@livekit/agents-plugin-deepgram 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.
- package/dist/index.cjs +1 -1
- package/dist/index.js +1 -1
- package/dist/stt.cjs +20 -4
- package/dist/stt.cjs.map +1 -1
- package/dist/stt.d.cts +2 -0
- package/dist/stt.d.ts +2 -0
- package/dist/stt.d.ts.map +1 -1
- package/dist/stt.js +22 -4
- package/dist/stt.js.map +1 -1
- package/dist/stt.test.cjs +11 -5
- package/dist/stt.test.cjs.map +1 -1
- package/dist/stt.test.js +12 -6
- package/dist/stt.test.js.map +1 -1
- package/dist/stt_v2.cjs +11 -3
- package/dist/stt_v2.cjs.map +1 -1
- package/dist/stt_v2.d.ts.map +1 -1
- package/dist/stt_v2.js +12 -3
- package/dist/stt_v2.js.map +1 -1
- package/dist/stt_v2.test.cjs +11 -5
- package/dist/stt_v2.test.cjs.map +1 -1
- package/dist/stt_v2.test.js +12 -6
- package/dist/stt_v2.test.js.map +1 -1
- package/dist/tts.cjs +6 -0
- package/dist/tts.cjs.map +1 -1
- package/dist/tts.d.cts +2 -0
- package/dist/tts.d.ts +2 -0
- package/dist/tts.d.ts.map +1 -1
- package/dist/tts.js +6 -0
- package/dist/tts.js.map +1 -1
- package/dist/tts.test.cjs +12 -4
- package/dist/tts.test.cjs.map +1 -1
- package/dist/tts.test.js +13 -5
- package/dist/tts.test.js.map +1 -1
- package/package.json +6 -5
- package/src/stt.test.ts +12 -6
- package/src/stt.ts +25 -4
- package/src/stt_v2.test.ts +12 -6
- package/src/stt_v2.ts +13 -3
- package/src/tts.test.ts +13 -5
- package/src/tts.ts +8 -0
package/dist/stt_v2.cjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/stt_v2.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n AudioByteStream,\n Event,\n calculateAudioDurationSeconds,\n createTimedString,\n log,\n stt,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport * as queryString from 'node:querystring';\nimport { WebSocket } from 'ws';\nimport { PeriodicCollector } from './_utils.js';\nimport type { V2Models } from './models.js';\n\nconst _CLOSE_MSG = JSON.stringify({ type: 'CloseStream' });\n\n// --- Configuration ---\n\n/**\n * Configuration options for STTv2 (Deepgram Flux model).\n */\nexport interface STTv2Options {\n apiKey?: string;\n model: V2Models | string;\n sampleRate: number;\n keyterms: string[];\n endpointUrl: string;\n language?: string;\n eagerEotThreshold?: number;\n eotThreshold?: number;\n eotTimeoutMs?: number;\n mipOptOut?: boolean;\n tags?: string[];\n}\n\nconst defaultSTTv2Options: Omit<STTv2Options, 'apiKey'> = {\n model: 'flux-general-en',\n sampleRate: 16000,\n keyterms: [],\n endpointUrl: 'wss://api.deepgram.com/v2/listen',\n language: 'en',\n mipOptOut: false,\n};\n\nfunction validateTags(tags: string[]): string[] {\n for (const tag of tags) {\n if (tag.length > 128) {\n throw new Error('tag must be no more than 128 characters');\n }\n }\n return tags;\n}\n\n/**\n * Deepgram STTv2 using the Flux model for streaming speech-to-text.\n *\n * This uses Deepgram's V2 API (`/v2/listen`) which provides turn-based\n * transcription with support for preemptive generation.\n *\n * @remarks\n * Key differences from STT (V1):\n * - Uses `TurnInfo` events instead of `SpeechStarted`/`Results`\n * - Supports `eagerEotThreshold` for preemptive LLM generation\n * - Sends `PREFLIGHT_TRANSCRIPT` events when eager end-of-turn is detected\n *\n * @example\n * ```typescript\n * import { STTv2 } from '@livekit/agents-plugin-deepgram';\n *\n * const stt = new STTv2({\n * model: 'flux-general-en',\n * eagerEotThreshold: 0.5, // Enable preemptive generation\n * });\n *\n * const stream = stt.stream();\n * stream.pushFrame(audioFrame);\n *\n * for await (const event of stream) {\n * if (event.type === SpeechEventType.FINAL_TRANSCRIPT) {\n * console.log(event.alternatives?.[0]?.text);\n * }\n * }\n * ```\n */\nexport class STTv2 extends stt.STT {\n readonly label = 'deepgram.STTv2';\n #opts: STTv2Options;\n #apiKey: string;\n #logger = log();\n\n /**\n * Create a new Deepgram STTv2 instance.\n *\n * @param opts - Configuration options\n * @param opts.apiKey - Deepgram API key (defaults to `DEEPGRAM_API_KEY` env var)\n * @param opts.model - Model to use (default: `flux-general-en`)\n * @param opts.eagerEotThreshold - Threshold (0.3-0.9) for preemptive generation\n * @param opts.eotThreshold - End-of-turn detection threshold (default: 0.7)\n * @param opts.eotTimeoutMs - End-of-turn timeout in ms (default: 3000)\n * @param opts.keyterms - List of key terms to improve recognition\n * @param opts.tags - Tags for usage reporting (max 128 chars each)\n *\n * @throws Error if no API key is provided\n */\n constructor(opts: Partial<STTv2Options> = {}) {\n super({\n streaming: true,\n interimResults: true,\n alignedTranscript: 'word',\n });\n\n this.#opts = { ...defaultSTTv2Options, ...opts };\n\n const apiKey = opts.apiKey || process.env.DEEPGRAM_API_KEY;\n if (!apiKey) {\n throw new Error('Deepgram API key is required');\n }\n this.#apiKey = apiKey;\n\n if (this.#opts.tags) {\n this.#opts.tags = validateTags(this.#opts.tags);\n }\n }\n\n /** The model being used for transcription */\n get model(): string {\n return this.#opts.model;\n }\n\n /** The STT provider name */\n get provider(): string {\n return 'Deepgram';\n }\n\n protected async _recognize(\n _frame: AudioFrame | AudioFrame[],\n _abortSignal?: AbortSignal,\n ): Promise<stt.SpeechEvent> {\n throw new Error('V2 API does not support non-streaming recognize. Use .stream()');\n }\n\n /**\n * Create a new streaming transcription session.\n *\n * @param options - Stream options\n * @returns A SpeechStream that emits transcription events\n */\n stream(options?: { connOptions?: APIConnectOptions }): stt.SpeechStream {\n const streamOpts = { ...this.#opts, apiKey: this.#apiKey };\n return new SpeechStreamv2(this, streamOpts, options?.connOptions);\n }\n\n /**\n * Update STT options. Changes will take effect on the next stream.\n *\n * @param opts - Partial options to update\n */\n updateOptions(opts: Partial<STTv2Options>) {\n this.#opts = { ...this.#opts, ...opts };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n this.#logger.debug('Updated STTv2 options');\n }\n}\n\n// --- Stream Implementation ---\n\nclass SpeechStreamv2 extends stt.SpeechStream {\n readonly label = 'deepgram.SpeechStreamv2';\n #opts: STTv2Options & { apiKey: string };\n #logger = log();\n #ws: WebSocket | null = null;\n\n #audioDurationCollector: PeriodicCollector<number>;\n #requestId = '';\n #speaking = false;\n\n // Parity: _reconnect_event - using existing Event class from @livekit/agents\n #reconnectEvent = new Event();\n\n constructor(\n sttInstance: STTv2,\n opts: STTv2Options & { apiKey: string },\n connOptions?: APIConnectOptions,\n ) {\n super(sttInstance, opts.sampleRate, connOptions);\n this.#opts = opts;\n\n this.#audioDurationCollector = new PeriodicCollector(\n (duration) => this.#onAudioDurationReport(duration),\n { duration: 5.0 },\n );\n }\n\n updateOptions(opts: Partial<STTv2Options>) {\n this.#logger.debug('Stream received option update', opts);\n this.#opts = { ...this.#opts, ...opts };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n\n // Trigger reconnection loop\n this.#reconnectEvent.set();\n }\n\n protected async run() {\n // Outer Loop: Handles reconnections (Configuration updates)\n while (!this.closed) {\n try {\n this.#reconnectEvent.clear();\n\n const url = this.#getDeepgramUrl();\n this.#logger.debug(`Connecting to Deepgram: ${url}`);\n\n this.#ws = new WebSocket(url, {\n headers: { Authorization: `Token ${this.#opts.apiKey}` },\n });\n\n // 1. Wait for Connection Open\n await new Promise<void>((resolve, reject) => {\n if (!this.#ws) return reject(new Error('WebSocket not initialized'));\n\n const onOpen = () => {\n this.#ws?.off('error', onError);\n resolve();\n };\n const onError = (err: Error) => {\n this.#ws?.off('open', onOpen);\n reject(err);\n };\n\n this.#ws.once('open', onOpen);\n this.#ws.once('error', onError);\n });\n\n // 2. Run Concurrent Tasks (Send & Receive)\n const sendPromise = this.#sendTask();\n const recvPromise = this.#recvTask();\n const reconnectWait = this.#reconnectEvent.wait();\n\n // 3. Race: Normal Completion vs Reconnect Signal\n const result = await Promise.race([\n Promise.all([sendPromise, recvPromise]),\n reconnectWait.then(() => 'RECONNECT'),\n ]);\n\n if (result === 'RECONNECT') {\n this.#logger.debug('Reconnecting stream due to option update...');\n // Close current socket; loop will restart and open a new one\n this.#ws.close();\n } else {\n // Normal finish (Stream ended or Error thrown)\n break;\n }\n } catch (error) {\n this.#logger.error('Deepgram stream error', { error });\n throw error; // Let Base Class handle retry logic\n } finally {\n if (this.#ws?.readyState === WebSocket.OPEN) {\n this.#ws.close();\n }\n }\n }\n this.close();\n }\n\n async #sendTask() {\n if (!this.#ws) return;\n\n // Buffer audio into 50ms chunks (Parity)\n const samples50ms = Math.floor(this.#opts.sampleRate / 20);\n const audioBstream = new AudioByteStream(this.#opts.sampleRate, 1, samples50ms);\n\n let hasEnded = false;\n\n // Manual Iterator to allow racing against Reconnect Signal\n const iterator = this.input[Symbol.asyncIterator]();\n\n while (true) {\n const nextPromise = iterator.next();\n // If reconnect signal fires, abort the wait\n const abortPromise = this.#reconnectEvent.wait().then(() => ({ abort: true }) as const);\n\n const result = await Promise.race([nextPromise, abortPromise]);\n\n // Check if we need to abort (Reconnect) or if stream is done\n if ('abort' in result || result.done) {\n if (!('abort' in result) && result.done) {\n // Normal stream end\n hasEnded = true;\n } else {\n // Reconnect triggered - break loop immediately\n break;\n }\n }\n\n // If we broke above, we don't process data. If not, 'result' is IteratorResult\n if (hasEnded && result.value === undefined) {\n // Process flush below\n } else if ('value' in result) {\n const data = result.value;\n const frames: AudioFrame[] = [];\n\n if (data === SpeechStreamv2.FLUSH_SENTINEL) {\n frames.push(...audioBstream.flush());\n hasEnded = true;\n } else {\n frames.push(...audioBstream.write((data as AudioFrame).data.buffer as ArrayBuffer));\n }\n\n for (const frame of frames) {\n this.#audioDurationCollector.push(calculateAudioDurationSeconds(frame));\n\n if (this.#ws!.readyState === WebSocket.OPEN) {\n this.#ws!.send(frame.data);\n }\n\n if (hasEnded) {\n this.#audioDurationCollector.flush();\n hasEnded = false;\n }\n }\n }\n\n if (hasEnded) break;\n }\n\n // Only send CloseStream if we are exiting normally (not reconnecting)\n if (!this.#reconnectEvent.isSet && this.#ws!.readyState === WebSocket.OPEN) {\n this.#logger.debug('Sending CloseStream message to Deepgram');\n this.#ws!.send(_CLOSE_MSG);\n }\n }\n\n async #recvTask() {\n if (!this.#ws) return;\n\n return new Promise<void>((resolve) => {\n if (!this.#ws) return resolve();\n\n this.#ws.on('message', (data: Buffer, isBinary: boolean) => {\n if (isBinary) {\n this.#logger.warn('Received unexpected binary message from Deepgram');\n return;\n }\n try {\n const msg = JSON.parse(data.toString());\n this.#processStreamEvent(msg);\n } catch (error) {\n this.#logger.error('Failed to parse Deepgram message', { error });\n }\n });\n\n this.#ws.on('close', (code, reason) => {\n this.#logger.debug(`Deepgram WebSocket closed: ${code} ${reason}`);\n resolve();\n });\n\n // Errors are caught by run() listener, resolve here to clean up task\n this.#ws.on('error', () => resolve());\n });\n }\n\n #processStreamEvent(data: Record<string, unknown>) {\n if (data.request_id) {\n this.#requestId = data.request_id as string;\n }\n\n if (data.type === 'TurnInfo') {\n const eventType = data.event;\n\n if (eventType === 'StartOfTurn') {\n if (this.#speaking) return;\n\n this.#speaking = true;\n this.queue.put({\n type: stt.SpeechEventType.START_OF_SPEECH,\n requestId: this.#requestId,\n });\n\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'Update') {\n if (!this.#speaking) return;\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'EagerEndOfTurn') {\n if (!this.#speaking) return;\n this.#sendTranscriptEvent(stt.SpeechEventType.PREFLIGHT_TRANSCRIPT, data);\n } else if (eventType === 'TurnResumed') {\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'EndOfTurn') {\n if (!this.#speaking) return;\n\n this.#speaking = false;\n this.#sendTranscriptEvent(stt.SpeechEventType.FINAL_TRANSCRIPT, data);\n\n this.queue.put({\n type: stt.SpeechEventType.END_OF_SPEECH,\n requestId: this.#requestId,\n });\n }\n } else if (data.type === 'Error') {\n this.#logger.warn('deepgram sent an error', { data });\n const desc = (data.description as string) || 'unknown error from deepgram';\n throw new Error(`Deepgram API Error: ${desc}`);\n }\n }\n\n #sendTranscriptEvent(eventType: stt.SpeechEventType, data: Record<string, unknown>) {\n const alts = parseTranscription(this.#opts.language || 'en', data, this.startTimeOffset);\n\n if (alts.length > 0) {\n this.queue.put({\n type: eventType,\n requestId: this.#requestId,\n alternatives: [alts[0]!, ...alts.slice(1)],\n });\n }\n }\n\n #onAudioDurationReport(duration: number) {\n const usageEvent: stt.SpeechEvent = {\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: {\n audioDuration: duration,\n },\n };\n this.queue.put(usageEvent);\n }\n\n #getDeepgramUrl(): string {\n const params: Record<string, string | string[]> = {\n model: this.#opts.model,\n sample_rate: this.#opts.sampleRate.toString(),\n encoding: 'linear16',\n mip_opt_out: String(this.#opts.mipOptOut),\n };\n\n // Note: v2 API does NOT include 'language' parameter\n if (this.#opts.eagerEotThreshold)\n params.eager_eot_threshold = this.#opts.eagerEotThreshold.toString();\n if (this.#opts.eotThreshold) params.eot_threshold = this.#opts.eotThreshold.toString();\n if (this.#opts.eotTimeoutMs) params.eot_timeout_ms = this.#opts.eotTimeoutMs.toString();\n\n if (this.#opts.keyterms.length > 0) params.keyterm = this.#opts.keyterms;\n if (this.#opts.tags && this.#opts.tags.length > 0) params.tag = this.#opts.tags;\n\n const baseUrl = this.#opts.endpointUrl.replace(/^http/, 'ws');\n const qs = queryString.stringify(params);\n return `${baseUrl}?${qs}`;\n }\n\n override close() {\n super.close();\n this.#ws?.close();\n }\n}\n\n// --- Helpers ---\n\nfunction parseTranscription(\n language: string,\n data: Record<string, unknown>,\n startTimeOffset: number,\n): stt.SpeechData[] {\n const transcript = data.transcript as string | undefined;\n const wordsData = (data.words as Array<Record<string, unknown>>) || [];\n\n if (!wordsData || wordsData.length === 0) {\n return [];\n }\n\n let confidence = 0;\n if (wordsData.length > 0) {\n const sum = wordsData.reduce((acc: number, w) => acc + ((w.confidence as number) || 0), 0);\n confidence = sum / wordsData.length;\n }\n\n const sd: stt.SpeechData = {\n language: language,\n startTime: ((data.audio_window_start as number) || 0) + startTimeOffset,\n endTime: ((data.audio_window_end as number) || 0) + startTimeOffset,\n confidence: confidence,\n text: transcript || '',\n // Note: Deepgram V2 (Flux) API does not provide word-level timing (start/end).\n // Words only contain 'word' and 'confidence' fields, so startTime/endTime will be 0.\n // See: https://developers.deepgram.com/docs/flux/nova-3-migration\n words: wordsData.map((word) =>\n createTimedString({\n text: (word.word as string) ?? '',\n startTime: ((word.start as number) ?? 0) + startTimeOffset,\n endTime: ((word.end as number) ?? 0) + startTimeOffset,\n confidence: (word.confidence as number) ?? 0.0,\n startTimeOffset,\n }),\n ),\n };\n\n return [sd];\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBAQO;AAEP,kBAA6B;AAC7B,gBAA0B;AAC1B,mBAAkC;AAGlC,MAAM,aAAa,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC;AAqBzD,MAAM,sBAAoD;AAAA,EACxD,OAAO;AAAA,EACP,YAAY;AAAA,EACZ,UAAU,CAAC;AAAA,EACX,aAAa;AAAA,EACb,UAAU;AAAA,EACV,WAAW;AACb;AAEA,SAAS,aAAa,MAA0B;AAC9C,aAAW,OAAO,MAAM;AACtB,QAAI,IAAI,SAAS,KAAK;AACpB,YAAM,IAAI,MAAM,yCAAyC;AAAA,IAC3D;AAAA,EACF;AACA,SAAO;AACT;AAiCO,MAAM,cAAc,kBAAI,IAAI;AAAA,EACxB,QAAQ;AAAA,EACjB;AAAA,EACA;AAAA,EACA,cAAU,mBAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAgBd,YAAY,OAA8B,CAAC,GAAG;AAC5C,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB;AAAA,MAChB,mBAAmB;AAAA,IACrB,CAAC;AAED,SAAK,QAAQ,EAAE,GAAG,qBAAqB,GAAG,KAAK;AAE/C,UAAM,SAAS,KAAK,UAAU,QAAQ,IAAI;AAC1C,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI,MAAM,8BAA8B;AAAA,IAChD;AACA,SAAK,UAAU;AAEf,QAAI,KAAK,MAAM,MAAM;AACnB,WAAK,MAAM,OAAO,aAAa,KAAK,MAAM,IAAI;AAAA,IAChD;AAAA,EACF;AAAA;AAAA,EAGA,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA;AAAA,EAGA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,MAAgB,WACd,QACA,cAC0B;AAC1B,UAAM,IAAI,MAAM,gEAAgE;AAAA,EAClF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,OAAO,SAAiE;AACtE,UAAM,aAAa,EAAE,GAAG,KAAK,OAAO,QAAQ,KAAK,QAAQ;AACzD,WAAO,IAAI,eAAe,MAAM,YAAY,mCAAS,WAAW;AAAA,EAClE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,cAAc,MAA6B;AACzC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AACvD,SAAK,QAAQ,MAAM,uBAAuB;AAAA,EAC5C;AACF;AAIA,MAAM,uBAAuB,kBAAI,aAAa;AAAA,EACnC,QAAQ;AAAA,EACjB;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,MAAwB;AAAA,EAExB;AAAA,EACA,aAAa;AAAA,EACb,YAAY;AAAA;AAAA,EAGZ,kBAAkB,IAAI,oBAAM;AAAA,EAE5B,YACE,aACA,MACA,aACA;AACA,UAAM,aAAa,KAAK,YAAY,WAAW;AAC/C,SAAK,QAAQ;AAEb,SAAK,0BAA0B,IAAI;AAAA,MACjC,CAAC,aAAa,KAAK,uBAAuB,QAAQ;AAAA,MAClD,EAAE,UAAU,EAAI;AAAA,IAClB;AAAA,EACF;AAAA,EAEA,cAAc,MAA6B;AACzC,SAAK,QAAQ,MAAM,iCAAiC,IAAI;AACxD,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AAGvD,SAAK,gBAAgB,IAAI;AAAA,EAC3B;AAAA,EAEA,MAAgB,MAAM;AA9MxB;AAgNI,WAAO,CAAC,KAAK,QAAQ;AACnB,UAAI;AACF,aAAK,gBAAgB,MAAM;AAE3B,cAAM,MAAM,KAAK,gBAAgB;AACjC,aAAK,QAAQ,MAAM,2BAA2B,GAAG,EAAE;AAEnD,aAAK,MAAM,IAAI,oBAAU,KAAK;AAAA,UAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,QACzD,CAAC;AAGD,cAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,cAAI,CAAC,KAAK,IAAK,QAAO,OAAO,IAAI,MAAM,2BAA2B,CAAC;AAEnE,gBAAM,SAAS,MAAM;AA/N/B,gBAAAA;AAgOY,aAAAA,MAAA,KAAK,QAAL,gBAAAA,IAAU,IAAI,SAAS;AACvB,oBAAQ;AAAA,UACV;AACA,gBAAM,UAAU,CAAC,QAAe;AAnO1C,gBAAAA;AAoOY,aAAAA,MAAA,KAAK,QAAL,gBAAAA,IAAU,IAAI,QAAQ;AACtB,mBAAO,GAAG;AAAA,UACZ;AAEA,eAAK,IAAI,KAAK,QAAQ,MAAM;AAC5B,eAAK,IAAI,KAAK,SAAS,OAAO;AAAA,QAChC,CAAC;AAGD,cAAM,cAAc,KAAK,UAAU;AACnC,cAAM,cAAc,KAAK,UAAU;AACnC,cAAM,gBAAgB,KAAK,gBAAgB,KAAK;AAGhD,cAAM,SAAS,MAAM,QAAQ,KAAK;AAAA,UAChC,QAAQ,IAAI,CAAC,aAAa,WAAW,CAAC;AAAA,UACtC,cAAc,KAAK,MAAM,WAAW;AAAA,QACtC,CAAC;AAED,YAAI,WAAW,aAAa;AAC1B,eAAK,QAAQ,MAAM,6CAA6C;AAEhE,eAAK,IAAI,MAAM;AAAA,QACjB,OAAO;AAEL;AAAA,QACF;AAAA,MACF,SAAS,OAAO;AACd,aAAK,QAAQ,MAAM,yBAAyB,EAAE,MAAM,CAAC;AACrD,cAAM;AAAA,MACR,UAAE;AACA,cAAI,UAAK,QAAL,mBAAU,gBAAe,oBAAU,MAAM;AAC3C,eAAK,IAAI,MAAM;AAAA,QACjB;AAAA,MACF;AAAA,IACF;AACA,SAAK,MAAM;AAAA,EACb;AAAA,EAEA,MAAM,YAAY;AAChB,QAAI,CAAC,KAAK,IAAK;AAGf,UAAM,cAAc,KAAK,MAAM,KAAK,MAAM,aAAa,EAAE;AACzD,UAAM,eAAe,IAAI,8BAAgB,KAAK,MAAM,YAAY,GAAG,WAAW;AAE9E,QAAI,WAAW;AAGf,UAAM,WAAW,KAAK,MAAM,OAAO,aAAa,EAAE;AAElD,WAAO,MAAM;AACX,YAAM,cAAc,SAAS,KAAK;AAElC,YAAM,eAAe,KAAK,gBAAgB,KAAK,EAAE,KAAK,OAAO,EAAE,OAAO,KAAK,EAAW;AAEtF,YAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,aAAa,YAAY,CAAC;AAG7D,UAAI,WAAW,UAAU,OAAO,MAAM;AACpC,YAAI,EAAE,WAAW,WAAW,OAAO,MAAM;AAEvC,qBAAW;AAAA,QACb,OAAO;AAEL;AAAA,QACF;AAAA,MACF;AAGA,UAAI,YAAY,OAAO,UAAU,QAAW;AAAA,MAE5C,WAAW,WAAW,QAAQ;AAC5B,cAAM,OAAO,OAAO;AACpB,cAAM,SAAuB,CAAC;AAE9B,YAAI,SAAS,eAAe,gBAAgB;AAC1C,iBAAO,KAAK,GAAG,aAAa,MAAM,CAAC;AACnC,qBAAW;AAAA,QACb,OAAO;AACL,iBAAO,KAAK,GAAG,aAAa,MAAO,KAAoB,KAAK,MAAqB,CAAC;AAAA,QACpF;AAEA,mBAAW,SAAS,QAAQ;AAC1B,eAAK,wBAAwB,SAAK,6CAA8B,KAAK,CAAC;AAEtE,cAAI,KAAK,IAAK,eAAe,oBAAU,MAAM;AAC3C,iBAAK,IAAK,KAAK,MAAM,IAAI;AAAA,UAC3B;AAEA,cAAI,UAAU;AACZ,iBAAK,wBAAwB,MAAM;AACnC,uBAAW;AAAA,UACb;AAAA,QACF;AAAA,MACF;AAEA,UAAI,SAAU;AAAA,IAChB;AAGA,QAAI,CAAC,KAAK,gBAAgB,SAAS,KAAK,IAAK,eAAe,oBAAU,MAAM;AAC1E,WAAK,QAAQ,MAAM,yCAAyC;AAC5D,WAAK,IAAK,KAAK,UAAU;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,MAAM,YAAY;AAChB,QAAI,CAAC,KAAK,IAAK;AAEf,WAAO,IAAI,QAAc,CAAC,YAAY;AACpC,UAAI,CAAC,KAAK,IAAK,QAAO,QAAQ;AAE9B,WAAK,IAAI,GAAG,WAAW,CAAC,MAAc,aAAsB;AAC1D,YAAI,UAAU;AACZ,eAAK,QAAQ,KAAK,kDAAkD;AACpE;AAAA,QACF;AACA,YAAI;AACF,gBAAM,MAAM,KAAK,MAAM,KAAK,SAAS,CAAC;AACtC,eAAK,oBAAoB,GAAG;AAAA,QAC9B,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,oCAAoC,EAAE,MAAM,CAAC;AAAA,QAClE;AAAA,MACF,CAAC;AAED,WAAK,IAAI,GAAG,SAAS,CAAC,MAAM,WAAW;AACrC,aAAK,QAAQ,MAAM,8BAA8B,IAAI,IAAI,MAAM,EAAE;AACjE,gBAAQ;AAAA,MACV,CAAC;AAGD,WAAK,IAAI,GAAG,SAAS,MAAM,QAAQ,CAAC;AAAA,IACtC,CAAC;AAAA,EACH;AAAA,EAEA,oBAAoB,MAA+B;AACjD,QAAI,KAAK,YAAY;AACnB,WAAK,aAAa,KAAK;AAAA,IACzB;AAEA,QAAI,KAAK,SAAS,YAAY;AAC5B,YAAM,YAAY,KAAK;AAEvB,UAAI,cAAc,eAAe;AAC/B,YAAI,KAAK,UAAW;AAEpB,aAAK,YAAY;AACjB,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,kBAAI,gBAAgB;AAAA,UAC1B,WAAW,KAAK;AAAA,QAClB,CAAC;AAED,aAAK,qBAAqB,kBAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,UAAU;AACjC,YAAI,CAAC,KAAK,UAAW;AACrB,aAAK,qBAAqB,kBAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,kBAAkB;AACzC,YAAI,CAAC,KAAK,UAAW;AACrB,aAAK,qBAAqB,kBAAI,gBAAgB,sBAAsB,IAAI;AAAA,MAC1E,WAAW,cAAc,eAAe;AACtC,aAAK,qBAAqB,kBAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,aAAa;AACpC,YAAI,CAAC,KAAK,UAAW;AAErB,aAAK,YAAY;AACjB,aAAK,qBAAqB,kBAAI,gBAAgB,kBAAkB,IAAI;AAEpE,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,kBAAI,gBAAgB;AAAA,UAC1B,WAAW,KAAK;AAAA,QAClB,CAAC;AAAA,MACH;AAAA,IACF,WAAW,KAAK,SAAS,SAAS;AAChC,WAAK,QAAQ,KAAK,0BAA0B,EAAE,KAAK,CAAC;AACpD,YAAM,OAAQ,KAAK,eAA0B;AAC7C,YAAM,IAAI,MAAM,uBAAuB,IAAI,EAAE;AAAA,IAC/C;AAAA,EACF;AAAA,EAEA,qBAAqB,WAAgC,MAA+B;AAClF,UAAM,OAAO,mBAAmB,KAAK,MAAM,YAAY,MAAM,MAAM,KAAK,eAAe;AAEvF,QAAI,KAAK,SAAS,GAAG;AACnB,WAAK,MAAM,IAAI;AAAA,QACb,MAAM;AAAA,QACN,WAAW,KAAK;AAAA,QAChB,cAAc,CAAC,KAAK,CAAC,GAAI,GAAG,KAAK,MAAM,CAAC,CAAC;AAAA,MAC3C,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EAEA,uBAAuB,UAAkB;AACvC,UAAM,aAA8B;AAAA,MAClC,MAAM,kBAAI,gBAAgB;AAAA,MAC1B,WAAW,KAAK;AAAA,MAChB,kBAAkB;AAAA,QAChB,eAAe;AAAA,MACjB;AAAA,IACF;AACA,SAAK,MAAM,IAAI,UAAU;AAAA,EAC3B;AAAA,EAEA,kBAA0B;AACxB,UAAM,SAA4C;AAAA,MAChD,OAAO,KAAK,MAAM;AAAA,MAClB,aAAa,KAAK,MAAM,WAAW,SAAS;AAAA,MAC5C,UAAU;AAAA,MACV,aAAa,OAAO,KAAK,MAAM,SAAS;AAAA,IAC1C;AAGA,QAAI,KAAK,MAAM;AACb,aAAO,sBAAsB,KAAK,MAAM,kBAAkB,SAAS;AACrE,QAAI,KAAK,MAAM,aAAc,QAAO,gBAAgB,KAAK,MAAM,aAAa,SAAS;AACrF,QAAI,KAAK,MAAM,aAAc,QAAO,iBAAiB,KAAK,MAAM,aAAa,SAAS;AAEtF,QAAI,KAAK,MAAM,SAAS,SAAS,EAAG,QAAO,UAAU,KAAK,MAAM;AAChE,QAAI,KAAK,MAAM,QAAQ,KAAK,MAAM,KAAK,SAAS,EAAG,QAAO,MAAM,KAAK,MAAM;AAE3E,UAAM,UAAU,KAAK,MAAM,YAAY,QAAQ,SAAS,IAAI;AAC5D,UAAM,KAAK,YAAY,UAAU,MAAM;AACvC,WAAO,GAAG,OAAO,IAAI,EAAE;AAAA,EACzB;AAAA,EAES,QAAQ;AArcnB;AAscI,UAAM,MAAM;AACZ,eAAK,QAAL,mBAAU;AAAA,EACZ;AACF;AAIA,SAAS,mBACP,UACA,MACA,iBACkB;AAClB,QAAM,aAAa,KAAK;AACxB,QAAM,YAAa,KAAK,SAA4C,CAAC;AAErE,MAAI,CAAC,aAAa,UAAU,WAAW,GAAG;AACxC,WAAO,CAAC;AAAA,EACV;AAEA,MAAI,aAAa;AACjB,MAAI,UAAU,SAAS,GAAG;AACxB,UAAM,MAAM,UAAU,OAAO,CAAC,KAAa,MAAM,OAAQ,EAAE,cAAyB,IAAI,CAAC;AACzF,iBAAa,MAAM,UAAU;AAAA,EAC/B;AAEA,QAAM,KAAqB;AAAA,IACzB;AAAA,IACA,YAAa,KAAK,sBAAiC,KAAK;AAAA,IACxD,UAAW,KAAK,oBAA+B,KAAK;AAAA,IACpD;AAAA,IACA,MAAM,cAAc;AAAA;AAAA;AAAA;AAAA,IAIpB,OAAO,UAAU;AAAA,MAAI,CAAC,aACpB,iCAAkB;AAAA,QAChB,MAAO,KAAK,QAAmB;AAAA,QAC/B,YAAa,KAAK,SAAoB,KAAK;AAAA,QAC3C,UAAW,KAAK,OAAkB,KAAK;AAAA,QACvC,YAAa,KAAK,cAAyB;AAAA,QAC3C;AAAA,MACF,CAAC;AAAA,IACH;AAAA,EACF;AAEA,SAAO,CAAC,EAAE;AACZ;","names":["_a"]}
|
|
1
|
+
{"version":3,"sources":["../src/stt_v2.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n AudioByteStream,\n Event,\n calculateAudioDurationSeconds,\n createTimedString,\n log,\n normalizeLanguage,\n stt,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport * as queryString from 'node:querystring';\nimport { WebSocket } from 'ws';\nimport { PeriodicCollector } from './_utils.js';\nimport type { V2Models } from './models.js';\n\nconst _CLOSE_MSG = JSON.stringify({ type: 'CloseStream' });\n\n// --- Configuration ---\n\n/**\n * Configuration options for STTv2 (Deepgram Flux model).\n */\nexport interface STTv2Options {\n apiKey?: string;\n model: V2Models | string;\n sampleRate: number;\n keyterms: string[];\n endpointUrl: string;\n language?: string;\n eagerEotThreshold?: number;\n eotThreshold?: number;\n eotTimeoutMs?: number;\n mipOptOut?: boolean;\n tags?: string[];\n}\n\nconst defaultSTTv2Options: Omit<STTv2Options, 'apiKey'> = {\n model: 'flux-general-en',\n sampleRate: 16000,\n keyterms: [],\n endpointUrl: 'wss://api.deepgram.com/v2/listen',\n language: 'en',\n mipOptOut: false,\n};\n\nfunction validateTags(tags: string[]): string[] {\n for (const tag of tags) {\n if (tag.length > 128) {\n throw new Error('tag must be no more than 128 characters');\n }\n }\n return tags;\n}\n\n/**\n * Deepgram STTv2 using the Flux model for streaming speech-to-text.\n *\n * This uses Deepgram's V2 API (`/v2/listen`) which provides turn-based\n * transcription with support for preemptive generation.\n *\n * @remarks\n * Key differences from STT (V1):\n * - Uses `TurnInfo` events instead of `SpeechStarted`/`Results`\n * - Supports `eagerEotThreshold` for preemptive LLM generation\n * - Sends `PREFLIGHT_TRANSCRIPT` events when eager end-of-turn is detected\n *\n * @example\n * ```typescript\n * import { STTv2 } from '@livekit/agents-plugin-deepgram';\n *\n * const stt = new STTv2({\n * model: 'flux-general-en',\n * eagerEotThreshold: 0.5, // Enable preemptive generation\n * });\n *\n * const stream = stt.stream();\n * stream.pushFrame(audioFrame);\n *\n * for await (const event of stream) {\n * if (event.type === SpeechEventType.FINAL_TRANSCRIPT) {\n * console.log(event.alternatives?.[0]?.text);\n * }\n * }\n * ```\n */\nexport class STTv2 extends stt.STT {\n readonly label = 'deepgram.STTv2';\n #opts: STTv2Options;\n #apiKey: string;\n #logger = log();\n\n /**\n * Create a new Deepgram STTv2 instance.\n *\n * @param opts - Configuration options\n * @param opts.apiKey - Deepgram API key (defaults to `DEEPGRAM_API_KEY` env var)\n * @param opts.model - Model to use (default: `flux-general-en`)\n * @param opts.eagerEotThreshold - Threshold (0.3-0.9) for preemptive generation\n * @param opts.eotThreshold - End-of-turn detection threshold (default: 0.7)\n * @param opts.eotTimeoutMs - End-of-turn timeout in ms (default: 3000)\n * @param opts.keyterms - List of key terms to improve recognition\n * @param opts.tags - Tags for usage reporting (max 128 chars each)\n *\n * @throws Error if no API key is provided\n */\n constructor(opts: Partial<STTv2Options> = {}) {\n super({\n streaming: true,\n interimResults: true,\n alignedTranscript: 'word',\n });\n\n this.#opts = {\n ...defaultSTTv2Options,\n ...opts,\n language: opts.language ? normalizeLanguage(opts.language) : defaultSTTv2Options.language,\n };\n\n const apiKey = opts.apiKey || process.env.DEEPGRAM_API_KEY;\n if (!apiKey) {\n throw new Error('Deepgram API key is required');\n }\n this.#apiKey = apiKey;\n\n if (this.#opts.tags) {\n this.#opts.tags = validateTags(this.#opts.tags);\n }\n }\n\n /** The model being used for transcription */\n get model(): string {\n return this.#opts.model;\n }\n\n /** The STT provider name */\n get provider(): string {\n return 'Deepgram';\n }\n\n protected async _recognize(\n _frame: AudioFrame | AudioFrame[],\n _abortSignal?: AbortSignal,\n ): Promise<stt.SpeechEvent> {\n throw new Error('V2 API does not support non-streaming recognize. Use .stream()');\n }\n\n /**\n * Create a new streaming transcription session.\n *\n * @param options - Stream options\n * @returns A SpeechStream that emits transcription events\n */\n stream(options?: { connOptions?: APIConnectOptions }): stt.SpeechStream {\n const streamOpts = { ...this.#opts, apiKey: this.#apiKey };\n return new SpeechStreamv2(this, streamOpts, options?.connOptions);\n }\n\n /**\n * Update STT options. Changes will take effect on the next stream.\n *\n * @param opts - Partial options to update\n */\n updateOptions(opts: Partial<STTv2Options>) {\n this.#opts = { ...this.#opts, ...opts };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n this.#logger.debug('Updated STTv2 options');\n }\n}\n\n// --- Stream Implementation ---\n\nclass SpeechStreamv2 extends stt.SpeechStream {\n readonly label = 'deepgram.SpeechStreamv2';\n #opts: STTv2Options & { apiKey: string };\n #logger = log();\n #ws: WebSocket | null = null;\n\n #audioDurationCollector: PeriodicCollector<number>;\n #requestId = '';\n #speaking = false;\n\n // Parity: _reconnect_event - using existing Event class from @livekit/agents\n #reconnectEvent = new Event();\n\n constructor(\n sttInstance: STTv2,\n opts: STTv2Options & { apiKey: string },\n connOptions?: APIConnectOptions,\n ) {\n super(sttInstance, opts.sampleRate, connOptions);\n this.#opts = opts;\n\n this.#audioDurationCollector = new PeriodicCollector(\n (duration) => this.#onAudioDurationReport(duration),\n { duration: 5.0 },\n );\n }\n\n updateOptions(opts: Partial<STTv2Options>) {\n this.#logger.debug('Stream received option update', opts);\n this.#opts = {\n ...this.#opts,\n ...opts,\n language:\n opts.language !== undefined ? normalizeLanguage(opts.language) : this.#opts.language,\n };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n\n // Trigger reconnection loop\n this.#reconnectEvent.set();\n }\n\n protected async run() {\n // Outer Loop: Handles reconnections (Configuration updates)\n while (!this.closed) {\n try {\n this.#reconnectEvent.clear();\n\n const url = this.#getDeepgramUrl();\n this.#logger.debug(`Connecting to Deepgram: ${url}`);\n\n this.#ws = new WebSocket(url, {\n headers: { Authorization: `Token ${this.#opts.apiKey}` },\n });\n\n // 1. Wait for Connection Open\n await new Promise<void>((resolve, reject) => {\n if (!this.#ws) return reject(new Error('WebSocket not initialized'));\n\n const onOpen = () => {\n this.#ws?.off('error', onError);\n resolve();\n };\n const onError = (err: Error) => {\n this.#ws?.off('open', onOpen);\n reject(err);\n };\n\n this.#ws.once('open', onOpen);\n this.#ws.once('error', onError);\n });\n\n // 2. Run Concurrent Tasks (Send & Receive)\n const sendPromise = this.#sendTask();\n const recvPromise = this.#recvTask();\n const reconnectWait = this.#reconnectEvent.wait();\n\n // 3. Race: Normal Completion vs Reconnect Signal\n const result = await Promise.race([\n Promise.all([sendPromise, recvPromise]),\n reconnectWait.then(() => 'RECONNECT'),\n ]);\n\n if (result === 'RECONNECT') {\n this.#logger.debug('Reconnecting stream due to option update...');\n // Close current socket; loop will restart and open a new one\n this.#ws.close();\n } else {\n // Normal finish (Stream ended or Error thrown)\n break;\n }\n } catch (error) {\n this.#logger.error('Deepgram stream error', { error });\n throw error; // Let Base Class handle retry logic\n } finally {\n if (this.#ws?.readyState === WebSocket.OPEN) {\n this.#ws.close();\n }\n }\n }\n this.close();\n }\n\n async #sendTask() {\n if (!this.#ws) return;\n\n // Buffer audio into 50ms chunks (Parity)\n const samples50ms = Math.floor(this.#opts.sampleRate / 20);\n const audioBstream = new AudioByteStream(this.#opts.sampleRate, 1, samples50ms);\n\n let hasEnded = false;\n\n // Manual Iterator to allow racing against Reconnect Signal\n const iterator = this.input[Symbol.asyncIterator]();\n\n while (true) {\n const nextPromise = iterator.next();\n // If reconnect signal fires, abort the wait\n const abortPromise = this.#reconnectEvent.wait().then(() => ({ abort: true }) as const);\n\n const result = await Promise.race([nextPromise, abortPromise]);\n\n // Check if we need to abort (Reconnect) or if stream is done\n if ('abort' in result || result.done) {\n if (!('abort' in result) && result.done) {\n // Normal stream end\n hasEnded = true;\n } else {\n // Reconnect triggered - break loop immediately\n break;\n }\n }\n\n // If we broke above, we don't process data. If not, 'result' is IteratorResult\n if (hasEnded && result.value === undefined) {\n // Process flush below\n } else if ('value' in result) {\n const data = result.value;\n const frames: AudioFrame[] = [];\n\n if (data === SpeechStreamv2.FLUSH_SENTINEL) {\n frames.push(...audioBstream.flush());\n hasEnded = true;\n } else {\n frames.push(...audioBstream.write((data as AudioFrame).data.buffer as ArrayBuffer));\n }\n\n for (const frame of frames) {\n this.#audioDurationCollector.push(calculateAudioDurationSeconds(frame));\n\n if (this.#ws!.readyState === WebSocket.OPEN) {\n this.#ws!.send(frame.data);\n }\n\n if (hasEnded) {\n this.#audioDurationCollector.flush();\n hasEnded = false;\n }\n }\n }\n\n if (hasEnded) break;\n }\n\n // Only send CloseStream if we are exiting normally (not reconnecting)\n if (!this.#reconnectEvent.isSet && this.#ws!.readyState === WebSocket.OPEN) {\n this.#logger.debug('Sending CloseStream message to Deepgram');\n this.#ws!.send(_CLOSE_MSG);\n }\n }\n\n async #recvTask() {\n if (!this.#ws) return;\n\n return new Promise<void>((resolve) => {\n if (!this.#ws) return resolve();\n\n this.#ws.on('message', (data: Buffer, isBinary: boolean) => {\n if (isBinary) {\n this.#logger.warn('Received unexpected binary message from Deepgram');\n return;\n }\n try {\n const msg = JSON.parse(data.toString());\n this.#processStreamEvent(msg);\n } catch (error) {\n this.#logger.error('Failed to parse Deepgram message', { error });\n }\n });\n\n this.#ws.on('close', (code, reason) => {\n this.#logger.debug(`Deepgram WebSocket closed: ${code} ${reason}`);\n resolve();\n });\n\n // Errors are caught by run() listener, resolve here to clean up task\n this.#ws.on('error', () => resolve());\n });\n }\n\n #processStreamEvent(data: Record<string, unknown>) {\n if (data.request_id) {\n this.#requestId = data.request_id as string;\n }\n\n if (data.type === 'TurnInfo') {\n const eventType = data.event;\n\n if (eventType === 'StartOfTurn') {\n if (this.#speaking) return;\n\n this.#speaking = true;\n this.queue.put({\n type: stt.SpeechEventType.START_OF_SPEECH,\n requestId: this.#requestId,\n });\n\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'Update') {\n if (!this.#speaking) return;\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'EagerEndOfTurn') {\n if (!this.#speaking) return;\n this.#sendTranscriptEvent(stt.SpeechEventType.PREFLIGHT_TRANSCRIPT, data);\n } else if (eventType === 'TurnResumed') {\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'EndOfTurn') {\n if (!this.#speaking) return;\n\n this.#speaking = false;\n this.#sendTranscriptEvent(stt.SpeechEventType.FINAL_TRANSCRIPT, data);\n\n this.queue.put({\n type: stt.SpeechEventType.END_OF_SPEECH,\n requestId: this.#requestId,\n });\n }\n } else if (data.type === 'Error') {\n this.#logger.warn('deepgram sent an error', { data });\n const desc = (data.description as string) || 'unknown error from deepgram';\n throw new Error(`Deepgram API Error: ${desc}`);\n }\n }\n\n #sendTranscriptEvent(eventType: stt.SpeechEventType, data: Record<string, unknown>) {\n const alts = parseTranscription(this.#opts.language || 'en', data, this.startTimeOffset);\n\n if (alts.length > 0) {\n this.queue.put({\n type: eventType,\n requestId: this.#requestId,\n alternatives: [alts[0]!, ...alts.slice(1)],\n });\n }\n }\n\n #onAudioDurationReport(duration: number) {\n const usageEvent: stt.SpeechEvent = {\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: {\n audioDuration: duration,\n },\n };\n this.queue.put(usageEvent);\n }\n\n #getDeepgramUrl(): string {\n const params: Record<string, string | string[]> = {\n model: this.#opts.model,\n sample_rate: this.#opts.sampleRate.toString(),\n encoding: 'linear16',\n mip_opt_out: String(this.#opts.mipOptOut),\n };\n\n // Note: v2 API does NOT include 'language' parameter\n if (this.#opts.eagerEotThreshold)\n params.eager_eot_threshold = this.#opts.eagerEotThreshold.toString();\n if (this.#opts.eotThreshold) params.eot_threshold = this.#opts.eotThreshold.toString();\n if (this.#opts.eotTimeoutMs) params.eot_timeout_ms = this.#opts.eotTimeoutMs.toString();\n\n if (this.#opts.keyterms.length > 0) params.keyterm = this.#opts.keyterms;\n if (this.#opts.tags && this.#opts.tags.length > 0) params.tag = this.#opts.tags;\n\n const baseUrl = this.#opts.endpointUrl.replace(/^http/, 'ws');\n const qs = queryString.stringify(params);\n return `${baseUrl}?${qs}`;\n }\n\n override close() {\n super.close();\n this.#ws?.close();\n }\n}\n\n// --- Helpers ---\n\nfunction parseTranscription(\n language: string,\n data: Record<string, unknown>,\n startTimeOffset: number,\n): stt.SpeechData[] {\n const transcript = data.transcript as string | undefined;\n const wordsData = (data.words as Array<Record<string, unknown>>) || [];\n\n if (!wordsData || wordsData.length === 0) {\n return [];\n }\n\n let confidence = 0;\n if (wordsData.length > 0) {\n const sum = wordsData.reduce((acc: number, w) => acc + ((w.confidence as number) || 0), 0);\n confidence = sum / wordsData.length;\n }\n\n const sd: stt.SpeechData = {\n language: normalizeLanguage(language),\n startTime: ((data.audio_window_start as number) || 0) + startTimeOffset,\n endTime: ((data.audio_window_end as number) || 0) + startTimeOffset,\n confidence: confidence,\n text: transcript || '',\n // Note: Deepgram V2 (Flux) API does not provide word-level timing (start/end).\n // Words only contain 'word' and 'confidence' fields, so startTime/endTime will be 0.\n // See: https://developers.deepgram.com/docs/flux/nova-3-migration\n words: wordsData.map((word) =>\n createTimedString({\n text: (word.word as string) ?? '',\n startTime: ((word.start as number) ?? 0) + startTimeOffset,\n endTime: ((word.end as number) ?? 0) + startTimeOffset,\n confidence: (word.confidence as number) ?? 0.0,\n startTimeOffset,\n }),\n ),\n };\n\n return [sd];\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBASO;AAEP,kBAA6B;AAC7B,gBAA0B;AAC1B,mBAAkC;AAGlC,MAAM,aAAa,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC;AAqBzD,MAAM,sBAAoD;AAAA,EACxD,OAAO;AAAA,EACP,YAAY;AAAA,EACZ,UAAU,CAAC;AAAA,EACX,aAAa;AAAA,EACb,UAAU;AAAA,EACV,WAAW;AACb;AAEA,SAAS,aAAa,MAA0B;AAC9C,aAAW,OAAO,MAAM;AACtB,QAAI,IAAI,SAAS,KAAK;AACpB,YAAM,IAAI,MAAM,yCAAyC;AAAA,IAC3D;AAAA,EACF;AACA,SAAO;AACT;AAiCO,MAAM,cAAc,kBAAI,IAAI;AAAA,EACxB,QAAQ;AAAA,EACjB;AAAA,EACA;AAAA,EACA,cAAU,mBAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAgBd,YAAY,OAA8B,CAAC,GAAG;AAC5C,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB;AAAA,MAChB,mBAAmB;AAAA,IACrB,CAAC;AAED,SAAK,QAAQ;AAAA,MACX,GAAG;AAAA,MACH,GAAG;AAAA,MACH,UAAU,KAAK,eAAW,iCAAkB,KAAK,QAAQ,IAAI,oBAAoB;AAAA,IACnF;AAEA,UAAM,SAAS,KAAK,UAAU,QAAQ,IAAI;AAC1C,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI,MAAM,8BAA8B;AAAA,IAChD;AACA,SAAK,UAAU;AAEf,QAAI,KAAK,MAAM,MAAM;AACnB,WAAK,MAAM,OAAO,aAAa,KAAK,MAAM,IAAI;AAAA,IAChD;AAAA,EACF;AAAA;AAAA,EAGA,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA;AAAA,EAGA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,MAAgB,WACd,QACA,cAC0B;AAC1B,UAAM,IAAI,MAAM,gEAAgE;AAAA,EAClF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,OAAO,SAAiE;AACtE,UAAM,aAAa,EAAE,GAAG,KAAK,OAAO,QAAQ,KAAK,QAAQ;AACzD,WAAO,IAAI,eAAe,MAAM,YAAY,mCAAS,WAAW;AAAA,EAClE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,cAAc,MAA6B;AACzC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AACvD,SAAK,QAAQ,MAAM,uBAAuB;AAAA,EAC5C;AACF;AAIA,MAAM,uBAAuB,kBAAI,aAAa;AAAA,EACnC,QAAQ;AAAA,EACjB;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,MAAwB;AAAA,EAExB;AAAA,EACA,aAAa;AAAA,EACb,YAAY;AAAA;AAAA,EAGZ,kBAAkB,IAAI,oBAAM;AAAA,EAE5B,YACE,aACA,MACA,aACA;AACA,UAAM,aAAa,KAAK,YAAY,WAAW;AAC/C,SAAK,QAAQ;AAEb,SAAK,0BAA0B,IAAI;AAAA,MACjC,CAAC,aAAa,KAAK,uBAAuB,QAAQ;AAAA,MAClD,EAAE,UAAU,EAAI;AAAA,IAClB;AAAA,EACF;AAAA,EAEA,cAAc,MAA6B;AACzC,SAAK,QAAQ,MAAM,iCAAiC,IAAI;AACxD,SAAK,QAAQ;AAAA,MACX,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,MACH,UACE,KAAK,aAAa,aAAY,iCAAkB,KAAK,QAAQ,IAAI,KAAK,MAAM;AAAA,IAChF;AACA,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AAGvD,SAAK,gBAAgB,IAAI;AAAA,EAC3B;AAAA,EAEA,MAAgB,MAAM;AAxNxB;AA0NI,WAAO,CAAC,KAAK,QAAQ;AACnB,UAAI;AACF,aAAK,gBAAgB,MAAM;AAE3B,cAAM,MAAM,KAAK,gBAAgB;AACjC,aAAK,QAAQ,MAAM,2BAA2B,GAAG,EAAE;AAEnD,aAAK,MAAM,IAAI,oBAAU,KAAK;AAAA,UAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,QACzD,CAAC;AAGD,cAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,cAAI,CAAC,KAAK,IAAK,QAAO,OAAO,IAAI,MAAM,2BAA2B,CAAC;AAEnE,gBAAM,SAAS,MAAM;AAzO/B,gBAAAA;AA0OY,aAAAA,MAAA,KAAK,QAAL,gBAAAA,IAAU,IAAI,SAAS;AACvB,oBAAQ;AAAA,UACV;AACA,gBAAM,UAAU,CAAC,QAAe;AA7O1C,gBAAAA;AA8OY,aAAAA,MAAA,KAAK,QAAL,gBAAAA,IAAU,IAAI,QAAQ;AACtB,mBAAO,GAAG;AAAA,UACZ;AAEA,eAAK,IAAI,KAAK,QAAQ,MAAM;AAC5B,eAAK,IAAI,KAAK,SAAS,OAAO;AAAA,QAChC,CAAC;AAGD,cAAM,cAAc,KAAK,UAAU;AACnC,cAAM,cAAc,KAAK,UAAU;AACnC,cAAM,gBAAgB,KAAK,gBAAgB,KAAK;AAGhD,cAAM,SAAS,MAAM,QAAQ,KAAK;AAAA,UAChC,QAAQ,IAAI,CAAC,aAAa,WAAW,CAAC;AAAA,UACtC,cAAc,KAAK,MAAM,WAAW;AAAA,QACtC,CAAC;AAED,YAAI,WAAW,aAAa;AAC1B,eAAK,QAAQ,MAAM,6CAA6C;AAEhE,eAAK,IAAI,MAAM;AAAA,QACjB,OAAO;AAEL;AAAA,QACF;AAAA,MACF,SAAS,OAAO;AACd,aAAK,QAAQ,MAAM,yBAAyB,EAAE,MAAM,CAAC;AACrD,cAAM;AAAA,MACR,UAAE;AACA,cAAI,UAAK,QAAL,mBAAU,gBAAe,oBAAU,MAAM;AAC3C,eAAK,IAAI,MAAM;AAAA,QACjB;AAAA,MACF;AAAA,IACF;AACA,SAAK,MAAM;AAAA,EACb;AAAA,EAEA,MAAM,YAAY;AAChB,QAAI,CAAC,KAAK,IAAK;AAGf,UAAM,cAAc,KAAK,MAAM,KAAK,MAAM,aAAa,EAAE;AACzD,UAAM,eAAe,IAAI,8BAAgB,KAAK,MAAM,YAAY,GAAG,WAAW;AAE9E,QAAI,WAAW;AAGf,UAAM,WAAW,KAAK,MAAM,OAAO,aAAa,EAAE;AAElD,WAAO,MAAM;AACX,YAAM,cAAc,SAAS,KAAK;AAElC,YAAM,eAAe,KAAK,gBAAgB,KAAK,EAAE,KAAK,OAAO,EAAE,OAAO,KAAK,EAAW;AAEtF,YAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,aAAa,YAAY,CAAC;AAG7D,UAAI,WAAW,UAAU,OAAO,MAAM;AACpC,YAAI,EAAE,WAAW,WAAW,OAAO,MAAM;AAEvC,qBAAW;AAAA,QACb,OAAO;AAEL;AAAA,QACF;AAAA,MACF;AAGA,UAAI,YAAY,OAAO,UAAU,QAAW;AAAA,MAE5C,WAAW,WAAW,QAAQ;AAC5B,cAAM,OAAO,OAAO;AACpB,cAAM,SAAuB,CAAC;AAE9B,YAAI,SAAS,eAAe,gBAAgB;AAC1C,iBAAO,KAAK,GAAG,aAAa,MAAM,CAAC;AACnC,qBAAW;AAAA,QACb,OAAO;AACL,iBAAO,KAAK,GAAG,aAAa,MAAO,KAAoB,KAAK,MAAqB,CAAC;AAAA,QACpF;AAEA,mBAAW,SAAS,QAAQ;AAC1B,eAAK,wBAAwB,SAAK,6CAA8B,KAAK,CAAC;AAEtE,cAAI,KAAK,IAAK,eAAe,oBAAU,MAAM;AAC3C,iBAAK,IAAK,KAAK,MAAM,IAAI;AAAA,UAC3B;AAEA,cAAI,UAAU;AACZ,iBAAK,wBAAwB,MAAM;AACnC,uBAAW;AAAA,UACb;AAAA,QACF;AAAA,MACF;AAEA,UAAI,SAAU;AAAA,IAChB;AAGA,QAAI,CAAC,KAAK,gBAAgB,SAAS,KAAK,IAAK,eAAe,oBAAU,MAAM;AAC1E,WAAK,QAAQ,MAAM,yCAAyC;AAC5D,WAAK,IAAK,KAAK,UAAU;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,MAAM,YAAY;AAChB,QAAI,CAAC,KAAK,IAAK;AAEf,WAAO,IAAI,QAAc,CAAC,YAAY;AACpC,UAAI,CAAC,KAAK,IAAK,QAAO,QAAQ;AAE9B,WAAK,IAAI,GAAG,WAAW,CAAC,MAAc,aAAsB;AAC1D,YAAI,UAAU;AACZ,eAAK,QAAQ,KAAK,kDAAkD;AACpE;AAAA,QACF;AACA,YAAI;AACF,gBAAM,MAAM,KAAK,MAAM,KAAK,SAAS,CAAC;AACtC,eAAK,oBAAoB,GAAG;AAAA,QAC9B,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,oCAAoC,EAAE,MAAM,CAAC;AAAA,QAClE;AAAA,MACF,CAAC;AAED,WAAK,IAAI,GAAG,SAAS,CAAC,MAAM,WAAW;AACrC,aAAK,QAAQ,MAAM,8BAA8B,IAAI,IAAI,MAAM,EAAE;AACjE,gBAAQ;AAAA,MACV,CAAC;AAGD,WAAK,IAAI,GAAG,SAAS,MAAM,QAAQ,CAAC;AAAA,IACtC,CAAC;AAAA,EACH;AAAA,EAEA,oBAAoB,MAA+B;AACjD,QAAI,KAAK,YAAY;AACnB,WAAK,aAAa,KAAK;AAAA,IACzB;AAEA,QAAI,KAAK,SAAS,YAAY;AAC5B,YAAM,YAAY,KAAK;AAEvB,UAAI,cAAc,eAAe;AAC/B,YAAI,KAAK,UAAW;AAEpB,aAAK,YAAY;AACjB,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,kBAAI,gBAAgB;AAAA,UAC1B,WAAW,KAAK;AAAA,QAClB,CAAC;AAED,aAAK,qBAAqB,kBAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,UAAU;AACjC,YAAI,CAAC,KAAK,UAAW;AACrB,aAAK,qBAAqB,kBAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,kBAAkB;AACzC,YAAI,CAAC,KAAK,UAAW;AACrB,aAAK,qBAAqB,kBAAI,gBAAgB,sBAAsB,IAAI;AAAA,MAC1E,WAAW,cAAc,eAAe;AACtC,aAAK,qBAAqB,kBAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,aAAa;AACpC,YAAI,CAAC,KAAK,UAAW;AAErB,aAAK,YAAY;AACjB,aAAK,qBAAqB,kBAAI,gBAAgB,kBAAkB,IAAI;AAEpE,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,kBAAI,gBAAgB;AAAA,UAC1B,WAAW,KAAK;AAAA,QAClB,CAAC;AAAA,MACH;AAAA,IACF,WAAW,KAAK,SAAS,SAAS;AAChC,WAAK,QAAQ,KAAK,0BAA0B,EAAE,KAAK,CAAC;AACpD,YAAM,OAAQ,KAAK,eAA0B;AAC7C,YAAM,IAAI,MAAM,uBAAuB,IAAI,EAAE;AAAA,IAC/C;AAAA,EACF;AAAA,EAEA,qBAAqB,WAAgC,MAA+B;AAClF,UAAM,OAAO,mBAAmB,KAAK,MAAM,YAAY,MAAM,MAAM,KAAK,eAAe;AAEvF,QAAI,KAAK,SAAS,GAAG;AACnB,WAAK,MAAM,IAAI;AAAA,QACb,MAAM;AAAA,QACN,WAAW,KAAK;AAAA,QAChB,cAAc,CAAC,KAAK,CAAC,GAAI,GAAG,KAAK,MAAM,CAAC,CAAC;AAAA,MAC3C,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EAEA,uBAAuB,UAAkB;AACvC,UAAM,aAA8B;AAAA,MAClC,MAAM,kBAAI,gBAAgB;AAAA,MAC1B,WAAW,KAAK;AAAA,MAChB,kBAAkB;AAAA,QAChB,eAAe;AAAA,MACjB;AAAA,IACF;AACA,SAAK,MAAM,IAAI,UAAU;AAAA,EAC3B;AAAA,EAEA,kBAA0B;AACxB,UAAM,SAA4C;AAAA,MAChD,OAAO,KAAK,MAAM;AAAA,MAClB,aAAa,KAAK,MAAM,WAAW,SAAS;AAAA,MAC5C,UAAU;AAAA,MACV,aAAa,OAAO,KAAK,MAAM,SAAS;AAAA,IAC1C;AAGA,QAAI,KAAK,MAAM;AACb,aAAO,sBAAsB,KAAK,MAAM,kBAAkB,SAAS;AACrE,QAAI,KAAK,MAAM,aAAc,QAAO,gBAAgB,KAAK,MAAM,aAAa,SAAS;AACrF,QAAI,KAAK,MAAM,aAAc,QAAO,iBAAiB,KAAK,MAAM,aAAa,SAAS;AAEtF,QAAI,KAAK,MAAM,SAAS,SAAS,EAAG,QAAO,UAAU,KAAK,MAAM;AAChE,QAAI,KAAK,MAAM,QAAQ,KAAK,MAAM,KAAK,SAAS,EAAG,QAAO,MAAM,KAAK,MAAM;AAE3E,UAAM,UAAU,KAAK,MAAM,YAAY,QAAQ,SAAS,IAAI;AAC5D,UAAM,KAAK,YAAY,UAAU,MAAM;AACvC,WAAO,GAAG,OAAO,IAAI,EAAE;AAAA,EACzB;AAAA,EAES,QAAQ;AA/cnB;AAgdI,UAAM,MAAM;AACZ,eAAK,QAAL,mBAAU;AAAA,EACZ;AACF;AAIA,SAAS,mBACP,UACA,MACA,iBACkB;AAClB,QAAM,aAAa,KAAK;AACxB,QAAM,YAAa,KAAK,SAA4C,CAAC;AAErE,MAAI,CAAC,aAAa,UAAU,WAAW,GAAG;AACxC,WAAO,CAAC;AAAA,EACV;AAEA,MAAI,aAAa;AACjB,MAAI,UAAU,SAAS,GAAG;AACxB,UAAM,MAAM,UAAU,OAAO,CAAC,KAAa,MAAM,OAAQ,EAAE,cAAyB,IAAI,CAAC;AACzF,iBAAa,MAAM,UAAU;AAAA,EAC/B;AAEA,QAAM,KAAqB;AAAA,IACzB,cAAU,iCAAkB,QAAQ;AAAA,IACpC,YAAa,KAAK,sBAAiC,KAAK;AAAA,IACxD,UAAW,KAAK,oBAA+B,KAAK;AAAA,IACpD;AAAA,IACA,MAAM,cAAc;AAAA;AAAA;AAAA;AAAA,IAIpB,OAAO,UAAU;AAAA,MAAI,CAAC,aACpB,iCAAkB;AAAA,QAChB,MAAO,KAAK,QAAmB;AAAA,QAC/B,YAAa,KAAK,SAAoB,KAAK;AAAA,QAC3C,UAAW,KAAK,OAAkB,KAAK;AAAA,QACvC,YAAa,KAAK,cAAyB;AAAA,QAC3C;AAAA,MACF,CAAC;AAAA,IACH;AAAA,EACF;AAEA,SAAO,CAAC,EAAE;AACZ;","names":["_a"]}
|
package/dist/stt_v2.d.ts.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"stt_v2.d.ts","sourceRoot":"","sources":["../src/stt_v2.ts"],"names":[],"mappings":"AAGA,OAAO,EACL,KAAK,iBAAiB,
|
|
1
|
+
{"version":3,"file":"stt_v2.d.ts","sourceRoot":"","sources":["../src/stt_v2.ts"],"names":[],"mappings":"AAGA,OAAO,EACL,KAAK,iBAAiB,EAOtB,GAAG,EACJ,MAAM,iBAAiB,CAAC;AACzB,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AAIpD,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,aAAa,CAAC;AAM5C;;GAEG;AACH,MAAM,WAAW,YAAY;IAC3B,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,KAAK,EAAE,QAAQ,GAAG,MAAM,CAAC;IACzB,UAAU,EAAE,MAAM,CAAC;IACnB,QAAQ,EAAE,MAAM,EAAE,CAAC;IACnB,WAAW,EAAE,MAAM,CAAC;IACpB,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,iBAAiB,CAAC,EAAE,MAAM,CAAC;IAC3B,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,SAAS,CAAC,EAAE,OAAO,CAAC;IACpB,IAAI,CAAC,EAAE,MAAM,EAAE,CAAC;CACjB;AAoBD;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA8BG;AACH,qBAAa,KAAM,SAAQ,GAAG,CAAC,GAAG;;IAChC,QAAQ,CAAC,KAAK,oBAAoB;IAKlC;;;;;;;;;;;;;OAaG;gBACS,IAAI,GAAE,OAAO,CAAC,YAAY,CAAM;IAwB5C,6CAA6C;IAC7C,IAAI,KAAK,IAAI,MAAM,CAElB;IAED,4BAA4B;IAC5B,IAAI,QAAQ,IAAI,MAAM,CAErB;cAEe,UAAU,CACxB,MAAM,EAAE,UAAU,GAAG,UAAU,EAAE,EACjC,YAAY,CAAC,EAAE,WAAW,GACzB,OAAO,CAAC,GAAG,CAAC,WAAW,CAAC;IAI3B;;;;;OAKG;IACH,MAAM,CAAC,OAAO,CAAC,EAAE;QAAE,WAAW,CAAC,EAAE,iBAAiB,CAAA;KAAE,GAAG,GAAG,CAAC,YAAY;IAKvE;;;;OAIG;IACH,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,YAAY,CAAC;CAK1C"}
|
package/dist/stt_v2.js
CHANGED
|
@@ -4,6 +4,7 @@ import {
|
|
|
4
4
|
calculateAudioDurationSeconds,
|
|
5
5
|
createTimedString,
|
|
6
6
|
log,
|
|
7
|
+
normalizeLanguage,
|
|
7
8
|
stt
|
|
8
9
|
} from "@livekit/agents";
|
|
9
10
|
import * as queryString from "node:querystring";
|
|
@@ -51,7 +52,11 @@ class STTv2 extends stt.STT {
|
|
|
51
52
|
interimResults: true,
|
|
52
53
|
alignedTranscript: "word"
|
|
53
54
|
});
|
|
54
|
-
this.#opts = {
|
|
55
|
+
this.#opts = {
|
|
56
|
+
...defaultSTTv2Options,
|
|
57
|
+
...opts,
|
|
58
|
+
language: opts.language ? normalizeLanguage(opts.language) : defaultSTTv2Options.language
|
|
59
|
+
};
|
|
55
60
|
const apiKey = opts.apiKey || process.env.DEEPGRAM_API_KEY;
|
|
56
61
|
if (!apiKey) {
|
|
57
62
|
throw new Error("Deepgram API key is required");
|
|
@@ -113,7 +118,11 @@ class SpeechStreamv2 extends stt.SpeechStream {
|
|
|
113
118
|
}
|
|
114
119
|
updateOptions(opts) {
|
|
115
120
|
this.#logger.debug("Stream received option update", opts);
|
|
116
|
-
this.#opts = {
|
|
121
|
+
this.#opts = {
|
|
122
|
+
...this.#opts,
|
|
123
|
+
...opts,
|
|
124
|
+
language: opts.language !== void 0 ? normalizeLanguage(opts.language) : this.#opts.language
|
|
125
|
+
};
|
|
117
126
|
if (opts.tags) this.#opts.tags = validateTags(opts.tags);
|
|
118
127
|
this.#reconnectEvent.set();
|
|
119
128
|
}
|
|
@@ -326,7 +335,7 @@ function parseTranscription(language, data, startTimeOffset) {
|
|
|
326
335
|
confidence = sum / wordsData.length;
|
|
327
336
|
}
|
|
328
337
|
const sd = {
|
|
329
|
-
language,
|
|
338
|
+
language: normalizeLanguage(language),
|
|
330
339
|
startTime: (data.audio_window_start || 0) + startTimeOffset,
|
|
331
340
|
endTime: (data.audio_window_end || 0) + startTimeOffset,
|
|
332
341
|
confidence,
|
package/dist/stt_v2.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/stt_v2.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n AudioByteStream,\n Event,\n calculateAudioDurationSeconds,\n createTimedString,\n log,\n stt,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport * as queryString from 'node:querystring';\nimport { WebSocket } from 'ws';\nimport { PeriodicCollector } from './_utils.js';\nimport type { V2Models } from './models.js';\n\nconst _CLOSE_MSG = JSON.stringify({ type: 'CloseStream' });\n\n// --- Configuration ---\n\n/**\n * Configuration options for STTv2 (Deepgram Flux model).\n */\nexport interface STTv2Options {\n apiKey?: string;\n model: V2Models | string;\n sampleRate: number;\n keyterms: string[];\n endpointUrl: string;\n language?: string;\n eagerEotThreshold?: number;\n eotThreshold?: number;\n eotTimeoutMs?: number;\n mipOptOut?: boolean;\n tags?: string[];\n}\n\nconst defaultSTTv2Options: Omit<STTv2Options, 'apiKey'> = {\n model: 'flux-general-en',\n sampleRate: 16000,\n keyterms: [],\n endpointUrl: 'wss://api.deepgram.com/v2/listen',\n language: 'en',\n mipOptOut: false,\n};\n\nfunction validateTags(tags: string[]): string[] {\n for (const tag of tags) {\n if (tag.length > 128) {\n throw new Error('tag must be no more than 128 characters');\n }\n }\n return tags;\n}\n\n/**\n * Deepgram STTv2 using the Flux model for streaming speech-to-text.\n *\n * This uses Deepgram's V2 API (`/v2/listen`) which provides turn-based\n * transcription with support for preemptive generation.\n *\n * @remarks\n * Key differences from STT (V1):\n * - Uses `TurnInfo` events instead of `SpeechStarted`/`Results`\n * - Supports `eagerEotThreshold` for preemptive LLM generation\n * - Sends `PREFLIGHT_TRANSCRIPT` events when eager end-of-turn is detected\n *\n * @example\n * ```typescript\n * import { STTv2 } from '@livekit/agents-plugin-deepgram';\n *\n * const stt = new STTv2({\n * model: 'flux-general-en',\n * eagerEotThreshold: 0.5, // Enable preemptive generation\n * });\n *\n * const stream = stt.stream();\n * stream.pushFrame(audioFrame);\n *\n * for await (const event of stream) {\n * if (event.type === SpeechEventType.FINAL_TRANSCRIPT) {\n * console.log(event.alternatives?.[0]?.text);\n * }\n * }\n * ```\n */\nexport class STTv2 extends stt.STT {\n readonly label = 'deepgram.STTv2';\n #opts: STTv2Options;\n #apiKey: string;\n #logger = log();\n\n /**\n * Create a new Deepgram STTv2 instance.\n *\n * @param opts - Configuration options\n * @param opts.apiKey - Deepgram API key (defaults to `DEEPGRAM_API_KEY` env var)\n * @param opts.model - Model to use (default: `flux-general-en`)\n * @param opts.eagerEotThreshold - Threshold (0.3-0.9) for preemptive generation\n * @param opts.eotThreshold - End-of-turn detection threshold (default: 0.7)\n * @param opts.eotTimeoutMs - End-of-turn timeout in ms (default: 3000)\n * @param opts.keyterms - List of key terms to improve recognition\n * @param opts.tags - Tags for usage reporting (max 128 chars each)\n *\n * @throws Error if no API key is provided\n */\n constructor(opts: Partial<STTv2Options> = {}) {\n super({\n streaming: true,\n interimResults: true,\n alignedTranscript: 'word',\n });\n\n this.#opts = { ...defaultSTTv2Options, ...opts };\n\n const apiKey = opts.apiKey || process.env.DEEPGRAM_API_KEY;\n if (!apiKey) {\n throw new Error('Deepgram API key is required');\n }\n this.#apiKey = apiKey;\n\n if (this.#opts.tags) {\n this.#opts.tags = validateTags(this.#opts.tags);\n }\n }\n\n /** The model being used for transcription */\n get model(): string {\n return this.#opts.model;\n }\n\n /** The STT provider name */\n get provider(): string {\n return 'Deepgram';\n }\n\n protected async _recognize(\n _frame: AudioFrame | AudioFrame[],\n _abortSignal?: AbortSignal,\n ): Promise<stt.SpeechEvent> {\n throw new Error('V2 API does not support non-streaming recognize. Use .stream()');\n }\n\n /**\n * Create a new streaming transcription session.\n *\n * @param options - Stream options\n * @returns A SpeechStream that emits transcription events\n */\n stream(options?: { connOptions?: APIConnectOptions }): stt.SpeechStream {\n const streamOpts = { ...this.#opts, apiKey: this.#apiKey };\n return new SpeechStreamv2(this, streamOpts, options?.connOptions);\n }\n\n /**\n * Update STT options. Changes will take effect on the next stream.\n *\n * @param opts - Partial options to update\n */\n updateOptions(opts: Partial<STTv2Options>) {\n this.#opts = { ...this.#opts, ...opts };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n this.#logger.debug('Updated STTv2 options');\n }\n}\n\n// --- Stream Implementation ---\n\nclass SpeechStreamv2 extends stt.SpeechStream {\n readonly label = 'deepgram.SpeechStreamv2';\n #opts: STTv2Options & { apiKey: string };\n #logger = log();\n #ws: WebSocket | null = null;\n\n #audioDurationCollector: PeriodicCollector<number>;\n #requestId = '';\n #speaking = false;\n\n // Parity: _reconnect_event - using existing Event class from @livekit/agents\n #reconnectEvent = new Event();\n\n constructor(\n sttInstance: STTv2,\n opts: STTv2Options & { apiKey: string },\n connOptions?: APIConnectOptions,\n ) {\n super(sttInstance, opts.sampleRate, connOptions);\n this.#opts = opts;\n\n this.#audioDurationCollector = new PeriodicCollector(\n (duration) => this.#onAudioDurationReport(duration),\n { duration: 5.0 },\n );\n }\n\n updateOptions(opts: Partial<STTv2Options>) {\n this.#logger.debug('Stream received option update', opts);\n this.#opts = { ...this.#opts, ...opts };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n\n // Trigger reconnection loop\n this.#reconnectEvent.set();\n }\n\n protected async run() {\n // Outer Loop: Handles reconnections (Configuration updates)\n while (!this.closed) {\n try {\n this.#reconnectEvent.clear();\n\n const url = this.#getDeepgramUrl();\n this.#logger.debug(`Connecting to Deepgram: ${url}`);\n\n this.#ws = new WebSocket(url, {\n headers: { Authorization: `Token ${this.#opts.apiKey}` },\n });\n\n // 1. Wait for Connection Open\n await new Promise<void>((resolve, reject) => {\n if (!this.#ws) return reject(new Error('WebSocket not initialized'));\n\n const onOpen = () => {\n this.#ws?.off('error', onError);\n resolve();\n };\n const onError = (err: Error) => {\n this.#ws?.off('open', onOpen);\n reject(err);\n };\n\n this.#ws.once('open', onOpen);\n this.#ws.once('error', onError);\n });\n\n // 2. Run Concurrent Tasks (Send & Receive)\n const sendPromise = this.#sendTask();\n const recvPromise = this.#recvTask();\n const reconnectWait = this.#reconnectEvent.wait();\n\n // 3. Race: Normal Completion vs Reconnect Signal\n const result = await Promise.race([\n Promise.all([sendPromise, recvPromise]),\n reconnectWait.then(() => 'RECONNECT'),\n ]);\n\n if (result === 'RECONNECT') {\n this.#logger.debug('Reconnecting stream due to option update...');\n // Close current socket; loop will restart and open a new one\n this.#ws.close();\n } else {\n // Normal finish (Stream ended or Error thrown)\n break;\n }\n } catch (error) {\n this.#logger.error('Deepgram stream error', { error });\n throw error; // Let Base Class handle retry logic\n } finally {\n if (this.#ws?.readyState === WebSocket.OPEN) {\n this.#ws.close();\n }\n }\n }\n this.close();\n }\n\n async #sendTask() {\n if (!this.#ws) return;\n\n // Buffer audio into 50ms chunks (Parity)\n const samples50ms = Math.floor(this.#opts.sampleRate / 20);\n const audioBstream = new AudioByteStream(this.#opts.sampleRate, 1, samples50ms);\n\n let hasEnded = false;\n\n // Manual Iterator to allow racing against Reconnect Signal\n const iterator = this.input[Symbol.asyncIterator]();\n\n while (true) {\n const nextPromise = iterator.next();\n // If reconnect signal fires, abort the wait\n const abortPromise = this.#reconnectEvent.wait().then(() => ({ abort: true }) as const);\n\n const result = await Promise.race([nextPromise, abortPromise]);\n\n // Check if we need to abort (Reconnect) or if stream is done\n if ('abort' in result || result.done) {\n if (!('abort' in result) && result.done) {\n // Normal stream end\n hasEnded = true;\n } else {\n // Reconnect triggered - break loop immediately\n break;\n }\n }\n\n // If we broke above, we don't process data. If not, 'result' is IteratorResult\n if (hasEnded && result.value === undefined) {\n // Process flush below\n } else if ('value' in result) {\n const data = result.value;\n const frames: AudioFrame[] = [];\n\n if (data === SpeechStreamv2.FLUSH_SENTINEL) {\n frames.push(...audioBstream.flush());\n hasEnded = true;\n } else {\n frames.push(...audioBstream.write((data as AudioFrame).data.buffer as ArrayBuffer));\n }\n\n for (const frame of frames) {\n this.#audioDurationCollector.push(calculateAudioDurationSeconds(frame));\n\n if (this.#ws!.readyState === WebSocket.OPEN) {\n this.#ws!.send(frame.data);\n }\n\n if (hasEnded) {\n this.#audioDurationCollector.flush();\n hasEnded = false;\n }\n }\n }\n\n if (hasEnded) break;\n }\n\n // Only send CloseStream if we are exiting normally (not reconnecting)\n if (!this.#reconnectEvent.isSet && this.#ws!.readyState === WebSocket.OPEN) {\n this.#logger.debug('Sending CloseStream message to Deepgram');\n this.#ws!.send(_CLOSE_MSG);\n }\n }\n\n async #recvTask() {\n if (!this.#ws) return;\n\n return new Promise<void>((resolve) => {\n if (!this.#ws) return resolve();\n\n this.#ws.on('message', (data: Buffer, isBinary: boolean) => {\n if (isBinary) {\n this.#logger.warn('Received unexpected binary message from Deepgram');\n return;\n }\n try {\n const msg = JSON.parse(data.toString());\n this.#processStreamEvent(msg);\n } catch (error) {\n this.#logger.error('Failed to parse Deepgram message', { error });\n }\n });\n\n this.#ws.on('close', (code, reason) => {\n this.#logger.debug(`Deepgram WebSocket closed: ${code} ${reason}`);\n resolve();\n });\n\n // Errors are caught by run() listener, resolve here to clean up task\n this.#ws.on('error', () => resolve());\n });\n }\n\n #processStreamEvent(data: Record<string, unknown>) {\n if (data.request_id) {\n this.#requestId = data.request_id as string;\n }\n\n if (data.type === 'TurnInfo') {\n const eventType = data.event;\n\n if (eventType === 'StartOfTurn') {\n if (this.#speaking) return;\n\n this.#speaking = true;\n this.queue.put({\n type: stt.SpeechEventType.START_OF_SPEECH,\n requestId: this.#requestId,\n });\n\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'Update') {\n if (!this.#speaking) return;\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'EagerEndOfTurn') {\n if (!this.#speaking) return;\n this.#sendTranscriptEvent(stt.SpeechEventType.PREFLIGHT_TRANSCRIPT, data);\n } else if (eventType === 'TurnResumed') {\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'EndOfTurn') {\n if (!this.#speaking) return;\n\n this.#speaking = false;\n this.#sendTranscriptEvent(stt.SpeechEventType.FINAL_TRANSCRIPT, data);\n\n this.queue.put({\n type: stt.SpeechEventType.END_OF_SPEECH,\n requestId: this.#requestId,\n });\n }\n } else if (data.type === 'Error') {\n this.#logger.warn('deepgram sent an error', { data });\n const desc = (data.description as string) || 'unknown error from deepgram';\n throw new Error(`Deepgram API Error: ${desc}`);\n }\n }\n\n #sendTranscriptEvent(eventType: stt.SpeechEventType, data: Record<string, unknown>) {\n const alts = parseTranscription(this.#opts.language || 'en', data, this.startTimeOffset);\n\n if (alts.length > 0) {\n this.queue.put({\n type: eventType,\n requestId: this.#requestId,\n alternatives: [alts[0]!, ...alts.slice(1)],\n });\n }\n }\n\n #onAudioDurationReport(duration: number) {\n const usageEvent: stt.SpeechEvent = {\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: {\n audioDuration: duration,\n },\n };\n this.queue.put(usageEvent);\n }\n\n #getDeepgramUrl(): string {\n const params: Record<string, string | string[]> = {\n model: this.#opts.model,\n sample_rate: this.#opts.sampleRate.toString(),\n encoding: 'linear16',\n mip_opt_out: String(this.#opts.mipOptOut),\n };\n\n // Note: v2 API does NOT include 'language' parameter\n if (this.#opts.eagerEotThreshold)\n params.eager_eot_threshold = this.#opts.eagerEotThreshold.toString();\n if (this.#opts.eotThreshold) params.eot_threshold = this.#opts.eotThreshold.toString();\n if (this.#opts.eotTimeoutMs) params.eot_timeout_ms = this.#opts.eotTimeoutMs.toString();\n\n if (this.#opts.keyterms.length > 0) params.keyterm = this.#opts.keyterms;\n if (this.#opts.tags && this.#opts.tags.length > 0) params.tag = this.#opts.tags;\n\n const baseUrl = this.#opts.endpointUrl.replace(/^http/, 'ws');\n const qs = queryString.stringify(params);\n return `${baseUrl}?${qs}`;\n }\n\n override close() {\n super.close();\n this.#ws?.close();\n }\n}\n\n// --- Helpers ---\n\nfunction parseTranscription(\n language: string,\n data: Record<string, unknown>,\n startTimeOffset: number,\n): stt.SpeechData[] {\n const transcript = data.transcript as string | undefined;\n const wordsData = (data.words as Array<Record<string, unknown>>) || [];\n\n if (!wordsData || wordsData.length === 0) {\n return [];\n }\n\n let confidence = 0;\n if (wordsData.length > 0) {\n const sum = wordsData.reduce((acc: number, w) => acc + ((w.confidence as number) || 0), 0);\n confidence = sum / wordsData.length;\n }\n\n const sd: stt.SpeechData = {\n language: language,\n startTime: ((data.audio_window_start as number) || 0) + startTimeOffset,\n endTime: ((data.audio_window_end as number) || 0) + startTimeOffset,\n confidence: confidence,\n text: transcript || '',\n // Note: Deepgram V2 (Flux) API does not provide word-level timing (start/end).\n // Words only contain 'word' and 'confidence' fields, so startTime/endTime will be 0.\n // See: https://developers.deepgram.com/docs/flux/nova-3-migration\n words: wordsData.map((word) =>\n createTimedString({\n text: (word.word as string) ?? '',\n startTime: ((word.start as number) ?? 0) + startTimeOffset,\n endTime: ((word.end as number) ?? 0) + startTimeOffset,\n confidence: (word.confidence as number) ?? 0.0,\n startTimeOffset,\n }),\n ),\n };\n\n return [sd];\n}\n"],"mappings":"AAGA;AAAA,EAEE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,YAAY,iBAAiB;AAC7B,SAAS,iBAAiB;AAC1B,SAAS,yBAAyB;AAGlC,MAAM,aAAa,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC;AAqBzD,MAAM,sBAAoD;AAAA,EACxD,OAAO;AAAA,EACP,YAAY;AAAA,EACZ,UAAU,CAAC;AAAA,EACX,aAAa;AAAA,EACb,UAAU;AAAA,EACV,WAAW;AACb;AAEA,SAAS,aAAa,MAA0B;AAC9C,aAAW,OAAO,MAAM;AACtB,QAAI,IAAI,SAAS,KAAK;AACpB,YAAM,IAAI,MAAM,yCAAyC;AAAA,IAC3D;AAAA,EACF;AACA,SAAO;AACT;AAiCO,MAAM,cAAc,IAAI,IAAI;AAAA,EACxB,QAAQ;AAAA,EACjB;AAAA,EACA;AAAA,EACA,UAAU,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAgBd,YAAY,OAA8B,CAAC,GAAG;AAC5C,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB;AAAA,MAChB,mBAAmB;AAAA,IACrB,CAAC;AAED,SAAK,QAAQ,EAAE,GAAG,qBAAqB,GAAG,KAAK;AAE/C,UAAM,SAAS,KAAK,UAAU,QAAQ,IAAI;AAC1C,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI,MAAM,8BAA8B;AAAA,IAChD;AACA,SAAK,UAAU;AAEf,QAAI,KAAK,MAAM,MAAM;AACnB,WAAK,MAAM,OAAO,aAAa,KAAK,MAAM,IAAI;AAAA,IAChD;AAAA,EACF;AAAA;AAAA,EAGA,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA;AAAA,EAGA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,MAAgB,WACd,QACA,cAC0B;AAC1B,UAAM,IAAI,MAAM,gEAAgE;AAAA,EAClF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,OAAO,SAAiE;AACtE,UAAM,aAAa,EAAE,GAAG,KAAK,OAAO,QAAQ,KAAK,QAAQ;AACzD,WAAO,IAAI,eAAe,MAAM,YAAY,mCAAS,WAAW;AAAA,EAClE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,cAAc,MAA6B;AACzC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AACvD,SAAK,QAAQ,MAAM,uBAAuB;AAAA,EAC5C;AACF;AAIA,MAAM,uBAAuB,IAAI,aAAa;AAAA,EACnC,QAAQ;AAAA,EACjB;AAAA,EACA,UAAU,IAAI;AAAA,EACd,MAAwB;AAAA,EAExB;AAAA,EACA,aAAa;AAAA,EACb,YAAY;AAAA;AAAA,EAGZ,kBAAkB,IAAI,MAAM;AAAA,EAE5B,YACE,aACA,MACA,aACA;AACA,UAAM,aAAa,KAAK,YAAY,WAAW;AAC/C,SAAK,QAAQ;AAEb,SAAK,0BAA0B,IAAI;AAAA,MACjC,CAAC,aAAa,KAAK,uBAAuB,QAAQ;AAAA,MAClD,EAAE,UAAU,EAAI;AAAA,IAClB;AAAA,EACF;AAAA,EAEA,cAAc,MAA6B;AACzC,SAAK,QAAQ,MAAM,iCAAiC,IAAI;AACxD,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AAGvD,SAAK,gBAAgB,IAAI;AAAA,EAC3B;AAAA,EAEA,MAAgB,MAAM;AA9MxB;AAgNI,WAAO,CAAC,KAAK,QAAQ;AACnB,UAAI;AACF,aAAK,gBAAgB,MAAM;AAE3B,cAAM,MAAM,KAAK,gBAAgB;AACjC,aAAK,QAAQ,MAAM,2BAA2B,GAAG,EAAE;AAEnD,aAAK,MAAM,IAAI,UAAU,KAAK;AAAA,UAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,QACzD,CAAC;AAGD,cAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,cAAI,CAAC,KAAK,IAAK,QAAO,OAAO,IAAI,MAAM,2BAA2B,CAAC;AAEnE,gBAAM,SAAS,MAAM;AA/N/B,gBAAAA;AAgOY,aAAAA,MAAA,KAAK,QAAL,gBAAAA,IAAU,IAAI,SAAS;AACvB,oBAAQ;AAAA,UACV;AACA,gBAAM,UAAU,CAAC,QAAe;AAnO1C,gBAAAA;AAoOY,aAAAA,MAAA,KAAK,QAAL,gBAAAA,IAAU,IAAI,QAAQ;AACtB,mBAAO,GAAG;AAAA,UACZ;AAEA,eAAK,IAAI,KAAK,QAAQ,MAAM;AAC5B,eAAK,IAAI,KAAK,SAAS,OAAO;AAAA,QAChC,CAAC;AAGD,cAAM,cAAc,KAAK,UAAU;AACnC,cAAM,cAAc,KAAK,UAAU;AACnC,cAAM,gBAAgB,KAAK,gBAAgB,KAAK;AAGhD,cAAM,SAAS,MAAM,QAAQ,KAAK;AAAA,UAChC,QAAQ,IAAI,CAAC,aAAa,WAAW,CAAC;AAAA,UACtC,cAAc,KAAK,MAAM,WAAW;AAAA,QACtC,CAAC;AAED,YAAI,WAAW,aAAa;AAC1B,eAAK,QAAQ,MAAM,6CAA6C;AAEhE,eAAK,IAAI,MAAM;AAAA,QACjB,OAAO;AAEL;AAAA,QACF;AAAA,MACF,SAAS,OAAO;AACd,aAAK,QAAQ,MAAM,yBAAyB,EAAE,MAAM,CAAC;AACrD,cAAM;AAAA,MACR,UAAE;AACA,cAAI,UAAK,QAAL,mBAAU,gBAAe,UAAU,MAAM;AAC3C,eAAK,IAAI,MAAM;AAAA,QACjB;AAAA,MACF;AAAA,IACF;AACA,SAAK,MAAM;AAAA,EACb;AAAA,EAEA,MAAM,YAAY;AAChB,QAAI,CAAC,KAAK,IAAK;AAGf,UAAM,cAAc,KAAK,MAAM,KAAK,MAAM,aAAa,EAAE;AACzD,UAAM,eAAe,IAAI,gBAAgB,KAAK,MAAM,YAAY,GAAG,WAAW;AAE9E,QAAI,WAAW;AAGf,UAAM,WAAW,KAAK,MAAM,OAAO,aAAa,EAAE;AAElD,WAAO,MAAM;AACX,YAAM,cAAc,SAAS,KAAK;AAElC,YAAM,eAAe,KAAK,gBAAgB,KAAK,EAAE,KAAK,OAAO,EAAE,OAAO,KAAK,EAAW;AAEtF,YAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,aAAa,YAAY,CAAC;AAG7D,UAAI,WAAW,UAAU,OAAO,MAAM;AACpC,YAAI,EAAE,WAAW,WAAW,OAAO,MAAM;AAEvC,qBAAW;AAAA,QACb,OAAO;AAEL;AAAA,QACF;AAAA,MACF;AAGA,UAAI,YAAY,OAAO,UAAU,QAAW;AAAA,MAE5C,WAAW,WAAW,QAAQ;AAC5B,cAAM,OAAO,OAAO;AACpB,cAAM,SAAuB,CAAC;AAE9B,YAAI,SAAS,eAAe,gBAAgB;AAC1C,iBAAO,KAAK,GAAG,aAAa,MAAM,CAAC;AACnC,qBAAW;AAAA,QACb,OAAO;AACL,iBAAO,KAAK,GAAG,aAAa,MAAO,KAAoB,KAAK,MAAqB,CAAC;AAAA,QACpF;AAEA,mBAAW,SAAS,QAAQ;AAC1B,eAAK,wBAAwB,KAAK,8BAA8B,KAAK,CAAC;AAEtE,cAAI,KAAK,IAAK,eAAe,UAAU,MAAM;AAC3C,iBAAK,IAAK,KAAK,MAAM,IAAI;AAAA,UAC3B;AAEA,cAAI,UAAU;AACZ,iBAAK,wBAAwB,MAAM;AACnC,uBAAW;AAAA,UACb;AAAA,QACF;AAAA,MACF;AAEA,UAAI,SAAU;AAAA,IAChB;AAGA,QAAI,CAAC,KAAK,gBAAgB,SAAS,KAAK,IAAK,eAAe,UAAU,MAAM;AAC1E,WAAK,QAAQ,MAAM,yCAAyC;AAC5D,WAAK,IAAK,KAAK,UAAU;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,MAAM,YAAY;AAChB,QAAI,CAAC,KAAK,IAAK;AAEf,WAAO,IAAI,QAAc,CAAC,YAAY;AACpC,UAAI,CAAC,KAAK,IAAK,QAAO,QAAQ;AAE9B,WAAK,IAAI,GAAG,WAAW,CAAC,MAAc,aAAsB;AAC1D,YAAI,UAAU;AACZ,eAAK,QAAQ,KAAK,kDAAkD;AACpE;AAAA,QACF;AACA,YAAI;AACF,gBAAM,MAAM,KAAK,MAAM,KAAK,SAAS,CAAC;AACtC,eAAK,oBAAoB,GAAG;AAAA,QAC9B,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,oCAAoC,EAAE,MAAM,CAAC;AAAA,QAClE;AAAA,MACF,CAAC;AAED,WAAK,IAAI,GAAG,SAAS,CAAC,MAAM,WAAW;AACrC,aAAK,QAAQ,MAAM,8BAA8B,IAAI,IAAI,MAAM,EAAE;AACjE,gBAAQ;AAAA,MACV,CAAC;AAGD,WAAK,IAAI,GAAG,SAAS,MAAM,QAAQ,CAAC;AAAA,IACtC,CAAC;AAAA,EACH;AAAA,EAEA,oBAAoB,MAA+B;AACjD,QAAI,KAAK,YAAY;AACnB,WAAK,aAAa,KAAK;AAAA,IACzB;AAEA,QAAI,KAAK,SAAS,YAAY;AAC5B,YAAM,YAAY,KAAK;AAEvB,UAAI,cAAc,eAAe;AAC/B,YAAI,KAAK,UAAW;AAEpB,aAAK,YAAY;AACjB,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,IAAI,gBAAgB;AAAA,UAC1B,WAAW,KAAK;AAAA,QAClB,CAAC;AAED,aAAK,qBAAqB,IAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,UAAU;AACjC,YAAI,CAAC,KAAK,UAAW;AACrB,aAAK,qBAAqB,IAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,kBAAkB;AACzC,YAAI,CAAC,KAAK,UAAW;AACrB,aAAK,qBAAqB,IAAI,gBAAgB,sBAAsB,IAAI;AAAA,MAC1E,WAAW,cAAc,eAAe;AACtC,aAAK,qBAAqB,IAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,aAAa;AACpC,YAAI,CAAC,KAAK,UAAW;AAErB,aAAK,YAAY;AACjB,aAAK,qBAAqB,IAAI,gBAAgB,kBAAkB,IAAI;AAEpE,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,IAAI,gBAAgB;AAAA,UAC1B,WAAW,KAAK;AAAA,QAClB,CAAC;AAAA,MACH;AAAA,IACF,WAAW,KAAK,SAAS,SAAS;AAChC,WAAK,QAAQ,KAAK,0BAA0B,EAAE,KAAK,CAAC;AACpD,YAAM,OAAQ,KAAK,eAA0B;AAC7C,YAAM,IAAI,MAAM,uBAAuB,IAAI,EAAE;AAAA,IAC/C;AAAA,EACF;AAAA,EAEA,qBAAqB,WAAgC,MAA+B;AAClF,UAAM,OAAO,mBAAmB,KAAK,MAAM,YAAY,MAAM,MAAM,KAAK,eAAe;AAEvF,QAAI,KAAK,SAAS,GAAG;AACnB,WAAK,MAAM,IAAI;AAAA,QACb,MAAM;AAAA,QACN,WAAW,KAAK;AAAA,QAChB,cAAc,CAAC,KAAK,CAAC,GAAI,GAAG,KAAK,MAAM,CAAC,CAAC;AAAA,MAC3C,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EAEA,uBAAuB,UAAkB;AACvC,UAAM,aAA8B;AAAA,MAClC,MAAM,IAAI,gBAAgB;AAAA,MAC1B,WAAW,KAAK;AAAA,MAChB,kBAAkB;AAAA,QAChB,eAAe;AAAA,MACjB;AAAA,IACF;AACA,SAAK,MAAM,IAAI,UAAU;AAAA,EAC3B;AAAA,EAEA,kBAA0B;AACxB,UAAM,SAA4C;AAAA,MAChD,OAAO,KAAK,MAAM;AAAA,MAClB,aAAa,KAAK,MAAM,WAAW,SAAS;AAAA,MAC5C,UAAU;AAAA,MACV,aAAa,OAAO,KAAK,MAAM,SAAS;AAAA,IAC1C;AAGA,QAAI,KAAK,MAAM;AACb,aAAO,sBAAsB,KAAK,MAAM,kBAAkB,SAAS;AACrE,QAAI,KAAK,MAAM,aAAc,QAAO,gBAAgB,KAAK,MAAM,aAAa,SAAS;AACrF,QAAI,KAAK,MAAM,aAAc,QAAO,iBAAiB,KAAK,MAAM,aAAa,SAAS;AAEtF,QAAI,KAAK,MAAM,SAAS,SAAS,EAAG,QAAO,UAAU,KAAK,MAAM;AAChE,QAAI,KAAK,MAAM,QAAQ,KAAK,MAAM,KAAK,SAAS,EAAG,QAAO,MAAM,KAAK,MAAM;AAE3E,UAAM,UAAU,KAAK,MAAM,YAAY,QAAQ,SAAS,IAAI;AAC5D,UAAM,KAAK,YAAY,UAAU,MAAM;AACvC,WAAO,GAAG,OAAO,IAAI,EAAE;AAAA,EACzB;AAAA,EAES,QAAQ;AArcnB;AAscI,UAAM,MAAM;AACZ,eAAK,QAAL,mBAAU;AAAA,EACZ;AACF;AAIA,SAAS,mBACP,UACA,MACA,iBACkB;AAClB,QAAM,aAAa,KAAK;AACxB,QAAM,YAAa,KAAK,SAA4C,CAAC;AAErE,MAAI,CAAC,aAAa,UAAU,WAAW,GAAG;AACxC,WAAO,CAAC;AAAA,EACV;AAEA,MAAI,aAAa;AACjB,MAAI,UAAU,SAAS,GAAG;AACxB,UAAM,MAAM,UAAU,OAAO,CAAC,KAAa,MAAM,OAAQ,EAAE,cAAyB,IAAI,CAAC;AACzF,iBAAa,MAAM,UAAU;AAAA,EAC/B;AAEA,QAAM,KAAqB;AAAA,IACzB;AAAA,IACA,YAAa,KAAK,sBAAiC,KAAK;AAAA,IACxD,UAAW,KAAK,oBAA+B,KAAK;AAAA,IACpD;AAAA,IACA,MAAM,cAAc;AAAA;AAAA;AAAA;AAAA,IAIpB,OAAO,UAAU;AAAA,MAAI,CAAC,SACpB,kBAAkB;AAAA,QAChB,MAAO,KAAK,QAAmB;AAAA,QAC/B,YAAa,KAAK,SAAoB,KAAK;AAAA,QAC3C,UAAW,KAAK,OAAkB,KAAK;AAAA,QACvC,YAAa,KAAK,cAAyB;AAAA,QAC3C;AAAA,MACF,CAAC;AAAA,IACH;AAAA,EACF;AAEA,SAAO,CAAC,EAAE;AACZ;","names":["_a"]}
|
|
1
|
+
{"version":3,"sources":["../src/stt_v2.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n type APIConnectOptions,\n AudioByteStream,\n Event,\n calculateAudioDurationSeconds,\n createTimedString,\n log,\n normalizeLanguage,\n stt,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport * as queryString from 'node:querystring';\nimport { WebSocket } from 'ws';\nimport { PeriodicCollector } from './_utils.js';\nimport type { V2Models } from './models.js';\n\nconst _CLOSE_MSG = JSON.stringify({ type: 'CloseStream' });\n\n// --- Configuration ---\n\n/**\n * Configuration options for STTv2 (Deepgram Flux model).\n */\nexport interface STTv2Options {\n apiKey?: string;\n model: V2Models | string;\n sampleRate: number;\n keyterms: string[];\n endpointUrl: string;\n language?: string;\n eagerEotThreshold?: number;\n eotThreshold?: number;\n eotTimeoutMs?: number;\n mipOptOut?: boolean;\n tags?: string[];\n}\n\nconst defaultSTTv2Options: Omit<STTv2Options, 'apiKey'> = {\n model: 'flux-general-en',\n sampleRate: 16000,\n keyterms: [],\n endpointUrl: 'wss://api.deepgram.com/v2/listen',\n language: 'en',\n mipOptOut: false,\n};\n\nfunction validateTags(tags: string[]): string[] {\n for (const tag of tags) {\n if (tag.length > 128) {\n throw new Error('tag must be no more than 128 characters');\n }\n }\n return tags;\n}\n\n/**\n * Deepgram STTv2 using the Flux model for streaming speech-to-text.\n *\n * This uses Deepgram's V2 API (`/v2/listen`) which provides turn-based\n * transcription with support for preemptive generation.\n *\n * @remarks\n * Key differences from STT (V1):\n * - Uses `TurnInfo` events instead of `SpeechStarted`/`Results`\n * - Supports `eagerEotThreshold` for preemptive LLM generation\n * - Sends `PREFLIGHT_TRANSCRIPT` events when eager end-of-turn is detected\n *\n * @example\n * ```typescript\n * import { STTv2 } from '@livekit/agents-plugin-deepgram';\n *\n * const stt = new STTv2({\n * model: 'flux-general-en',\n * eagerEotThreshold: 0.5, // Enable preemptive generation\n * });\n *\n * const stream = stt.stream();\n * stream.pushFrame(audioFrame);\n *\n * for await (const event of stream) {\n * if (event.type === SpeechEventType.FINAL_TRANSCRIPT) {\n * console.log(event.alternatives?.[0]?.text);\n * }\n * }\n * ```\n */\nexport class STTv2 extends stt.STT {\n readonly label = 'deepgram.STTv2';\n #opts: STTv2Options;\n #apiKey: string;\n #logger = log();\n\n /**\n * Create a new Deepgram STTv2 instance.\n *\n * @param opts - Configuration options\n * @param opts.apiKey - Deepgram API key (defaults to `DEEPGRAM_API_KEY` env var)\n * @param opts.model - Model to use (default: `flux-general-en`)\n * @param opts.eagerEotThreshold - Threshold (0.3-0.9) for preemptive generation\n * @param opts.eotThreshold - End-of-turn detection threshold (default: 0.7)\n * @param opts.eotTimeoutMs - End-of-turn timeout in ms (default: 3000)\n * @param opts.keyterms - List of key terms to improve recognition\n * @param opts.tags - Tags for usage reporting (max 128 chars each)\n *\n * @throws Error if no API key is provided\n */\n constructor(opts: Partial<STTv2Options> = {}) {\n super({\n streaming: true,\n interimResults: true,\n alignedTranscript: 'word',\n });\n\n this.#opts = {\n ...defaultSTTv2Options,\n ...opts,\n language: opts.language ? normalizeLanguage(opts.language) : defaultSTTv2Options.language,\n };\n\n const apiKey = opts.apiKey || process.env.DEEPGRAM_API_KEY;\n if (!apiKey) {\n throw new Error('Deepgram API key is required');\n }\n this.#apiKey = apiKey;\n\n if (this.#opts.tags) {\n this.#opts.tags = validateTags(this.#opts.tags);\n }\n }\n\n /** The model being used for transcription */\n get model(): string {\n return this.#opts.model;\n }\n\n /** The STT provider name */\n get provider(): string {\n return 'Deepgram';\n }\n\n protected async _recognize(\n _frame: AudioFrame | AudioFrame[],\n _abortSignal?: AbortSignal,\n ): Promise<stt.SpeechEvent> {\n throw new Error('V2 API does not support non-streaming recognize. Use .stream()');\n }\n\n /**\n * Create a new streaming transcription session.\n *\n * @param options - Stream options\n * @returns A SpeechStream that emits transcription events\n */\n stream(options?: { connOptions?: APIConnectOptions }): stt.SpeechStream {\n const streamOpts = { ...this.#opts, apiKey: this.#apiKey };\n return new SpeechStreamv2(this, streamOpts, options?.connOptions);\n }\n\n /**\n * Update STT options. Changes will take effect on the next stream.\n *\n * @param opts - Partial options to update\n */\n updateOptions(opts: Partial<STTv2Options>) {\n this.#opts = { ...this.#opts, ...opts };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n this.#logger.debug('Updated STTv2 options');\n }\n}\n\n// --- Stream Implementation ---\n\nclass SpeechStreamv2 extends stt.SpeechStream {\n readonly label = 'deepgram.SpeechStreamv2';\n #opts: STTv2Options & { apiKey: string };\n #logger = log();\n #ws: WebSocket | null = null;\n\n #audioDurationCollector: PeriodicCollector<number>;\n #requestId = '';\n #speaking = false;\n\n // Parity: _reconnect_event - using existing Event class from @livekit/agents\n #reconnectEvent = new Event();\n\n constructor(\n sttInstance: STTv2,\n opts: STTv2Options & { apiKey: string },\n connOptions?: APIConnectOptions,\n ) {\n super(sttInstance, opts.sampleRate, connOptions);\n this.#opts = opts;\n\n this.#audioDurationCollector = new PeriodicCollector(\n (duration) => this.#onAudioDurationReport(duration),\n { duration: 5.0 },\n );\n }\n\n updateOptions(opts: Partial<STTv2Options>) {\n this.#logger.debug('Stream received option update', opts);\n this.#opts = {\n ...this.#opts,\n ...opts,\n language:\n opts.language !== undefined ? normalizeLanguage(opts.language) : this.#opts.language,\n };\n if (opts.tags) this.#opts.tags = validateTags(opts.tags);\n\n // Trigger reconnection loop\n this.#reconnectEvent.set();\n }\n\n protected async run() {\n // Outer Loop: Handles reconnections (Configuration updates)\n while (!this.closed) {\n try {\n this.#reconnectEvent.clear();\n\n const url = this.#getDeepgramUrl();\n this.#logger.debug(`Connecting to Deepgram: ${url}`);\n\n this.#ws = new WebSocket(url, {\n headers: { Authorization: `Token ${this.#opts.apiKey}` },\n });\n\n // 1. Wait for Connection Open\n await new Promise<void>((resolve, reject) => {\n if (!this.#ws) return reject(new Error('WebSocket not initialized'));\n\n const onOpen = () => {\n this.#ws?.off('error', onError);\n resolve();\n };\n const onError = (err: Error) => {\n this.#ws?.off('open', onOpen);\n reject(err);\n };\n\n this.#ws.once('open', onOpen);\n this.#ws.once('error', onError);\n });\n\n // 2. Run Concurrent Tasks (Send & Receive)\n const sendPromise = this.#sendTask();\n const recvPromise = this.#recvTask();\n const reconnectWait = this.#reconnectEvent.wait();\n\n // 3. Race: Normal Completion vs Reconnect Signal\n const result = await Promise.race([\n Promise.all([sendPromise, recvPromise]),\n reconnectWait.then(() => 'RECONNECT'),\n ]);\n\n if (result === 'RECONNECT') {\n this.#logger.debug('Reconnecting stream due to option update...');\n // Close current socket; loop will restart and open a new one\n this.#ws.close();\n } else {\n // Normal finish (Stream ended or Error thrown)\n break;\n }\n } catch (error) {\n this.#logger.error('Deepgram stream error', { error });\n throw error; // Let Base Class handle retry logic\n } finally {\n if (this.#ws?.readyState === WebSocket.OPEN) {\n this.#ws.close();\n }\n }\n }\n this.close();\n }\n\n async #sendTask() {\n if (!this.#ws) return;\n\n // Buffer audio into 50ms chunks (Parity)\n const samples50ms = Math.floor(this.#opts.sampleRate / 20);\n const audioBstream = new AudioByteStream(this.#opts.sampleRate, 1, samples50ms);\n\n let hasEnded = false;\n\n // Manual Iterator to allow racing against Reconnect Signal\n const iterator = this.input[Symbol.asyncIterator]();\n\n while (true) {\n const nextPromise = iterator.next();\n // If reconnect signal fires, abort the wait\n const abortPromise = this.#reconnectEvent.wait().then(() => ({ abort: true }) as const);\n\n const result = await Promise.race([nextPromise, abortPromise]);\n\n // Check if we need to abort (Reconnect) or if stream is done\n if ('abort' in result || result.done) {\n if (!('abort' in result) && result.done) {\n // Normal stream end\n hasEnded = true;\n } else {\n // Reconnect triggered - break loop immediately\n break;\n }\n }\n\n // If we broke above, we don't process data. If not, 'result' is IteratorResult\n if (hasEnded && result.value === undefined) {\n // Process flush below\n } else if ('value' in result) {\n const data = result.value;\n const frames: AudioFrame[] = [];\n\n if (data === SpeechStreamv2.FLUSH_SENTINEL) {\n frames.push(...audioBstream.flush());\n hasEnded = true;\n } else {\n frames.push(...audioBstream.write((data as AudioFrame).data.buffer as ArrayBuffer));\n }\n\n for (const frame of frames) {\n this.#audioDurationCollector.push(calculateAudioDurationSeconds(frame));\n\n if (this.#ws!.readyState === WebSocket.OPEN) {\n this.#ws!.send(frame.data);\n }\n\n if (hasEnded) {\n this.#audioDurationCollector.flush();\n hasEnded = false;\n }\n }\n }\n\n if (hasEnded) break;\n }\n\n // Only send CloseStream if we are exiting normally (not reconnecting)\n if (!this.#reconnectEvent.isSet && this.#ws!.readyState === WebSocket.OPEN) {\n this.#logger.debug('Sending CloseStream message to Deepgram');\n this.#ws!.send(_CLOSE_MSG);\n }\n }\n\n async #recvTask() {\n if (!this.#ws) return;\n\n return new Promise<void>((resolve) => {\n if (!this.#ws) return resolve();\n\n this.#ws.on('message', (data: Buffer, isBinary: boolean) => {\n if (isBinary) {\n this.#logger.warn('Received unexpected binary message from Deepgram');\n return;\n }\n try {\n const msg = JSON.parse(data.toString());\n this.#processStreamEvent(msg);\n } catch (error) {\n this.#logger.error('Failed to parse Deepgram message', { error });\n }\n });\n\n this.#ws.on('close', (code, reason) => {\n this.#logger.debug(`Deepgram WebSocket closed: ${code} ${reason}`);\n resolve();\n });\n\n // Errors are caught by run() listener, resolve here to clean up task\n this.#ws.on('error', () => resolve());\n });\n }\n\n #processStreamEvent(data: Record<string, unknown>) {\n if (data.request_id) {\n this.#requestId = data.request_id as string;\n }\n\n if (data.type === 'TurnInfo') {\n const eventType = data.event;\n\n if (eventType === 'StartOfTurn') {\n if (this.#speaking) return;\n\n this.#speaking = true;\n this.queue.put({\n type: stt.SpeechEventType.START_OF_SPEECH,\n requestId: this.#requestId,\n });\n\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'Update') {\n if (!this.#speaking) return;\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'EagerEndOfTurn') {\n if (!this.#speaking) return;\n this.#sendTranscriptEvent(stt.SpeechEventType.PREFLIGHT_TRANSCRIPT, data);\n } else if (eventType === 'TurnResumed') {\n this.#sendTranscriptEvent(stt.SpeechEventType.INTERIM_TRANSCRIPT, data);\n } else if (eventType === 'EndOfTurn') {\n if (!this.#speaking) return;\n\n this.#speaking = false;\n this.#sendTranscriptEvent(stt.SpeechEventType.FINAL_TRANSCRIPT, data);\n\n this.queue.put({\n type: stt.SpeechEventType.END_OF_SPEECH,\n requestId: this.#requestId,\n });\n }\n } else if (data.type === 'Error') {\n this.#logger.warn('deepgram sent an error', { data });\n const desc = (data.description as string) || 'unknown error from deepgram';\n throw new Error(`Deepgram API Error: ${desc}`);\n }\n }\n\n #sendTranscriptEvent(eventType: stt.SpeechEventType, data: Record<string, unknown>) {\n const alts = parseTranscription(this.#opts.language || 'en', data, this.startTimeOffset);\n\n if (alts.length > 0) {\n this.queue.put({\n type: eventType,\n requestId: this.#requestId,\n alternatives: [alts[0]!, ...alts.slice(1)],\n });\n }\n }\n\n #onAudioDurationReport(duration: number) {\n const usageEvent: stt.SpeechEvent = {\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n requestId: this.#requestId,\n recognitionUsage: {\n audioDuration: duration,\n },\n };\n this.queue.put(usageEvent);\n }\n\n #getDeepgramUrl(): string {\n const params: Record<string, string | string[]> = {\n model: this.#opts.model,\n sample_rate: this.#opts.sampleRate.toString(),\n encoding: 'linear16',\n mip_opt_out: String(this.#opts.mipOptOut),\n };\n\n // Note: v2 API does NOT include 'language' parameter\n if (this.#opts.eagerEotThreshold)\n params.eager_eot_threshold = this.#opts.eagerEotThreshold.toString();\n if (this.#opts.eotThreshold) params.eot_threshold = this.#opts.eotThreshold.toString();\n if (this.#opts.eotTimeoutMs) params.eot_timeout_ms = this.#opts.eotTimeoutMs.toString();\n\n if (this.#opts.keyterms.length > 0) params.keyterm = this.#opts.keyterms;\n if (this.#opts.tags && this.#opts.tags.length > 0) params.tag = this.#opts.tags;\n\n const baseUrl = this.#opts.endpointUrl.replace(/^http/, 'ws');\n const qs = queryString.stringify(params);\n return `${baseUrl}?${qs}`;\n }\n\n override close() {\n super.close();\n this.#ws?.close();\n }\n}\n\n// --- Helpers ---\n\nfunction parseTranscription(\n language: string,\n data: Record<string, unknown>,\n startTimeOffset: number,\n): stt.SpeechData[] {\n const transcript = data.transcript as string | undefined;\n const wordsData = (data.words as Array<Record<string, unknown>>) || [];\n\n if (!wordsData || wordsData.length === 0) {\n return [];\n }\n\n let confidence = 0;\n if (wordsData.length > 0) {\n const sum = wordsData.reduce((acc: number, w) => acc + ((w.confidence as number) || 0), 0);\n confidence = sum / wordsData.length;\n }\n\n const sd: stt.SpeechData = {\n language: normalizeLanguage(language),\n startTime: ((data.audio_window_start as number) || 0) + startTimeOffset,\n endTime: ((data.audio_window_end as number) || 0) + startTimeOffset,\n confidence: confidence,\n text: transcript || '',\n // Note: Deepgram V2 (Flux) API does not provide word-level timing (start/end).\n // Words only contain 'word' and 'confidence' fields, so startTime/endTime will be 0.\n // See: https://developers.deepgram.com/docs/flux/nova-3-migration\n words: wordsData.map((word) =>\n createTimedString({\n text: (word.word as string) ?? '',\n startTime: ((word.start as number) ?? 0) + startTimeOffset,\n endTime: ((word.end as number) ?? 0) + startTimeOffset,\n confidence: (word.confidence as number) ?? 0.0,\n startTimeOffset,\n }),\n ),\n };\n\n return [sd];\n}\n"],"mappings":"AAGA;AAAA,EAEE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,YAAY,iBAAiB;AAC7B,SAAS,iBAAiB;AAC1B,SAAS,yBAAyB;AAGlC,MAAM,aAAa,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC;AAqBzD,MAAM,sBAAoD;AAAA,EACxD,OAAO;AAAA,EACP,YAAY;AAAA,EACZ,UAAU,CAAC;AAAA,EACX,aAAa;AAAA,EACb,UAAU;AAAA,EACV,WAAW;AACb;AAEA,SAAS,aAAa,MAA0B;AAC9C,aAAW,OAAO,MAAM;AACtB,QAAI,IAAI,SAAS,KAAK;AACpB,YAAM,IAAI,MAAM,yCAAyC;AAAA,IAC3D;AAAA,EACF;AACA,SAAO;AACT;AAiCO,MAAM,cAAc,IAAI,IAAI;AAAA,EACxB,QAAQ;AAAA,EACjB;AAAA,EACA;AAAA,EACA,UAAU,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAgBd,YAAY,OAA8B,CAAC,GAAG;AAC5C,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB;AAAA,MAChB,mBAAmB;AAAA,IACrB,CAAC;AAED,SAAK,QAAQ;AAAA,MACX,GAAG;AAAA,MACH,GAAG;AAAA,MACH,UAAU,KAAK,WAAW,kBAAkB,KAAK,QAAQ,IAAI,oBAAoB;AAAA,IACnF;AAEA,UAAM,SAAS,KAAK,UAAU,QAAQ,IAAI;AAC1C,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI,MAAM,8BAA8B;AAAA,IAChD;AACA,SAAK,UAAU;AAEf,QAAI,KAAK,MAAM,MAAM;AACnB,WAAK,MAAM,OAAO,aAAa,KAAK,MAAM,IAAI;AAAA,IAChD;AAAA,EACF;AAAA;AAAA,EAGA,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA;AAAA,EAGA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,MAAgB,WACd,QACA,cAC0B;AAC1B,UAAM,IAAI,MAAM,gEAAgE;AAAA,EAClF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,OAAO,SAAiE;AACtE,UAAM,aAAa,EAAE,GAAG,KAAK,OAAO,QAAQ,KAAK,QAAQ;AACzD,WAAO,IAAI,eAAe,MAAM,YAAY,mCAAS,WAAW;AAAA,EAClE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,cAAc,MAA6B;AACzC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AACvD,SAAK,QAAQ,MAAM,uBAAuB;AAAA,EAC5C;AACF;AAIA,MAAM,uBAAuB,IAAI,aAAa;AAAA,EACnC,QAAQ;AAAA,EACjB;AAAA,EACA,UAAU,IAAI;AAAA,EACd,MAAwB;AAAA,EAExB;AAAA,EACA,aAAa;AAAA,EACb,YAAY;AAAA;AAAA,EAGZ,kBAAkB,IAAI,MAAM;AAAA,EAE5B,YACE,aACA,MACA,aACA;AACA,UAAM,aAAa,KAAK,YAAY,WAAW;AAC/C,SAAK,QAAQ;AAEb,SAAK,0BAA0B,IAAI;AAAA,MACjC,CAAC,aAAa,KAAK,uBAAuB,QAAQ;AAAA,MAClD,EAAE,UAAU,EAAI;AAAA,IAClB;AAAA,EACF;AAAA,EAEA,cAAc,MAA6B;AACzC,SAAK,QAAQ,MAAM,iCAAiC,IAAI;AACxD,SAAK,QAAQ;AAAA,MACX,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,MACH,UACE,KAAK,aAAa,SAAY,kBAAkB,KAAK,QAAQ,IAAI,KAAK,MAAM;AAAA,IAChF;AACA,QAAI,KAAK,KAAM,MAAK,MAAM,OAAO,aAAa,KAAK,IAAI;AAGvD,SAAK,gBAAgB,IAAI;AAAA,EAC3B;AAAA,EAEA,MAAgB,MAAM;AAxNxB;AA0NI,WAAO,CAAC,KAAK,QAAQ;AACnB,UAAI;AACF,aAAK,gBAAgB,MAAM;AAE3B,cAAM,MAAM,KAAK,gBAAgB;AACjC,aAAK,QAAQ,MAAM,2BAA2B,GAAG,EAAE;AAEnD,aAAK,MAAM,IAAI,UAAU,KAAK;AAAA,UAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,QACzD,CAAC;AAGD,cAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,cAAI,CAAC,KAAK,IAAK,QAAO,OAAO,IAAI,MAAM,2BAA2B,CAAC;AAEnE,gBAAM,SAAS,MAAM;AAzO/B,gBAAAA;AA0OY,aAAAA,MAAA,KAAK,QAAL,gBAAAA,IAAU,IAAI,SAAS;AACvB,oBAAQ;AAAA,UACV;AACA,gBAAM,UAAU,CAAC,QAAe;AA7O1C,gBAAAA;AA8OY,aAAAA,MAAA,KAAK,QAAL,gBAAAA,IAAU,IAAI,QAAQ;AACtB,mBAAO,GAAG;AAAA,UACZ;AAEA,eAAK,IAAI,KAAK,QAAQ,MAAM;AAC5B,eAAK,IAAI,KAAK,SAAS,OAAO;AAAA,QAChC,CAAC;AAGD,cAAM,cAAc,KAAK,UAAU;AACnC,cAAM,cAAc,KAAK,UAAU;AACnC,cAAM,gBAAgB,KAAK,gBAAgB,KAAK;AAGhD,cAAM,SAAS,MAAM,QAAQ,KAAK;AAAA,UAChC,QAAQ,IAAI,CAAC,aAAa,WAAW,CAAC;AAAA,UACtC,cAAc,KAAK,MAAM,WAAW;AAAA,QACtC,CAAC;AAED,YAAI,WAAW,aAAa;AAC1B,eAAK,QAAQ,MAAM,6CAA6C;AAEhE,eAAK,IAAI,MAAM;AAAA,QACjB,OAAO;AAEL;AAAA,QACF;AAAA,MACF,SAAS,OAAO;AACd,aAAK,QAAQ,MAAM,yBAAyB,EAAE,MAAM,CAAC;AACrD,cAAM;AAAA,MACR,UAAE;AACA,cAAI,UAAK,QAAL,mBAAU,gBAAe,UAAU,MAAM;AAC3C,eAAK,IAAI,MAAM;AAAA,QACjB;AAAA,MACF;AAAA,IACF;AACA,SAAK,MAAM;AAAA,EACb;AAAA,EAEA,MAAM,YAAY;AAChB,QAAI,CAAC,KAAK,IAAK;AAGf,UAAM,cAAc,KAAK,MAAM,KAAK,MAAM,aAAa,EAAE;AACzD,UAAM,eAAe,IAAI,gBAAgB,KAAK,MAAM,YAAY,GAAG,WAAW;AAE9E,QAAI,WAAW;AAGf,UAAM,WAAW,KAAK,MAAM,OAAO,aAAa,EAAE;AAElD,WAAO,MAAM;AACX,YAAM,cAAc,SAAS,KAAK;AAElC,YAAM,eAAe,KAAK,gBAAgB,KAAK,EAAE,KAAK,OAAO,EAAE,OAAO,KAAK,EAAW;AAEtF,YAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,aAAa,YAAY,CAAC;AAG7D,UAAI,WAAW,UAAU,OAAO,MAAM;AACpC,YAAI,EAAE,WAAW,WAAW,OAAO,MAAM;AAEvC,qBAAW;AAAA,QACb,OAAO;AAEL;AAAA,QACF;AAAA,MACF;AAGA,UAAI,YAAY,OAAO,UAAU,QAAW;AAAA,MAE5C,WAAW,WAAW,QAAQ;AAC5B,cAAM,OAAO,OAAO;AACpB,cAAM,SAAuB,CAAC;AAE9B,YAAI,SAAS,eAAe,gBAAgB;AAC1C,iBAAO,KAAK,GAAG,aAAa,MAAM,CAAC;AACnC,qBAAW;AAAA,QACb,OAAO;AACL,iBAAO,KAAK,GAAG,aAAa,MAAO,KAAoB,KAAK,MAAqB,CAAC;AAAA,QACpF;AAEA,mBAAW,SAAS,QAAQ;AAC1B,eAAK,wBAAwB,KAAK,8BAA8B,KAAK,CAAC;AAEtE,cAAI,KAAK,IAAK,eAAe,UAAU,MAAM;AAC3C,iBAAK,IAAK,KAAK,MAAM,IAAI;AAAA,UAC3B;AAEA,cAAI,UAAU;AACZ,iBAAK,wBAAwB,MAAM;AACnC,uBAAW;AAAA,UACb;AAAA,QACF;AAAA,MACF;AAEA,UAAI,SAAU;AAAA,IAChB;AAGA,QAAI,CAAC,KAAK,gBAAgB,SAAS,KAAK,IAAK,eAAe,UAAU,MAAM;AAC1E,WAAK,QAAQ,MAAM,yCAAyC;AAC5D,WAAK,IAAK,KAAK,UAAU;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,MAAM,YAAY;AAChB,QAAI,CAAC,KAAK,IAAK;AAEf,WAAO,IAAI,QAAc,CAAC,YAAY;AACpC,UAAI,CAAC,KAAK,IAAK,QAAO,QAAQ;AAE9B,WAAK,IAAI,GAAG,WAAW,CAAC,MAAc,aAAsB;AAC1D,YAAI,UAAU;AACZ,eAAK,QAAQ,KAAK,kDAAkD;AACpE;AAAA,QACF;AACA,YAAI;AACF,gBAAM,MAAM,KAAK,MAAM,KAAK,SAAS,CAAC;AACtC,eAAK,oBAAoB,GAAG;AAAA,QAC9B,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,oCAAoC,EAAE,MAAM,CAAC;AAAA,QAClE;AAAA,MACF,CAAC;AAED,WAAK,IAAI,GAAG,SAAS,CAAC,MAAM,WAAW;AACrC,aAAK,QAAQ,MAAM,8BAA8B,IAAI,IAAI,MAAM,EAAE;AACjE,gBAAQ;AAAA,MACV,CAAC;AAGD,WAAK,IAAI,GAAG,SAAS,MAAM,QAAQ,CAAC;AAAA,IACtC,CAAC;AAAA,EACH;AAAA,EAEA,oBAAoB,MAA+B;AACjD,QAAI,KAAK,YAAY;AACnB,WAAK,aAAa,KAAK;AAAA,IACzB;AAEA,QAAI,KAAK,SAAS,YAAY;AAC5B,YAAM,YAAY,KAAK;AAEvB,UAAI,cAAc,eAAe;AAC/B,YAAI,KAAK,UAAW;AAEpB,aAAK,YAAY;AACjB,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,IAAI,gBAAgB;AAAA,UAC1B,WAAW,KAAK;AAAA,QAClB,CAAC;AAED,aAAK,qBAAqB,IAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,UAAU;AACjC,YAAI,CAAC,KAAK,UAAW;AACrB,aAAK,qBAAqB,IAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,kBAAkB;AACzC,YAAI,CAAC,KAAK,UAAW;AACrB,aAAK,qBAAqB,IAAI,gBAAgB,sBAAsB,IAAI;AAAA,MAC1E,WAAW,cAAc,eAAe;AACtC,aAAK,qBAAqB,IAAI,gBAAgB,oBAAoB,IAAI;AAAA,MACxE,WAAW,cAAc,aAAa;AACpC,YAAI,CAAC,KAAK,UAAW;AAErB,aAAK,YAAY;AACjB,aAAK,qBAAqB,IAAI,gBAAgB,kBAAkB,IAAI;AAEpE,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,IAAI,gBAAgB;AAAA,UAC1B,WAAW,KAAK;AAAA,QAClB,CAAC;AAAA,MACH;AAAA,IACF,WAAW,KAAK,SAAS,SAAS;AAChC,WAAK,QAAQ,KAAK,0BAA0B,EAAE,KAAK,CAAC;AACpD,YAAM,OAAQ,KAAK,eAA0B;AAC7C,YAAM,IAAI,MAAM,uBAAuB,IAAI,EAAE;AAAA,IAC/C;AAAA,EACF;AAAA,EAEA,qBAAqB,WAAgC,MAA+B;AAClF,UAAM,OAAO,mBAAmB,KAAK,MAAM,YAAY,MAAM,MAAM,KAAK,eAAe;AAEvF,QAAI,KAAK,SAAS,GAAG;AACnB,WAAK,MAAM,IAAI;AAAA,QACb,MAAM;AAAA,QACN,WAAW,KAAK;AAAA,QAChB,cAAc,CAAC,KAAK,CAAC,GAAI,GAAG,KAAK,MAAM,CAAC,CAAC;AAAA,MAC3C,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EAEA,uBAAuB,UAAkB;AACvC,UAAM,aAA8B;AAAA,MAClC,MAAM,IAAI,gBAAgB;AAAA,MAC1B,WAAW,KAAK;AAAA,MAChB,kBAAkB;AAAA,QAChB,eAAe;AAAA,MACjB;AAAA,IACF;AACA,SAAK,MAAM,IAAI,UAAU;AAAA,EAC3B;AAAA,EAEA,kBAA0B;AACxB,UAAM,SAA4C;AAAA,MAChD,OAAO,KAAK,MAAM;AAAA,MAClB,aAAa,KAAK,MAAM,WAAW,SAAS;AAAA,MAC5C,UAAU;AAAA,MACV,aAAa,OAAO,KAAK,MAAM,SAAS;AAAA,IAC1C;AAGA,QAAI,KAAK,MAAM;AACb,aAAO,sBAAsB,KAAK,MAAM,kBAAkB,SAAS;AACrE,QAAI,KAAK,MAAM,aAAc,QAAO,gBAAgB,KAAK,MAAM,aAAa,SAAS;AACrF,QAAI,KAAK,MAAM,aAAc,QAAO,iBAAiB,KAAK,MAAM,aAAa,SAAS;AAEtF,QAAI,KAAK,MAAM,SAAS,SAAS,EAAG,QAAO,UAAU,KAAK,MAAM;AAChE,QAAI,KAAK,MAAM,QAAQ,KAAK,MAAM,KAAK,SAAS,EAAG,QAAO,MAAM,KAAK,MAAM;AAE3E,UAAM,UAAU,KAAK,MAAM,YAAY,QAAQ,SAAS,IAAI;AAC5D,UAAM,KAAK,YAAY,UAAU,MAAM;AACvC,WAAO,GAAG,OAAO,IAAI,EAAE;AAAA,EACzB;AAAA,EAES,QAAQ;AA/cnB;AAgdI,UAAM,MAAM;AACZ,eAAK,QAAL,mBAAU;AAAA,EACZ;AACF;AAIA,SAAS,mBACP,UACA,MACA,iBACkB;AAClB,QAAM,aAAa,KAAK;AACxB,QAAM,YAAa,KAAK,SAA4C,CAAC;AAErE,MAAI,CAAC,aAAa,UAAU,WAAW,GAAG;AACxC,WAAO,CAAC;AAAA,EACV;AAEA,MAAI,aAAa;AACjB,MAAI,UAAU,SAAS,GAAG;AACxB,UAAM,MAAM,UAAU,OAAO,CAAC,KAAa,MAAM,OAAQ,EAAE,cAAyB,IAAI,CAAC;AACzF,iBAAa,MAAM,UAAU;AAAA,EAC/B;AAEA,QAAM,KAAqB;AAAA,IACzB,UAAU,kBAAkB,QAAQ;AAAA,IACpC,YAAa,KAAK,sBAAiC,KAAK;AAAA,IACxD,UAAW,KAAK,oBAA+B,KAAK;AAAA,IACpD;AAAA,IACA,MAAM,cAAc;AAAA;AAAA;AAAA;AAAA,IAIpB,OAAO,UAAU;AAAA,MAAI,CAAC,SACpB,kBAAkB;AAAA,QAChB,MAAO,KAAK,QAAmB;AAAA,QAC/B,YAAa,KAAK,SAAoB,KAAK;AAAA,QAC3C,UAAW,KAAK,OAAkB,KAAK;AAAA,QACvC,YAAa,KAAK,cAAyB;AAAA,QAC3C;AAAA,MACF,CAAC;AAAA,IACH;AAAA,EACF;AAEA,SAAO,CAAC,EAAE;AACZ;","names":["_a"]}
|
package/dist/stt_v2.test.cjs
CHANGED
|
@@ -1,11 +1,17 @@
|
|
|
1
1
|
"use strict";
|
|
2
|
-
var import_agents = require("@livekit/agents");
|
|
3
2
|
var import_agents_plugin_silero = require("@livekit/agents-plugin-silero");
|
|
4
3
|
var import_agents_plugins_test = require("@livekit/agents-plugins-test");
|
|
5
4
|
var import_vitest = require("vitest");
|
|
6
5
|
var import_stt_v2 = require("./stt_v2.cjs");
|
|
7
|
-
|
|
8
|
-
|
|
9
|
-
|
|
10
|
-
});
|
|
6
|
+
const hasDeepgramApiKey = Boolean(process.env.DEEPGRAM_API_KEY);
|
|
7
|
+
if (hasDeepgramApiKey) {
|
|
8
|
+
(0, import_vitest.describe)("Deepgram STTv2 (Flux)", async () => {
|
|
9
|
+
await (0, import_agents_plugins_test.stt)(new import_stt_v2.STTv2(), await import_agents_plugin_silero.VAD.load(), { nonStreaming: false });
|
|
10
|
+
});
|
|
11
|
+
} else {
|
|
12
|
+
(0, import_vitest.describe)("Deepgram STTv2 (Flux)", () => {
|
|
13
|
+
import_vitest.it.skip("requires DEEPGRAM_API_KEY", () => {
|
|
14
|
+
});
|
|
15
|
+
});
|
|
16
|
+
}
|
|
11
17
|
//# sourceMappingURL=stt_v2.test.cjs.map
|
package/dist/stt_v2.test.cjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/stt_v2.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {
|
|
1
|
+
{"version":3,"sources":["../src/stt_v2.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 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 { STTv2 } from './stt_v2.js';\n\nconst hasDeepgramApiKey = Boolean(process.env.DEEPGRAM_API_KEY);\n\nif (hasDeepgramApiKey) {\n describe('Deepgram STTv2 (Flux)', async () => {\n await stt(new STTv2(), await VAD.load(), { nonStreaming: false });\n });\n} else {\n describe('Deepgram STTv2 (Flux)', () => {\n it.skip('requires DEEPGRAM_API_KEY', () => {});\n });\n}\n"],"mappings":";AAGA,kCAAoB;AACpB,iCAAoB;AACpB,oBAA6B;AAC7B,oBAAsB;AAEtB,MAAM,oBAAoB,QAAQ,QAAQ,IAAI,gBAAgB;AAE9D,IAAI,mBAAmB;AACrB,8BAAS,yBAAyB,YAAY;AAC5C,cAAM,gCAAI,IAAI,oBAAM,GAAG,MAAM,gCAAI,KAAK,GAAG,EAAE,cAAc,MAAM,CAAC;AAAA,EAClE,CAAC;AACH,OAAO;AACL,8BAAS,yBAAyB,MAAM;AACtC,qBAAG,KAAK,6BAA6B,MAAM;AAAA,IAAC,CAAC;AAAA,EAC/C,CAAC;AACH;","names":[]}
|
package/dist/stt_v2.test.js
CHANGED
|
@@ -1,10 +1,16 @@
|
|
|
1
|
-
import { initializeLogger } from "@livekit/agents";
|
|
2
1
|
import { VAD } from "@livekit/agents-plugin-silero";
|
|
3
2
|
import { stt } from "@livekit/agents-plugins-test";
|
|
4
|
-
import { describe } from "vitest";
|
|
3
|
+
import { describe, it } from "vitest";
|
|
5
4
|
import { STTv2 } from "./stt_v2.js";
|
|
6
|
-
|
|
7
|
-
|
|
8
|
-
|
|
9
|
-
});
|
|
5
|
+
const hasDeepgramApiKey = Boolean(process.env.DEEPGRAM_API_KEY);
|
|
6
|
+
if (hasDeepgramApiKey) {
|
|
7
|
+
describe("Deepgram STTv2 (Flux)", async () => {
|
|
8
|
+
await stt(new STTv2(), await VAD.load(), { nonStreaming: false });
|
|
9
|
+
});
|
|
10
|
+
} else {
|
|
11
|
+
describe("Deepgram STTv2 (Flux)", () => {
|
|
12
|
+
it.skip("requires DEEPGRAM_API_KEY", () => {
|
|
13
|
+
});
|
|
14
|
+
});
|
|
15
|
+
}
|
|
10
16
|
//# sourceMappingURL=stt_v2.test.js.map
|
package/dist/stt_v2.test.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/stt_v2.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {
|
|
1
|
+
{"version":3,"sources":["../src/stt_v2.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 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 { STTv2 } from './stt_v2.js';\n\nconst hasDeepgramApiKey = Boolean(process.env.DEEPGRAM_API_KEY);\n\nif (hasDeepgramApiKey) {\n describe('Deepgram STTv2 (Flux)', async () => {\n await stt(new STTv2(), await VAD.load(), { nonStreaming: false });\n });\n} else {\n describe('Deepgram STTv2 (Flux)', () => {\n it.skip('requires DEEPGRAM_API_KEY', () => {});\n });\n}\n"],"mappings":"AAGA,SAAS,WAAW;AACpB,SAAS,WAAW;AACpB,SAAS,UAAU,UAAU;AAC7B,SAAS,aAAa;AAEtB,MAAM,oBAAoB,QAAQ,QAAQ,IAAI,gBAAgB;AAE9D,IAAI,mBAAmB;AACrB,WAAS,yBAAyB,YAAY;AAC5C,UAAM,IAAI,IAAI,MAAM,GAAG,MAAM,IAAI,KAAK,GAAG,EAAE,cAAc,MAAM,CAAC;AAAA,EAClE,CAAC;AACH,OAAO;AACL,WAAS,yBAAyB,MAAM;AACtC,OAAG,KAAK,6BAA6B,MAAM;AAAA,IAAC,CAAC;AAAA,EAC/C,CAAC;AACH;","names":[]}
|
package/dist/tts.cjs
CHANGED
|
@@ -45,6 +45,12 @@ const defaultTTSOptions = {
|
|
|
45
45
|
class TTS extends import_agents.tts.TTS {
|
|
46
46
|
opts;
|
|
47
47
|
label = "deepgram.TTS";
|
|
48
|
+
get model() {
|
|
49
|
+
return this.opts.model;
|
|
50
|
+
}
|
|
51
|
+
get provider() {
|
|
52
|
+
return "Deepgram";
|
|
53
|
+
}
|
|
48
54
|
constructor(opts = {}) {
|
|
49
55
|
var _a;
|
|
50
56
|
super(opts.sampleRate || defaultTTSOptions.sampleRate, NUM_CHANNELS, {
|
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 log,\n shortuuid,\n tokenize,\n tts,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { request } from 'node:https';\nimport { type RawData, WebSocket } from 'ws';\nimport type { TTSEncoding, TTSModels } from './models.js';\n\nconst AUTHORIZATION_HEADER = 'Authorization';\nconst NUM_CHANNELS = 1;\nconst MIN_SENTENCE_LENGTH = 8;\n\nexport interface TTSOptions {\n model: TTSModels | string;\n encoding: TTSEncoding;\n sampleRate: number;\n apiKey?: string;\n baseUrl?: string;\n sentenceTokenizer: tokenize.SentenceTokenizer;\n capabilities: tts.TTSCapabilities;\n}\n\nconst defaultTTSOptions: TTSOptions = {\n model: 'aura-asteria-en',\n encoding: 'linear16',\n sampleRate: 24000,\n apiKey: process.env.DEEPGRAM_API_KEY,\n baseUrl: 'https://api.deepgram.com',\n capabilities: {\n streaming: true,\n },\n sentenceTokenizer: new tokenize.basic.SentenceTokenizer({\n minSentenceLength: MIN_SENTENCE_LENGTH,\n }),\n};\n\nexport class TTS extends tts.TTS {\n private opts: TTSOptions;\n label = 'deepgram.TTS';\n\n constructor(opts: Partial<TTSOptions> = {}) {\n super(opts.sampleRate || defaultTTSOptions.sampleRate, NUM_CHANNELS, {\n streaming: opts.capabilities?.streaming ?? defaultTTSOptions.capabilities.streaming,\n });\n\n this.opts = {\n ...defaultTTSOptions,\n ...opts,\n };\n\n if (this.opts.apiKey === undefined) {\n throw new Error(\n 'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n );\n }\n }\n\n synthesize(\n text: string,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ): tts.ChunkedStream {\n return new ChunkedStream(this, text, this.opts, connOptions, abortSignal);\n }\n\n stream(): tts.SynthesizeStream {\n return new SynthesizeStream(this, this.opts);\n }\n}\n\nexport class ChunkedStream extends tts.ChunkedStream {\n label = 'deepgram.ChunkedStream';\n #logger = log();\n private opts: TTSOptions;\n private text: string;\n\n constructor(\n tts: TTS,\n text: string,\n opts: TTSOptions,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ) {\n super(text, tts, connOptions, abortSignal);\n this.text = text;\n this.opts = opts;\n }\n\n protected async run() {\n const requestId = shortuuid();\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n const json = { text: this.text };\n const url = new URL(`${this.opts.baseUrl!}/v1/speak`);\n url.searchParams.append('sample_rate', this.opts.sampleRate.toString());\n url.searchParams.append('model', this.opts.model);\n url.searchParams.append('encoding', this.opts.encoding);\n\n await new Promise<void>((resolve, reject) => {\n const req = request(\n {\n hostname: url.hostname,\n port: 443,\n path: url.pathname + url.search,\n method: 'POST',\n headers: {\n [AUTHORIZATION_HEADER]: `Token ${this.opts.apiKey!}`,\n 'Content-Type': 'application/json',\n },\n signal: this.abortSignal,\n },\n (res) => {\n if (res.statusCode !== 200) {\n reject(\n new Error(`Deepgram TTS HTTP request failed: ${res.statusCode} ${res.statusMessage}`),\n );\n return;\n }\n\n res.on('data', (chunk) => {\n for (const frame of bstream.write(chunk)) {\n if (!this.queue.closed) {\n this.queue.put({\n requestId,\n frame,\n final: false,\n segmentId: requestId,\n });\n }\n }\n });\n\n res.on('error', (err) => {\n if (err.message === 'aborted') return;\n this.#logger.error({ err }, 'Deepgram TTS response error');\n });\n\n res.on('close', () => {\n for (const frame of bstream.flush()) {\n if (!this.queue.closed) {\n this.queue.put({\n requestId,\n frame,\n final: false,\n segmentId: requestId,\n });\n }\n }\n if (!this.queue.closed) {\n this.queue.close();\n }\n resolve();\n });\n },\n );\n\n req.on('error', (err) => {\n if (err.name === 'AbortError') return;\n this.#logger.error({ err }, 'Deepgram TTS request error');\n });\n\n req.on('close', () => resolve());\n req.write(JSON.stringify(json));\n req.end();\n });\n }\n}\n\nexport class SynthesizeStream extends tts.SynthesizeStream {\n private opts: TTSOptions;\n private tokenizer: tokenize.SentenceStream;\n label = 'deepgram.SynthesizeStream';\n\n private static readonly FLUSH_MSG = JSON.stringify({ type: 'Flush' });\n private static readonly CLOSE_MSG = JSON.stringify({ type: 'Close' });\n\n constructor(tts: TTS, opts: TTSOptions) {\n super(tts);\n this.opts = opts;\n this.tokenizer = opts.sentenceTokenizer.stream();\n }\n\n private async closeWebSocket(ws: WebSocket): Promise<void> {\n try {\n // Send Flush and Close messages to ensure Deepgram processes all remaining audio\n // and properly terminates the session, preventing lingering TTS sessions\n if (ws.readyState === WebSocket.OPEN) {\n ws.send(SynthesizeStream.FLUSH_MSG);\n ws.send(SynthesizeStream.CLOSE_MSG);\n\n // Wait for server acknowledgment to prevent race conditions and ensure\n // proper cleanup, avoiding 429 Too Many Requests errors from lingering sessions\n try {\n await new Promise<void>((resolve, _reject) => {\n const timeout = setTimeout(() => {\n resolve();\n }, 1000);\n\n ws.once('message', () => {\n clearTimeout(timeout);\n resolve();\n });\n\n ws.once('close', () => {\n clearTimeout(timeout);\n resolve();\n });\n\n ws.once('error', () => {\n clearTimeout(timeout);\n resolve();\n });\n });\n } catch (e) {\n // Ignore timeout or other errors during close sequence\n }\n }\n } catch (e) {\n console.warn(`Error during WebSocket close sequence: ${e}`);\n } finally {\n if (ws.readyState === WebSocket.OPEN || ws.readyState === WebSocket.CONNECTING) {\n ws.close();\n }\n }\n }\n\n protected async run() {\n const requestId = shortuuid();\n const segmentId = shortuuid();\n\n const wsUrl = this.opts.baseUrl!.replace(/^http/, 'ws');\n const url = new URL(`${wsUrl}/v1/speak`);\n url.searchParams.append('sample_rate', this.opts.sampleRate.toString());\n url.searchParams.append('model', this.opts.model);\n url.searchParams.append('encoding', this.opts.encoding);\n\n const ws = new WebSocket(url, {\n headers: {\n [AUTHORIZATION_HEADER]: `Token ${this.opts.apiKey!}`,\n },\n });\n\n await new Promise((resolve, reject) => {\n ws.on('open', resolve);\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n });\n\n const inputTask = async () => {\n for await (const data of this.input) {\n if (data === SynthesizeStream.FLUSH_SENTINEL) {\n this.tokenizer.flush();\n continue;\n }\n this.tokenizer.pushText(data);\n }\n this.tokenizer.endInput();\n this.tokenizer.close();\n };\n\n const sendTask = async () => {\n for await (const event of this.tokenizer) {\n if (this.abortController.signal.aborted) break;\n\n let text = event.token;\n if (!text.endsWith(' ')) {\n text += ' ';\n }\n\n const message = JSON.stringify({\n type: 'Speak',\n text: text,\n });\n\n ws.send(message);\n }\n\n if (!this.abortController.signal.aborted) {\n ws.send(SynthesizeStream.FLUSH_MSG);\n }\n };\n\n const recvTask = async () => {\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n let finalReceived = false;\n let timeout: NodeJS.Timeout | null = null;\n let lastFrame: AudioFrame | undefined;\n\n const sendLastFrame = (segmentId: string, final: boolean) => {\n if (lastFrame && !this.queue.closed) {\n this.queue.put({ requestId, segmentId, frame: lastFrame, final });\n lastFrame = undefined;\n }\n };\n\n const clearMessageTimeout = () => {\n if (timeout) {\n clearTimeout(timeout);\n timeout = null;\n }\n };\n\n return new Promise<void>((resolve, reject) => {\n ws.on('message', (data: RawData, isBinary: boolean) => {\n clearMessageTimeout();\n\n if (!isBinary) {\n const message = JSON.parse(data.toString());\n if (message.type === 'Flushed') {\n finalReceived = true;\n clearMessageTimeout();\n for (const frame of bstream.flush()) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n sendLastFrame(segmentId, true);\n\n if (!this.queue.closed) {\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n }\n resolve();\n }\n\n return;\n }\n\n const buffer =\n data instanceof Buffer\n ? data.buffer.slice(data.byteOffset, data.byteOffset + data.byteLength)\n : (data as ArrayBuffer);\n for (const frame of bstream.write(buffer as ArrayBuffer)) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n });\n\n ws.on('close', (_code, _reason) => {\n if (!finalReceived) {\n for (const frame of bstream.flush()) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n sendLastFrame(segmentId, true);\n\n if (!this.queue.closed) {\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n }\n }\n resolve();\n });\n\n ws.on('error', (error) => {\n clearMessageTimeout();\n reject(error);\n });\n });\n };\n\n try {\n await Promise.all([inputTask(), sendTask(), recvTask()]);\n } catch (e) {\n throw new Error(`failed in main task: ${e}`);\n } finally {\n await this.closeWebSocket(ws);\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBAOO;AAEP,wBAAwB;AACxB,gBAAwC;AAGxC,MAAM,uBAAuB;AAC7B,MAAM,eAAe;AACrB,MAAM,sBAAsB;AAY5B,MAAM,oBAAgC;AAAA,EACpC,OAAO;AAAA,EACP,UAAU;AAAA,EACV,YAAY;AAAA,EACZ,QAAQ,QAAQ,IAAI;AAAA,EACpB,SAAS;AAAA,EACT,cAAc;AAAA,IACZ,WAAW;AAAA,EACb;AAAA,EACA,mBAAmB,IAAI,uBAAS,MAAM,kBAAkB;AAAA,IACtD,mBAAmB;AAAA,EACrB,CAAC;AACH;AAEO,MAAM,YAAY,kBAAI,IAAI;AAAA,EACvB;AAAA,EACR,QAAQ;AAAA,EAER,YAAY,OAA4B,CAAC,GAAG;AAhD9C;AAiDI,UAAM,KAAK,cAAc,kBAAkB,YAAY,cAAc;AAAA,MACnE,aAAW,UAAK,iBAAL,mBAAmB,cAAa,kBAAkB,aAAa;AAAA,IAC5E,CAAC;AAED,SAAK,OAAO;AAAA,MACV,GAAG;AAAA,MACH,GAAG;AAAA,IACL;AAEA,QAAI,KAAK,KAAK,WAAW,QAAW;AAClC,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,WACE,MACA,aACA,aACmB;AACnB,WAAO,IAAI,cAAc,MAAM,MAAM,KAAK,MAAM,aAAa,WAAW;AAAA,EAC1E;AAAA,EAEA,SAA+B;AAC7B,WAAO,IAAI,iBAAiB,MAAM,KAAK,IAAI;AAAA,EAC7C;AACF;AAEO,MAAM,sBAAsB,kBAAI,cAAc;AAAA,EACnD,QAAQ;AAAA,EACR,cAAU,mBAAI;AAAA,EACN;AAAA,EACA;AAAA,EAER,YACEA,MACA,MACA,MACA,aACA,aACA;AACA,UAAM,MAAMA,MAAK,aAAa,WAAW;AACzC,SAAK,OAAO;AACZ,SAAK,OAAO;AAAA,EACd;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,gBAAY,yBAAU;AAC5B,UAAM,UAAU,IAAI,8BAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,UAAM,OAAO,EAAE,MAAM,KAAK,KAAK;AAC/B,UAAM,MAAM,IAAI,IAAI,GAAG,KAAK,KAAK,OAAQ,WAAW;AACpD,QAAI,aAAa,OAAO,eAAe,KAAK,KAAK,WAAW,SAAS,CAAC;AACtE,QAAI,aAAa,OAAO,SAAS,KAAK,KAAK,KAAK;AAChD,QAAI,aAAa,OAAO,YAAY,KAAK,KAAK,QAAQ;AAEtD,UAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,YAAM,UAAM;AAAA,QACV;AAAA,UACE,UAAU,IAAI;AAAA,UACd,MAAM;AAAA,UACN,MAAM,IAAI,WAAW,IAAI;AAAA,UACzB,QAAQ;AAAA,UACR,SAAS;AAAA,YACP,CAAC,oBAAoB,GAAG,SAAS,KAAK,KAAK,MAAO;AAAA,YAClD,gBAAgB;AAAA,UAClB;AAAA,UACA,QAAQ,KAAK;AAAA,QACf;AAAA,QACA,CAAC,QAAQ;AACP,cAAI,IAAI,eAAe,KAAK;AAC1B;AAAA,cACE,IAAI,MAAM,qCAAqC,IAAI,UAAU,IAAI,IAAI,aAAa,EAAE;AAAA,YACtF;AACA;AAAA,UACF;AAEA,cAAI,GAAG,QAAQ,CAAC,UAAU;AACxB,uBAAW,SAAS,QAAQ,MAAM,KAAK,GAAG;AACxC,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI;AAAA,kBACb;AAAA,kBACA;AAAA,kBACA,OAAO;AAAA,kBACP,WAAW;AAAA,gBACb,CAAC;AAAA,cACH;AAAA,YACF;AAAA,UACF,CAAC;AAED,cAAI,GAAG,SAAS,CAAC,QAAQ;AACvB,gBAAI,IAAI,YAAY,UAAW;AAC/B,iBAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,6BAA6B;AAAA,UAC3D,CAAC;AAED,cAAI,GAAG,SAAS,MAAM;AACpB,uBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI;AAAA,kBACb;AAAA,kBACA;AAAA,kBACA,OAAO;AAAA,kBACP,WAAW;AAAA,gBACb,CAAC;AAAA,cACH;AAAA,YACF;AACA,gBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,mBAAK,MAAM,MAAM;AAAA,YACnB;AACA,oBAAQ;AAAA,UACV,CAAC;AAAA,QACH;AAAA,MACF;AAEA,UAAI,GAAG,SAAS,CAAC,QAAQ;AACvB,YAAI,IAAI,SAAS,aAAc;AAC/B,aAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,4BAA4B;AAAA,MAC1D,CAAC;AAED,UAAI,GAAG,SAAS,MAAM,QAAQ,CAAC;AAC/B,UAAI,MAAM,KAAK,UAAU,IAAI,CAAC;AAC9B,UAAI,IAAI;AAAA,IACV,CAAC;AAAA,EACH;AACF;AAEO,MAAM,yBAAyB,kBAAI,iBAAiB;AAAA,EACjD;AAAA,EACA;AAAA,EACR,QAAQ;AAAA,EAER,OAAwB,YAAY,KAAK,UAAU,EAAE,MAAM,QAAQ,CAAC;AAAA,EACpE,OAAwB,YAAY,KAAK,UAAU,EAAE,MAAM,QAAQ,CAAC;AAAA,EAEpE,YAAYA,MAAU,MAAkB;AACtC,UAAMA,IAAG;AACT,SAAK,OAAO;AACZ,SAAK,YAAY,KAAK,kBAAkB,OAAO;AAAA,EACjD;AAAA,EAEA,MAAc,eAAe,IAA8B;AACzD,QAAI;AAGF,UAAI,GAAG,eAAe,oBAAU,MAAM;AACpC,WAAG,KAAK,iBAAiB,SAAS;AAClC,WAAG,KAAK,iBAAiB,SAAS;AAIlC,YAAI;AACF,gBAAM,IAAI,QAAc,CAAC,SAAS,YAAY;AAC5C,kBAAM,UAAU,WAAW,MAAM;AAC/B,sBAAQ;AAAA,YACV,GAAG,GAAI;AAEP,eAAG,KAAK,WAAW,MAAM;AACvB,2BAAa,OAAO;AACpB,sBAAQ;AAAA,YACV,CAAC;AAED,eAAG,KAAK,SAAS,MAAM;AACrB,2BAAa,OAAO;AACpB,sBAAQ;AAAA,YACV,CAAC;AAED,eAAG,KAAK,SAAS,MAAM;AACrB,2BAAa,OAAO;AACpB,sBAAQ;AAAA,YACV,CAAC;AAAA,UACH,CAAC;AAAA,QACH,SAAS,GAAG;AAAA,QAEZ;AAAA,MACF;AAAA,IACF,SAAS,GAAG;AACV,cAAQ,KAAK,0CAA0C,CAAC,EAAE;AAAA,IAC5D,UAAE;AACA,UAAI,GAAG,eAAe,oBAAU,QAAQ,GAAG,eAAe,oBAAU,YAAY;AAC9E,WAAG,MAAM;AAAA,MACX;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,gBAAY,yBAAU;AAC5B,UAAM,gBAAY,yBAAU;AAE5B,UAAM,QAAQ,KAAK,KAAK,QAAS,QAAQ,SAAS,IAAI;AACtD,UAAM,MAAM,IAAI,IAAI,GAAG,KAAK,WAAW;AACvC,QAAI,aAAa,OAAO,eAAe,KAAK,KAAK,WAAW,SAAS,CAAC;AACtE,QAAI,aAAa,OAAO,SAAS,KAAK,KAAK,KAAK;AAChD,QAAI,aAAa,OAAO,YAAY,KAAK,KAAK,QAAQ;AAEtD,UAAM,KAAK,IAAI,oBAAU,KAAK;AAAA,MAC5B,SAAS;AAAA,QACP,CAAC,oBAAoB,GAAG,SAAS,KAAK,KAAK,MAAO;AAAA,MACpD;AAAA,IACF,CAAC;AAED,UAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,SAAG,GAAG,QAAQ,OAAO;AACrB,SAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,SAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,IAC/D,CAAC;AAED,UAAM,YAAY,YAAY;AAC5B,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI,SAAS,iBAAiB,gBAAgB;AAC5C,eAAK,UAAU,MAAM;AACrB;AAAA,QACF;AACA,aAAK,UAAU,SAAS,IAAI;AAAA,MAC9B;AACA,WAAK,UAAU,SAAS;AACxB,WAAK,UAAU,MAAM;AAAA,IACvB;AAEA,UAAM,WAAW,YAAY;AAC3B,uBAAiB,SAAS,KAAK,WAAW;AACxC,YAAI,KAAK,gBAAgB,OAAO,QAAS;AAEzC,YAAI,OAAO,MAAM;AACjB,YAAI,CAAC,KAAK,SAAS,GAAG,GAAG;AACvB,kBAAQ;AAAA,QACV;AAEA,cAAM,UAAU,KAAK,UAAU;AAAA,UAC7B,MAAM;AAAA,UACN;AAAA,QACF,CAAC;AAED,WAAG,KAAK,OAAO;AAAA,MACjB;AAEA,UAAI,CAAC,KAAK,gBAAgB,OAAO,SAAS;AACxC,WAAG,KAAK,iBAAiB,SAAS;AAAA,MACpC;AAAA,IACF;AAEA,UAAM,WAAW,YAAY;AAC3B,YAAM,UAAU,IAAI,8BAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,UAAI,gBAAgB;AACpB,UAAI,UAAiC;AACrC,UAAI;AAEJ,YAAM,gBAAgB,CAACC,YAAmB,UAAmB;AAC3D,YAAI,aAAa,CAAC,KAAK,MAAM,QAAQ;AACnC,eAAK,MAAM,IAAI,EAAE,WAAW,WAAAA,YAAW,OAAO,WAAW,MAAM,CAAC;AAChE,sBAAY;AAAA,QACd;AAAA,MACF;AAEA,YAAM,sBAAsB,MAAM;AAChC,YAAI,SAAS;AACX,uBAAa,OAAO;AACpB,oBAAU;AAAA,QACZ;AAAA,MACF;AAEA,aAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,WAAG,GAAG,WAAW,CAAC,MAAe,aAAsB;AACrD,8BAAoB;AAEpB,cAAI,CAAC,UAAU;AACb,kBAAM,UAAU,KAAK,MAAM,KAAK,SAAS,CAAC;AAC1C,gBAAI,QAAQ,SAAS,WAAW;AAC9B,8BAAgB;AAChB,kCAAoB;AACpB,yBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,8BAAc,WAAW,KAAK;AAC9B,4BAAY;AAAA,cACd;AACA,4BAAc,WAAW,IAAI;AAE7B,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI,iBAAiB,aAAa;AAAA,cAC/C;AACA,sBAAQ;AAAA,YACV;AAEA;AAAA,UACF;AAEA,gBAAM,SACJ,gBAAgB,SACZ,KAAK,OAAO,MAAM,KAAK,YAAY,KAAK,aAAa,KAAK,UAAU,IACnE;AACP,qBAAW,SAAS,QAAQ,MAAM,MAAqB,GAAG;AACxD,0BAAc,WAAW,KAAK;AAC9B,wBAAY;AAAA,UACd;AAAA,QACF,CAAC;AAED,WAAG,GAAG,SAAS,CAAC,OAAO,YAAY;AACjC,cAAI,CAAC,eAAe;AAClB,uBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,4BAAc,WAAW,KAAK;AAC9B,0BAAY;AAAA,YACd;AACA,0BAAc,WAAW,IAAI;AAE7B,gBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,mBAAK,MAAM,IAAI,iBAAiB,aAAa;AAAA,YAC/C;AAAA,UACF;AACA,kBAAQ;AAAA,QACV,CAAC;AAED,WAAG,GAAG,SAAS,CAAC,UAAU;AACxB,8BAAoB;AACpB,iBAAO,KAAK;AAAA,QACd,CAAC;AAAA,MACH,CAAC;AAAA,IACH;AAEA,QAAI;AACF,YAAM,QAAQ,IAAI,CAAC,UAAU,GAAG,SAAS,GAAG,SAAS,CAAC,CAAC;AAAA,IACzD,SAAS,GAAG;AACV,YAAM,IAAI,MAAM,wBAAwB,CAAC,EAAE;AAAA,IAC7C,UAAE;AACA,YAAM,KAAK,eAAe,EAAE;AAAA,IAC9B;AAAA,EACF;AACF;","names":["tts","segmentId"]}
|
|
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 log,\n shortuuid,\n tokenize,\n tts,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { request } from 'node:https';\nimport { type RawData, WebSocket } from 'ws';\nimport type { TTSEncoding, TTSModels } from './models.js';\n\nconst AUTHORIZATION_HEADER = 'Authorization';\nconst NUM_CHANNELS = 1;\nconst MIN_SENTENCE_LENGTH = 8;\n\nexport interface TTSOptions {\n model: TTSModels | string;\n encoding: TTSEncoding;\n sampleRate: number;\n apiKey?: string;\n baseUrl?: string;\n sentenceTokenizer: tokenize.SentenceTokenizer;\n capabilities: tts.TTSCapabilities;\n}\n\nconst defaultTTSOptions: TTSOptions = {\n model: 'aura-asteria-en',\n encoding: 'linear16',\n sampleRate: 24000,\n apiKey: process.env.DEEPGRAM_API_KEY,\n baseUrl: 'https://api.deepgram.com',\n capabilities: {\n streaming: true,\n },\n sentenceTokenizer: new tokenize.basic.SentenceTokenizer({\n minSentenceLength: MIN_SENTENCE_LENGTH,\n }),\n};\n\nexport class TTS extends tts.TTS {\n private opts: TTSOptions;\n label = 'deepgram.TTS';\n\n get model(): string {\n return this.opts.model;\n }\n\n get provider(): string {\n return 'Deepgram';\n }\n\n constructor(opts: Partial<TTSOptions> = {}) {\n super(opts.sampleRate || defaultTTSOptions.sampleRate, NUM_CHANNELS, {\n streaming: opts.capabilities?.streaming ?? defaultTTSOptions.capabilities.streaming,\n });\n\n this.opts = {\n ...defaultTTSOptions,\n ...opts,\n };\n\n if (this.opts.apiKey === undefined) {\n throw new Error(\n 'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n );\n }\n }\n\n synthesize(\n text: string,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ): tts.ChunkedStream {\n return new ChunkedStream(this, text, this.opts, connOptions, abortSignal);\n }\n\n stream(): tts.SynthesizeStream {\n return new SynthesizeStream(this, this.opts);\n }\n}\n\nexport class ChunkedStream extends tts.ChunkedStream {\n label = 'deepgram.ChunkedStream';\n #logger = log();\n private opts: TTSOptions;\n private text: string;\n\n constructor(\n tts: TTS,\n text: string,\n opts: TTSOptions,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ) {\n super(text, tts, connOptions, abortSignal);\n this.text = text;\n this.opts = opts;\n }\n\n protected async run() {\n const requestId = shortuuid();\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n const json = { text: this.text };\n const url = new URL(`${this.opts.baseUrl!}/v1/speak`);\n url.searchParams.append('sample_rate', this.opts.sampleRate.toString());\n url.searchParams.append('model', this.opts.model);\n url.searchParams.append('encoding', this.opts.encoding);\n\n await new Promise<void>((resolve, reject) => {\n const req = request(\n {\n hostname: url.hostname,\n port: 443,\n path: url.pathname + url.search,\n method: 'POST',\n headers: {\n [AUTHORIZATION_HEADER]: `Token ${this.opts.apiKey!}`,\n 'Content-Type': 'application/json',\n },\n signal: this.abortSignal,\n },\n (res) => {\n if (res.statusCode !== 200) {\n reject(\n new Error(`Deepgram TTS HTTP request failed: ${res.statusCode} ${res.statusMessage}`),\n );\n return;\n }\n\n res.on('data', (chunk) => {\n for (const frame of bstream.write(chunk)) {\n if (!this.queue.closed) {\n this.queue.put({\n requestId,\n frame,\n final: false,\n segmentId: requestId,\n });\n }\n }\n });\n\n res.on('error', (err) => {\n if (err.message === 'aborted') return;\n this.#logger.error({ err }, 'Deepgram TTS response error');\n });\n\n res.on('close', () => {\n for (const frame of bstream.flush()) {\n if (!this.queue.closed) {\n this.queue.put({\n requestId,\n frame,\n final: false,\n segmentId: requestId,\n });\n }\n }\n if (!this.queue.closed) {\n this.queue.close();\n }\n resolve();\n });\n },\n );\n\n req.on('error', (err) => {\n if (err.name === 'AbortError') return;\n this.#logger.error({ err }, 'Deepgram TTS request error');\n });\n\n req.on('close', () => resolve());\n req.write(JSON.stringify(json));\n req.end();\n });\n }\n}\n\nexport class SynthesizeStream extends tts.SynthesizeStream {\n private opts: TTSOptions;\n private tokenizer: tokenize.SentenceStream;\n label = 'deepgram.SynthesizeStream';\n\n private static readonly FLUSH_MSG = JSON.stringify({ type: 'Flush' });\n private static readonly CLOSE_MSG = JSON.stringify({ type: 'Close' });\n\n constructor(tts: TTS, opts: TTSOptions) {\n super(tts);\n this.opts = opts;\n this.tokenizer = opts.sentenceTokenizer.stream();\n }\n\n private async closeWebSocket(ws: WebSocket): Promise<void> {\n try {\n // Send Flush and Close messages to ensure Deepgram processes all remaining audio\n // and properly terminates the session, preventing lingering TTS sessions\n if (ws.readyState === WebSocket.OPEN) {\n ws.send(SynthesizeStream.FLUSH_MSG);\n ws.send(SynthesizeStream.CLOSE_MSG);\n\n // Wait for server acknowledgment to prevent race conditions and ensure\n // proper cleanup, avoiding 429 Too Many Requests errors from lingering sessions\n try {\n await new Promise<void>((resolve, _reject) => {\n const timeout = setTimeout(() => {\n resolve();\n }, 1000);\n\n ws.once('message', () => {\n clearTimeout(timeout);\n resolve();\n });\n\n ws.once('close', () => {\n clearTimeout(timeout);\n resolve();\n });\n\n ws.once('error', () => {\n clearTimeout(timeout);\n resolve();\n });\n });\n } catch (e) {\n // Ignore timeout or other errors during close sequence\n }\n }\n } catch (e) {\n console.warn(`Error during WebSocket close sequence: ${e}`);\n } finally {\n if (ws.readyState === WebSocket.OPEN || ws.readyState === WebSocket.CONNECTING) {\n ws.close();\n }\n }\n }\n\n protected async run() {\n const requestId = shortuuid();\n const segmentId = shortuuid();\n\n const wsUrl = this.opts.baseUrl!.replace(/^http/, 'ws');\n const url = new URL(`${wsUrl}/v1/speak`);\n url.searchParams.append('sample_rate', this.opts.sampleRate.toString());\n url.searchParams.append('model', this.opts.model);\n url.searchParams.append('encoding', this.opts.encoding);\n\n const ws = new WebSocket(url, {\n headers: {\n [AUTHORIZATION_HEADER]: `Token ${this.opts.apiKey!}`,\n },\n });\n\n await new Promise((resolve, reject) => {\n ws.on('open', resolve);\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n });\n\n const inputTask = async () => {\n for await (const data of this.input) {\n if (data === SynthesizeStream.FLUSH_SENTINEL) {\n this.tokenizer.flush();\n continue;\n }\n this.tokenizer.pushText(data);\n }\n this.tokenizer.endInput();\n this.tokenizer.close();\n };\n\n const sendTask = async () => {\n for await (const event of this.tokenizer) {\n if (this.abortController.signal.aborted) break;\n\n let text = event.token;\n if (!text.endsWith(' ')) {\n text += ' ';\n }\n\n const message = JSON.stringify({\n type: 'Speak',\n text: text,\n });\n\n ws.send(message);\n }\n\n if (!this.abortController.signal.aborted) {\n ws.send(SynthesizeStream.FLUSH_MSG);\n }\n };\n\n const recvTask = async () => {\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n let finalReceived = false;\n let timeout: NodeJS.Timeout | null = null;\n let lastFrame: AudioFrame | undefined;\n\n const sendLastFrame = (segmentId: string, final: boolean) => {\n if (lastFrame && !this.queue.closed) {\n this.queue.put({ requestId, segmentId, frame: lastFrame, final });\n lastFrame = undefined;\n }\n };\n\n const clearMessageTimeout = () => {\n if (timeout) {\n clearTimeout(timeout);\n timeout = null;\n }\n };\n\n return new Promise<void>((resolve, reject) => {\n ws.on('message', (data: RawData, isBinary: boolean) => {\n clearMessageTimeout();\n\n if (!isBinary) {\n const message = JSON.parse(data.toString());\n if (message.type === 'Flushed') {\n finalReceived = true;\n clearMessageTimeout();\n for (const frame of bstream.flush()) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n sendLastFrame(segmentId, true);\n\n if (!this.queue.closed) {\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n }\n resolve();\n }\n\n return;\n }\n\n const buffer =\n data instanceof Buffer\n ? data.buffer.slice(data.byteOffset, data.byteOffset + data.byteLength)\n : (data as ArrayBuffer);\n for (const frame of bstream.write(buffer as ArrayBuffer)) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n });\n\n ws.on('close', (_code, _reason) => {\n if (!finalReceived) {\n for (const frame of bstream.flush()) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n sendLastFrame(segmentId, true);\n\n if (!this.queue.closed) {\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n }\n }\n resolve();\n });\n\n ws.on('error', (error) => {\n clearMessageTimeout();\n reject(error);\n });\n });\n };\n\n try {\n await Promise.all([inputTask(), sendTask(), recvTask()]);\n } catch (e) {\n throw new Error(`failed in main task: ${e}`);\n } finally {\n await this.closeWebSocket(ws);\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBAOO;AAEP,wBAAwB;AACxB,gBAAwC;AAGxC,MAAM,uBAAuB;AAC7B,MAAM,eAAe;AACrB,MAAM,sBAAsB;AAY5B,MAAM,oBAAgC;AAAA,EACpC,OAAO;AAAA,EACP,UAAU;AAAA,EACV,YAAY;AAAA,EACZ,QAAQ,QAAQ,IAAI;AAAA,EACpB,SAAS;AAAA,EACT,cAAc;AAAA,IACZ,WAAW;AAAA,EACb;AAAA,EACA,mBAAmB,IAAI,uBAAS,MAAM,kBAAkB;AAAA,IACtD,mBAAmB;AAAA,EACrB,CAAC;AACH;AAEO,MAAM,YAAY,kBAAI,IAAI;AAAA,EACvB;AAAA,EACR,QAAQ;AAAA,EAER,IAAI,QAAgB;AAClB,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,YAAY,OAA4B,CAAC,GAAG;AAxD9C;AAyDI,UAAM,KAAK,cAAc,kBAAkB,YAAY,cAAc;AAAA,MACnE,aAAW,UAAK,iBAAL,mBAAmB,cAAa,kBAAkB,aAAa;AAAA,IAC5E,CAAC;AAED,SAAK,OAAO;AAAA,MACV,GAAG;AAAA,MACH,GAAG;AAAA,IACL;AAEA,QAAI,KAAK,KAAK,WAAW,QAAW;AAClC,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,WACE,MACA,aACA,aACmB;AACnB,WAAO,IAAI,cAAc,MAAM,MAAM,KAAK,MAAM,aAAa,WAAW;AAAA,EAC1E;AAAA,EAEA,SAA+B;AAC7B,WAAO,IAAI,iBAAiB,MAAM,KAAK,IAAI;AAAA,EAC7C;AACF;AAEO,MAAM,sBAAsB,kBAAI,cAAc;AAAA,EACnD,QAAQ;AAAA,EACR,cAAU,mBAAI;AAAA,EACN;AAAA,EACA;AAAA,EAER,YACEA,MACA,MACA,MACA,aACA,aACA;AACA,UAAM,MAAMA,MAAK,aAAa,WAAW;AACzC,SAAK,OAAO;AACZ,SAAK,OAAO;AAAA,EACd;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,gBAAY,yBAAU;AAC5B,UAAM,UAAU,IAAI,8BAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,UAAM,OAAO,EAAE,MAAM,KAAK,KAAK;AAC/B,UAAM,MAAM,IAAI,IAAI,GAAG,KAAK,KAAK,OAAQ,WAAW;AACpD,QAAI,aAAa,OAAO,eAAe,KAAK,KAAK,WAAW,SAAS,CAAC;AACtE,QAAI,aAAa,OAAO,SAAS,KAAK,KAAK,KAAK;AAChD,QAAI,aAAa,OAAO,YAAY,KAAK,KAAK,QAAQ;AAEtD,UAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,YAAM,UAAM;AAAA,QACV;AAAA,UACE,UAAU,IAAI;AAAA,UACd,MAAM;AAAA,UACN,MAAM,IAAI,WAAW,IAAI;AAAA,UACzB,QAAQ;AAAA,UACR,SAAS;AAAA,YACP,CAAC,oBAAoB,GAAG,SAAS,KAAK,KAAK,MAAO;AAAA,YAClD,gBAAgB;AAAA,UAClB;AAAA,UACA,QAAQ,KAAK;AAAA,QACf;AAAA,QACA,CAAC,QAAQ;AACP,cAAI,IAAI,eAAe,KAAK;AAC1B;AAAA,cACE,IAAI,MAAM,qCAAqC,IAAI,UAAU,IAAI,IAAI,aAAa,EAAE;AAAA,YACtF;AACA;AAAA,UACF;AAEA,cAAI,GAAG,QAAQ,CAAC,UAAU;AACxB,uBAAW,SAAS,QAAQ,MAAM,KAAK,GAAG;AACxC,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI;AAAA,kBACb;AAAA,kBACA;AAAA,kBACA,OAAO;AAAA,kBACP,WAAW;AAAA,gBACb,CAAC;AAAA,cACH;AAAA,YACF;AAAA,UACF,CAAC;AAED,cAAI,GAAG,SAAS,CAAC,QAAQ;AACvB,gBAAI,IAAI,YAAY,UAAW;AAC/B,iBAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,6BAA6B;AAAA,UAC3D,CAAC;AAED,cAAI,GAAG,SAAS,MAAM;AACpB,uBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI;AAAA,kBACb;AAAA,kBACA;AAAA,kBACA,OAAO;AAAA,kBACP,WAAW;AAAA,gBACb,CAAC;AAAA,cACH;AAAA,YACF;AACA,gBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,mBAAK,MAAM,MAAM;AAAA,YACnB;AACA,oBAAQ;AAAA,UACV,CAAC;AAAA,QACH;AAAA,MACF;AAEA,UAAI,GAAG,SAAS,CAAC,QAAQ;AACvB,YAAI,IAAI,SAAS,aAAc;AAC/B,aAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,4BAA4B;AAAA,MAC1D,CAAC;AAED,UAAI,GAAG,SAAS,MAAM,QAAQ,CAAC;AAC/B,UAAI,MAAM,KAAK,UAAU,IAAI,CAAC;AAC9B,UAAI,IAAI;AAAA,IACV,CAAC;AAAA,EACH;AACF;AAEO,MAAM,yBAAyB,kBAAI,iBAAiB;AAAA,EACjD;AAAA,EACA;AAAA,EACR,QAAQ;AAAA,EAER,OAAwB,YAAY,KAAK,UAAU,EAAE,MAAM,QAAQ,CAAC;AAAA,EACpE,OAAwB,YAAY,KAAK,UAAU,EAAE,MAAM,QAAQ,CAAC;AAAA,EAEpE,YAAYA,MAAU,MAAkB;AACtC,UAAMA,IAAG;AACT,SAAK,OAAO;AACZ,SAAK,YAAY,KAAK,kBAAkB,OAAO;AAAA,EACjD;AAAA,EAEA,MAAc,eAAe,IAA8B;AACzD,QAAI;AAGF,UAAI,GAAG,eAAe,oBAAU,MAAM;AACpC,WAAG,KAAK,iBAAiB,SAAS;AAClC,WAAG,KAAK,iBAAiB,SAAS;AAIlC,YAAI;AACF,gBAAM,IAAI,QAAc,CAAC,SAAS,YAAY;AAC5C,kBAAM,UAAU,WAAW,MAAM;AAC/B,sBAAQ;AAAA,YACV,GAAG,GAAI;AAEP,eAAG,KAAK,WAAW,MAAM;AACvB,2BAAa,OAAO;AACpB,sBAAQ;AAAA,YACV,CAAC;AAED,eAAG,KAAK,SAAS,MAAM;AACrB,2BAAa,OAAO;AACpB,sBAAQ;AAAA,YACV,CAAC;AAED,eAAG,KAAK,SAAS,MAAM;AACrB,2BAAa,OAAO;AACpB,sBAAQ;AAAA,YACV,CAAC;AAAA,UACH,CAAC;AAAA,QACH,SAAS,GAAG;AAAA,QAEZ;AAAA,MACF;AAAA,IACF,SAAS,GAAG;AACV,cAAQ,KAAK,0CAA0C,CAAC,EAAE;AAAA,IAC5D,UAAE;AACA,UAAI,GAAG,eAAe,oBAAU,QAAQ,GAAG,eAAe,oBAAU,YAAY;AAC9E,WAAG,MAAM;AAAA,MACX;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,gBAAY,yBAAU;AAC5B,UAAM,gBAAY,yBAAU;AAE5B,UAAM,QAAQ,KAAK,KAAK,QAAS,QAAQ,SAAS,IAAI;AACtD,UAAM,MAAM,IAAI,IAAI,GAAG,KAAK,WAAW;AACvC,QAAI,aAAa,OAAO,eAAe,KAAK,KAAK,WAAW,SAAS,CAAC;AACtE,QAAI,aAAa,OAAO,SAAS,KAAK,KAAK,KAAK;AAChD,QAAI,aAAa,OAAO,YAAY,KAAK,KAAK,QAAQ;AAEtD,UAAM,KAAK,IAAI,oBAAU,KAAK;AAAA,MAC5B,SAAS;AAAA,QACP,CAAC,oBAAoB,GAAG,SAAS,KAAK,KAAK,MAAO;AAAA,MACpD;AAAA,IACF,CAAC;AAED,UAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,SAAG,GAAG,QAAQ,OAAO;AACrB,SAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,SAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,IAC/D,CAAC;AAED,UAAM,YAAY,YAAY;AAC5B,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI,SAAS,iBAAiB,gBAAgB;AAC5C,eAAK,UAAU,MAAM;AACrB;AAAA,QACF;AACA,aAAK,UAAU,SAAS,IAAI;AAAA,MAC9B;AACA,WAAK,UAAU,SAAS;AACxB,WAAK,UAAU,MAAM;AAAA,IACvB;AAEA,UAAM,WAAW,YAAY;AAC3B,uBAAiB,SAAS,KAAK,WAAW;AACxC,YAAI,KAAK,gBAAgB,OAAO,QAAS;AAEzC,YAAI,OAAO,MAAM;AACjB,YAAI,CAAC,KAAK,SAAS,GAAG,GAAG;AACvB,kBAAQ;AAAA,QACV;AAEA,cAAM,UAAU,KAAK,UAAU;AAAA,UAC7B,MAAM;AAAA,UACN;AAAA,QACF,CAAC;AAED,WAAG,KAAK,OAAO;AAAA,MACjB;AAEA,UAAI,CAAC,KAAK,gBAAgB,OAAO,SAAS;AACxC,WAAG,KAAK,iBAAiB,SAAS;AAAA,MACpC;AAAA,IACF;AAEA,UAAM,WAAW,YAAY;AAC3B,YAAM,UAAU,IAAI,8BAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,UAAI,gBAAgB;AACpB,UAAI,UAAiC;AACrC,UAAI;AAEJ,YAAM,gBAAgB,CAACC,YAAmB,UAAmB;AAC3D,YAAI,aAAa,CAAC,KAAK,MAAM,QAAQ;AACnC,eAAK,MAAM,IAAI,EAAE,WAAW,WAAAA,YAAW,OAAO,WAAW,MAAM,CAAC;AAChE,sBAAY;AAAA,QACd;AAAA,MACF;AAEA,YAAM,sBAAsB,MAAM;AAChC,YAAI,SAAS;AACX,uBAAa,OAAO;AACpB,oBAAU;AAAA,QACZ;AAAA,MACF;AAEA,aAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,WAAG,GAAG,WAAW,CAAC,MAAe,aAAsB;AACrD,8BAAoB;AAEpB,cAAI,CAAC,UAAU;AACb,kBAAM,UAAU,KAAK,MAAM,KAAK,SAAS,CAAC;AAC1C,gBAAI,QAAQ,SAAS,WAAW;AAC9B,8BAAgB;AAChB,kCAAoB;AACpB,yBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,8BAAc,WAAW,KAAK;AAC9B,4BAAY;AAAA,cACd;AACA,4BAAc,WAAW,IAAI;AAE7B,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI,iBAAiB,aAAa;AAAA,cAC/C;AACA,sBAAQ;AAAA,YACV;AAEA;AAAA,UACF;AAEA,gBAAM,SACJ,gBAAgB,SACZ,KAAK,OAAO,MAAM,KAAK,YAAY,KAAK,aAAa,KAAK,UAAU,IACnE;AACP,qBAAW,SAAS,QAAQ,MAAM,MAAqB,GAAG;AACxD,0BAAc,WAAW,KAAK;AAC9B,wBAAY;AAAA,UACd;AAAA,QACF,CAAC;AAED,WAAG,GAAG,SAAS,CAAC,OAAO,YAAY;AACjC,cAAI,CAAC,eAAe;AAClB,uBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,4BAAc,WAAW,KAAK;AAC9B,0BAAY;AAAA,YACd;AACA,0BAAc,WAAW,IAAI;AAE7B,gBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,mBAAK,MAAM,IAAI,iBAAiB,aAAa;AAAA,YAC/C;AAAA,UACF;AACA,kBAAQ;AAAA,QACV,CAAC;AAED,WAAG,GAAG,SAAS,CAAC,UAAU;AACxB,8BAAoB;AACpB,iBAAO,KAAK;AAAA,QACd,CAAC;AAAA,MACH,CAAC;AAAA,IACH;AAEA,QAAI;AACF,YAAM,QAAQ,IAAI,CAAC,UAAU,GAAG,SAAS,GAAG,SAAS,CAAC,CAAC;AAAA,IACzD,SAAS,GAAG;AACV,YAAM,IAAI,MAAM,wBAAwB,CAAC,EAAE;AAAA,IAC7C,UAAE;AACA,YAAM,KAAK,eAAe,EAAE;AAAA,IAC9B;AAAA,EACF;AACF;","names":["tts","segmentId"]}
|
package/dist/tts.d.cts
CHANGED
|
@@ -12,6 +12,8 @@ export interface TTSOptions {
|
|
|
12
12
|
export declare class TTS extends tts.TTS {
|
|
13
13
|
private opts;
|
|
14
14
|
label: string;
|
|
15
|
+
get model(): string;
|
|
16
|
+
get provider(): string;
|
|
15
17
|
constructor(opts?: Partial<TTSOptions>);
|
|
16
18
|
synthesize(text: string, connOptions?: APIConnectOptions, abortSignal?: AbortSignal): tts.ChunkedStream;
|
|
17
19
|
stream(): tts.SynthesizeStream;
|
package/dist/tts.d.ts
CHANGED
|
@@ -12,6 +12,8 @@ export interface TTSOptions {
|
|
|
12
12
|
export declare class TTS extends tts.TTS {
|
|
13
13
|
private opts;
|
|
14
14
|
label: string;
|
|
15
|
+
get model(): string;
|
|
16
|
+
get provider(): string;
|
|
15
17
|
constructor(opts?: Partial<TTSOptions>);
|
|
16
18
|
synthesize(text: string, connOptions?: APIConnectOptions, abortSignal?: AbortSignal): tts.ChunkedStream;
|
|
17
19
|
stream(): tts.SynthesizeStream;
|
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,EAItB,QAAQ,EACR,GAAG,EACJ,MAAM,iBAAiB,CAAC;AAIzB,OAAO,KAAK,EAAE,WAAW,EAAE,SAAS,EAAE,MAAM,aAAa,CAAC;AAM1D,MAAM,WAAW,UAAU;IACzB,KAAK,EAAE,SAAS,GAAG,MAAM,CAAC;IAC1B,QAAQ,EAAE,WAAW,CAAC;IACtB,UAAU,EAAE,MAAM,CAAC;IACnB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,iBAAiB,EAAE,QAAQ,CAAC,iBAAiB,CAAC;IAC9C,YAAY,EAAE,GAAG,CAAC,eAAe,CAAC;CACnC;AAgBD,qBAAa,GAAI,SAAQ,GAAG,CAAC,GAAG;IAC9B,OAAO,CAAC,IAAI,CAAa;IACzB,KAAK,SAAkB;
|
|
1
|
+
{"version":3,"file":"tts.d.ts","sourceRoot":"","sources":["../src/tts.ts"],"names":[],"mappings":"AAGA,OAAO,EACL,KAAK,iBAAiB,EAItB,QAAQ,EACR,GAAG,EACJ,MAAM,iBAAiB,CAAC;AAIzB,OAAO,KAAK,EAAE,WAAW,EAAE,SAAS,EAAE,MAAM,aAAa,CAAC;AAM1D,MAAM,WAAW,UAAU;IACzB,KAAK,EAAE,SAAS,GAAG,MAAM,CAAC;IAC1B,QAAQ,EAAE,WAAW,CAAC;IACtB,UAAU,EAAE,MAAM,CAAC;IACnB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,iBAAiB,EAAE,QAAQ,CAAC,iBAAiB,CAAC;IAC9C,YAAY,EAAE,GAAG,CAAC,eAAe,CAAC;CACnC;AAgBD,qBAAa,GAAI,SAAQ,GAAG,CAAC,GAAG;IAC9B,OAAO,CAAC,IAAI,CAAa;IACzB,KAAK,SAAkB;IAEvB,IAAI,KAAK,IAAI,MAAM,CAElB;IAED,IAAI,QAAQ,IAAI,MAAM,CAErB;gBAEW,IAAI,GAAE,OAAO,CAAC,UAAU,CAAM;IAiB1C,UAAU,CACR,IAAI,EAAE,MAAM,EACZ,WAAW,CAAC,EAAE,iBAAiB,EAC/B,WAAW,CAAC,EAAE,WAAW,GACxB,GAAG,CAAC,aAAa;IAIpB,MAAM,IAAI,GAAG,CAAC,gBAAgB;CAG/B;AAED,qBAAa,aAAc,SAAQ,GAAG,CAAC,aAAa;;IAClD,KAAK,SAA4B;IAEjC,OAAO,CAAC,IAAI,CAAa;IACzB,OAAO,CAAC,IAAI,CAAS;gBAGnB,GAAG,EAAE,GAAG,EACR,IAAI,EAAE,MAAM,EACZ,IAAI,EAAE,UAAU,EAChB,WAAW,CAAC,EAAE,iBAAiB,EAC/B,WAAW,CAAC,EAAE,WAAW;cAOX,GAAG;CA6EpB;AAED,qBAAa,gBAAiB,SAAQ,GAAG,CAAC,gBAAgB;IACxD,OAAO,CAAC,IAAI,CAAa;IACzB,OAAO,CAAC,SAAS,CAA0B;IAC3C,KAAK,SAA+B;IAEpC,OAAO,CAAC,MAAM,CAAC,QAAQ,CAAC,SAAS,CAAqC;IACtE,OAAO,CAAC,MAAM,CAAC,QAAQ,CAAC,SAAS,CAAqC;gBAE1D,GAAG,EAAE,GAAG,EAAE,IAAI,EAAE,UAAU;YAMxB,cAAc;cA4CZ,GAAG;CA4IpB"}
|
package/dist/tts.js
CHANGED
|
@@ -26,6 +26,12 @@ const defaultTTSOptions = {
|
|
|
26
26
|
class TTS extends tts.TTS {
|
|
27
27
|
opts;
|
|
28
28
|
label = "deepgram.TTS";
|
|
29
|
+
get model() {
|
|
30
|
+
return this.opts.model;
|
|
31
|
+
}
|
|
32
|
+
get provider() {
|
|
33
|
+
return "Deepgram";
|
|
34
|
+
}
|
|
29
35
|
constructor(opts = {}) {
|
|
30
36
|
var _a;
|
|
31
37
|
super(opts.sampleRate || defaultTTSOptions.sampleRate, NUM_CHANNELS, {
|
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 log,\n shortuuid,\n tokenize,\n tts,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { request } from 'node:https';\nimport { type RawData, WebSocket } from 'ws';\nimport type { TTSEncoding, TTSModels } from './models.js';\n\nconst AUTHORIZATION_HEADER = 'Authorization';\nconst NUM_CHANNELS = 1;\nconst MIN_SENTENCE_LENGTH = 8;\n\nexport interface TTSOptions {\n model: TTSModels | string;\n encoding: TTSEncoding;\n sampleRate: number;\n apiKey?: string;\n baseUrl?: string;\n sentenceTokenizer: tokenize.SentenceTokenizer;\n capabilities: tts.TTSCapabilities;\n}\n\nconst defaultTTSOptions: TTSOptions = {\n model: 'aura-asteria-en',\n encoding: 'linear16',\n sampleRate: 24000,\n apiKey: process.env.DEEPGRAM_API_KEY,\n baseUrl: 'https://api.deepgram.com',\n capabilities: {\n streaming: true,\n },\n sentenceTokenizer: new tokenize.basic.SentenceTokenizer({\n minSentenceLength: MIN_SENTENCE_LENGTH,\n }),\n};\n\nexport class TTS extends tts.TTS {\n private opts: TTSOptions;\n label = 'deepgram.TTS';\n\n constructor(opts: Partial<TTSOptions> = {}) {\n super(opts.sampleRate || defaultTTSOptions.sampleRate, NUM_CHANNELS, {\n streaming: opts.capabilities?.streaming ?? defaultTTSOptions.capabilities.streaming,\n });\n\n this.opts = {\n ...defaultTTSOptions,\n ...opts,\n };\n\n if (this.opts.apiKey === undefined) {\n throw new Error(\n 'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n );\n }\n }\n\n synthesize(\n text: string,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ): tts.ChunkedStream {\n return new ChunkedStream(this, text, this.opts, connOptions, abortSignal);\n }\n\n stream(): tts.SynthesizeStream {\n return new SynthesizeStream(this, this.opts);\n }\n}\n\nexport class ChunkedStream extends tts.ChunkedStream {\n label = 'deepgram.ChunkedStream';\n #logger = log();\n private opts: TTSOptions;\n private text: string;\n\n constructor(\n tts: TTS,\n text: string,\n opts: TTSOptions,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ) {\n super(text, tts, connOptions, abortSignal);\n this.text = text;\n this.opts = opts;\n }\n\n protected async run() {\n const requestId = shortuuid();\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n const json = { text: this.text };\n const url = new URL(`${this.opts.baseUrl!}/v1/speak`);\n url.searchParams.append('sample_rate', this.opts.sampleRate.toString());\n url.searchParams.append('model', this.opts.model);\n url.searchParams.append('encoding', this.opts.encoding);\n\n await new Promise<void>((resolve, reject) => {\n const req = request(\n {\n hostname: url.hostname,\n port: 443,\n path: url.pathname + url.search,\n method: 'POST',\n headers: {\n [AUTHORIZATION_HEADER]: `Token ${this.opts.apiKey!}`,\n 'Content-Type': 'application/json',\n },\n signal: this.abortSignal,\n },\n (res) => {\n if (res.statusCode !== 200) {\n reject(\n new Error(`Deepgram TTS HTTP request failed: ${res.statusCode} ${res.statusMessage}`),\n );\n return;\n }\n\n res.on('data', (chunk) => {\n for (const frame of bstream.write(chunk)) {\n if (!this.queue.closed) {\n this.queue.put({\n requestId,\n frame,\n final: false,\n segmentId: requestId,\n });\n }\n }\n });\n\n res.on('error', (err) => {\n if (err.message === 'aborted') return;\n this.#logger.error({ err }, 'Deepgram TTS response error');\n });\n\n res.on('close', () => {\n for (const frame of bstream.flush()) {\n if (!this.queue.closed) {\n this.queue.put({\n requestId,\n frame,\n final: false,\n segmentId: requestId,\n });\n }\n }\n if (!this.queue.closed) {\n this.queue.close();\n }\n resolve();\n });\n },\n );\n\n req.on('error', (err) => {\n if (err.name === 'AbortError') return;\n this.#logger.error({ err }, 'Deepgram TTS request error');\n });\n\n req.on('close', () => resolve());\n req.write(JSON.stringify(json));\n req.end();\n });\n }\n}\n\nexport class SynthesizeStream extends tts.SynthesizeStream {\n private opts: TTSOptions;\n private tokenizer: tokenize.SentenceStream;\n label = 'deepgram.SynthesizeStream';\n\n private static readonly FLUSH_MSG = JSON.stringify({ type: 'Flush' });\n private static readonly CLOSE_MSG = JSON.stringify({ type: 'Close' });\n\n constructor(tts: TTS, opts: TTSOptions) {\n super(tts);\n this.opts = opts;\n this.tokenizer = opts.sentenceTokenizer.stream();\n }\n\n private async closeWebSocket(ws: WebSocket): Promise<void> {\n try {\n // Send Flush and Close messages to ensure Deepgram processes all remaining audio\n // and properly terminates the session, preventing lingering TTS sessions\n if (ws.readyState === WebSocket.OPEN) {\n ws.send(SynthesizeStream.FLUSH_MSG);\n ws.send(SynthesizeStream.CLOSE_MSG);\n\n // Wait for server acknowledgment to prevent race conditions and ensure\n // proper cleanup, avoiding 429 Too Many Requests errors from lingering sessions\n try {\n await new Promise<void>((resolve, _reject) => {\n const timeout = setTimeout(() => {\n resolve();\n }, 1000);\n\n ws.once('message', () => {\n clearTimeout(timeout);\n resolve();\n });\n\n ws.once('close', () => {\n clearTimeout(timeout);\n resolve();\n });\n\n ws.once('error', () => {\n clearTimeout(timeout);\n resolve();\n });\n });\n } catch (e) {\n // Ignore timeout or other errors during close sequence\n }\n }\n } catch (e) {\n console.warn(`Error during WebSocket close sequence: ${e}`);\n } finally {\n if (ws.readyState === WebSocket.OPEN || ws.readyState === WebSocket.CONNECTING) {\n ws.close();\n }\n }\n }\n\n protected async run() {\n const requestId = shortuuid();\n const segmentId = shortuuid();\n\n const wsUrl = this.opts.baseUrl!.replace(/^http/, 'ws');\n const url = new URL(`${wsUrl}/v1/speak`);\n url.searchParams.append('sample_rate', this.opts.sampleRate.toString());\n url.searchParams.append('model', this.opts.model);\n url.searchParams.append('encoding', this.opts.encoding);\n\n const ws = new WebSocket(url, {\n headers: {\n [AUTHORIZATION_HEADER]: `Token ${this.opts.apiKey!}`,\n },\n });\n\n await new Promise((resolve, reject) => {\n ws.on('open', resolve);\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n });\n\n const inputTask = async () => {\n for await (const data of this.input) {\n if (data === SynthesizeStream.FLUSH_SENTINEL) {\n this.tokenizer.flush();\n continue;\n }\n this.tokenizer.pushText(data);\n }\n this.tokenizer.endInput();\n this.tokenizer.close();\n };\n\n const sendTask = async () => {\n for await (const event of this.tokenizer) {\n if (this.abortController.signal.aborted) break;\n\n let text = event.token;\n if (!text.endsWith(' ')) {\n text += ' ';\n }\n\n const message = JSON.stringify({\n type: 'Speak',\n text: text,\n });\n\n ws.send(message);\n }\n\n if (!this.abortController.signal.aborted) {\n ws.send(SynthesizeStream.FLUSH_MSG);\n }\n };\n\n const recvTask = async () => {\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n let finalReceived = false;\n let timeout: NodeJS.Timeout | null = null;\n let lastFrame: AudioFrame | undefined;\n\n const sendLastFrame = (segmentId: string, final: boolean) => {\n if (lastFrame && !this.queue.closed) {\n this.queue.put({ requestId, segmentId, frame: lastFrame, final });\n lastFrame = undefined;\n }\n };\n\n const clearMessageTimeout = () => {\n if (timeout) {\n clearTimeout(timeout);\n timeout = null;\n }\n };\n\n return new Promise<void>((resolve, reject) => {\n ws.on('message', (data: RawData, isBinary: boolean) => {\n clearMessageTimeout();\n\n if (!isBinary) {\n const message = JSON.parse(data.toString());\n if (message.type === 'Flushed') {\n finalReceived = true;\n clearMessageTimeout();\n for (const frame of bstream.flush()) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n sendLastFrame(segmentId, true);\n\n if (!this.queue.closed) {\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n }\n resolve();\n }\n\n return;\n }\n\n const buffer =\n data instanceof Buffer\n ? data.buffer.slice(data.byteOffset, data.byteOffset + data.byteLength)\n : (data as ArrayBuffer);\n for (const frame of bstream.write(buffer as ArrayBuffer)) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n });\n\n ws.on('close', (_code, _reason) => {\n if (!finalReceived) {\n for (const frame of bstream.flush()) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n sendLastFrame(segmentId, true);\n\n if (!this.queue.closed) {\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n }\n }\n resolve();\n });\n\n ws.on('error', (error) => {\n clearMessageTimeout();\n reject(error);\n });\n });\n };\n\n try {\n await Promise.all([inputTask(), sendTask(), recvTask()]);\n } catch (e) {\n throw new Error(`failed in main task: ${e}`);\n } finally {\n await this.closeWebSocket(ws);\n }\n }\n}\n"],"mappings":"AAGA;AAAA,EAEE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,eAAe;AACxB,SAAuB,iBAAiB;AAGxC,MAAM,uBAAuB;AAC7B,MAAM,eAAe;AACrB,MAAM,sBAAsB;AAY5B,MAAM,oBAAgC;AAAA,EACpC,OAAO;AAAA,EACP,UAAU;AAAA,EACV,YAAY;AAAA,EACZ,QAAQ,QAAQ,IAAI;AAAA,EACpB,SAAS;AAAA,EACT,cAAc;AAAA,IACZ,WAAW;AAAA,EACb;AAAA,EACA,mBAAmB,IAAI,SAAS,MAAM,kBAAkB;AAAA,IACtD,mBAAmB;AAAA,EACrB,CAAC;AACH;AAEO,MAAM,YAAY,IAAI,IAAI;AAAA,EACvB;AAAA,EACR,QAAQ;AAAA,EAER,YAAY,OAA4B,CAAC,GAAG;AAhD9C;AAiDI,UAAM,KAAK,cAAc,kBAAkB,YAAY,cAAc;AAAA,MACnE,aAAW,UAAK,iBAAL,mBAAmB,cAAa,kBAAkB,aAAa;AAAA,IAC5E,CAAC;AAED,SAAK,OAAO;AAAA,MACV,GAAG;AAAA,MACH,GAAG;AAAA,IACL;AAEA,QAAI,KAAK,KAAK,WAAW,QAAW;AAClC,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,WACE,MACA,aACA,aACmB;AACnB,WAAO,IAAI,cAAc,MAAM,MAAM,KAAK,MAAM,aAAa,WAAW;AAAA,EAC1E;AAAA,EAEA,SAA+B;AAC7B,WAAO,IAAI,iBAAiB,MAAM,KAAK,IAAI;AAAA,EAC7C;AACF;AAEO,MAAM,sBAAsB,IAAI,cAAc;AAAA,EACnD,QAAQ;AAAA,EACR,UAAU,IAAI;AAAA,EACN;AAAA,EACA;AAAA,EAER,YACEA,MACA,MACA,MACA,aACA,aACA;AACA,UAAM,MAAMA,MAAK,aAAa,WAAW;AACzC,SAAK,OAAO;AACZ,SAAK,OAAO;AAAA,EACd;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,YAAY,UAAU;AAC5B,UAAM,UAAU,IAAI,gBAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,UAAM,OAAO,EAAE,MAAM,KAAK,KAAK;AAC/B,UAAM,MAAM,IAAI,IAAI,GAAG,KAAK,KAAK,OAAQ,WAAW;AACpD,QAAI,aAAa,OAAO,eAAe,KAAK,KAAK,WAAW,SAAS,CAAC;AACtE,QAAI,aAAa,OAAO,SAAS,KAAK,KAAK,KAAK;AAChD,QAAI,aAAa,OAAO,YAAY,KAAK,KAAK,QAAQ;AAEtD,UAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,YAAM,MAAM;AAAA,QACV;AAAA,UACE,UAAU,IAAI;AAAA,UACd,MAAM;AAAA,UACN,MAAM,IAAI,WAAW,IAAI;AAAA,UACzB,QAAQ;AAAA,UACR,SAAS;AAAA,YACP,CAAC,oBAAoB,GAAG,SAAS,KAAK,KAAK,MAAO;AAAA,YAClD,gBAAgB;AAAA,UAClB;AAAA,UACA,QAAQ,KAAK;AAAA,QACf;AAAA,QACA,CAAC,QAAQ;AACP,cAAI,IAAI,eAAe,KAAK;AAC1B;AAAA,cACE,IAAI,MAAM,qCAAqC,IAAI,UAAU,IAAI,IAAI,aAAa,EAAE;AAAA,YACtF;AACA;AAAA,UACF;AAEA,cAAI,GAAG,QAAQ,CAAC,UAAU;AACxB,uBAAW,SAAS,QAAQ,MAAM,KAAK,GAAG;AACxC,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI;AAAA,kBACb;AAAA,kBACA;AAAA,kBACA,OAAO;AAAA,kBACP,WAAW;AAAA,gBACb,CAAC;AAAA,cACH;AAAA,YACF;AAAA,UACF,CAAC;AAED,cAAI,GAAG,SAAS,CAAC,QAAQ;AACvB,gBAAI,IAAI,YAAY,UAAW;AAC/B,iBAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,6BAA6B;AAAA,UAC3D,CAAC;AAED,cAAI,GAAG,SAAS,MAAM;AACpB,uBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI;AAAA,kBACb;AAAA,kBACA;AAAA,kBACA,OAAO;AAAA,kBACP,WAAW;AAAA,gBACb,CAAC;AAAA,cACH;AAAA,YACF;AACA,gBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,mBAAK,MAAM,MAAM;AAAA,YACnB;AACA,oBAAQ;AAAA,UACV,CAAC;AAAA,QACH;AAAA,MACF;AAEA,UAAI,GAAG,SAAS,CAAC,QAAQ;AACvB,YAAI,IAAI,SAAS,aAAc;AAC/B,aAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,4BAA4B;AAAA,MAC1D,CAAC;AAED,UAAI,GAAG,SAAS,MAAM,QAAQ,CAAC;AAC/B,UAAI,MAAM,KAAK,UAAU,IAAI,CAAC;AAC9B,UAAI,IAAI;AAAA,IACV,CAAC;AAAA,EACH;AACF;AAEO,MAAM,yBAAyB,IAAI,iBAAiB;AAAA,EACjD;AAAA,EACA;AAAA,EACR,QAAQ;AAAA,EAER,OAAwB,YAAY,KAAK,UAAU,EAAE,MAAM,QAAQ,CAAC;AAAA,EACpE,OAAwB,YAAY,KAAK,UAAU,EAAE,MAAM,QAAQ,CAAC;AAAA,EAEpE,YAAYA,MAAU,MAAkB;AACtC,UAAMA,IAAG;AACT,SAAK,OAAO;AACZ,SAAK,YAAY,KAAK,kBAAkB,OAAO;AAAA,EACjD;AAAA,EAEA,MAAc,eAAe,IAA8B;AACzD,QAAI;AAGF,UAAI,GAAG,eAAe,UAAU,MAAM;AACpC,WAAG,KAAK,iBAAiB,SAAS;AAClC,WAAG,KAAK,iBAAiB,SAAS;AAIlC,YAAI;AACF,gBAAM,IAAI,QAAc,CAAC,SAAS,YAAY;AAC5C,kBAAM,UAAU,WAAW,MAAM;AAC/B,sBAAQ;AAAA,YACV,GAAG,GAAI;AAEP,eAAG,KAAK,WAAW,MAAM;AACvB,2BAAa,OAAO;AACpB,sBAAQ;AAAA,YACV,CAAC;AAED,eAAG,KAAK,SAAS,MAAM;AACrB,2BAAa,OAAO;AACpB,sBAAQ;AAAA,YACV,CAAC;AAED,eAAG,KAAK,SAAS,MAAM;AACrB,2BAAa,OAAO;AACpB,sBAAQ;AAAA,YACV,CAAC;AAAA,UACH,CAAC;AAAA,QACH,SAAS,GAAG;AAAA,QAEZ;AAAA,MACF;AAAA,IACF,SAAS,GAAG;AACV,cAAQ,KAAK,0CAA0C,CAAC,EAAE;AAAA,IAC5D,UAAE;AACA,UAAI,GAAG,eAAe,UAAU,QAAQ,GAAG,eAAe,UAAU,YAAY;AAC9E,WAAG,MAAM;AAAA,MACX;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,YAAY,UAAU;AAC5B,UAAM,YAAY,UAAU;AAE5B,UAAM,QAAQ,KAAK,KAAK,QAAS,QAAQ,SAAS,IAAI;AACtD,UAAM,MAAM,IAAI,IAAI,GAAG,KAAK,WAAW;AACvC,QAAI,aAAa,OAAO,eAAe,KAAK,KAAK,WAAW,SAAS,CAAC;AACtE,QAAI,aAAa,OAAO,SAAS,KAAK,KAAK,KAAK;AAChD,QAAI,aAAa,OAAO,YAAY,KAAK,KAAK,QAAQ;AAEtD,UAAM,KAAK,IAAI,UAAU,KAAK;AAAA,MAC5B,SAAS;AAAA,QACP,CAAC,oBAAoB,GAAG,SAAS,KAAK,KAAK,MAAO;AAAA,MACpD;AAAA,IACF,CAAC;AAED,UAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,SAAG,GAAG,QAAQ,OAAO;AACrB,SAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,SAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,IAC/D,CAAC;AAED,UAAM,YAAY,YAAY;AAC5B,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI,SAAS,iBAAiB,gBAAgB;AAC5C,eAAK,UAAU,MAAM;AACrB;AAAA,QACF;AACA,aAAK,UAAU,SAAS,IAAI;AAAA,MAC9B;AACA,WAAK,UAAU,SAAS;AACxB,WAAK,UAAU,MAAM;AAAA,IACvB;AAEA,UAAM,WAAW,YAAY;AAC3B,uBAAiB,SAAS,KAAK,WAAW;AACxC,YAAI,KAAK,gBAAgB,OAAO,QAAS;AAEzC,YAAI,OAAO,MAAM;AACjB,YAAI,CAAC,KAAK,SAAS,GAAG,GAAG;AACvB,kBAAQ;AAAA,QACV;AAEA,cAAM,UAAU,KAAK,UAAU;AAAA,UAC7B,MAAM;AAAA,UACN;AAAA,QACF,CAAC;AAED,WAAG,KAAK,OAAO;AAAA,MACjB;AAEA,UAAI,CAAC,KAAK,gBAAgB,OAAO,SAAS;AACxC,WAAG,KAAK,iBAAiB,SAAS;AAAA,MACpC;AAAA,IACF;AAEA,UAAM,WAAW,YAAY;AAC3B,YAAM,UAAU,IAAI,gBAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,UAAI,gBAAgB;AACpB,UAAI,UAAiC;AACrC,UAAI;AAEJ,YAAM,gBAAgB,CAACC,YAAmB,UAAmB;AAC3D,YAAI,aAAa,CAAC,KAAK,MAAM,QAAQ;AACnC,eAAK,MAAM,IAAI,EAAE,WAAW,WAAAA,YAAW,OAAO,WAAW,MAAM,CAAC;AAChE,sBAAY;AAAA,QACd;AAAA,MACF;AAEA,YAAM,sBAAsB,MAAM;AAChC,YAAI,SAAS;AACX,uBAAa,OAAO;AACpB,oBAAU;AAAA,QACZ;AAAA,MACF;AAEA,aAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,WAAG,GAAG,WAAW,CAAC,MAAe,aAAsB;AACrD,8BAAoB;AAEpB,cAAI,CAAC,UAAU;AACb,kBAAM,UAAU,KAAK,MAAM,KAAK,SAAS,CAAC;AAC1C,gBAAI,QAAQ,SAAS,WAAW;AAC9B,8BAAgB;AAChB,kCAAoB;AACpB,yBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,8BAAc,WAAW,KAAK;AAC9B,4BAAY;AAAA,cACd;AACA,4BAAc,WAAW,IAAI;AAE7B,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI,iBAAiB,aAAa;AAAA,cAC/C;AACA,sBAAQ;AAAA,YACV;AAEA;AAAA,UACF;AAEA,gBAAM,SACJ,gBAAgB,SACZ,KAAK,OAAO,MAAM,KAAK,YAAY,KAAK,aAAa,KAAK,UAAU,IACnE;AACP,qBAAW,SAAS,QAAQ,MAAM,MAAqB,GAAG;AACxD,0BAAc,WAAW,KAAK;AAC9B,wBAAY;AAAA,UACd;AAAA,QACF,CAAC;AAED,WAAG,GAAG,SAAS,CAAC,OAAO,YAAY;AACjC,cAAI,CAAC,eAAe;AAClB,uBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,4BAAc,WAAW,KAAK;AAC9B,0BAAY;AAAA,YACd;AACA,0BAAc,WAAW,IAAI;AAE7B,gBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,mBAAK,MAAM,IAAI,iBAAiB,aAAa;AAAA,YAC/C;AAAA,UACF;AACA,kBAAQ;AAAA,QACV,CAAC;AAED,WAAG,GAAG,SAAS,CAAC,UAAU;AACxB,8BAAoB;AACpB,iBAAO,KAAK;AAAA,QACd,CAAC;AAAA,MACH,CAAC;AAAA,IACH;AAEA,QAAI;AACF,YAAM,QAAQ,IAAI,CAAC,UAAU,GAAG,SAAS,GAAG,SAAS,CAAC,CAAC;AAAA,IACzD,SAAS,GAAG;AACV,YAAM,IAAI,MAAM,wBAAwB,CAAC,EAAE;AAAA,IAC7C,UAAE;AACA,YAAM,KAAK,eAAe,EAAE;AAAA,IAC9B;AAAA,EACF;AACF;","names":["tts","segmentId"]}
|
|
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 log,\n shortuuid,\n tokenize,\n tts,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { request } from 'node:https';\nimport { type RawData, WebSocket } from 'ws';\nimport type { TTSEncoding, TTSModels } from './models.js';\n\nconst AUTHORIZATION_HEADER = 'Authorization';\nconst NUM_CHANNELS = 1;\nconst MIN_SENTENCE_LENGTH = 8;\n\nexport interface TTSOptions {\n model: TTSModels | string;\n encoding: TTSEncoding;\n sampleRate: number;\n apiKey?: string;\n baseUrl?: string;\n sentenceTokenizer: tokenize.SentenceTokenizer;\n capabilities: tts.TTSCapabilities;\n}\n\nconst defaultTTSOptions: TTSOptions = {\n model: 'aura-asteria-en',\n encoding: 'linear16',\n sampleRate: 24000,\n apiKey: process.env.DEEPGRAM_API_KEY,\n baseUrl: 'https://api.deepgram.com',\n capabilities: {\n streaming: true,\n },\n sentenceTokenizer: new tokenize.basic.SentenceTokenizer({\n minSentenceLength: MIN_SENTENCE_LENGTH,\n }),\n};\n\nexport class TTS extends tts.TTS {\n private opts: TTSOptions;\n label = 'deepgram.TTS';\n\n get model(): string {\n return this.opts.model;\n }\n\n get provider(): string {\n return 'Deepgram';\n }\n\n constructor(opts: Partial<TTSOptions> = {}) {\n super(opts.sampleRate || defaultTTSOptions.sampleRate, NUM_CHANNELS, {\n streaming: opts.capabilities?.streaming ?? defaultTTSOptions.capabilities.streaming,\n });\n\n this.opts = {\n ...defaultTTSOptions,\n ...opts,\n };\n\n if (this.opts.apiKey === undefined) {\n throw new Error(\n 'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n );\n }\n }\n\n synthesize(\n text: string,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ): tts.ChunkedStream {\n return new ChunkedStream(this, text, this.opts, connOptions, abortSignal);\n }\n\n stream(): tts.SynthesizeStream {\n return new SynthesizeStream(this, this.opts);\n }\n}\n\nexport class ChunkedStream extends tts.ChunkedStream {\n label = 'deepgram.ChunkedStream';\n #logger = log();\n private opts: TTSOptions;\n private text: string;\n\n constructor(\n tts: TTS,\n text: string,\n opts: TTSOptions,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ) {\n super(text, tts, connOptions, abortSignal);\n this.text = text;\n this.opts = opts;\n }\n\n protected async run() {\n const requestId = shortuuid();\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n const json = { text: this.text };\n const url = new URL(`${this.opts.baseUrl!}/v1/speak`);\n url.searchParams.append('sample_rate', this.opts.sampleRate.toString());\n url.searchParams.append('model', this.opts.model);\n url.searchParams.append('encoding', this.opts.encoding);\n\n await new Promise<void>((resolve, reject) => {\n const req = request(\n {\n hostname: url.hostname,\n port: 443,\n path: url.pathname + url.search,\n method: 'POST',\n headers: {\n [AUTHORIZATION_HEADER]: `Token ${this.opts.apiKey!}`,\n 'Content-Type': 'application/json',\n },\n signal: this.abortSignal,\n },\n (res) => {\n if (res.statusCode !== 200) {\n reject(\n new Error(`Deepgram TTS HTTP request failed: ${res.statusCode} ${res.statusMessage}`),\n );\n return;\n }\n\n res.on('data', (chunk) => {\n for (const frame of bstream.write(chunk)) {\n if (!this.queue.closed) {\n this.queue.put({\n requestId,\n frame,\n final: false,\n segmentId: requestId,\n });\n }\n }\n });\n\n res.on('error', (err) => {\n if (err.message === 'aborted') return;\n this.#logger.error({ err }, 'Deepgram TTS response error');\n });\n\n res.on('close', () => {\n for (const frame of bstream.flush()) {\n if (!this.queue.closed) {\n this.queue.put({\n requestId,\n frame,\n final: false,\n segmentId: requestId,\n });\n }\n }\n if (!this.queue.closed) {\n this.queue.close();\n }\n resolve();\n });\n },\n );\n\n req.on('error', (err) => {\n if (err.name === 'AbortError') return;\n this.#logger.error({ err }, 'Deepgram TTS request error');\n });\n\n req.on('close', () => resolve());\n req.write(JSON.stringify(json));\n req.end();\n });\n }\n}\n\nexport class SynthesizeStream extends tts.SynthesizeStream {\n private opts: TTSOptions;\n private tokenizer: tokenize.SentenceStream;\n label = 'deepgram.SynthesizeStream';\n\n private static readonly FLUSH_MSG = JSON.stringify({ type: 'Flush' });\n private static readonly CLOSE_MSG = JSON.stringify({ type: 'Close' });\n\n constructor(tts: TTS, opts: TTSOptions) {\n super(tts);\n this.opts = opts;\n this.tokenizer = opts.sentenceTokenizer.stream();\n }\n\n private async closeWebSocket(ws: WebSocket): Promise<void> {\n try {\n // Send Flush and Close messages to ensure Deepgram processes all remaining audio\n // and properly terminates the session, preventing lingering TTS sessions\n if (ws.readyState === WebSocket.OPEN) {\n ws.send(SynthesizeStream.FLUSH_MSG);\n ws.send(SynthesizeStream.CLOSE_MSG);\n\n // Wait for server acknowledgment to prevent race conditions and ensure\n // proper cleanup, avoiding 429 Too Many Requests errors from lingering sessions\n try {\n await new Promise<void>((resolve, _reject) => {\n const timeout = setTimeout(() => {\n resolve();\n }, 1000);\n\n ws.once('message', () => {\n clearTimeout(timeout);\n resolve();\n });\n\n ws.once('close', () => {\n clearTimeout(timeout);\n resolve();\n });\n\n ws.once('error', () => {\n clearTimeout(timeout);\n resolve();\n });\n });\n } catch (e) {\n // Ignore timeout or other errors during close sequence\n }\n }\n } catch (e) {\n console.warn(`Error during WebSocket close sequence: ${e}`);\n } finally {\n if (ws.readyState === WebSocket.OPEN || ws.readyState === WebSocket.CONNECTING) {\n ws.close();\n }\n }\n }\n\n protected async run() {\n const requestId = shortuuid();\n const segmentId = shortuuid();\n\n const wsUrl = this.opts.baseUrl!.replace(/^http/, 'ws');\n const url = new URL(`${wsUrl}/v1/speak`);\n url.searchParams.append('sample_rate', this.opts.sampleRate.toString());\n url.searchParams.append('model', this.opts.model);\n url.searchParams.append('encoding', this.opts.encoding);\n\n const ws = new WebSocket(url, {\n headers: {\n [AUTHORIZATION_HEADER]: `Token ${this.opts.apiKey!}`,\n },\n });\n\n await new Promise((resolve, reject) => {\n ws.on('open', resolve);\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n });\n\n const inputTask = async () => {\n for await (const data of this.input) {\n if (data === SynthesizeStream.FLUSH_SENTINEL) {\n this.tokenizer.flush();\n continue;\n }\n this.tokenizer.pushText(data);\n }\n this.tokenizer.endInput();\n this.tokenizer.close();\n };\n\n const sendTask = async () => {\n for await (const event of this.tokenizer) {\n if (this.abortController.signal.aborted) break;\n\n let text = event.token;\n if (!text.endsWith(' ')) {\n text += ' ';\n }\n\n const message = JSON.stringify({\n type: 'Speak',\n text: text,\n });\n\n ws.send(message);\n }\n\n if (!this.abortController.signal.aborted) {\n ws.send(SynthesizeStream.FLUSH_MSG);\n }\n };\n\n const recvTask = async () => {\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n let finalReceived = false;\n let timeout: NodeJS.Timeout | null = null;\n let lastFrame: AudioFrame | undefined;\n\n const sendLastFrame = (segmentId: string, final: boolean) => {\n if (lastFrame && !this.queue.closed) {\n this.queue.put({ requestId, segmentId, frame: lastFrame, final });\n lastFrame = undefined;\n }\n };\n\n const clearMessageTimeout = () => {\n if (timeout) {\n clearTimeout(timeout);\n timeout = null;\n }\n };\n\n return new Promise<void>((resolve, reject) => {\n ws.on('message', (data: RawData, isBinary: boolean) => {\n clearMessageTimeout();\n\n if (!isBinary) {\n const message = JSON.parse(data.toString());\n if (message.type === 'Flushed') {\n finalReceived = true;\n clearMessageTimeout();\n for (const frame of bstream.flush()) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n sendLastFrame(segmentId, true);\n\n if (!this.queue.closed) {\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n }\n resolve();\n }\n\n return;\n }\n\n const buffer =\n data instanceof Buffer\n ? data.buffer.slice(data.byteOffset, data.byteOffset + data.byteLength)\n : (data as ArrayBuffer);\n for (const frame of bstream.write(buffer as ArrayBuffer)) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n });\n\n ws.on('close', (_code, _reason) => {\n if (!finalReceived) {\n for (const frame of bstream.flush()) {\n sendLastFrame(segmentId, false);\n lastFrame = frame;\n }\n sendLastFrame(segmentId, true);\n\n if (!this.queue.closed) {\n this.queue.put(SynthesizeStream.END_OF_STREAM);\n }\n }\n resolve();\n });\n\n ws.on('error', (error) => {\n clearMessageTimeout();\n reject(error);\n });\n });\n };\n\n try {\n await Promise.all([inputTask(), sendTask(), recvTask()]);\n } catch (e) {\n throw new Error(`failed in main task: ${e}`);\n } finally {\n await this.closeWebSocket(ws);\n }\n }\n}\n"],"mappings":"AAGA;AAAA,EAEE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,eAAe;AACxB,SAAuB,iBAAiB;AAGxC,MAAM,uBAAuB;AAC7B,MAAM,eAAe;AACrB,MAAM,sBAAsB;AAY5B,MAAM,oBAAgC;AAAA,EACpC,OAAO;AAAA,EACP,UAAU;AAAA,EACV,YAAY;AAAA,EACZ,QAAQ,QAAQ,IAAI;AAAA,EACpB,SAAS;AAAA,EACT,cAAc;AAAA,IACZ,WAAW;AAAA,EACb;AAAA,EACA,mBAAmB,IAAI,SAAS,MAAM,kBAAkB;AAAA,IACtD,mBAAmB;AAAA,EACrB,CAAC;AACH;AAEO,MAAM,YAAY,IAAI,IAAI;AAAA,EACvB;AAAA,EACR,QAAQ;AAAA,EAER,IAAI,QAAgB;AAClB,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,YAAY,OAA4B,CAAC,GAAG;AAxD9C;AAyDI,UAAM,KAAK,cAAc,kBAAkB,YAAY,cAAc;AAAA,MACnE,aAAW,UAAK,iBAAL,mBAAmB,cAAa,kBAAkB,aAAa;AAAA,IAC5E,CAAC;AAED,SAAK,OAAO;AAAA,MACV,GAAG;AAAA,MACH,GAAG;AAAA,IACL;AAEA,QAAI,KAAK,KAAK,WAAW,QAAW;AAClC,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,WACE,MACA,aACA,aACmB;AACnB,WAAO,IAAI,cAAc,MAAM,MAAM,KAAK,MAAM,aAAa,WAAW;AAAA,EAC1E;AAAA,EAEA,SAA+B;AAC7B,WAAO,IAAI,iBAAiB,MAAM,KAAK,IAAI;AAAA,EAC7C;AACF;AAEO,MAAM,sBAAsB,IAAI,cAAc;AAAA,EACnD,QAAQ;AAAA,EACR,UAAU,IAAI;AAAA,EACN;AAAA,EACA;AAAA,EAER,YACEA,MACA,MACA,MACA,aACA,aACA;AACA,UAAM,MAAMA,MAAK,aAAa,WAAW;AACzC,SAAK,OAAO;AACZ,SAAK,OAAO;AAAA,EACd;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,YAAY,UAAU;AAC5B,UAAM,UAAU,IAAI,gBAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,UAAM,OAAO,EAAE,MAAM,KAAK,KAAK;AAC/B,UAAM,MAAM,IAAI,IAAI,GAAG,KAAK,KAAK,OAAQ,WAAW;AACpD,QAAI,aAAa,OAAO,eAAe,KAAK,KAAK,WAAW,SAAS,CAAC;AACtE,QAAI,aAAa,OAAO,SAAS,KAAK,KAAK,KAAK;AAChD,QAAI,aAAa,OAAO,YAAY,KAAK,KAAK,QAAQ;AAEtD,UAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,YAAM,MAAM;AAAA,QACV;AAAA,UACE,UAAU,IAAI;AAAA,UACd,MAAM;AAAA,UACN,MAAM,IAAI,WAAW,IAAI;AAAA,UACzB,QAAQ;AAAA,UACR,SAAS;AAAA,YACP,CAAC,oBAAoB,GAAG,SAAS,KAAK,KAAK,MAAO;AAAA,YAClD,gBAAgB;AAAA,UAClB;AAAA,UACA,QAAQ,KAAK;AAAA,QACf;AAAA,QACA,CAAC,QAAQ;AACP,cAAI,IAAI,eAAe,KAAK;AAC1B;AAAA,cACE,IAAI,MAAM,qCAAqC,IAAI,UAAU,IAAI,IAAI,aAAa,EAAE;AAAA,YACtF;AACA;AAAA,UACF;AAEA,cAAI,GAAG,QAAQ,CAAC,UAAU;AACxB,uBAAW,SAAS,QAAQ,MAAM,KAAK,GAAG;AACxC,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI;AAAA,kBACb;AAAA,kBACA;AAAA,kBACA,OAAO;AAAA,kBACP,WAAW;AAAA,gBACb,CAAC;AAAA,cACH;AAAA,YACF;AAAA,UACF,CAAC;AAED,cAAI,GAAG,SAAS,CAAC,QAAQ;AACvB,gBAAI,IAAI,YAAY,UAAW;AAC/B,iBAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,6BAA6B;AAAA,UAC3D,CAAC;AAED,cAAI,GAAG,SAAS,MAAM;AACpB,uBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI;AAAA,kBACb;AAAA,kBACA;AAAA,kBACA,OAAO;AAAA,kBACP,WAAW;AAAA,gBACb,CAAC;AAAA,cACH;AAAA,YACF;AACA,gBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,mBAAK,MAAM,MAAM;AAAA,YACnB;AACA,oBAAQ;AAAA,UACV,CAAC;AAAA,QACH;AAAA,MACF;AAEA,UAAI,GAAG,SAAS,CAAC,QAAQ;AACvB,YAAI,IAAI,SAAS,aAAc;AAC/B,aAAK,QAAQ,MAAM,EAAE,IAAI,GAAG,4BAA4B;AAAA,MAC1D,CAAC;AAED,UAAI,GAAG,SAAS,MAAM,QAAQ,CAAC;AAC/B,UAAI,MAAM,KAAK,UAAU,IAAI,CAAC;AAC9B,UAAI,IAAI;AAAA,IACV,CAAC;AAAA,EACH;AACF;AAEO,MAAM,yBAAyB,IAAI,iBAAiB;AAAA,EACjD;AAAA,EACA;AAAA,EACR,QAAQ;AAAA,EAER,OAAwB,YAAY,KAAK,UAAU,EAAE,MAAM,QAAQ,CAAC;AAAA,EACpE,OAAwB,YAAY,KAAK,UAAU,EAAE,MAAM,QAAQ,CAAC;AAAA,EAEpE,YAAYA,MAAU,MAAkB;AACtC,UAAMA,IAAG;AACT,SAAK,OAAO;AACZ,SAAK,YAAY,KAAK,kBAAkB,OAAO;AAAA,EACjD;AAAA,EAEA,MAAc,eAAe,IAA8B;AACzD,QAAI;AAGF,UAAI,GAAG,eAAe,UAAU,MAAM;AACpC,WAAG,KAAK,iBAAiB,SAAS;AAClC,WAAG,KAAK,iBAAiB,SAAS;AAIlC,YAAI;AACF,gBAAM,IAAI,QAAc,CAAC,SAAS,YAAY;AAC5C,kBAAM,UAAU,WAAW,MAAM;AAC/B,sBAAQ;AAAA,YACV,GAAG,GAAI;AAEP,eAAG,KAAK,WAAW,MAAM;AACvB,2BAAa,OAAO;AACpB,sBAAQ;AAAA,YACV,CAAC;AAED,eAAG,KAAK,SAAS,MAAM;AACrB,2BAAa,OAAO;AACpB,sBAAQ;AAAA,YACV,CAAC;AAED,eAAG,KAAK,SAAS,MAAM;AACrB,2BAAa,OAAO;AACpB,sBAAQ;AAAA,YACV,CAAC;AAAA,UACH,CAAC;AAAA,QACH,SAAS,GAAG;AAAA,QAEZ;AAAA,MACF;AAAA,IACF,SAAS,GAAG;AACV,cAAQ,KAAK,0CAA0C,CAAC,EAAE;AAAA,IAC5D,UAAE;AACA,UAAI,GAAG,eAAe,UAAU,QAAQ,GAAG,eAAe,UAAU,YAAY;AAC9E,WAAG,MAAM;AAAA,MACX;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,YAAY,UAAU;AAC5B,UAAM,YAAY,UAAU;AAE5B,UAAM,QAAQ,KAAK,KAAK,QAAS,QAAQ,SAAS,IAAI;AACtD,UAAM,MAAM,IAAI,IAAI,GAAG,KAAK,WAAW;AACvC,QAAI,aAAa,OAAO,eAAe,KAAK,KAAK,WAAW,SAAS,CAAC;AACtE,QAAI,aAAa,OAAO,SAAS,KAAK,KAAK,KAAK;AAChD,QAAI,aAAa,OAAO,YAAY,KAAK,KAAK,QAAQ;AAEtD,UAAM,KAAK,IAAI,UAAU,KAAK;AAAA,MAC5B,SAAS;AAAA,QACP,CAAC,oBAAoB,GAAG,SAAS,KAAK,KAAK,MAAO;AAAA,MACpD;AAAA,IACF,CAAC;AAED,UAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,SAAG,GAAG,QAAQ,OAAO;AACrB,SAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,SAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,IAC/D,CAAC;AAED,UAAM,YAAY,YAAY;AAC5B,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI,SAAS,iBAAiB,gBAAgB;AAC5C,eAAK,UAAU,MAAM;AACrB;AAAA,QACF;AACA,aAAK,UAAU,SAAS,IAAI;AAAA,MAC9B;AACA,WAAK,UAAU,SAAS;AACxB,WAAK,UAAU,MAAM;AAAA,IACvB;AAEA,UAAM,WAAW,YAAY;AAC3B,uBAAiB,SAAS,KAAK,WAAW;AACxC,YAAI,KAAK,gBAAgB,OAAO,QAAS;AAEzC,YAAI,OAAO,MAAM;AACjB,YAAI,CAAC,KAAK,SAAS,GAAG,GAAG;AACvB,kBAAQ;AAAA,QACV;AAEA,cAAM,UAAU,KAAK,UAAU;AAAA,UAC7B,MAAM;AAAA,UACN;AAAA,QACF,CAAC;AAED,WAAG,KAAK,OAAO;AAAA,MACjB;AAEA,UAAI,CAAC,KAAK,gBAAgB,OAAO,SAAS;AACxC,WAAG,KAAK,iBAAiB,SAAS;AAAA,MACpC;AAAA,IACF;AAEA,UAAM,WAAW,YAAY;AAC3B,YAAM,UAAU,IAAI,gBAAgB,KAAK,KAAK,YAAY,YAAY;AACtE,UAAI,gBAAgB;AACpB,UAAI,UAAiC;AACrC,UAAI;AAEJ,YAAM,gBAAgB,CAACC,YAAmB,UAAmB;AAC3D,YAAI,aAAa,CAAC,KAAK,MAAM,QAAQ;AACnC,eAAK,MAAM,IAAI,EAAE,WAAW,WAAAA,YAAW,OAAO,WAAW,MAAM,CAAC;AAChE,sBAAY;AAAA,QACd;AAAA,MACF;AAEA,YAAM,sBAAsB,MAAM;AAChC,YAAI,SAAS;AACX,uBAAa,OAAO;AACpB,oBAAU;AAAA,QACZ;AAAA,MACF;AAEA,aAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,WAAG,GAAG,WAAW,CAAC,MAAe,aAAsB;AACrD,8BAAoB;AAEpB,cAAI,CAAC,UAAU;AACb,kBAAM,UAAU,KAAK,MAAM,KAAK,SAAS,CAAC;AAC1C,gBAAI,QAAQ,SAAS,WAAW;AAC9B,8BAAgB;AAChB,kCAAoB;AACpB,yBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,8BAAc,WAAW,KAAK;AAC9B,4BAAY;AAAA,cACd;AACA,4BAAc,WAAW,IAAI;AAE7B,kBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,qBAAK,MAAM,IAAI,iBAAiB,aAAa;AAAA,cAC/C;AACA,sBAAQ;AAAA,YACV;AAEA;AAAA,UACF;AAEA,gBAAM,SACJ,gBAAgB,SACZ,KAAK,OAAO,MAAM,KAAK,YAAY,KAAK,aAAa,KAAK,UAAU,IACnE;AACP,qBAAW,SAAS,QAAQ,MAAM,MAAqB,GAAG;AACxD,0BAAc,WAAW,KAAK;AAC9B,wBAAY;AAAA,UACd;AAAA,QACF,CAAC;AAED,WAAG,GAAG,SAAS,CAAC,OAAO,YAAY;AACjC,cAAI,CAAC,eAAe;AAClB,uBAAW,SAAS,QAAQ,MAAM,GAAG;AACnC,4BAAc,WAAW,KAAK;AAC9B,0BAAY;AAAA,YACd;AACA,0BAAc,WAAW,IAAI;AAE7B,gBAAI,CAAC,KAAK,MAAM,QAAQ;AACtB,mBAAK,MAAM,IAAI,iBAAiB,aAAa;AAAA,YAC/C;AAAA,UACF;AACA,kBAAQ;AAAA,QACV,CAAC;AAED,WAAG,GAAG,SAAS,CAAC,UAAU;AACxB,8BAAoB;AACpB,iBAAO,KAAK;AAAA,QACd,CAAC;AAAA,MACH,CAAC;AAAA,IACH;AAEA,QAAI;AACF,YAAM,QAAQ,IAAI,CAAC,UAAU,GAAG,SAAS,GAAG,SAAS,CAAC,CAAC;AAAA,IACzD,SAAS,GAAG;AACV,YAAM,IAAI,MAAM,wBAAwB,CAAC,EAAE;AAAA,IAC7C,UAAE;AACA,YAAM,KAAK,eAAe,EAAE;AAAA,IAC9B;AAAA,EACF;AACF;","names":["tts","segmentId"]}
|