frames = audioByteStream.write(buffer);\n\n let lastFrame: AudioFrame | undefined;\n const sendLastFrame = (segmentId: string, final: boolean) => {\n if (lastFrame) {\n this.queue.put({ requestId, segmentId, frame: lastFrame, final });\n lastFrame = undefined;\n }\n };\n\n for (const frame of frames) {\n sendLastFrame(requestId, false);\n lastFrame = frame;\n }\n sendLastFrame(requestId, true);\n\n this.queue.close();\n } catch (error) {\n if (error instanceof Error && error.name === 'AbortError') {\n return;\n }\n throw error;\n } finally {\n this.queue.close();\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBAAwE;AAExE,oBAAuB;AAGvB,MAAM,yBAAyB;AAC/B,MAAM,sBAAsB;AAY5B,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,OAAO;AAAA,EACP,OAAO;AAAA,EACP,OAAO;AACT;AAEO,MAAM,YAAY,kBAAI,IAAI;AAAA,EAC/B;AAAA,EACA;AAAA,EACA,QAAQ;AAAA,EACA,kBAAkB,IAAI,gBAAgB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAS9C,YAAY,OAA4B,mBAAmB;AACzD,UAAM,wBAAwB,qBAAqB,EAAE,WAAW,MAAM,CAAC;AAEvE,SAAK,QAAQ,EAAE,GAAG,mBAAmB,GAAG,KAAK;AAC7C,QAAI,KAAK,MAAM,WAAW,QAAW;AACnC,YAAM,IAAI,MAAM,0EAA0E;AAAA,IAC5F;AAEA,SAAK,UACH,KAAK,MAAM,UACX,IAAI,qBAAO;AAAA,MACT,SAAS,KAAK;AAAA,MACd,QAAQ,KAAK;AAAA,IACf,CAAC;AAAA,EACL;AAAA,EAEA,cAAc,MAAyE;AACrF,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAAA,EACxC;AAAA,EAEA,WACE,MACA,aACA,aACe;AACf,WAAO,IAAI;AAAA,MACT;AAAA,MACA;AAAA,MACA,KAAK,QAAQ,MAAM,OAAO;AAAA,QACxB;AAAA,UACE,OAAO;AAAA,UACP,OAAO,KAAK,MAAM;AAAA,UAClB,OAAO,KAAK,MAAM;AAAA,UAClB,cAAc,KAAK,MAAM;AAAA,UACzB,iBAAiB;AAAA,UACjB,OAAO,KAAK,MAAM;AAAA,QACpB;AAAA,QACA,EAAE,QAAQ,YAAY;AAAA,MACxB;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA,EAEA,SAA+B;AAC7B,UAAM,IAAI,MAAM,0CAA0C;AAAA,EAC5D;AAAA,EAEA,MAAM,QAAuB;AAC3B,SAAK,gBAAgB,MAAM;AAAA,EAC7B;AACF;AAEO,MAAM,sBAAsB,kBAAI,cAAc;AAAA,EACnD,QAAQ;AAAA,EACA;AAAA;AAAA,EAGR,YACEA,MACA,MACA,QACA,aACA,aACA;AACA,UAAM,MAAMA,MAAK,aAAa,WAAW;AACzC,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAgB,MAAM;AACpB,QAAI;AACF,YAAM,SAAS,MAAM,KAAK,OAAO,KAAK,CAAC,MAAM,EAAE,YAAY,CAAC;AAC5D,YAAM,gBAAY,yBAAU;AAC5B,YAAM,kBAAkB,IAAI,8BAAgB,wBAAwB,mBAAmB;AACvF,YAAM,SAAS,gBAAgB,MAAM,MAAM;AAE3C,UAAI;AACJ,YAAM,gBAAgB,CAAC,WAAmB,UAAmB;AAC3D,YAAI,WAAW;AACb,eAAK,MAAM,IAAI,EAAE,WAAW,WAAW,OAAO,WAAW,MAAM,CAAC;AAChE,sBAAY;AAAA,QACd;AAAA,MACF;AAEA,iBAAW,SAAS,QAAQ;AAC1B,sBAAc,WAAW,KAAK;AAC9B,oBAAY;AAAA,MACd;AACA,oBAAc,WAAW,IAAI;AAE7B,WAAK,MAAM,MAAM;AAAA,IACnB,SAAS,OAAO;AACd,UAAI,iBAAiB,SAAS,MAAM,SAAS,cAAc;AACzD;AAAA,MACF;AACA,YAAM;AAAA,IACR,UAAE;AACA,WAAK,MAAM,MAAM;AAAA,IACnB;AAAA,EACF;AACF;","names":["tts"]}
|
package/dist/tts.d.cts
CHANGED
|
@@ -1,4 +1,4 @@
|
|
|
1
|
-
import { tts } from '@livekit/agents';
|
|
1
|
+
import { type APIConnectOptions, tts } from '@livekit/agents';
|
|
2
2
|
import { OpenAI } from 'openai';
|
|
3
3
|
import type { TTSModels, TTSVoices } from './models.js';
|
|
4
4
|
export interface TTSOptions {
|
|
@@ -27,14 +27,14 @@ export declare class TTS extends tts.TTS {
|
|
|
27
27
|
voice?: TTSVoices;
|
|
28
28
|
speed?: number;
|
|
29
29
|
}): void;
|
|
30
|
-
synthesize(text: string): ChunkedStream;
|
|
30
|
+
synthesize(text: string, connOptions?: APIConnectOptions, abortSignal?: AbortSignal): ChunkedStream;
|
|
31
31
|
stream(): tts.SynthesizeStream;
|
|
32
32
|
close(): Promise<void>;
|
|
33
33
|
}
|
|
34
34
|
export declare class ChunkedStream extends tts.ChunkedStream {
|
|
35
35
|
label: string;
|
|
36
36
|
private stream;
|
|
37
|
-
constructor(tts: TTS, text: string, stream: Promise<any
|
|
37
|
+
constructor(tts: TTS, text: string, stream: Promise<any>, connOptions?: APIConnectOptions, abortSignal?: AbortSignal);
|
|
38
38
|
protected run(): Promise<void>;
|
|
39
39
|
}
|
|
40
40
|
//# sourceMappingURL=tts.d.ts.map
|
package/dist/tts.d.ts
CHANGED
|
@@ -1,4 +1,4 @@
|
|
|
1
|
-
import { tts } from '@livekit/agents';
|
|
1
|
+
import { type APIConnectOptions, tts } from '@livekit/agents';
|
|
2
2
|
import { OpenAI } from 'openai';
|
|
3
3
|
import type { TTSModels, TTSVoices } from './models.js';
|
|
4
4
|
export interface TTSOptions {
|
|
@@ -27,14 +27,14 @@ export declare class TTS extends tts.TTS {
|
|
|
27
27
|
voice?: TTSVoices;
|
|
28
28
|
speed?: number;
|
|
29
29
|
}): void;
|
|
30
|
-
synthesize(text: string): ChunkedStream;
|
|
30
|
+
synthesize(text: string, connOptions?: APIConnectOptions, abortSignal?: AbortSignal): ChunkedStream;
|
|
31
31
|
stream(): tts.SynthesizeStream;
|
|
32
32
|
close(): Promise<void>;
|
|
33
33
|
}
|
|
34
34
|
export declare class ChunkedStream extends tts.ChunkedStream {
|
|
35
35
|
label: string;
|
|
36
36
|
private stream;
|
|
37
|
-
constructor(tts: TTS, text: string, stream: Promise<any
|
|
37
|
+
constructor(tts: TTS, text: string, stream: Promise<any>, connOptions?: APIConnectOptions, abortSignal?: AbortSignal);
|
|
38
38
|
protected run(): Promise<void>;
|
|
39
39
|
}
|
|
40
40
|
//# sourceMappingURL=tts.d.ts.map
|
package/dist/tts.d.ts.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"tts.d.ts","sourceRoot":"","sources":["../src/tts.ts"],"names":[],"mappings":"AAGA,OAAO,EAA8B,GAAG,EAAE,MAAM,iBAAiB,CAAC;
|
|
1
|
+
{"version":3,"file":"tts.d.ts","sourceRoot":"","sources":["../src/tts.ts"],"names":[],"mappings":"AAGA,OAAO,EAAE,KAAK,iBAAiB,EAA8B,GAAG,EAAE,MAAM,iBAAiB,CAAC;AAE1F,OAAO,EAAE,MAAM,EAAE,MAAM,QAAQ,CAAC;AAChC,OAAO,KAAK,EAAE,SAAS,EAAE,SAAS,EAAE,MAAM,aAAa,CAAC;AAKxD,MAAM,WAAW,UAAU;IACzB,KAAK,EAAE,SAAS,GAAG,MAAM,CAAC;IAC1B,KAAK,EAAE,SAAS,CAAC;IACjB,KAAK,EAAE,MAAM,CAAC;IACd,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,MAAM,CAAC,EAAE,MAAM,CAAC;CACjB;AASD,qBAAa,GAAI,SAAQ,GAAG,CAAC,GAAG;;IAG9B,KAAK,SAAgB;IACrB,OAAO,CAAC,eAAe,CAAyB;IAEhD;;;;;;OAMG;gBACS,IAAI,GAAE,OAAO,CAAC,UAAU,CAAqB;IAgBzD,aAAa,CAAC,IAAI,EAAE;QAAE,KAAK,CAAC,EAAE,SAAS,GAAG,MAAM,CAAC;QAAC,KAAK,CAAC,EAAE,SAAS,CAAC;QAAC,KAAK,CAAC,EAAE,MAAM,CAAA;KAAE;IAIrF,UAAU,CACR,IAAI,EAAE,MAAM,EACZ,WAAW,CAAC,EAAE,iBAAiB,EAC/B,WAAW,CAAC,EAAE,WAAW,GACxB,aAAa;IAoBhB,MAAM,IAAI,GAAG,CAAC,gBAAgB;IAIxB,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;CAG7B;AAED,qBAAa,aAAc,SAAQ,GAAG,CAAC,aAAa;IAClD,KAAK,SAA0B;IAC/B,OAAO,CAAC,MAAM,CAAe;gBAI3B,GAAG,EAAE,GAAG,EACR,IAAI,EAAE,MAAM,EACZ,MAAM,EAAE,OAAO,CAAC,GAAG,CAAC,EACpB,WAAW,CAAC,EAAE,iBAAiB,EAC/B,WAAW,CAAC,EAAE,WAAW;cAMX,GAAG;CA+BpB"}
|
package/dist/tts.js
CHANGED
|
@@ -34,7 +34,7 @@ class TTS extends tts.TTS {
|
|
|
34
34
|
updateOptions(opts) {
|
|
35
35
|
this.#opts = { ...this.#opts, ...opts };
|
|
36
36
|
}
|
|
37
|
-
synthesize(text) {
|
|
37
|
+
synthesize(text, connOptions, abortSignal) {
|
|
38
38
|
return new ChunkedStream(
|
|
39
39
|
this,
|
|
40
40
|
text,
|
|
@@ -47,8 +47,10 @@ class TTS extends tts.TTS {
|
|
|
47
47
|
response_format: "pcm",
|
|
48
48
|
speed: this.#opts.speed
|
|
49
49
|
},
|
|
50
|
-
{ signal:
|
|
51
|
-
)
|
|
50
|
+
{ signal: abortSignal }
|
|
51
|
+
),
|
|
52
|
+
connOptions,
|
|
53
|
+
abortSignal
|
|
52
54
|
);
|
|
53
55
|
}
|
|
54
56
|
stream() {
|
|
@@ -62,8 +64,8 @@ class ChunkedStream extends tts.ChunkedStream {
|
|
|
62
64
|
label = "openai.ChunkedStream";
|
|
63
65
|
stream;
|
|
64
66
|
// set Promise<T> to any because OpenAI returns an annoying Response type
|
|
65
|
-
constructor(tts2, text, stream) {
|
|
66
|
-
super(text, tts2);
|
|
67
|
+
constructor(tts2, text, stream, connOptions, abortSignal) {
|
|
68
|
+
super(text, tts2, connOptions, abortSignal);
|
|
67
69
|
this.stream = stream;
|
|
68
70
|
}
|
|
69
71
|
async run() {
|
package/dist/tts.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { AudioByteStream, shortuuid, tts } from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { OpenAI } from 'openai';\nimport type { TTSModels, TTSVoices } from './models.js';\n\nconst OPENAI_TTS_SAMPLE_RATE = 24000;\nconst OPENAI_TTS_CHANNELS = 1;\n\nexport interface TTSOptions {\n model: TTSModels | string;\n voice: TTSVoices;\n speed: number;\n instructions?: string;\n baseURL?: string;\n client?: OpenAI;\n apiKey?: string;\n}\n\nconst defaultTTSOptions: TTSOptions = {\n apiKey: process.env.OPENAI_API_KEY,\n model: 'tts-1',\n voice: 'alloy',\n speed: 1,\n};\n\nexport class TTS extends tts.TTS {\n #opts: TTSOptions;\n #client: OpenAI;\n label = 'openai.TTS';\n private abortController = new AbortController();\n\n /**\n * Create a new instance of OpenAI TTS.\n *\n * @remarks\n * `apiKey` must be set to your OpenAI API key, either using the argument or by setting the\n * `OPENAI_API_KEY` environment variable.\n */\n constructor(opts: Partial<TTSOptions> = defaultTTSOptions) {\n super(OPENAI_TTS_SAMPLE_RATE, OPENAI_TTS_CHANNELS, { streaming: false });\n\n this.#opts = { ...defaultTTSOptions, ...opts };\n if (this.#opts.apiKey === undefined) {\n throw new Error('OpenAI API key is required, whether as an argument or as $OPENAI_API_KEY');\n }\n\n this.#client =\n this.#opts.client ||\n new OpenAI({\n baseURL: opts.baseURL,\n apiKey: opts.apiKey,\n });\n }\n\n updateOptions(opts: { model?: TTSModels | string; voice?: TTSVoices; speed?: number }) {\n this.#opts = { ...this.#opts, ...opts };\n }\n\n synthesize(text: string): ChunkedStream {\n return new ChunkedStream(\n this,\n text,\n this.#client.audio.speech.create(\n {\n input: text,\n model: this.#opts.model,\n voice: this.#opts.voice,\n instructions: this.#opts.instructions,\n response_format: 'pcm',\n speed: this.#opts.speed,\n },\n { signal:
|
|
1
|
+
{"version":3,"sources":["../src/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { type APIConnectOptions, AudioByteStream, shortuuid, tts } from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { OpenAI } from 'openai';\nimport type { TTSModels, TTSVoices } from './models.js';\n\nconst OPENAI_TTS_SAMPLE_RATE = 24000;\nconst OPENAI_TTS_CHANNELS = 1;\n\nexport interface TTSOptions {\n model: TTSModels | string;\n voice: TTSVoices;\n speed: number;\n instructions?: string;\n baseURL?: string;\n client?: OpenAI;\n apiKey?: string;\n}\n\nconst defaultTTSOptions: TTSOptions = {\n apiKey: process.env.OPENAI_API_KEY,\n model: 'tts-1',\n voice: 'alloy',\n speed: 1,\n};\n\nexport class TTS extends tts.TTS {\n #opts: TTSOptions;\n #client: OpenAI;\n label = 'openai.TTS';\n private abortController = new AbortController();\n\n /**\n * Create a new instance of OpenAI TTS.\n *\n * @remarks\n * `apiKey` must be set to your OpenAI API key, either using the argument or by setting the\n * `OPENAI_API_KEY` environment variable.\n */\n constructor(opts: Partial<TTSOptions> = defaultTTSOptions) {\n super(OPENAI_TTS_SAMPLE_RATE, OPENAI_TTS_CHANNELS, { streaming: false });\n\n this.#opts = { ...defaultTTSOptions, ...opts };\n if (this.#opts.apiKey === undefined) {\n throw new Error('OpenAI API key is required, whether as an argument or as $OPENAI_API_KEY');\n }\n\n this.#client =\n this.#opts.client ||\n new OpenAI({\n baseURL: opts.baseURL,\n apiKey: opts.apiKey,\n });\n }\n\n updateOptions(opts: { model?: TTSModels | string; voice?: TTSVoices; speed?: number }) {\n this.#opts = { ...this.#opts, ...opts };\n }\n\n synthesize(\n text: string,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ): ChunkedStream {\n return new ChunkedStream(\n this,\n text,\n this.#client.audio.speech.create(\n {\n input: text,\n model: this.#opts.model,\n voice: this.#opts.voice,\n instructions: this.#opts.instructions,\n response_format: 'pcm',\n speed: this.#opts.speed,\n },\n { signal: abortSignal },\n ),\n connOptions,\n abortSignal,\n );\n }\n\n stream(): tts.SynthesizeStream {\n throw new Error('Streaming is not supported on OpenAI TTS');\n }\n\n async close(): Promise<void> {\n this.abortController.abort();\n }\n}\n\nexport class ChunkedStream extends tts.ChunkedStream {\n label = 'openai.ChunkedStream';\n private stream: Promise<any>;\n\n // set Promise<T> to any because OpenAI returns an annoying Response type\n constructor(\n tts: TTS,\n text: string,\n stream: Promise<any>,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ) {\n super(text, tts, connOptions, abortSignal);\n this.stream = stream;\n }\n\n protected async run() {\n try {\n const buffer = await this.stream.then((r) => r.arrayBuffer());\n const requestId = shortuuid();\n const audioByteStream = new AudioByteStream(OPENAI_TTS_SAMPLE_RATE, OPENAI_TTS_CHANNELS);\n const frames = audioByteStream.write(buffer);\n\n let lastFrame: AudioFrame | undefined;\n const sendLastFrame = (segmentId: string, final: boolean) => {\n if (lastFrame) {\n this.queue.put({ requestId, segmentId, frame: lastFrame, final });\n lastFrame = undefined;\n }\n };\n\n for (const frame of frames) {\n sendLastFrame(requestId, false);\n lastFrame = frame;\n }\n sendLastFrame(requestId, true);\n\n this.queue.close();\n } catch (error) {\n if (error instanceof Error && error.name === 'AbortError') {\n return;\n }\n throw error;\n } finally {\n this.queue.close();\n }\n }\n}\n"],"mappings":"AAGA,SAAiC,iBAAiB,WAAW,WAAW;AAExE,SAAS,cAAc;AAGvB,MAAM,yBAAyB;AAC/B,MAAM,sBAAsB;AAY5B,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,OAAO;AAAA,EACP,OAAO;AAAA,EACP,OAAO;AACT;AAEO,MAAM,YAAY,IAAI,IAAI;AAAA,EAC/B;AAAA,EACA;AAAA,EACA,QAAQ;AAAA,EACA,kBAAkB,IAAI,gBAAgB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAS9C,YAAY,OAA4B,mBAAmB;AACzD,UAAM,wBAAwB,qBAAqB,EAAE,WAAW,MAAM,CAAC;AAEvE,SAAK,QAAQ,EAAE,GAAG,mBAAmB,GAAG,KAAK;AAC7C,QAAI,KAAK,MAAM,WAAW,QAAW;AACnC,YAAM,IAAI,MAAM,0EAA0E;AAAA,IAC5F;AAEA,SAAK,UACH,KAAK,MAAM,UACX,IAAI,OAAO;AAAA,MACT,SAAS,KAAK;AAAA,MACd,QAAQ,KAAK;AAAA,IACf,CAAC;AAAA,EACL;AAAA,EAEA,cAAc,MAAyE;AACrF,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAAA,EACxC;AAAA,EAEA,WACE,MACA,aACA,aACe;AACf,WAAO,IAAI;AAAA,MACT;AAAA,MACA;AAAA,MACA,KAAK,QAAQ,MAAM,OAAO;AAAA,QACxB;AAAA,UACE,OAAO;AAAA,UACP,OAAO,KAAK,MAAM;AAAA,UAClB,OAAO,KAAK,MAAM;AAAA,UAClB,cAAc,KAAK,MAAM;AAAA,UACzB,iBAAiB;AAAA,UACjB,OAAO,KAAK,MAAM;AAAA,QACpB;AAAA,QACA,EAAE,QAAQ,YAAY;AAAA,MACxB;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA,EAEA,SAA+B;AAC7B,UAAM,IAAI,MAAM,0CAA0C;AAAA,EAC5D;AAAA,EAEA,MAAM,QAAuB;AAC3B,SAAK,gBAAgB,MAAM;AAAA,EAC7B;AACF;AAEO,MAAM,sBAAsB,IAAI,cAAc;AAAA,EACnD,QAAQ;AAAA,EACA;AAAA;AAAA,EAGR,YACEA,MACA,MACA,QACA,aACA,aACA;AACA,UAAM,MAAMA,MAAK,aAAa,WAAW;AACzC,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAgB,MAAM;AACpB,QAAI;AACF,YAAM,SAAS,MAAM,KAAK,OAAO,KAAK,CAAC,MAAM,EAAE,YAAY,CAAC;AAC5D,YAAM,YAAY,UAAU;AAC5B,YAAM,kBAAkB,IAAI,gBAAgB,wBAAwB,mBAAmB;AACvF,YAAM,SAAS,gBAAgB,MAAM,MAAM;AAE3C,UAAI;AACJ,YAAM,gBAAgB,CAAC,WAAmB,UAAmB;AAC3D,YAAI,WAAW;AACb,eAAK,MAAM,IAAI,EAAE,WAAW,WAAW,OAAO,WAAW,MAAM,CAAC;AAChE,sBAAY;AAAA,QACd;AAAA,MACF;AAEA,iBAAW,SAAS,QAAQ;AAC1B,sBAAc,WAAW,KAAK;AAC9B,oBAAY;AAAA,MACd;AACA,oBAAc,WAAW,IAAI;AAE7B,WAAK,MAAM,MAAM;AAAA,IACnB,SAAS,OAAO;AACd,UAAI,iBAAiB,SAAS,MAAM,SAAS,cAAc;AACzD;AAAA,MACF;AACA,YAAM;AAAA,IACR,UAAE;AACA,WAAK,MAAM,MAAM;AAAA,IACnB;AAAA,EACF;AACF;","names":["tts"]}
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@livekit/agents-plugin-openai",
|
|
3
|
-
"version": "1.0.
|
|
3
|
+
"version": "1.0.27",
|
|
4
4
|
"description": "OpenAI plugin for LiveKit Node Agents",
|
|
5
5
|
"main": "dist/index.js",
|
|
6
6
|
"require": "dist/index.cjs",
|
|
@@ -30,9 +30,9 @@
|
|
|
30
30
|
"@types/ws": "^8.5.10",
|
|
31
31
|
"tsup": "^8.3.5",
|
|
32
32
|
"typescript": "^5.0.0",
|
|
33
|
-
"@livekit/agents
|
|
34
|
-
"@livekit/agents": "1.0.
|
|
35
|
-
"@livekit/agents-plugins-test": "1.0.
|
|
33
|
+
"@livekit/agents": "1.0.27",
|
|
34
|
+
"@livekit/agents-plugin-silero": "1.0.27",
|
|
35
|
+
"@livekit/agents-plugins-test": "1.0.27"
|
|
36
36
|
},
|
|
37
37
|
"dependencies": {
|
|
38
38
|
"@livekit/mutex": "^1.1.1",
|
|
@@ -41,7 +41,7 @@
|
|
|
41
41
|
},
|
|
42
42
|
"peerDependencies": {
|
|
43
43
|
"@livekit/rtc-node": "^0.13.12",
|
|
44
|
-
"@livekit/agents": "1.0.
|
|
44
|
+
"@livekit/agents": "1.0.27"
|
|
45
45
|
},
|
|
46
46
|
"scripts": {
|
|
47
47
|
"build": "tsup --onSuccess \"pnpm build:types\"",
|
package/src/tts.ts
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
// SPDX-FileCopyrightText: 2024 LiveKit, Inc.
|
|
2
2
|
//
|
|
3
3
|
// SPDX-License-Identifier: Apache-2.0
|
|
4
|
-
import { AudioByteStream, shortuuid, tts } from '@livekit/agents';
|
|
4
|
+
import { type APIConnectOptions, AudioByteStream, shortuuid, tts } from '@livekit/agents';
|
|
5
5
|
import type { AudioFrame } from '@livekit/rtc-node';
|
|
6
6
|
import { OpenAI } from 'openai';
|
|
7
7
|
import type { TTSModels, TTSVoices } from './models.js';
|
|
@@ -59,7 +59,11 @@ export class TTS extends tts.TTS {
|
|
|
59
59
|
this.#opts = { ...this.#opts, ...opts };
|
|
60
60
|
}
|
|
61
61
|
|
|
62
|
-
synthesize(
|
|
62
|
+
synthesize(
|
|
63
|
+
text: string,
|
|
64
|
+
connOptions?: APIConnectOptions,
|
|
65
|
+
abortSignal?: AbortSignal,
|
|
66
|
+
): ChunkedStream {
|
|
63
67
|
return new ChunkedStream(
|
|
64
68
|
this,
|
|
65
69
|
text,
|
|
@@ -72,8 +76,10 @@ export class TTS extends tts.TTS {
|
|
|
72
76
|
response_format: 'pcm',
|
|
73
77
|
speed: this.#opts.speed,
|
|
74
78
|
},
|
|
75
|
-
{ signal:
|
|
79
|
+
{ signal: abortSignal },
|
|
76
80
|
),
|
|
81
|
+
connOptions,
|
|
82
|
+
abortSignal,
|
|
77
83
|
);
|
|
78
84
|
}
|
|
79
85
|
|
|
@@ -91,8 +97,14 @@ export class ChunkedStream extends tts.ChunkedStream {
|
|
|
91
97
|
private stream: Promise<any>;
|
|
92
98
|
|
|
93
99
|
// set Promise<T> to any because OpenAI returns an annoying Response type
|
|
94
|
-
constructor(
|
|
95
|
-
|
|
100
|
+
constructor(
|
|
101
|
+
tts: TTS,
|
|
102
|
+
text: string,
|
|
103
|
+
stream: Promise<any>,
|
|
104
|
+
connOptions?: APIConnectOptions,
|
|
105
|
+
abortSignal?: AbortSignal,
|
|
106
|
+
) {
|
|
107
|
+
super(text, tts, connOptions, abortSignal);
|
|
96
108
|
this.stream = stream;
|
|
97
109
|
}
|
|
98
110
|
|