@livekit/agents-plugin-baseten 1.4.0 → 1.4.2
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/index.cjs +1 -1
- package/dist/index.js +1 -1
- package/dist/tts.cjs +2 -1
- package/dist/tts.cjs.map +1 -1
- package/dist/tts.d.ts.map +1 -1
- package/dist/tts.js +2 -1
- package/dist/tts.js.map +1 -1
- package/package.json +5 -5
- package/src/tts.ts +4 -1
package/dist/index.cjs
CHANGED
package/dist/index.js
CHANGED
package/dist/tts.cjs
CHANGED
|
@@ -66,7 +66,8 @@ class TTS extends import_agents.tts.TTS {
|
|
|
66
66
|
* call {@link ChunkedStream.stop} on the returned object.
|
|
67
67
|
*/
|
|
68
68
|
synthesize(text, connOptions, abortSignal) {
|
|
69
|
-
|
|
69
|
+
const signal = abortSignal ? AbortSignal.any([abortSignal, this.abortController.signal]) : this.abortController.signal;
|
|
70
|
+
return new ChunkedStream(this, text, this.opts, connOptions, signal);
|
|
70
71
|
}
|
|
71
72
|
stream() {
|
|
72
73
|
throw new Error("Streaming is not supported on Baseten TTS");
|
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 {\n type APIConnectOptions,\n AudioByteStream,\n shortuuid,\n tts,\n waitForAbort,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { BasetenTTSOptions } from './types.js';\n\nconst defaultTTSOptions: Partial<BasetenTTSOptions> = {\n voice: 'tara',\n language: 'en',\n temperature: 0.6,\n};\n\n/**\n * Baseten TTS implementation (streaming, 24kHz mono)\n */\nexport class TTS extends tts.TTS {\n private opts: BasetenTTSOptions;\n label = 'baseten.TTS';\n private abortController = new AbortController();\n constructor(opts: Partial<BasetenTTSOptions> = {}) {\n /**\n * Baseten audio is 24kHz mono.\n * The Orpheus model generates audio chunks that are processed as they arrive,\n * which reduces latency and improves agent responsiveness.\n */\n super(24000, 1, { streaming: false });\n\n // Apply defaults and environment fallbacks.\n const apiKey = opts.apiKey ?? process.env.BASETEN_API_KEY;\n const modelEndpoint = opts.modelEndpoint ?? process.env.BASETEN_MODEL_ENDPOINT;\n\n if (!apiKey) {\n throw new Error(\n 'Baseten API key is required, either pass it as `apiKey` or set $BASETEN_API_KEY',\n );\n }\n if (!modelEndpoint) {\n throw new Error(\n 'Baseten model endpoint is required, either pass it as `modelEndpoint` or set $BASETEN_MODEL_ENDPOINT',\n );\n }\n\n this.opts = {\n ...defaultTTSOptions,\n ...opts,\n apiKey,\n modelEndpoint,\n } as BasetenTTSOptions;\n }\n\n updateOptions(opts: Partial<Omit<BasetenTTSOptions, 'apiKey' | 'modelEndpoint'>>) {\n this.opts = {\n ...this.opts,\n ...opts,\n } as BasetenTTSOptions;\n }\n\n /**\n * Synthesize speech for a given piece of text. Returns a `ChunkedStream`\n * which will asynchronously fetch audio from Baseten and push frames into\n * LiveKit's playback pipeline. If you need to cancel synthesis you can\n * call {@link ChunkedStream.stop} on the returned object.\n */\n synthesize(\n text: string,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ): ChunkedStream {\n return new ChunkedStream(this, text, this.opts, connOptions,
|
|
1
|
+
{"version":3,"sources":["../src/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n AudioByteStream,\n shortuuid,\n tts,\n waitForAbort,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { BasetenTTSOptions } from './types.js';\n\nconst defaultTTSOptions: Partial<BasetenTTSOptions> = {\n voice: 'tara',\n language: 'en',\n temperature: 0.6,\n};\n\n/**\n * Baseten TTS implementation (streaming, 24kHz mono)\n */\nexport class TTS extends tts.TTS {\n private opts: BasetenTTSOptions;\n label = 'baseten.TTS';\n private abortController = new AbortController();\n constructor(opts: Partial<BasetenTTSOptions> = {}) {\n /**\n * Baseten audio is 24kHz mono.\n * The Orpheus model generates audio chunks that are processed as they arrive,\n * which reduces latency and improves agent responsiveness.\n */\n super(24000, 1, { streaming: false });\n\n // Apply defaults and environment fallbacks.\n const apiKey = opts.apiKey ?? process.env.BASETEN_API_KEY;\n const modelEndpoint = opts.modelEndpoint ?? process.env.BASETEN_MODEL_ENDPOINT;\n\n if (!apiKey) {\n throw new Error(\n 'Baseten API key is required, either pass it as `apiKey` or set $BASETEN_API_KEY',\n );\n }\n if (!modelEndpoint) {\n throw new Error(\n 'Baseten model endpoint is required, either pass it as `modelEndpoint` or set $BASETEN_MODEL_ENDPOINT',\n );\n }\n\n this.opts = {\n ...defaultTTSOptions,\n ...opts,\n apiKey,\n modelEndpoint,\n } as BasetenTTSOptions;\n }\n\n updateOptions(opts: Partial<Omit<BasetenTTSOptions, 'apiKey' | 'modelEndpoint'>>) {\n this.opts = {\n ...this.opts,\n ...opts,\n } as BasetenTTSOptions;\n }\n\n /**\n * Synthesize speech for a given piece of text. Returns a `ChunkedStream`\n * which will asynchronously fetch audio from Baseten and push frames into\n * LiveKit's playback pipeline. If you need to cancel synthesis you can\n * call {@link ChunkedStream.stop} on the returned object.\n */\n synthesize(\n text: string,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ): ChunkedStream {\n const signal = abortSignal\n ? AbortSignal.any([abortSignal, this.abortController.signal])\n : this.abortController.signal;\n return new ChunkedStream(this, text, this.opts, connOptions, signal);\n }\n\n stream(): tts.SynthesizeStream {\n throw new Error('Streaming is not supported on Baseten TTS');\n }\n\n async close(): Promise<void> {\n this.abortController.abort();\n }\n}\n\n/**\n * Internal helper that performs the actual HTTP request and converts the\n * response into audio frames. It inherits from `tts.ChunkedStream` to\n * integrate with LiveKit's event and cancellation framework.\n *\n * This implementation streams audio chunks as they arrive from the Baseten\n * model endpoint, processing them incrementally instead of waiting for the\n * complete response.\n */\nexport class ChunkedStream extends tts.ChunkedStream {\n label = 'baseten.ChunkedStream';\n private readonly opts: BasetenTTSOptions;\n\n constructor(\n tts: TTS,\n text: string,\n opts: BasetenTTSOptions,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ) {\n super(text, tts, connOptions, abortSignal);\n this.opts = opts;\n }\n\n /**\n * Execute the synthesis request. This method is automatically invoked\n * by the base class when the stream starts. It performs a POST request\n * to the configured `modelEndpoint` with the input text and optional\n * parameters. Audio chunks are streamed as they arrive and transformed\n * into a sequence of `AudioFrame` objects that are enqueued immediately\n * for playback.\n */\n protected async run() {\n const { apiKey, modelEndpoint, voice, language, temperature, maxTokens } = this.opts;\n const payload: Record<string, unknown> = {\n prompt: this.inputText,\n };\n if (voice) payload.voice = voice;\n if (language) payload.language = language;\n if (temperature !== undefined) payload.temperature = temperature;\n if (maxTokens !== undefined) payload.max_tokens = maxTokens;\n\n const headers: Record<string, string> = {\n Authorization: `Api-Key ${apiKey}`,\n 'Content-Type': 'application/json',\n };\n\n const response = await fetch(modelEndpoint, {\n method: 'POST',\n headers,\n body: JSON.stringify(payload),\n signal: this.abortSignal,\n });\n\n if (!response.ok) {\n let errText: string;\n try {\n errText = await response.text();\n } catch {\n errText = response.statusText;\n }\n throw new Error(`Baseten TTS request failed: ${response.status} ${errText}`);\n }\n\n // Stream the response body as chunks arrive\n if (!response.body) {\n throw new Error('Response body is not available for streaming');\n }\n\n const requestId = shortuuid();\n const audioByteStream = new AudioByteStream(24000, 1);\n const reader = response.body.getReader();\n\n try {\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 // waitForAbort internally sets up an abort listener on the abort signal\n // we need to put it outside loop to avoid constant re-registration of the listener\n const abortPromise = waitForAbort(this.abortSignal);\n\n while (!this.abortSignal.aborted) {\n const result = await Promise.race([reader.read(), abortPromise]);\n\n if (result === undefined) break; // aborted\n\n const { done, value } = result;\n\n if (done) {\n break;\n }\n\n // Process the chunk and convert to audio frames\n // Convert Uint8Array to ArrayBuffer for AudioByteStream\n const frames = audioByteStream.write(value.buffer);\n\n for (const frame of frames) {\n sendLastFrame(requestId, false);\n lastFrame = frame;\n }\n }\n\n // Send the final frame\n sendLastFrame(requestId, true);\n } finally {\n reader.releaseLock();\n this.queue.close();\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBAMO;AAIP,MAAM,oBAAgD;AAAA,EACpD,OAAO;AAAA,EACP,UAAU;AAAA,EACV,aAAa;AACf;AAKO,MAAM,YAAY,kBAAI,IAAI;AAAA,EACvB;AAAA,EACR,QAAQ;AAAA,EACA,kBAAkB,IAAI,gBAAgB;AAAA,EAC9C,YAAY,OAAmC,CAAC,GAAG;AAMjD,UAAM,MAAO,GAAG,EAAE,WAAW,MAAM,CAAC;AAGpC,UAAM,SAAS,KAAK,UAAU,QAAQ,IAAI;AAC1C,UAAM,gBAAgB,KAAK,iBAAiB,QAAQ,IAAI;AAExD,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,eAAe;AAClB,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,OAAO;AAAA,MACV,GAAG;AAAA,MACH,GAAG;AAAA,MACH;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA,EAEA,cAAc,MAAoE;AAChF,SAAK,OAAO;AAAA,MACV,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,IACL;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,WACE,MACA,aACA,aACe;AACf,UAAM,SAAS,cACX,YAAY,IAAI,CAAC,aAAa,KAAK,gBAAgB,MAAM,CAAC,IAC1D,KAAK,gBAAgB;AACzB,WAAO,IAAI,cAAc,MAAM,MAAM,KAAK,MAAM,aAAa,MAAM;AAAA,EACrE;AAAA,EAEA,SAA+B;AAC7B,UAAM,IAAI,MAAM,2CAA2C;AAAA,EAC7D;AAAA,EAEA,MAAM,QAAuB;AAC3B,SAAK,gBAAgB,MAAM;AAAA,EAC7B;AACF;AAWO,MAAM,sBAAsB,kBAAI,cAAc;AAAA,EACnD,QAAQ;AAAA,EACS;AAAA,EAEjB,YACEA,MACA,MACA,MACA,aACA,aACA;AACA,UAAM,MAAMA,MAAK,aAAa,WAAW;AACzC,SAAK,OAAO;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUA,MAAgB,MAAM;AACpB,UAAM,EAAE,QAAQ,eAAe,OAAO,UAAU,aAAa,UAAU,IAAI,KAAK;AAChF,UAAM,UAAmC;AAAA,MACvC,QAAQ,KAAK;AAAA,IACf;AACA,QAAI,MAAO,SAAQ,QAAQ;AAC3B,QAAI,SAAU,SAAQ,WAAW;AACjC,QAAI,gBAAgB,OAAW,SAAQ,cAAc;AACrD,QAAI,cAAc,OAAW,SAAQ,aAAa;AAElD,UAAM,UAAkC;AAAA,MACtC,eAAe,WAAW,MAAM;AAAA,MAChC,gBAAgB;AAAA,IAClB;AAEA,UAAM,WAAW,MAAM,MAAM,eAAe;AAAA,MAC1C,QAAQ;AAAA,MACR;AAAA,MACA,MAAM,KAAK,UAAU,OAAO;AAAA,MAC5B,QAAQ,KAAK;AAAA,IACf,CAAC;AAED,QAAI,CAAC,SAAS,IAAI;AAChB,UAAI;AACJ,UAAI;AACF,kBAAU,MAAM,SAAS,KAAK;AAAA,MAChC,QAAQ;AACN,kBAAU,SAAS;AAAA,MACrB;AACA,YAAM,IAAI,MAAM,+BAA+B,SAAS,MAAM,IAAI,OAAO,EAAE;AAAA,IAC7E;AAGA,QAAI,CAAC,SAAS,MAAM;AAClB,YAAM,IAAI,MAAM,8CAA8C;AAAA,IAChE;AAEA,UAAM,gBAAY,yBAAU;AAC5B,UAAM,kBAAkB,IAAI,8BAAgB,MAAO,CAAC;AACpD,UAAM,SAAS,SAAS,KAAK,UAAU;AAEvC,QAAI;AACF,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;AAIA,YAAM,mBAAe,4BAAa,KAAK,WAAW;AAElD,aAAO,CAAC,KAAK,YAAY,SAAS;AAChC,cAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,OAAO,KAAK,GAAG,YAAY,CAAC;AAE/D,YAAI,WAAW,OAAW;AAE1B,cAAM,EAAE,MAAM,MAAM,IAAI;AAExB,YAAI,MAAM;AACR;AAAA,QACF;AAIA,cAAM,SAAS,gBAAgB,MAAM,MAAM,MAAM;AAEjD,mBAAW,SAAS,QAAQ;AAC1B,wBAAc,WAAW,KAAK;AAC9B,sBAAY;AAAA,QACd;AAAA,MACF;AAGA,oBAAc,WAAW,IAAI;AAAA,IAC/B,UAAE;AACA,aAAO,YAAY;AACnB,WAAK,MAAM,MAAM;AAAA,IACnB;AAAA,EACF;AACF;","names":["tts"]}
|
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,EACL,KAAK,iBAAiB,EAGtB,GAAG,EAEJ,MAAM,iBAAiB,CAAC;AAEzB,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,YAAY,CAAC;AAQpD;;GAEG;AACH,qBAAa,GAAI,SAAQ,GAAG,CAAC,GAAG;IAC9B,OAAO,CAAC,IAAI,CAAoB;IAChC,KAAK,SAAiB;IACtB,OAAO,CAAC,eAAe,CAAyB;gBACpC,IAAI,GAAE,OAAO,CAAC,iBAAiB,CAAM;IA+BjD,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,iBAAiB,EAAE,QAAQ,GAAG,eAAe,CAAC,CAAC;IAOhF;;;;;OAKG;IACH,UAAU,CACR,IAAI,EAAE,MAAM,EACZ,WAAW,CAAC,EAAE,iBAAiB,EAC/B,WAAW,CAAC,EAAE,WAAW,GACxB,aAAa;
|
|
1
|
+
{"version":3,"file":"tts.d.ts","sourceRoot":"","sources":["../src/tts.ts"],"names":[],"mappings":"AAGA,OAAO,EACL,KAAK,iBAAiB,EAGtB,GAAG,EAEJ,MAAM,iBAAiB,CAAC;AAEzB,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,YAAY,CAAC;AAQpD;;GAEG;AACH,qBAAa,GAAI,SAAQ,GAAG,CAAC,GAAG;IAC9B,OAAO,CAAC,IAAI,CAAoB;IAChC,KAAK,SAAiB;IACtB,OAAO,CAAC,eAAe,CAAyB;gBACpC,IAAI,GAAE,OAAO,CAAC,iBAAiB,CAAM;IA+BjD,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,iBAAiB,EAAE,QAAQ,GAAG,eAAe,CAAC,CAAC;IAOhF;;;;;OAKG;IACH,UAAU,CACR,IAAI,EAAE,MAAM,EACZ,WAAW,CAAC,EAAE,iBAAiB,EAC/B,WAAW,CAAC,EAAE,WAAW,GACxB,aAAa;IAOhB,MAAM,IAAI,GAAG,CAAC,gBAAgB;IAIxB,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;CAG7B;AAED;;;;;;;;GAQG;AACH,qBAAa,aAAc,SAAQ,GAAG,CAAC,aAAa;IAClD,KAAK,SAA2B;IAChC,OAAO,CAAC,QAAQ,CAAC,IAAI,CAAoB;gBAGvC,GAAG,EAAE,GAAG,EACR,IAAI,EAAE,MAAM,EACZ,IAAI,EAAE,iBAAiB,EACvB,WAAW,CAAC,EAAE,iBAAiB,EAC/B,WAAW,CAAC,EAAE,WAAW;IAM3B;;;;;;;OAOG;cACa,GAAG;CAkFpB"}
|
package/dist/tts.js
CHANGED
|
@@ -47,7 +47,8 @@ class TTS extends tts.TTS {
|
|
|
47
47
|
* call {@link ChunkedStream.stop} on the returned object.
|
|
48
48
|
*/
|
|
49
49
|
synthesize(text, connOptions, abortSignal) {
|
|
50
|
-
|
|
50
|
+
const signal = abortSignal ? AbortSignal.any([abortSignal, this.abortController.signal]) : this.abortController.signal;
|
|
51
|
+
return new ChunkedStream(this, text, this.opts, connOptions, signal);
|
|
51
52
|
}
|
|
52
53
|
stream() {
|
|
53
54
|
throw new Error("Streaming is not supported on Baseten TTS");
|
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 {\n type APIConnectOptions,\n AudioByteStream,\n shortuuid,\n tts,\n waitForAbort,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { BasetenTTSOptions } from './types.js';\n\nconst defaultTTSOptions: Partial<BasetenTTSOptions> = {\n voice: 'tara',\n language: 'en',\n temperature: 0.6,\n};\n\n/**\n * Baseten TTS implementation (streaming, 24kHz mono)\n */\nexport class TTS extends tts.TTS {\n private opts: BasetenTTSOptions;\n label = 'baseten.TTS';\n private abortController = new AbortController();\n constructor(opts: Partial<BasetenTTSOptions> = {}) {\n /**\n * Baseten audio is 24kHz mono.\n * The Orpheus model generates audio chunks that are processed as they arrive,\n * which reduces latency and improves agent responsiveness.\n */\n super(24000, 1, { streaming: false });\n\n // Apply defaults and environment fallbacks.\n const apiKey = opts.apiKey ?? process.env.BASETEN_API_KEY;\n const modelEndpoint = opts.modelEndpoint ?? process.env.BASETEN_MODEL_ENDPOINT;\n\n if (!apiKey) {\n throw new Error(\n 'Baseten API key is required, either pass it as `apiKey` or set $BASETEN_API_KEY',\n );\n }\n if (!modelEndpoint) {\n throw new Error(\n 'Baseten model endpoint is required, either pass it as `modelEndpoint` or set $BASETEN_MODEL_ENDPOINT',\n );\n }\n\n this.opts = {\n ...defaultTTSOptions,\n ...opts,\n apiKey,\n modelEndpoint,\n } as BasetenTTSOptions;\n }\n\n updateOptions(opts: Partial<Omit<BasetenTTSOptions, 'apiKey' | 'modelEndpoint'>>) {\n this.opts = {\n ...this.opts,\n ...opts,\n } as BasetenTTSOptions;\n }\n\n /**\n * Synthesize speech for a given piece of text. Returns a `ChunkedStream`\n * which will asynchronously fetch audio from Baseten and push frames into\n * LiveKit's playback pipeline. If you need to cancel synthesis you can\n * call {@link ChunkedStream.stop} on the returned object.\n */\n synthesize(\n text: string,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ): ChunkedStream {\n return new ChunkedStream(this, text, this.opts, connOptions,
|
|
1
|
+
{"version":3,"sources":["../src/tts.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n AudioByteStream,\n shortuuid,\n tts,\n waitForAbort,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { BasetenTTSOptions } from './types.js';\n\nconst defaultTTSOptions: Partial<BasetenTTSOptions> = {\n voice: 'tara',\n language: 'en',\n temperature: 0.6,\n};\n\n/**\n * Baseten TTS implementation (streaming, 24kHz mono)\n */\nexport class TTS extends tts.TTS {\n private opts: BasetenTTSOptions;\n label = 'baseten.TTS';\n private abortController = new AbortController();\n constructor(opts: Partial<BasetenTTSOptions> = {}) {\n /**\n * Baseten audio is 24kHz mono.\n * The Orpheus model generates audio chunks that are processed as they arrive,\n * which reduces latency and improves agent responsiveness.\n */\n super(24000, 1, { streaming: false });\n\n // Apply defaults and environment fallbacks.\n const apiKey = opts.apiKey ?? process.env.BASETEN_API_KEY;\n const modelEndpoint = opts.modelEndpoint ?? process.env.BASETEN_MODEL_ENDPOINT;\n\n if (!apiKey) {\n throw new Error(\n 'Baseten API key is required, either pass it as `apiKey` or set $BASETEN_API_KEY',\n );\n }\n if (!modelEndpoint) {\n throw new Error(\n 'Baseten model endpoint is required, either pass it as `modelEndpoint` or set $BASETEN_MODEL_ENDPOINT',\n );\n }\n\n this.opts = {\n ...defaultTTSOptions,\n ...opts,\n apiKey,\n modelEndpoint,\n } as BasetenTTSOptions;\n }\n\n updateOptions(opts: Partial<Omit<BasetenTTSOptions, 'apiKey' | 'modelEndpoint'>>) {\n this.opts = {\n ...this.opts,\n ...opts,\n } as BasetenTTSOptions;\n }\n\n /**\n * Synthesize speech for a given piece of text. Returns a `ChunkedStream`\n * which will asynchronously fetch audio from Baseten and push frames into\n * LiveKit's playback pipeline. If you need to cancel synthesis you can\n * call {@link ChunkedStream.stop} on the returned object.\n */\n synthesize(\n text: string,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ): ChunkedStream {\n const signal = abortSignal\n ? AbortSignal.any([abortSignal, this.abortController.signal])\n : this.abortController.signal;\n return new ChunkedStream(this, text, this.opts, connOptions, signal);\n }\n\n stream(): tts.SynthesizeStream {\n throw new Error('Streaming is not supported on Baseten TTS');\n }\n\n async close(): Promise<void> {\n this.abortController.abort();\n }\n}\n\n/**\n * Internal helper that performs the actual HTTP request and converts the\n * response into audio frames. It inherits from `tts.ChunkedStream` to\n * integrate with LiveKit's event and cancellation framework.\n *\n * This implementation streams audio chunks as they arrive from the Baseten\n * model endpoint, processing them incrementally instead of waiting for the\n * complete response.\n */\nexport class ChunkedStream extends tts.ChunkedStream {\n label = 'baseten.ChunkedStream';\n private readonly opts: BasetenTTSOptions;\n\n constructor(\n tts: TTS,\n text: string,\n opts: BasetenTTSOptions,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ) {\n super(text, tts, connOptions, abortSignal);\n this.opts = opts;\n }\n\n /**\n * Execute the synthesis request. This method is automatically invoked\n * by the base class when the stream starts. It performs a POST request\n * to the configured `modelEndpoint` with the input text and optional\n * parameters. Audio chunks are streamed as they arrive and transformed\n * into a sequence of `AudioFrame` objects that are enqueued immediately\n * for playback.\n */\n protected async run() {\n const { apiKey, modelEndpoint, voice, language, temperature, maxTokens } = this.opts;\n const payload: Record<string, unknown> = {\n prompt: this.inputText,\n };\n if (voice) payload.voice = voice;\n if (language) payload.language = language;\n if (temperature !== undefined) payload.temperature = temperature;\n if (maxTokens !== undefined) payload.max_tokens = maxTokens;\n\n const headers: Record<string, string> = {\n Authorization: `Api-Key ${apiKey}`,\n 'Content-Type': 'application/json',\n };\n\n const response = await fetch(modelEndpoint, {\n method: 'POST',\n headers,\n body: JSON.stringify(payload),\n signal: this.abortSignal,\n });\n\n if (!response.ok) {\n let errText: string;\n try {\n errText = await response.text();\n } catch {\n errText = response.statusText;\n }\n throw new Error(`Baseten TTS request failed: ${response.status} ${errText}`);\n }\n\n // Stream the response body as chunks arrive\n if (!response.body) {\n throw new Error('Response body is not available for streaming');\n }\n\n const requestId = shortuuid();\n const audioByteStream = new AudioByteStream(24000, 1);\n const reader = response.body.getReader();\n\n try {\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 // waitForAbort internally sets up an abort listener on the abort signal\n // we need to put it outside loop to avoid constant re-registration of the listener\n const abortPromise = waitForAbort(this.abortSignal);\n\n while (!this.abortSignal.aborted) {\n const result = await Promise.race([reader.read(), abortPromise]);\n\n if (result === undefined) break; // aborted\n\n const { done, value } = result;\n\n if (done) {\n break;\n }\n\n // Process the chunk and convert to audio frames\n // Convert Uint8Array to ArrayBuffer for AudioByteStream\n const frames = audioByteStream.write(value.buffer);\n\n for (const frame of frames) {\n sendLastFrame(requestId, false);\n lastFrame = frame;\n }\n }\n\n // Send the final frame\n sendLastFrame(requestId, true);\n } finally {\n reader.releaseLock();\n this.queue.close();\n }\n }\n}\n"],"mappings":"AAGA;AAAA,EAEE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAIP,MAAM,oBAAgD;AAAA,EACpD,OAAO;AAAA,EACP,UAAU;AAAA,EACV,aAAa;AACf;AAKO,MAAM,YAAY,IAAI,IAAI;AAAA,EACvB;AAAA,EACR,QAAQ;AAAA,EACA,kBAAkB,IAAI,gBAAgB;AAAA,EAC9C,YAAY,OAAmC,CAAC,GAAG;AAMjD,UAAM,MAAO,GAAG,EAAE,WAAW,MAAM,CAAC;AAGpC,UAAM,SAAS,KAAK,UAAU,QAAQ,IAAI;AAC1C,UAAM,gBAAgB,KAAK,iBAAiB,QAAQ,IAAI;AAExD,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,eAAe;AAClB,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,OAAO;AAAA,MACV,GAAG;AAAA,MACH,GAAG;AAAA,MACH;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA,EAEA,cAAc,MAAoE;AAChF,SAAK,OAAO;AAAA,MACV,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,IACL;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,WACE,MACA,aACA,aACe;AACf,UAAM,SAAS,cACX,YAAY,IAAI,CAAC,aAAa,KAAK,gBAAgB,MAAM,CAAC,IAC1D,KAAK,gBAAgB;AACzB,WAAO,IAAI,cAAc,MAAM,MAAM,KAAK,MAAM,aAAa,MAAM;AAAA,EACrE;AAAA,EAEA,SAA+B;AAC7B,UAAM,IAAI,MAAM,2CAA2C;AAAA,EAC7D;AAAA,EAEA,MAAM,QAAuB;AAC3B,SAAK,gBAAgB,MAAM;AAAA,EAC7B;AACF;AAWO,MAAM,sBAAsB,IAAI,cAAc;AAAA,EACnD,QAAQ;AAAA,EACS;AAAA,EAEjB,YACEA,MACA,MACA,MACA,aACA,aACA;AACA,UAAM,MAAMA,MAAK,aAAa,WAAW;AACzC,SAAK,OAAO;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUA,MAAgB,MAAM;AACpB,UAAM,EAAE,QAAQ,eAAe,OAAO,UAAU,aAAa,UAAU,IAAI,KAAK;AAChF,UAAM,UAAmC;AAAA,MACvC,QAAQ,KAAK;AAAA,IACf;AACA,QAAI,MAAO,SAAQ,QAAQ;AAC3B,QAAI,SAAU,SAAQ,WAAW;AACjC,QAAI,gBAAgB,OAAW,SAAQ,cAAc;AACrD,QAAI,cAAc,OAAW,SAAQ,aAAa;AAElD,UAAM,UAAkC;AAAA,MACtC,eAAe,WAAW,MAAM;AAAA,MAChC,gBAAgB;AAAA,IAClB;AAEA,UAAM,WAAW,MAAM,MAAM,eAAe;AAAA,MAC1C,QAAQ;AAAA,MACR;AAAA,MACA,MAAM,KAAK,UAAU,OAAO;AAAA,MAC5B,QAAQ,KAAK;AAAA,IACf,CAAC;AAED,QAAI,CAAC,SAAS,IAAI;AAChB,UAAI;AACJ,UAAI;AACF,kBAAU,MAAM,SAAS,KAAK;AAAA,MAChC,QAAQ;AACN,kBAAU,SAAS;AAAA,MACrB;AACA,YAAM,IAAI,MAAM,+BAA+B,SAAS,MAAM,IAAI,OAAO,EAAE;AAAA,IAC7E;AAGA,QAAI,CAAC,SAAS,MAAM;AAClB,YAAM,IAAI,MAAM,8CAA8C;AAAA,IAChE;AAEA,UAAM,YAAY,UAAU;AAC5B,UAAM,kBAAkB,IAAI,gBAAgB,MAAO,CAAC;AACpD,UAAM,SAAS,SAAS,KAAK,UAAU;AAEvC,QAAI;AACF,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;AAIA,YAAM,eAAe,aAAa,KAAK,WAAW;AAElD,aAAO,CAAC,KAAK,YAAY,SAAS;AAChC,cAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,OAAO,KAAK,GAAG,YAAY,CAAC;AAE/D,YAAI,WAAW,OAAW;AAE1B,cAAM,EAAE,MAAM,MAAM,IAAI;AAExB,YAAI,MAAM;AACR;AAAA,QACF;AAIA,cAAM,SAAS,gBAAgB,MAAM,MAAM,MAAM;AAEjD,mBAAW,SAAS,QAAQ;AAC1B,wBAAc,WAAW,KAAK;AAC9B,sBAAY;AAAA,QACd;AAAA,MACF;AAGA,oBAAc,WAAW,IAAI;AAAA,IAC/B,UAAE;AACA,aAAO,YAAY;AACnB,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-baseten",
|
|
3
|
-
"version": "1.4.
|
|
3
|
+
"version": "1.4.2",
|
|
4
4
|
"description": "Baseten plugin for LiveKit Node Agents",
|
|
5
5
|
"main": "dist/index.js",
|
|
6
6
|
"require": "dist/index.cjs",
|
|
@@ -40,9 +40,9 @@
|
|
|
40
40
|
"@types/ws": "^8.5.10",
|
|
41
41
|
"tsx": "^4.7.0",
|
|
42
42
|
"typescript": "^5.9.3",
|
|
43
|
-
"@livekit/agents": "1.4.
|
|
44
|
-
"@livekit/agents-plugin-silero": "1.4.
|
|
45
|
-
"@livekit/agents-plugins-test": "1.4.
|
|
43
|
+
"@livekit/agents": "1.4.2",
|
|
44
|
+
"@livekit/agents-plugin-silero": "1.4.2",
|
|
45
|
+
"@livekit/agents-plugins-test": "1.4.2"
|
|
46
46
|
},
|
|
47
47
|
"dependencies": {
|
|
48
48
|
"dotenv": "^17.2.3",
|
|
@@ -51,7 +51,7 @@
|
|
|
51
51
|
},
|
|
52
52
|
"peerDependencies": {
|
|
53
53
|
"@livekit/rtc-node": "^0.13.27",
|
|
54
|
-
"@livekit/agents": "1.4.
|
|
54
|
+
"@livekit/agents": "1.4.2"
|
|
55
55
|
},
|
|
56
56
|
"scripts": {
|
|
57
57
|
"build": "tsup --onSuccess \"pnpm build:types\"",
|
package/src/tts.ts
CHANGED
|
@@ -73,7 +73,10 @@ export class TTS extends tts.TTS {
|
|
|
73
73
|
connOptions?: APIConnectOptions,
|
|
74
74
|
abortSignal?: AbortSignal,
|
|
75
75
|
): ChunkedStream {
|
|
76
|
-
|
|
76
|
+
const signal = abortSignal
|
|
77
|
+
? AbortSignal.any([abortSignal, this.abortController.signal])
|
|
78
|
+
: this.abortController.signal;
|
|
79
|
+
return new ChunkedStream(this, text, this.opts, connOptions, signal);
|
|
77
80
|
}
|
|
78
81
|
|
|
79
82
|
stream(): tts.SynthesizeStream {
|