@livekit/agents-plugin-openai 1.0.50 → 1.1.0

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.
Files changed (118) hide show
  1. package/dist/index.cjs +5 -2
  2. package/dist/index.cjs.map +1 -1
  3. package/dist/index.d.cts +1 -0
  4. package/dist/index.d.ts +1 -0
  5. package/dist/index.d.ts.map +1 -1
  6. package/dist/index.js +4 -2
  7. package/dist/index.js.map +1 -1
  8. package/dist/llm.cjs +8 -0
  9. package/dist/llm.cjs.map +1 -1
  10. package/dist/llm.d.cts +1 -0
  11. package/dist/llm.d.ts +1 -0
  12. package/dist/llm.d.ts.map +1 -1
  13. package/dist/llm.js +8 -0
  14. package/dist/llm.js.map +1 -1
  15. package/dist/llm.test.cjs +31 -16
  16. package/dist/llm.test.cjs.map +1 -1
  17. package/dist/llm.test.js +32 -17
  18. package/dist/llm.test.js.map +1 -1
  19. package/dist/realtime/api_proto.cjs.map +1 -1
  20. package/dist/realtime/api_proto.d.cts +7 -3
  21. package/dist/realtime/api_proto.d.ts +7 -3
  22. package/dist/realtime/api_proto.d.ts.map +1 -1
  23. package/dist/realtime/api_proto.js.map +1 -1
  24. package/dist/realtime/realtime_model.cjs +46 -22
  25. package/dist/realtime/realtime_model.cjs.map +1 -1
  26. package/dist/realtime/realtime_model.d.cts +2 -1
  27. package/dist/realtime/realtime_model.d.ts +2 -1
  28. package/dist/realtime/realtime_model.d.ts.map +1 -1
  29. package/dist/realtime/realtime_model.js +46 -22
  30. package/dist/realtime/realtime_model.js.map +1 -1
  31. package/dist/realtime/realtime_model.test.cjs +104 -14
  32. package/dist/realtime/realtime_model.test.cjs.map +1 -1
  33. package/dist/realtime/realtime_model.test.js +104 -14
  34. package/dist/realtime/realtime_model.test.js.map +1 -1
  35. package/dist/realtime/realtime_model_beta.cjs +40 -22
  36. package/dist/realtime/realtime_model_beta.cjs.map +1 -1
  37. package/dist/realtime/realtime_model_beta.d.ts.map +1 -1
  38. package/dist/realtime/realtime_model_beta.js +40 -22
  39. package/dist/realtime/realtime_model_beta.js.map +1 -1
  40. package/dist/responses/llm.cjs +71 -16
  41. package/dist/responses/llm.cjs.map +1 -1
  42. package/dist/responses/llm.d.cts +10 -25
  43. package/dist/responses/llm.d.ts +10 -25
  44. package/dist/responses/llm.d.ts.map +1 -1
  45. package/dist/responses/llm.js +71 -14
  46. package/dist/responses/llm.js.map +1 -1
  47. package/dist/responses/llm.test.cjs +32 -17
  48. package/dist/responses/llm.test.cjs.map +1 -1
  49. package/dist/responses/llm.test.js +33 -18
  50. package/dist/responses/llm.test.js.map +1 -1
  51. package/dist/stt.cjs +18 -3
  52. package/dist/stt.cjs.map +1 -1
  53. package/dist/stt.d.cts +2 -0
  54. package/dist/stt.d.ts +2 -0
  55. package/dist/stt.d.ts.map +1 -1
  56. package/dist/stt.js +19 -4
  57. package/dist/stt.js.map +1 -1
  58. package/dist/stt.test.cjs +11 -3
  59. package/dist/stt.test.cjs.map +1 -1
  60. package/dist/stt.test.js +12 -4
  61. package/dist/stt.test.js.map +1 -1
  62. package/dist/tts.cjs +11 -0
  63. package/dist/tts.cjs.map +1 -1
  64. package/dist/tts.d.cts +2 -0
  65. package/dist/tts.d.ts +2 -0
  66. package/dist/tts.d.ts.map +1 -1
  67. package/dist/tts.js +11 -0
  68. package/dist/tts.js.map +1 -1
  69. package/dist/tts.test.cjs +11 -3
  70. package/dist/tts.test.cjs.map +1 -1
  71. package/dist/tts.test.js +12 -4
  72. package/dist/tts.test.js.map +1 -1
  73. package/dist/ws/index.cjs +29 -0
  74. package/dist/ws/index.cjs.map +1 -0
  75. package/dist/ws/index.d.cts +3 -0
  76. package/dist/ws/index.d.ts +3 -0
  77. package/dist/ws/index.d.ts.map +1 -0
  78. package/dist/ws/index.js +5 -0
  79. package/dist/ws/index.js.map +1 -0
  80. package/dist/ws/llm.cjs +502 -0
  81. package/dist/ws/llm.cjs.map +1 -0
  82. package/dist/ws/llm.d.cts +74 -0
  83. package/dist/ws/llm.d.ts +74 -0
  84. package/dist/ws/llm.d.ts.map +1 -0
  85. package/dist/ws/llm.js +485 -0
  86. package/dist/ws/llm.js.map +1 -0
  87. package/dist/ws/llm.test.cjs +26 -0
  88. package/dist/ws/llm.test.cjs.map +1 -0
  89. package/dist/ws/llm.test.d.cts +2 -0
  90. package/dist/ws/llm.test.d.ts +2 -0
  91. package/dist/ws/llm.test.d.ts.map +1 -0
  92. package/dist/ws/llm.test.js +25 -0
  93. package/dist/ws/llm.test.js.map +1 -0
  94. package/dist/ws/types.cjs +128 -0
  95. package/dist/ws/types.cjs.map +1 -0
  96. package/dist/ws/types.d.cts +167 -0
  97. package/dist/ws/types.d.ts +167 -0
  98. package/dist/ws/types.d.ts.map +1 -0
  99. package/dist/ws/types.js +95 -0
  100. package/dist/ws/types.js.map +1 -0
  101. package/package.json +6 -5
  102. package/src/index.ts +1 -0
  103. package/src/llm.test.ts +31 -17
  104. package/src/llm.ts +9 -0
  105. package/src/realtime/api_proto.ts +8 -2
  106. package/src/realtime/realtime_model.test.ts +129 -14
  107. package/src/realtime/realtime_model.ts +51 -26
  108. package/src/realtime/realtime_model_beta.ts +42 -25
  109. package/src/responses/llm.test.ts +32 -18
  110. package/src/responses/llm.ts +105 -19
  111. package/src/stt.test.ts +12 -4
  112. package/src/stt.ts +21 -4
  113. package/src/tts.test.ts +12 -4
  114. package/src/tts.ts +13 -0
  115. package/src/ws/index.ts +17 -0
  116. package/src/ws/llm.test.ts +30 -0
  117. package/src/ws/llm.ts +665 -0
  118. package/src/ws/types.ts +131 -0
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 { type AudioBuffer, mergeFrames, stt } from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { OpenAI } from 'openai';\nimport type { GroqAudioModels, WhisperModels } from './models.js';\n\nexport interface STTOptions {\n apiKey?: string;\n language: string;\n prompt?: string;\n detectLanguage: boolean;\n model: WhisperModels | string;\n baseURL?: string;\n client?: OpenAI;\n}\n\nconst defaultSTTOptions: STTOptions = {\n apiKey: process.env.OPENAI_API_KEY,\n language: 'en',\n detectLanguage: false,\n model: 'whisper-1',\n};\n\nexport class STT extends stt.STT {\n #opts: STTOptions;\n #client: OpenAI;\n label = 'openai.STT';\n\n /**\n * Create a new instance of OpenAI STT.\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<STTOptions> = defaultSTTOptions) {\n super({ streaming: false, interimResults: false, alignedTranscript: false });\n\n this.#opts = { ...defaultSTTOptions, ...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: this.#opts.baseURL,\n apiKey: this.#opts.apiKey,\n });\n }\n\n /**\n * Create a new instance of Groq STT.\n *\n * @remarks\n * `apiKey` must be set to your Groq API key, either using the argument or by setting the\n * `GROQ_API_KEY` environment variable.\n */\n static withGroq(\n opts: Partial<{\n model: string | GroqAudioModels;\n apiKey?: string;\n baseURL?: string;\n client: OpenAI;\n language: string;\n detectLanguage: boolean;\n }> = {},\n ): STT {\n opts.apiKey = opts.apiKey || process.env.GROQ_API_KEY;\n if (opts.apiKey === undefined) {\n throw new Error('Groq API key is required, whether as an argument or as $GROQ_API_KEY');\n }\n\n return new STT({\n model: 'whisper-large-v3-turbo',\n baseURL: 'https://api.groq.com/openai/v1',\n ...opts,\n });\n }\n\n /**\n * Create a new instance of OVHcloud AI Endpoints STT.\n *\n * @remarks\n * `apiKey` must be set to your OVHcloud AI Endpoints API key, either using the argument or by setting the\n * `OVHCLOUD_API_KEY` environment variable.\n */\n static withOVHcloud(\n opts: Partial<{\n model: string;\n apiKey?: string;\n baseURL?: string;\n client: OpenAI;\n language: string;\n detectLanguage: boolean;\n }> = {},\n ): STT {\n opts.apiKey = opts.apiKey || process.env.OVHCLOUD_API_KEY;\n if (opts.apiKey === undefined) {\n throw new Error(\n 'OVHcloud AI Endpoints API key is required, whether as an argument or as $OVHCLOUD_API_KEY',\n );\n }\n\n return new STT({\n model: 'whisper-large-v3-turbo',\n baseURL: 'https://oai.endpoints.kepler.ai.cloud.ovh.net/v1',\n ...opts,\n });\n }\n\n #sanitizeOptions(language?: string): STTOptions {\n if (language) {\n return { ...this.#opts, language };\n } else {\n return this.#opts;\n }\n }\n\n #createWav(frame: AudioFrame): Buffer {\n const bitsPerSample = 16;\n const byteRate = (frame.sampleRate * frame.channels * bitsPerSample) / 8;\n const blockAlign = (frame.channels * bitsPerSample) / 8;\n\n const header = Buffer.alloc(44);\n header.write('RIFF', 0);\n header.writeUInt32LE(36 + frame.data.byteLength, 4);\n header.write('WAVE', 8);\n header.write('fmt ', 12);\n header.writeUInt32LE(16, 16);\n header.writeUInt16LE(1, 20);\n header.writeUInt16LE(frame.channels, 22);\n header.writeUInt32LE(frame.sampleRate, 24);\n header.writeUInt32LE(byteRate, 28);\n header.writeUInt16LE(blockAlign, 32);\n header.writeUInt16LE(16, 34);\n header.write('data', 36);\n header.writeUInt32LE(frame.data.byteLength, 40);\n return Buffer.concat([header, Buffer.from(frame.data.buffer)]);\n }\n\n async _recognize(buffer: AudioBuffer, abortSignal?: AbortSignal): Promise<stt.SpeechEvent> {\n const config = this.#sanitizeOptions();\n buffer = mergeFrames(buffer);\n const wavBuffer = this.#createWav(buffer);\n const file = new File([new Uint8Array(wavBuffer)], 'audio.wav', { type: 'audio/wav' });\n\n const resp = await this.#client.audio.transcriptions.create(\n {\n file,\n model: this.#opts.model,\n language: config.language,\n prompt: config.prompt,\n response_format: 'json',\n },\n {\n signal: abortSignal,\n },\n );\n\n return {\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n alternatives: [\n {\n text: resp.text || '',\n language: config.language || '',\n startTime: 0,\n endTime: 0,\n confidence: 0,\n },\n ],\n };\n }\n\n /** This method throws an error; streaming is unsupported on OpenAI STT. */\n stream(): stt.SpeechStream {\n throw new Error('Streaming is not supported on OpenAI STT');\n }\n}\n"],"mappings":"AAGA,SAA2B,aAAa,WAAW;AAEnD,SAAS,cAAc;AAavB,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,gBAAgB;AAAA,EAChB,OAAO;AACT;AAEO,MAAM,YAAY,IAAI,IAAI;AAAA,EAC/B;AAAA,EACA;AAAA,EACA,QAAQ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASR,YAAY,OAA4B,mBAAmB;AACzD,UAAM,EAAE,WAAW,OAAO,gBAAgB,OAAO,mBAAmB,MAAM,CAAC;AAE3E,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,MAAM;AAAA,MACpB,QAAQ,KAAK,MAAM;AAAA,IACrB,CAAC;AAAA,EACL;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,OAAO,SACL,OAOK,CAAC,GACD;AACL,SAAK,SAAS,KAAK,UAAU,QAAQ,IAAI;AACzC,QAAI,KAAK,WAAW,QAAW;AAC7B,YAAM,IAAI,MAAM,sEAAsE;AAAA,IACxF;AAEA,WAAO,IAAI,IAAI;AAAA,MACb,OAAO;AAAA,MACP,SAAS;AAAA,MACT,GAAG;AAAA,IACL,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,OAAO,aACL,OAOK,CAAC,GACD;AACL,SAAK,SAAS,KAAK,UAAU,QAAQ,IAAI;AACzC,QAAI,KAAK,WAAW,QAAW;AAC7B,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,WAAO,IAAI,IAAI;AAAA,MACb,OAAO;AAAA,MACP,SAAS;AAAA,MACT,GAAG;AAAA,IACL,CAAC;AAAA,EACH;AAAA,EAEA,iBAAiB,UAA+B;AAC9C,QAAI,UAAU;AACZ,aAAO,EAAE,GAAG,KAAK,OAAO,SAAS;AAAA,IACnC,OAAO;AACL,aAAO,KAAK;AAAA,IACd;AAAA,EACF;AAAA,EAEA,WAAW,OAA2B;AACpC,UAAM,gBAAgB;AACtB,UAAM,WAAY,MAAM,aAAa,MAAM,WAAW,gBAAiB;AACvE,UAAM,aAAc,MAAM,WAAW,gBAAiB;AAEtD,UAAM,SAAS,OAAO,MAAM,EAAE;AAC9B,WAAO,MAAM,QAAQ,CAAC;AACtB,WAAO,cAAc,KAAK,MAAM,KAAK,YAAY,CAAC;AAClD,WAAO,MAAM,QAAQ,CAAC;AACtB,WAAO,MAAM,QAAQ,EAAE;AACvB,WAAO,cAAc,IAAI,EAAE;AAC3B,WAAO,cAAc,GAAG,EAAE;AAC1B,WAAO,cAAc,MAAM,UAAU,EAAE;AACvC,WAAO,cAAc,MAAM,YAAY,EAAE;AACzC,WAAO,cAAc,UAAU,EAAE;AACjC,WAAO,cAAc,YAAY,EAAE;AACnC,WAAO,cAAc,IAAI,EAAE;AAC3B,WAAO,MAAM,QAAQ,EAAE;AACvB,WAAO,cAAc,MAAM,KAAK,YAAY,EAAE;AAC9C,WAAO,OAAO,OAAO,CAAC,QAAQ,OAAO,KAAK,MAAM,KAAK,MAAM,CAAC,CAAC;AAAA,EAC/D;AAAA,EAEA,MAAM,WAAW,QAAqB,aAAqD;AACzF,UAAM,SAAS,KAAK,iBAAiB;AACrC,aAAS,YAAY,MAAM;AAC3B,UAAM,YAAY,KAAK,WAAW,MAAM;AACxC,UAAM,OAAO,IAAI,KAAK,CAAC,IAAI,WAAW,SAAS,CAAC,GAAG,aAAa,EAAE,MAAM,YAAY,CAAC;AAErF,UAAM,OAAO,MAAM,KAAK,QAAQ,MAAM,eAAe;AAAA,MACnD;AAAA,QACE;AAAA,QACA,OAAO,KAAK,MAAM;AAAA,QAClB,UAAU,OAAO;AAAA,QACjB,QAAQ,OAAO;AAAA,QACf,iBAAiB;AAAA,MACnB;AAAA,MACA;AAAA,QACE,QAAQ;AAAA,MACV;AAAA,IACF;AAEA,WAAO;AAAA,MACL,MAAM,IAAI,gBAAgB;AAAA,MAC1B,cAAc;AAAA,QACZ;AAAA,UACE,MAAM,KAAK,QAAQ;AAAA,UACnB,UAAU,OAAO,YAAY;AAAA,UAC7B,WAAW;AAAA,UACX,SAAS;AAAA,UACT,YAAY;AAAA,QACd;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA;AAAA,EAGA,SAA2B;AACzB,UAAM,IAAI,MAAM,0CAA0C;AAAA,EAC5D;AACF;","names":[]}
1
+ {"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { type AudioBuffer, mergeFrames, normalizeLanguage, stt } from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { OpenAI } from 'openai';\nimport type { GroqAudioModels, WhisperModels } from './models.js';\n\nexport interface STTOptions {\n apiKey?: string;\n language: string;\n prompt?: string;\n detectLanguage: boolean;\n model: WhisperModels | string;\n baseURL?: string;\n client?: OpenAI;\n}\n\nconst defaultSTTOptions: STTOptions = {\n apiKey: process.env.OPENAI_API_KEY,\n language: 'en',\n detectLanguage: false,\n model: 'whisper-1',\n};\n\nexport class STT extends stt.STT {\n #opts: STTOptions;\n #client: OpenAI;\n label = 'openai.STT';\n\n get model(): string {\n return this.#opts.model;\n }\n\n get provider(): string {\n try {\n const url = new URL(this.#client.baseURL);\n return url.host;\n } catch {\n return 'api.openai.com';\n }\n }\n\n /**\n * Create a new instance of OpenAI STT.\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<STTOptions> = defaultSTTOptions) {\n super({ streaming: false, interimResults: false, alignedTranscript: false });\n\n this.#opts = {\n ...defaultSTTOptions,\n ...opts,\n language: normalizeLanguage(opts.language ?? defaultSTTOptions.language),\n };\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: this.#opts.baseURL,\n apiKey: this.#opts.apiKey,\n });\n }\n\n /**\n * Create a new instance of Groq STT.\n *\n * @remarks\n * `apiKey` must be set to your Groq API key, either using the argument or by setting the\n * `GROQ_API_KEY` environment variable.\n */\n static withGroq(\n opts: Partial<{\n model: string | GroqAudioModels;\n apiKey?: string;\n baseURL?: string;\n client: OpenAI;\n language: string;\n detectLanguage: boolean;\n }> = {},\n ): STT {\n opts.apiKey = opts.apiKey || process.env.GROQ_API_KEY;\n if (opts.apiKey === undefined) {\n throw new Error('Groq API key is required, whether as an argument or as $GROQ_API_KEY');\n }\n\n return new STT({\n model: 'whisper-large-v3-turbo',\n baseURL: 'https://api.groq.com/openai/v1',\n ...opts,\n });\n }\n\n /**\n * Create a new instance of OVHcloud AI Endpoints STT.\n *\n * @remarks\n * `apiKey` must be set to your OVHcloud AI Endpoints API key, either using the argument or by setting the\n * `OVHCLOUD_API_KEY` environment variable.\n */\n static withOVHcloud(\n opts: Partial<{\n model: string;\n apiKey?: string;\n baseURL?: string;\n client: OpenAI;\n language: string;\n detectLanguage: boolean;\n }> = {},\n ): STT {\n opts.apiKey = opts.apiKey || process.env.OVHCLOUD_API_KEY;\n if (opts.apiKey === undefined) {\n throw new Error(\n 'OVHcloud AI Endpoints API key is required, whether as an argument or as $OVHCLOUD_API_KEY',\n );\n }\n\n return new STT({\n model: 'whisper-large-v3-turbo',\n baseURL: 'https://oai.endpoints.kepler.ai.cloud.ovh.net/v1',\n ...opts,\n });\n }\n\n #sanitizeOptions(language?: string): STTOptions {\n if (language) {\n return { ...this.#opts, language: normalizeLanguage(language) };\n } else {\n return this.#opts;\n }\n }\n\n #createWav(frame: AudioFrame): Buffer {\n const bitsPerSample = 16;\n const byteRate = (frame.sampleRate * frame.channels * bitsPerSample) / 8;\n const blockAlign = (frame.channels * bitsPerSample) / 8;\n\n const header = Buffer.alloc(44);\n header.write('RIFF', 0);\n header.writeUInt32LE(36 + frame.data.byteLength, 4);\n header.write('WAVE', 8);\n header.write('fmt ', 12);\n header.writeUInt32LE(16, 16);\n header.writeUInt16LE(1, 20);\n header.writeUInt16LE(frame.channels, 22);\n header.writeUInt32LE(frame.sampleRate, 24);\n header.writeUInt32LE(byteRate, 28);\n header.writeUInt16LE(blockAlign, 32);\n header.writeUInt16LE(16, 34);\n header.write('data', 36);\n header.writeUInt32LE(frame.data.byteLength, 40);\n return Buffer.concat([header, Buffer.from(frame.data.buffer)]);\n }\n\n async _recognize(buffer: AudioBuffer, abortSignal?: AbortSignal): Promise<stt.SpeechEvent> {\n const config = this.#sanitizeOptions();\n buffer = mergeFrames(buffer);\n const wavBuffer = this.#createWav(buffer);\n const file = new File([new Uint8Array(wavBuffer)], 'audio.wav', { type: 'audio/wav' });\n\n const resp = await this.#client.audio.transcriptions.create(\n {\n file,\n model: this.#opts.model,\n language: config.language,\n prompt: config.prompt,\n response_format: 'json',\n },\n {\n signal: abortSignal,\n },\n );\n\n return {\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n alternatives: [\n {\n text: resp.text || '',\n language: normalizeLanguage(config.language || ''),\n startTime: 0,\n endTime: 0,\n confidence: 0,\n },\n ],\n };\n }\n\n /** This method throws an error; streaming is unsupported on OpenAI STT. */\n stream(): stt.SpeechStream {\n throw new Error('Streaming is not supported on OpenAI STT');\n }\n}\n"],"mappings":"AAGA,SAA2B,aAAa,mBAAmB,WAAW;AAEtE,SAAS,cAAc;AAavB,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,gBAAgB;AAAA,EAChB,OAAO;AACT;AAEO,MAAM,YAAY,IAAI,IAAI;AAAA,EAC/B;AAAA,EACA;AAAA,EACA,QAAQ;AAAA,EAER,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAI,WAAmB;AACrB,QAAI;AACF,YAAM,MAAM,IAAI,IAAI,KAAK,QAAQ,OAAO;AACxC,aAAO,IAAI;AAAA,IACb,QAAQ;AACN,aAAO;AAAA,IACT;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,YAAY,OAA4B,mBAAmB;AACzD,UAAM,EAAE,WAAW,OAAO,gBAAgB,OAAO,mBAAmB,MAAM,CAAC;AAE3E,SAAK,QAAQ;AAAA,MACX,GAAG;AAAA,MACH,GAAG;AAAA,MACH,UAAU,kBAAkB,KAAK,YAAY,kBAAkB,QAAQ;AAAA,IACzE;AACA,QAAI,KAAK,MAAM,WAAW,QAAW;AACnC,YAAM,IAAI,MAAM,0EAA0E;AAAA,IAC5F;AAEA,SAAK,UACH,KAAK,MAAM,UACX,IAAI,OAAO;AAAA,MACT,SAAS,KAAK,MAAM;AAAA,MACpB,QAAQ,KAAK,MAAM;AAAA,IACrB,CAAC;AAAA,EACL;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,OAAO,SACL,OAOK,CAAC,GACD;AACL,SAAK,SAAS,KAAK,UAAU,QAAQ,IAAI;AACzC,QAAI,KAAK,WAAW,QAAW;AAC7B,YAAM,IAAI,MAAM,sEAAsE;AAAA,IACxF;AAEA,WAAO,IAAI,IAAI;AAAA,MACb,OAAO;AAAA,MACP,SAAS;AAAA,MACT,GAAG;AAAA,IACL,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,OAAO,aACL,OAOK,CAAC,GACD;AACL,SAAK,SAAS,KAAK,UAAU,QAAQ,IAAI;AACzC,QAAI,KAAK,WAAW,QAAW;AAC7B,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,WAAO,IAAI,IAAI;AAAA,MACb,OAAO;AAAA,MACP,SAAS;AAAA,MACT,GAAG;AAAA,IACL,CAAC;AAAA,EACH;AAAA,EAEA,iBAAiB,UAA+B;AAC9C,QAAI,UAAU;AACZ,aAAO,EAAE,GAAG,KAAK,OAAO,UAAU,kBAAkB,QAAQ,EAAE;AAAA,IAChE,OAAO;AACL,aAAO,KAAK;AAAA,IACd;AAAA,EACF;AAAA,EAEA,WAAW,OAA2B;AACpC,UAAM,gBAAgB;AACtB,UAAM,WAAY,MAAM,aAAa,MAAM,WAAW,gBAAiB;AACvE,UAAM,aAAc,MAAM,WAAW,gBAAiB;AAEtD,UAAM,SAAS,OAAO,MAAM,EAAE;AAC9B,WAAO,MAAM,QAAQ,CAAC;AACtB,WAAO,cAAc,KAAK,MAAM,KAAK,YAAY,CAAC;AAClD,WAAO,MAAM,QAAQ,CAAC;AACtB,WAAO,MAAM,QAAQ,EAAE;AACvB,WAAO,cAAc,IAAI,EAAE;AAC3B,WAAO,cAAc,GAAG,EAAE;AAC1B,WAAO,cAAc,MAAM,UAAU,EAAE;AACvC,WAAO,cAAc,MAAM,YAAY,EAAE;AACzC,WAAO,cAAc,UAAU,EAAE;AACjC,WAAO,cAAc,YAAY,EAAE;AACnC,WAAO,cAAc,IAAI,EAAE;AAC3B,WAAO,MAAM,QAAQ,EAAE;AACvB,WAAO,cAAc,MAAM,KAAK,YAAY,EAAE;AAC9C,WAAO,OAAO,OAAO,CAAC,QAAQ,OAAO,KAAK,MAAM,KAAK,MAAM,CAAC,CAAC;AAAA,EAC/D;AAAA,EAEA,MAAM,WAAW,QAAqB,aAAqD;AACzF,UAAM,SAAS,KAAK,iBAAiB;AACrC,aAAS,YAAY,MAAM;AAC3B,UAAM,YAAY,KAAK,WAAW,MAAM;AACxC,UAAM,OAAO,IAAI,KAAK,CAAC,IAAI,WAAW,SAAS,CAAC,GAAG,aAAa,EAAE,MAAM,YAAY,CAAC;AAErF,UAAM,OAAO,MAAM,KAAK,QAAQ,MAAM,eAAe;AAAA,MACnD;AAAA,QACE;AAAA,QACA,OAAO,KAAK,MAAM;AAAA,QAClB,UAAU,OAAO;AAAA,QACjB,QAAQ,OAAO;AAAA,QACf,iBAAiB;AAAA,MACnB;AAAA,MACA;AAAA,QACE,QAAQ;AAAA,MACV;AAAA,IACF;AAEA,WAAO;AAAA,MACL,MAAM,IAAI,gBAAgB;AAAA,MAC1B,cAAc;AAAA,QACZ;AAAA,UACE,MAAM,KAAK,QAAQ;AAAA,UACnB,UAAU,kBAAkB,OAAO,YAAY,EAAE;AAAA,UACjD,WAAW;AAAA,UACX,SAAS;AAAA,UACT,YAAY;AAAA,QACd;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA;AAAA,EAGA,SAA2B;AACzB,UAAM,IAAI,MAAM,0CAA0C;AAAA,EAC5D;AACF;","names":[]}
package/dist/stt.test.cjs CHANGED
@@ -3,7 +3,15 @@ var import_agents_plugin_silero = require("@livekit/agents-plugin-silero");
3
3
  var import_agents_plugins_test = require("@livekit/agents-plugins-test");
4
4
  var import_vitest = require("vitest");
5
5
  var import_stt = require("./stt.cjs");
6
- (0, import_vitest.describe)("OpenAI", async () => {
7
- await (0, import_agents_plugins_test.stt)(new import_stt.STT(), await import_agents_plugin_silero.VAD.load(), { streaming: false });
8
- });
6
+ const hasOpenAIApiKey = Boolean(process.env.OPENAI_API_KEY);
7
+ if (hasOpenAIApiKey) {
8
+ (0, import_vitest.describe)("OpenAI", async () => {
9
+ await (0, import_agents_plugins_test.stt)(new import_stt.STT(), await import_agents_plugin_silero.VAD.load(), { streaming: false });
10
+ });
11
+ } else {
12
+ (0, import_vitest.describe)("OpenAI", () => {
13
+ import_vitest.it.skip("requires OPENAI_API_KEY", () => {
14
+ });
15
+ });
16
+ }
9
17
  //# sourceMappingURL=stt.test.cjs.map
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/stt.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { VAD } from '@livekit/agents-plugin-silero';\nimport { stt } from '@livekit/agents-plugins-test';\nimport { describe } from 'vitest';\nimport { STT } from './stt.js';\n\ndescribe('OpenAI', async () => {\n await stt(new STT(), await VAD.load(), { streaming: false });\n});\n"],"mappings":";AAGA,kCAAoB;AACpB,iCAAoB;AACpB,oBAAyB;AACzB,iBAAoB;AAAA,IAEpB,wBAAS,UAAU,YAAY;AAC7B,YAAM,gCAAI,IAAI,eAAI,GAAG,MAAM,gCAAI,KAAK,GAAG,EAAE,WAAW,MAAM,CAAC;AAC7D,CAAC;","names":[]}
1
+ {"version":3,"sources":["../src/stt.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { VAD } from '@livekit/agents-plugin-silero';\nimport { stt } from '@livekit/agents-plugins-test';\nimport { describe, it } from 'vitest';\nimport { STT } from './stt.js';\n\nconst hasOpenAIApiKey = Boolean(process.env.OPENAI_API_KEY);\n\nif (hasOpenAIApiKey) {\n describe('OpenAI', async () => {\n await stt(new STT(), await VAD.load(), { streaming: false });\n });\n} else {\n describe('OpenAI', () => {\n it.skip('requires OPENAI_API_KEY', () => {});\n });\n}\n"],"mappings":";AAGA,kCAAoB;AACpB,iCAAoB;AACpB,oBAA6B;AAC7B,iBAAoB;AAEpB,MAAM,kBAAkB,QAAQ,QAAQ,IAAI,cAAc;AAE1D,IAAI,iBAAiB;AACnB,8BAAS,UAAU,YAAY;AAC7B,cAAM,gCAAI,IAAI,eAAI,GAAG,MAAM,gCAAI,KAAK,GAAG,EAAE,WAAW,MAAM,CAAC;AAAA,EAC7D,CAAC;AACH,OAAO;AACL,8BAAS,UAAU,MAAM;AACvB,qBAAG,KAAK,2BAA2B,MAAM;AAAA,IAAC,CAAC;AAAA,EAC7C,CAAC;AACH;","names":[]}
package/dist/stt.test.js CHANGED
@@ -1,8 +1,16 @@
1
1
  import { VAD } from "@livekit/agents-plugin-silero";
2
2
  import { stt } from "@livekit/agents-plugins-test";
3
- import { describe } from "vitest";
3
+ import { describe, it } from "vitest";
4
4
  import { STT } from "./stt.js";
5
- describe("OpenAI", async () => {
6
- await stt(new STT(), await VAD.load(), { streaming: false });
7
- });
5
+ const hasOpenAIApiKey = Boolean(process.env.OPENAI_API_KEY);
6
+ if (hasOpenAIApiKey) {
7
+ describe("OpenAI", async () => {
8
+ await stt(new STT(), await VAD.load(), { streaming: false });
9
+ });
10
+ } else {
11
+ describe("OpenAI", () => {
12
+ it.skip("requires OPENAI_API_KEY", () => {
13
+ });
14
+ });
15
+ }
8
16
  //# sourceMappingURL=stt.test.js.map
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/stt.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { VAD } from '@livekit/agents-plugin-silero';\nimport { stt } from '@livekit/agents-plugins-test';\nimport { describe } from 'vitest';\nimport { STT } from './stt.js';\n\ndescribe('OpenAI', async () => {\n await stt(new STT(), await VAD.load(), { streaming: false });\n});\n"],"mappings":"AAGA,SAAS,WAAW;AACpB,SAAS,WAAW;AACpB,SAAS,gBAAgB;AACzB,SAAS,WAAW;AAEpB,SAAS,UAAU,YAAY;AAC7B,QAAM,IAAI,IAAI,IAAI,GAAG,MAAM,IAAI,KAAK,GAAG,EAAE,WAAW,MAAM,CAAC;AAC7D,CAAC;","names":[]}
1
+ {"version":3,"sources":["../src/stt.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { VAD } from '@livekit/agents-plugin-silero';\nimport { stt } from '@livekit/agents-plugins-test';\nimport { describe, it } from 'vitest';\nimport { STT } from './stt.js';\n\nconst hasOpenAIApiKey = Boolean(process.env.OPENAI_API_KEY);\n\nif (hasOpenAIApiKey) {\n describe('OpenAI', async () => {\n await stt(new STT(), await VAD.load(), { streaming: false });\n });\n} else {\n describe('OpenAI', () => {\n it.skip('requires OPENAI_API_KEY', () => {});\n });\n}\n"],"mappings":"AAGA,SAAS,WAAW;AACpB,SAAS,WAAW;AACpB,SAAS,UAAU,UAAU;AAC7B,SAAS,WAAW;AAEpB,MAAM,kBAAkB,QAAQ,QAAQ,IAAI,cAAc;AAE1D,IAAI,iBAAiB;AACnB,WAAS,UAAU,YAAY;AAC7B,UAAM,IAAI,IAAI,IAAI,GAAG,MAAM,IAAI,KAAK,GAAG,EAAE,WAAW,MAAM,CAAC;AAAA,EAC7D,CAAC;AACH,OAAO;AACL,WAAS,UAAU,MAAM;AACvB,OAAG,KAAK,2BAA2B,MAAM;AAAA,IAAC,CAAC;AAAA,EAC7C,CAAC;AACH;","names":[]}
package/dist/tts.cjs CHANGED
@@ -37,6 +37,17 @@ class TTS extends import_agents.tts.TTS {
37
37
  #client;
38
38
  label = "openai.TTS";
39
39
  abortController = new AbortController();
40
+ get model() {
41
+ return this.#opts.model;
42
+ }
43
+ get provider() {
44
+ try {
45
+ const url = new URL(this.#client.baseURL);
46
+ return url.host;
47
+ } catch {
48
+ return "api.openai.com";
49
+ }
50
+ }
40
51
  /**
41
52
  * Create a new instance of OpenAI TTS.
42
53
  *
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 { 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 && !this.#opts.client) {\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: this.#opts.baseURL,\n apiKey: this.#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,UAAa,CAAC,KAAK,MAAM,QAAQ;AACzD,YAAM,IAAI,MAAM,0EAA0E;AAAA,IAC5F;AAEA,SAAK,UACH,KAAK,MAAM,UACX,IAAI,qBAAO;AAAA,MACT,SAAS,KAAK,MAAM;AAAA,MACpB,QAAQ,KAAK,MAAM;AAAA,IACrB,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"]}
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 get model(): string {\n return this.#opts.model;\n }\n\n get provider(): string {\n try {\n const url = new URL(this.#client.baseURL);\n return url.host;\n } catch {\n return 'api.openai.com';\n }\n }\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 && !this.#opts.client) {\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: this.#opts.baseURL,\n apiKey: this.#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,EAE9C,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAI,WAAmB;AACrB,QAAI;AACF,YAAM,MAAM,IAAI,IAAI,KAAK,QAAQ,OAAO;AACxC,aAAO,IAAI;AAAA,IACb,QAAQ;AACN,aAAO;AAAA,IACT;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,YAAY,OAA4B,mBAAmB;AACzD,UAAM,wBAAwB,qBAAqB,EAAE,WAAW,MAAM,CAAC;AAEvE,SAAK,QAAQ,EAAE,GAAG,mBAAmB,GAAG,KAAK;AAC7C,QAAI,KAAK,MAAM,WAAW,UAAa,CAAC,KAAK,MAAM,QAAQ;AACzD,YAAM,IAAI,MAAM,0EAA0E;AAAA,IAC5F;AAEA,SAAK,UACH,KAAK,MAAM,UACX,IAAI,qBAAO;AAAA,MACT,SAAS,KAAK,MAAM;AAAA,MACpB,QAAQ,KAAK,MAAM;AAAA,IACrB,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
@@ -14,6 +14,8 @@ export declare class TTS extends tts.TTS {
14
14
  #private;
15
15
  label: string;
16
16
  private abortController;
17
+ get model(): string;
18
+ get provider(): string;
17
19
  /**
18
20
  * Create a new instance of OpenAI TTS.
19
21
  *
package/dist/tts.d.ts CHANGED
@@ -14,6 +14,8 @@ export declare class TTS extends tts.TTS {
14
14
  #private;
15
15
  label: string;
16
16
  private abortController;
17
+ get model(): string;
18
+ get provider(): string;
17
19
  /**
18
20
  * Create a new instance of OpenAI TTS.
19
21
  *
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,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"}
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,IAAI,KAAK,IAAI,MAAM,CAElB;IAED,IAAI,QAAQ,IAAI,MAAM,CAOrB;IAED;;;;;;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
@@ -13,6 +13,17 @@ class TTS extends tts.TTS {
13
13
  #client;
14
14
  label = "openai.TTS";
15
15
  abortController = new AbortController();
16
+ get model() {
17
+ return this.#opts.model;
18
+ }
19
+ get provider() {
20
+ try {
21
+ const url = new URL(this.#client.baseURL);
22
+ return url.host;
23
+ } catch {
24
+ return "api.openai.com";
25
+ }
26
+ }
16
27
  /**
17
28
  * Create a new instance of OpenAI TTS.
18
29
  *
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 { 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 && !this.#opts.client) {\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: this.#opts.baseURL,\n apiKey: this.#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,UAAa,CAAC,KAAK,MAAM,QAAQ;AACzD,YAAM,IAAI,MAAM,0EAA0E;AAAA,IAC5F;AAEA,SAAK,UACH,KAAK,MAAM,UACX,IAAI,OAAO;AAAA,MACT,SAAS,KAAK,MAAM;AAAA,MACpB,QAAQ,KAAK,MAAM;AAAA,IACrB,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"]}
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 get model(): string {\n return this.#opts.model;\n }\n\n get provider(): string {\n try {\n const url = new URL(this.#client.baseURL);\n return url.host;\n } catch {\n return 'api.openai.com';\n }\n }\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 && !this.#opts.client) {\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: this.#opts.baseURL,\n apiKey: this.#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,EAE9C,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAI,WAAmB;AACrB,QAAI;AACF,YAAM,MAAM,IAAI,IAAI,KAAK,QAAQ,OAAO;AACxC,aAAO,IAAI;AAAA,IACb,QAAQ;AACN,aAAO;AAAA,IACT;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,YAAY,OAA4B,mBAAmB;AACzD,UAAM,wBAAwB,qBAAqB,EAAE,WAAW,MAAM,CAAC;AAEvE,SAAK,QAAQ,EAAE,GAAG,mBAAmB,GAAG,KAAK;AAC7C,QAAI,KAAK,MAAM,WAAW,UAAa,CAAC,KAAK,MAAM,QAAQ;AACzD,YAAM,IAAI,MAAM,0EAA0E;AAAA,IAC5F;AAEA,SAAK,UACH,KAAK,MAAM,UACX,IAAI,OAAO;AAAA,MACT,SAAS,KAAK,MAAM;AAAA,MACpB,QAAQ,KAAK,MAAM;AAAA,IACrB,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/dist/tts.test.cjs CHANGED
@@ -3,7 +3,15 @@ var import_agents_plugins_test = require("@livekit/agents-plugins-test");
3
3
  var import_vitest = require("vitest");
4
4
  var import_stt = require("./stt.cjs");
5
5
  var import_tts = require("./tts.cjs");
6
- (0, import_vitest.describe)("OpenAI", async () => {
7
- await (0, import_agents_plugins_test.tts)(new import_tts.TTS(), new import_stt.STT(), { streaming: false });
8
- });
6
+ const hasOpenAIApiKey = Boolean(process.env.OPENAI_API_KEY);
7
+ if (hasOpenAIApiKey) {
8
+ (0, import_vitest.describe)("OpenAI", async () => {
9
+ await (0, import_agents_plugins_test.tts)(new import_tts.TTS(), new import_stt.STT(), { streaming: false });
10
+ });
11
+ } else {
12
+ (0, import_vitest.describe)("OpenAI", () => {
13
+ import_vitest.it.skip("requires OPENAI_API_KEY", () => {
14
+ });
15
+ });
16
+ }
9
17
  //# sourceMappingURL=tts.test.cjs.map
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/tts.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { tts } from '@livekit/agents-plugins-test';\nimport { describe } from 'vitest';\nimport { STT } from './stt.js';\nimport { TTS } from './tts.js';\n\ndescribe('OpenAI', async () => {\n await tts(new TTS(), new STT(), { streaming: false });\n});\n"],"mappings":";AAGA,iCAAoB;AACpB,oBAAyB;AACzB,iBAAoB;AACpB,iBAAoB;AAAA,IAEpB,wBAAS,UAAU,YAAY;AAC7B,YAAM,gCAAI,IAAI,eAAI,GAAG,IAAI,eAAI,GAAG,EAAE,WAAW,MAAM,CAAC;AACtD,CAAC;","names":[]}
1
+ {"version":3,"sources":["../src/tts.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { tts } from '@livekit/agents-plugins-test';\nimport { describe, it } from 'vitest';\nimport { STT } from './stt.js';\nimport { TTS } from './tts.js';\n\nconst hasOpenAIApiKey = Boolean(process.env.OPENAI_API_KEY);\n\nif (hasOpenAIApiKey) {\n describe('OpenAI', async () => {\n await tts(new TTS(), new STT(), { streaming: false });\n });\n} else {\n describe('OpenAI', () => {\n it.skip('requires OPENAI_API_KEY', () => {});\n });\n}\n"],"mappings":";AAGA,iCAAoB;AACpB,oBAA6B;AAC7B,iBAAoB;AACpB,iBAAoB;AAEpB,MAAM,kBAAkB,QAAQ,QAAQ,IAAI,cAAc;AAE1D,IAAI,iBAAiB;AACnB,8BAAS,UAAU,YAAY;AAC7B,cAAM,gCAAI,IAAI,eAAI,GAAG,IAAI,eAAI,GAAG,EAAE,WAAW,MAAM,CAAC;AAAA,EACtD,CAAC;AACH,OAAO;AACL,8BAAS,UAAU,MAAM;AACvB,qBAAG,KAAK,2BAA2B,MAAM;AAAA,IAAC,CAAC;AAAA,EAC7C,CAAC;AACH;","names":[]}
package/dist/tts.test.js CHANGED
@@ -1,8 +1,16 @@
1
1
  import { tts } from "@livekit/agents-plugins-test";
2
- import { describe } from "vitest";
2
+ import { describe, it } from "vitest";
3
3
  import { STT } from "./stt.js";
4
4
  import { TTS } from "./tts.js";
5
- describe("OpenAI", async () => {
6
- await tts(new TTS(), new STT(), { streaming: false });
7
- });
5
+ const hasOpenAIApiKey = Boolean(process.env.OPENAI_API_KEY);
6
+ if (hasOpenAIApiKey) {
7
+ describe("OpenAI", async () => {
8
+ await tts(new TTS(), new STT(), { streaming: false });
9
+ });
10
+ } else {
11
+ describe("OpenAI", () => {
12
+ it.skip("requires OPENAI_API_KEY", () => {
13
+ });
14
+ });
15
+ }
8
16
  //# sourceMappingURL=tts.test.js.map
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/tts.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { tts } from '@livekit/agents-plugins-test';\nimport { describe } from 'vitest';\nimport { STT } from './stt.js';\nimport { TTS } from './tts.js';\n\ndescribe('OpenAI', async () => {\n await tts(new TTS(), new STT(), { streaming: false });\n});\n"],"mappings":"AAGA,SAAS,WAAW;AACpB,SAAS,gBAAgB;AACzB,SAAS,WAAW;AACpB,SAAS,WAAW;AAEpB,SAAS,UAAU,YAAY;AAC7B,QAAM,IAAI,IAAI,IAAI,GAAG,IAAI,IAAI,GAAG,EAAE,WAAW,MAAM,CAAC;AACtD,CAAC;","names":[]}
1
+ {"version":3,"sources":["../src/tts.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { tts } from '@livekit/agents-plugins-test';\nimport { describe, it } from 'vitest';\nimport { STT } from './stt.js';\nimport { TTS } from './tts.js';\n\nconst hasOpenAIApiKey = Boolean(process.env.OPENAI_API_KEY);\n\nif (hasOpenAIApiKey) {\n describe('OpenAI', async () => {\n await tts(new TTS(), new STT(), { streaming: false });\n });\n} else {\n describe('OpenAI', () => {\n it.skip('requires OPENAI_API_KEY', () => {});\n });\n}\n"],"mappings":"AAGA,SAAS,WAAW;AACpB,SAAS,UAAU,UAAU;AAC7B,SAAS,WAAW;AACpB,SAAS,WAAW;AAEpB,MAAM,kBAAkB,QAAQ,QAAQ,IAAI,cAAc;AAE1D,IAAI,iBAAiB;AACnB,WAAS,UAAU,YAAY;AAC7B,UAAM,IAAI,IAAI,IAAI,GAAG,IAAI,IAAI,GAAG,EAAE,WAAW,MAAM,CAAC;AAAA,EACtD,CAAC;AACH,OAAO;AACL,WAAS,UAAU,MAAM;AACvB,OAAG,KAAK,2BAA2B,MAAM;AAAA,IAAC,CAAC;AAAA,EAC7C,CAAC;AACH;","names":[]}
@@ -0,0 +1,29 @@
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 ws_exports = {};
20
+ __export(ws_exports, {
21
+ ResponsesWebSocket: () => import_llm.ResponsesWebSocket
22
+ });
23
+ module.exports = __toCommonJS(ws_exports);
24
+ var import_llm = require("./llm.cjs");
25
+ // Annotate the CommonJS export names for ESM import in node:
26
+ 0 && (module.exports = {
27
+ ResponsesWebSocket
28
+ });
29
+ //# sourceMappingURL=index.cjs.map
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["../../src/ws/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\nexport { ResponsesWebSocket } from './llm.js';\nexport type {\n WsErrorEvent,\n WsFunctionCallItem,\n WsOutputItem,\n WsOutputItemDoneEvent,\n WsOutputTextDeltaEvent,\n WsResponseCompletedEvent,\n WsResponseCreateEvent,\n WsResponseCreatedEvent,\n WsResponseFailedEvent,\n WsServerEvent,\n} from './types.js';\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,iBAAmC;","names":[]}
@@ -0,0 +1,3 @@
1
+ export { ResponsesWebSocket } from './llm.js';
2
+ export type { WsErrorEvent, WsFunctionCallItem, WsOutputItem, WsOutputItemDoneEvent, WsOutputTextDeltaEvent, WsResponseCompletedEvent, WsResponseCreateEvent, WsResponseCreatedEvent, WsResponseFailedEvent, WsServerEvent, } from './types.js';
3
+ //# sourceMappingURL=index.d.ts.map
@@ -0,0 +1,3 @@
1
+ export { ResponsesWebSocket } from './llm.js';
2
+ export type { WsErrorEvent, WsFunctionCallItem, WsOutputItem, WsOutputItemDoneEvent, WsOutputTextDeltaEvent, WsResponseCompletedEvent, WsResponseCreateEvent, WsResponseCreatedEvent, WsResponseFailedEvent, WsServerEvent, } from './types.js';
3
+ //# sourceMappingURL=index.d.ts.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/ws/index.ts"],"names":[],"mappings":"AAIA,OAAO,EAAE,kBAAkB,EAAE,MAAM,UAAU,CAAC;AAC9C,YAAY,EACV,YAAY,EACZ,kBAAkB,EAClB,YAAY,EACZ,qBAAqB,EACrB,sBAAsB,EACtB,wBAAwB,EACxB,qBAAqB,EACrB,sBAAsB,EACtB,qBAAqB,EACrB,aAAa,GACd,MAAM,YAAY,CAAC"}
@@ -0,0 +1,5 @@
1
+ import { ResponsesWebSocket } from "./llm.js";
2
+ export {
3
+ ResponsesWebSocket
4
+ };
5
+ //# sourceMappingURL=index.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["../../src/ws/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\nexport { ResponsesWebSocket } from './llm.js';\nexport type {\n WsErrorEvent,\n WsFunctionCallItem,\n WsOutputItem,\n WsOutputItemDoneEvent,\n WsOutputTextDeltaEvent,\n WsResponseCompletedEvent,\n WsResponseCreateEvent,\n WsResponseCreatedEvent,\n WsResponseFailedEvent,\n WsServerEvent,\n} from './types.js';\n"],"mappings":"AAIA,SAAS,0BAA0B;","names":[]}