@livekit/agents-plugin-deepgram 1.6.3 → 1.7.0
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 -8
- package/dist/stt.cjs.map +1 -1
- package/dist/stt.d.ts.map +1 -1
- package/dist/stt.js +20 -8
- package/dist/stt.js.map +1 -1
- package/dist/stt_logging.test.cjs +69 -0
- package/dist/stt_logging.test.cjs.map +1 -0
- package/dist/stt_logging.test.d.cts +2 -0
- package/dist/stt_logging.test.d.ts +2 -0
- package/dist/stt_logging.test.d.ts.map +1 -0
- package/dist/stt_logging.test.js +68 -0
- package/dist/stt_logging.test.js.map +1 -0
- package/dist/stt_v2.cjs +2 -2
- package/dist/stt_v2.cjs.map +1 -1
- package/dist/stt_v2.js +2 -2
- package/dist/stt_v2.js.map +1 -1
- package/dist/tts.cjs +1 -1
- package/dist/tts.cjs.map +1 -1
- package/dist/tts.js +1 -1
- package/dist/tts.js.map +1 -1
- package/package.json +9 -9
- package/src/stt.ts +28 -5
- package/src/stt_logging.test.ts +78 -0
- package/src/stt_v2.ts +2 -2
- package/src/tts.ts +1 -1
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 APIConnectionError,\n APIError,\n APIStatusError,\n APITimeoutError,\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 bitRate?: number | null;\n speed?: number;\n mipOptOut: boolean;\n apiKey?: string;\n baseUrl?: string;\n sentenceTokenizer: tokenize.SentenceTokenizer;\n capabilities: tts.TTSCapabilities;\n}\n\nconst defaultTTSOptions: TTSOptions = {\n model: 'aura-2-andromeda-en',\n encoding: 'linear16',\n sampleRate: 24000,\n bitRate: null,\n mipOptOut: false,\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 override get sampleRate(): number {\n return this.opts.sampleRate;\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 if (this.opts.speed !== undefined && (this.opts.speed < 0.7 || this.opts.speed > 1.5)) {\n throw new Error(`Deepgram TTS speed must be between 0.7 and 1.5, got ${this.opts.speed}`);\n }\n }\n\n synthesize(\n text: string,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ): tts.ChunkedStream {\n return new ChunkedStream(this, text, this.opts, connOptions, abortSignal);\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): tts.SynthesizeStream {\n return new SynthesizeStream(this, this.opts, options?.connOptions);\n }\n\n updateOptions(\n opts: Partial<\n Pick<TTSOptions, 'model' | 'encoding' | 'sampleRate' | 'bitRate' | 'speed' | 'mipOptOut'>\n >,\n ) {\n this.opts = {\n ...this.opts,\n ...opts,\n };\n }\n}\n\nexport class ChunkedStream extends tts.ChunkedStream {\n label = 'deepgram.ChunkedStream';\n #logger = log();\n private opts: TTSOptions;\n private text: string;\n\n constructor(\n tts: TTS,\n text: string,\n opts: TTSOptions,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ) {\n super(text, tts, connOptions, abortSignal);\n this.text = text;\n this.opts = opts;\n }\n\n protected async run() {\n const requestId = shortuuid();\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n const json = { text: this.text };\n const url = new URL(`${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 url.searchParams.append('container', 'none');\n url.searchParams.append('mip_opt_out', String(this.opts.mipOptOut));\n if (this.opts.bitRate !== undefined && this.opts.bitRate !== null) {\n url.searchParams.append('bit_rate', this.opts.bitRate.toString());\n }\n if (this.opts.speed !== undefined) {\n url.searchParams.append('speed', this.opts.speed.toString());\n }\n\n await new Promise<void>((resolve, reject) => {\n let settled = false;\n const settle = (fn: () => void) => {\n if (!settled) {\n settled = true;\n fn();\n }\n };\n\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 settle(() =>\n reject(\n new Error(\n `Deepgram TTS HTTP request failed: ${res.statusCode} ${res.statusMessage}`,\n ),\n ),\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 settle(() => reject(err));\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 settle(() => 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 settle(() => reject(err));\n });\n\n req.on('close', () => settle(() => 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 #logger = log();\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, connOptions?: APIConnectOptions) {\n super(tts, connOptions);\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 url.searchParams.append('mip_opt_out', String(this.opts.mipOptOut));\n if (this.opts.bitRate !== undefined && this.opts.bitRate !== null) {\n url.searchParams.append('bit_rate', this.opts.bitRate.toString());\n }\n if (this.opts.speed !== undefined) {\n url.searchParams.append('speed', this.opts.speed.toString());\n }\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 let markInputSent: () => void = () => {};\n const inputSent = new Promise<void>((resolve) => {\n markInputSent = resolve;\n });\n\n const sendTask = async () => {\n try {\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 this.markStarted();\n ws.send(message);\n markInputSent();\n }\n\n if (!this.abortController.signal.aborted) {\n ws.send(SynthesizeStream.FLUSH_MSG);\n markInputSent();\n }\n } finally {\n markInputSent();\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 const resetMessageTimeout = (reject: (reason?: unknown) => void) => {\n clearMessageTimeout();\n timeout = setTimeout(() => {\n reject(new APITimeoutError({ message: 'Deepgram TTS recv idle timeout' }));\n }, this.connOptions.timeoutMs);\n };\n\n await inputSent;\n if (this.abortController.signal.aborted) return;\n\n return new Promise<void>((resolve, reject) => {\n resetMessageTimeout(reject);\n\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 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 return;\n } else if (message.type === 'Warning') {\n this.#logger.warn(`Deepgram warning: ${message.warn_msg}`);\n } else if (message.type === 'Error' || message.type === 'error') {\n reject(new APIError('Deepgram TTS returned error', { body: message }));\n return;\n } else if (message.type !== 'Metadata') {\n this.#logger.warn({ message }, 'Unknown Deepgram message type');\n }\n\n resetMessageTimeout(reject);\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 resetMessageTimeout(reject);\n });\n\n ws.on('close', (code, reason) => {\n clearMessageTimeout();\n if (!finalReceived) {\n reject(\n new APIStatusError({\n message: 'Deepgram websocket connection closed unexpectedly',\n options: {\n statusCode: code || -1,\n body: { reason: reason.toString() },\n },\n }),\n );\n return;\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 if (this.abortController.signal.aborted) return;\n if (e instanceof APIError) throw e;\n throw new APIConnectionError({\n message: `Deepgram TTS WebSocket failed: ${(e as Error).message ?? 'unknown error'}`,\n });\n } finally {\n await this.closeWebSocket(ws);\n }\n }\n}\n"],"mappings":"AAGA;AAAA,EAEE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;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;AAe5B,MAAM,oBAAgC;AAAA,EACpC,OAAO;AAAA,EACP,UAAU;AAAA,EACV,YAAY;AAAA,EACZ,SAAS;AAAA,EACT,WAAW;AAAA,EACX,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,IAAa,aAAqB;AAChC,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA,EAEA,YAAY,OAA4B,CAAC,GAAG;AArE9C;AAsEI,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;AAEA,QAAI,KAAK,KAAK,UAAU,WAAc,KAAK,KAAK,QAAQ,OAAO,KAAK,KAAK,QAAQ,MAAM;AACrF,YAAM,IAAI,MAAM,uDAAuD,KAAK,KAAK,KAAK,EAAE;AAAA,IAC1F;AAAA,EACF;AAAA,EAEA,WACE,MACA,aACA,aACmB;AACnB,WAAO,IAAI,cAAc,MAAM,MAAM,KAAK,MAAM,aAAa,WAAW;AAAA,EAC1E;AAAA,EAEA,OAAO,SAAqE;AAC1E,WAAO,IAAI,iBAAiB,MAAM,KAAK,MAAM,mCAAS,WAAW;AAAA,EACnE;AAAA,EAEA,cACE,MAGA;AACA,SAAK,OAAO;AAAA,MACV,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,IACL;AAAA,EACF;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;AACtD,QAAI,aAAa,OAAO,aAAa,MAAM;AAC3C,QAAI,aAAa,OAAO,eAAe,OAAO,KAAK,KAAK,SAAS,CAAC;AAClE,QAAI,KAAK,KAAK,YAAY,UAAa,KAAK,KAAK,YAAY,MAAM;AACjE,UAAI,aAAa,OAAO,YAAY,KAAK,KAAK,QAAQ,SAAS,CAAC;AAAA,IAClE;AACA,QAAI,KAAK,KAAK,UAAU,QAAW;AACjC,UAAI,aAAa,OAAO,SAAS,KAAK,KAAK,MAAM,SAAS,CAAC;AAAA,IAC7D;AAEA,UAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,UAAI,UAAU;AACd,YAAM,SAAS,CAAC,OAAmB;AACjC,YAAI,CAAC,SAAS;AACZ,oBAAU;AACV,aAAG;AAAA,QACL;AAAA,MACF;AAEA,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,cAAO,MACL;AAAA,gBACE,IAAI;AAAA,kBACF,qCAAqC,IAAI,UAAU,IAAI,IAAI,aAAa;AAAA,gBAC1E;AAAA,cACF;AAAA,YACF;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;AACzD,mBAAO,MAAM,OAAO,GAAG,CAAC;AAAA,UAC1B,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,mBAAO,MAAM,QAAQ,CAAC;AAAA,UACxB,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;AACxD,eAAO,MAAM,OAAO,GAAG,CAAC;AAAA,MAC1B,CAAC;AAED,UAAI,GAAG,SAAS,MAAM,OAAO,MAAM,QAAQ,CAAC,CAAC;AAC7C,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,UAAU,IAAI;AAAA,EACd,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,aAAiC;AACvE,UAAMA,MAAK,WAAW;AACtB,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;AACtD,QAAI,aAAa,OAAO,eAAe,OAAO,KAAK,KAAK,SAAS,CAAC;AAClE,QAAI,KAAK,KAAK,YAAY,UAAa,KAAK,KAAK,YAAY,MAAM;AACjE,UAAI,aAAa,OAAO,YAAY,KAAK,KAAK,QAAQ,SAAS,CAAC;AAAA,IAClE;AACA,QAAI,KAAK,KAAK,UAAU,QAAW;AACjC,UAAI,aAAa,OAAO,SAAS,KAAK,KAAK,MAAM,SAAS,CAAC;AAAA,IAC7D;AAEA,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,QAAI,gBAA4B,MAAM;AAAA,IAAC;AACvC,UAAM,YAAY,IAAI,QAAc,CAAC,YAAY;AAC/C,sBAAgB;AAAA,IAClB,CAAC;AAED,UAAM,WAAW,YAAY;AAC3B,UAAI;AACF,yBAAiB,SAAS,KAAK,WAAW;AACxC,cAAI,KAAK,gBAAgB,OAAO,QAAS;AAEzC,cAAI,OAAO,MAAM;AACjB,cAAI,CAAC,KAAK,SAAS,GAAG,GAAG;AACvB,oBAAQ;AAAA,UACV;AAEA,gBAAM,UAAU,KAAK,UAAU;AAAA,YAC7B,MAAM;AAAA,YACN;AAAA,UACF,CAAC;AAED,eAAK,YAAY;AACjB,aAAG,KAAK,OAAO;AACf,wBAAc;AAAA,QAChB;AAEA,YAAI,CAAC,KAAK,gBAAgB,OAAO,SAAS;AACxC,aAAG,KAAK,iBAAiB,SAAS;AAClC,wBAAc;AAAA,QAChB;AAAA,MACF,UAAE;AACA,sBAAc;AAAA,MAChB;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,YAAM,sBAAsB,CAAC,WAAuC;AAClE,4BAAoB;AACpB,kBAAU,WAAW,MAAM;AACzB,iBAAO,IAAI,gBAAgB,EAAE,SAAS,iCAAiC,CAAC,CAAC;AAAA,QAC3E,GAAG,KAAK,YAAY,SAAS;AAAA,MAC/B;AAEA,YAAM;AACN,UAAI,KAAK,gBAAgB,OAAO,QAAS;AAEzC,aAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,4BAAoB,MAAM;AAE1B,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,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;AACR;AAAA,YACF,WAAW,QAAQ,SAAS,WAAW;AACrC,mBAAK,QAAQ,KAAK,qBAAqB,QAAQ,QAAQ,EAAE;AAAA,YAC3D,WAAW,QAAQ,SAAS,WAAW,QAAQ,SAAS,SAAS;AAC/D,qBAAO,IAAI,SAAS,+BAA+B,EAAE,MAAM,QAAQ,CAAC,CAAC;AACrE;AAAA,YACF,WAAW,QAAQ,SAAS,YAAY;AACtC,mBAAK,QAAQ,KAAK,EAAE,QAAQ,GAAG,+BAA+B;AAAA,YAChE;AAEA,gCAAoB,MAAM;AAC1B;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;AACA,8BAAoB,MAAM;AAAA,QAC5B,CAAC;AAED,WAAG,GAAG,SAAS,CAAC,MAAM,WAAW;AAC/B,8BAAoB;AACpB,cAAI,CAAC,eAAe;AAClB;AAAA,cACE,IAAI,eAAe;AAAA,gBACjB,SAAS;AAAA,gBACT,SAAS;AAAA,kBACP,YAAY,QAAQ;AAAA,kBACpB,MAAM,EAAE,QAAQ,OAAO,SAAS,EAAE;AAAA,gBACpC;AAAA,cACF,CAAC;AAAA,YACH;AACA;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,UAAI,KAAK,gBAAgB,OAAO,QAAS;AACzC,UAAI,aAAa,SAAU,OAAM;AACjC,YAAM,IAAI,mBAAmB;AAAA,QAC3B,SAAS,kCAAmC,EAAY,WAAW,eAAe;AAAA,MACpF,CAAC;AAAA,IACH,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 APIConnectionError,\n APIError,\n APIStatusError,\n APITimeoutError,\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 bitRate?: number | null;\n speed?: number;\n mipOptOut: boolean;\n apiKey?: string;\n baseUrl?: string;\n sentenceTokenizer: tokenize.SentenceTokenizer;\n capabilities: tts.TTSCapabilities;\n}\n\nconst defaultTTSOptions: TTSOptions = {\n model: 'aura-2-andromeda-en',\n encoding: 'linear16',\n sampleRate: 24000,\n bitRate: null,\n mipOptOut: false,\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 override get sampleRate(): number {\n return this.opts.sampleRate;\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 if (this.opts.speed !== undefined && (this.opts.speed < 0.7 || this.opts.speed > 1.5)) {\n throw new Error(`Deepgram TTS speed must be between 0.7 and 1.5, got ${this.opts.speed}`);\n }\n }\n\n synthesize(\n text: string,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ): tts.ChunkedStream {\n return new ChunkedStream(this, text, this.opts, connOptions, abortSignal);\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): tts.SynthesizeStream {\n return new SynthesizeStream(this, this.opts, options?.connOptions);\n }\n\n updateOptions(\n opts: Partial<\n Pick<TTSOptions, 'model' | 'encoding' | 'sampleRate' | 'bitRate' | 'speed' | 'mipOptOut'>\n >,\n ) {\n this.opts = {\n ...this.opts,\n ...opts,\n };\n }\n}\n\nexport class ChunkedStream extends tts.ChunkedStream {\n label = 'deepgram.ChunkedStream';\n #logger = log();\n private opts: TTSOptions;\n private text: string;\n\n constructor(\n tts: TTS,\n text: string,\n opts: TTSOptions,\n connOptions?: APIConnectOptions,\n abortSignal?: AbortSignal,\n ) {\n super(text, tts, connOptions, abortSignal);\n this.text = text;\n this.opts = opts;\n }\n\n protected async run() {\n const requestId = shortuuid();\n const bstream = new AudioByteStream(this.opts.sampleRate, NUM_CHANNELS);\n const json = { text: this.text };\n const url = new URL(`${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 url.searchParams.append('container', 'none');\n url.searchParams.append('mip_opt_out', String(this.opts.mipOptOut));\n if (this.opts.bitRate !== undefined && this.opts.bitRate !== null) {\n url.searchParams.append('bit_rate', this.opts.bitRate.toString());\n }\n if (this.opts.speed !== undefined) {\n url.searchParams.append('speed', this.opts.speed.toString());\n }\n\n await new Promise<void>((resolve, reject) => {\n let settled = false;\n const settle = (fn: () => void) => {\n if (!settled) {\n settled = true;\n fn();\n }\n };\n\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 settle(() =>\n reject(\n new Error(\n `Deepgram TTS HTTP request failed: ${res.statusCode} ${res.statusMessage}`,\n ),\n ),\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 settle(() => reject(err));\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 settle(() => 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 settle(() => reject(err));\n });\n\n req.on('close', () => settle(() => 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 #logger = log();\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, connOptions?: APIConnectOptions) {\n super(tts, connOptions);\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 url.searchParams.append('mip_opt_out', String(this.opts.mipOptOut));\n if (this.opts.bitRate !== undefined && this.opts.bitRate !== null) {\n url.searchParams.append('bit_rate', this.opts.bitRate.toString());\n }\n if (this.opts.speed !== undefined) {\n url.searchParams.append('speed', this.opts.speed.toString());\n }\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 let markInputSent: () => void = () => {};\n const inputSent = new Promise<void>((resolve) => {\n markInputSent = resolve;\n });\n\n const sendTask = async () => {\n try {\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 this.markStarted();\n ws.send(message);\n markInputSent();\n }\n\n if (!this.abortController.signal.aborted) {\n ws.send(SynthesizeStream.FLUSH_MSG);\n markInputSent();\n }\n } finally {\n markInputSent();\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 const resetMessageTimeout = (reject: (reason?: unknown) => void) => {\n clearMessageTimeout();\n timeout = setTimeout(() => {\n reject(new APITimeoutError({ message: 'Deepgram TTS recv idle timeout' }));\n }, this.connOptions.timeoutMs);\n };\n\n await inputSent;\n if (this.abortController.signal.aborted) return;\n\n return new Promise<void>((resolve, reject) => {\n resetMessageTimeout(reject);\n\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 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 return;\n } else if (message.type === 'Warning') {\n this.#logger.warn(`Deepgram warning: ${message.warn_msg}`);\n } else if (message.type === 'Error' || message.type === 'error') {\n reject(new APIError('Deepgram TTS returned error', { body: message }));\n return;\n } else if (message.type !== 'Metadata') {\n this.#logger.warn({ 'lk.pii.message': message }, 'Unknown Deepgram message type');\n }\n\n resetMessageTimeout(reject);\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 resetMessageTimeout(reject);\n });\n\n ws.on('close', (code, reason) => {\n clearMessageTimeout();\n if (!finalReceived) {\n reject(\n new APIStatusError({\n message: 'Deepgram websocket connection closed unexpectedly',\n options: {\n statusCode: code || -1,\n body: { reason: reason.toString() },\n },\n }),\n );\n return;\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 if (this.abortController.signal.aborted) return;\n if (e instanceof APIError) throw e;\n throw new APIConnectionError({\n message: `Deepgram TTS WebSocket failed: ${(e as Error).message ?? 'unknown error'}`,\n });\n } finally {\n await this.closeWebSocket(ws);\n }\n }\n}\n"],"mappings":"AAGA;AAAA,EAEE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;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;AAe5B,MAAM,oBAAgC;AAAA,EACpC,OAAO;AAAA,EACP,UAAU;AAAA,EACV,YAAY;AAAA,EACZ,SAAS;AAAA,EACT,WAAW;AAAA,EACX,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,IAAa,aAAqB;AAChC,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA,EAEA,YAAY,OAA4B,CAAC,GAAG;AArE9C;AAsEI,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;AAEA,QAAI,KAAK,KAAK,UAAU,WAAc,KAAK,KAAK,QAAQ,OAAO,KAAK,KAAK,QAAQ,MAAM;AACrF,YAAM,IAAI,MAAM,uDAAuD,KAAK,KAAK,KAAK,EAAE;AAAA,IAC1F;AAAA,EACF;AAAA,EAEA,WACE,MACA,aACA,aACmB;AACnB,WAAO,IAAI,cAAc,MAAM,MAAM,KAAK,MAAM,aAAa,WAAW;AAAA,EAC1E;AAAA,EAEA,OAAO,SAAqE;AAC1E,WAAO,IAAI,iBAAiB,MAAM,KAAK,MAAM,mCAAS,WAAW;AAAA,EACnE;AAAA,EAEA,cACE,MAGA;AACA,SAAK,OAAO;AAAA,MACV,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,IACL;AAAA,EACF;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;AACtD,QAAI,aAAa,OAAO,aAAa,MAAM;AAC3C,QAAI,aAAa,OAAO,eAAe,OAAO,KAAK,KAAK,SAAS,CAAC;AAClE,QAAI,KAAK,KAAK,YAAY,UAAa,KAAK,KAAK,YAAY,MAAM;AACjE,UAAI,aAAa,OAAO,YAAY,KAAK,KAAK,QAAQ,SAAS,CAAC;AAAA,IAClE;AACA,QAAI,KAAK,KAAK,UAAU,QAAW;AACjC,UAAI,aAAa,OAAO,SAAS,KAAK,KAAK,MAAM,SAAS,CAAC;AAAA,IAC7D;AAEA,UAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,UAAI,UAAU;AACd,YAAM,SAAS,CAAC,OAAmB;AACjC,YAAI,CAAC,SAAS;AACZ,oBAAU;AACV,aAAG;AAAA,QACL;AAAA,MACF;AAEA,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,cAAO,MACL;AAAA,gBACE,IAAI;AAAA,kBACF,qCAAqC,IAAI,UAAU,IAAI,IAAI,aAAa;AAAA,gBAC1E;AAAA,cACF;AAAA,YACF;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;AACzD,mBAAO,MAAM,OAAO,GAAG,CAAC;AAAA,UAC1B,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,mBAAO,MAAM,QAAQ,CAAC;AAAA,UACxB,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;AACxD,eAAO,MAAM,OAAO,GAAG,CAAC;AAAA,MAC1B,CAAC;AAED,UAAI,GAAG,SAAS,MAAM,OAAO,MAAM,QAAQ,CAAC,CAAC;AAC7C,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,UAAU,IAAI;AAAA,EACd,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,aAAiC;AACvE,UAAMA,MAAK,WAAW;AACtB,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;AACtD,QAAI,aAAa,OAAO,eAAe,OAAO,KAAK,KAAK,SAAS,CAAC;AAClE,QAAI,KAAK,KAAK,YAAY,UAAa,KAAK,KAAK,YAAY,MAAM;AACjE,UAAI,aAAa,OAAO,YAAY,KAAK,KAAK,QAAQ,SAAS,CAAC;AAAA,IAClE;AACA,QAAI,KAAK,KAAK,UAAU,QAAW;AACjC,UAAI,aAAa,OAAO,SAAS,KAAK,KAAK,MAAM,SAAS,CAAC;AAAA,IAC7D;AAEA,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,QAAI,gBAA4B,MAAM;AAAA,IAAC;AACvC,UAAM,YAAY,IAAI,QAAc,CAAC,YAAY;AAC/C,sBAAgB;AAAA,IAClB,CAAC;AAED,UAAM,WAAW,YAAY;AAC3B,UAAI;AACF,yBAAiB,SAAS,KAAK,WAAW;AACxC,cAAI,KAAK,gBAAgB,OAAO,QAAS;AAEzC,cAAI,OAAO,MAAM;AACjB,cAAI,CAAC,KAAK,SAAS,GAAG,GAAG;AACvB,oBAAQ;AAAA,UACV;AAEA,gBAAM,UAAU,KAAK,UAAU;AAAA,YAC7B,MAAM;AAAA,YACN;AAAA,UACF,CAAC;AAED,eAAK,YAAY;AACjB,aAAG,KAAK,OAAO;AACf,wBAAc;AAAA,QAChB;AAEA,YAAI,CAAC,KAAK,gBAAgB,OAAO,SAAS;AACxC,aAAG,KAAK,iBAAiB,SAAS;AAClC,wBAAc;AAAA,QAChB;AAAA,MACF,UAAE;AACA,sBAAc;AAAA,MAChB;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,YAAM,sBAAsB,CAAC,WAAuC;AAClE,4BAAoB;AACpB,kBAAU,WAAW,MAAM;AACzB,iBAAO,IAAI,gBAAgB,EAAE,SAAS,iCAAiC,CAAC,CAAC;AAAA,QAC3E,GAAG,KAAK,YAAY,SAAS;AAAA,MAC/B;AAEA,YAAM;AACN,UAAI,KAAK,gBAAgB,OAAO,QAAS;AAEzC,aAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,4BAAoB,MAAM;AAE1B,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,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;AACR;AAAA,YACF,WAAW,QAAQ,SAAS,WAAW;AACrC,mBAAK,QAAQ,KAAK,qBAAqB,QAAQ,QAAQ,EAAE;AAAA,YAC3D,WAAW,QAAQ,SAAS,WAAW,QAAQ,SAAS,SAAS;AAC/D,qBAAO,IAAI,SAAS,+BAA+B,EAAE,MAAM,QAAQ,CAAC,CAAC;AACrE;AAAA,YACF,WAAW,QAAQ,SAAS,YAAY;AACtC,mBAAK,QAAQ,KAAK,EAAE,kBAAkB,QAAQ,GAAG,+BAA+B;AAAA,YAClF;AAEA,gCAAoB,MAAM;AAC1B;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;AACA,8BAAoB,MAAM;AAAA,QAC5B,CAAC;AAED,WAAG,GAAG,SAAS,CAAC,MAAM,WAAW;AAC/B,8BAAoB;AACpB,cAAI,CAAC,eAAe;AAClB;AAAA,cACE,IAAI,eAAe;AAAA,gBACjB,SAAS;AAAA,gBACT,SAAS;AAAA,kBACP,YAAY,QAAQ;AAAA,kBACpB,MAAM,EAAE,QAAQ,OAAO,SAAS,EAAE;AAAA,gBACpC;AAAA,cACF,CAAC;AAAA,YACH;AACA;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,UAAI,KAAK,gBAAgB,OAAO,QAAS;AACzC,UAAI,aAAa,SAAU,OAAM;AACjC,YAAM,IAAI,mBAAmB;AAAA,QAC3B,SAAS,kCAAmC,EAAY,WAAW,eAAe;AAAA,MACpF,CAAC;AAAA,IACH,UAAE;AACA,YAAM,KAAK,eAAe,EAAE;AAAA,IAC9B;AAAA,EACF;AACF;","names":["tts","segmentId"]}
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@livekit/agents-plugin-deepgram",
|
|
3
|
-
"version": "1.
|
|
3
|
+
"version": "1.7.0",
|
|
4
4
|
"description": "Deepgram plugin for LiveKit Agents for Node.js",
|
|
5
5
|
"main": "dist/index.js",
|
|
6
6
|
"require": "dist/index.cjs",
|
|
@@ -28,22 +28,22 @@
|
|
|
28
28
|
"README.md"
|
|
29
29
|
],
|
|
30
30
|
"devDependencies": {
|
|
31
|
-
"@livekit/rtc-node": "^0.13.
|
|
32
|
-
"@microsoft/api-extractor": "^7.
|
|
31
|
+
"@livekit/rtc-node": "^0.13.34",
|
|
32
|
+
"@microsoft/api-extractor": "^7.58.12",
|
|
33
33
|
"@types/ws": "^8.5.10",
|
|
34
34
|
"tsup": "^8.3.5",
|
|
35
35
|
"typescript": "^5.0.0",
|
|
36
|
-
"@livekit/agents": "1.
|
|
37
|
-
"@livekit/agents-plugin-openai": "^1.
|
|
38
|
-
"@livekit/agents-plugin-silero": "1.
|
|
39
|
-
"@livekit/agents-plugins-test": "1.
|
|
36
|
+
"@livekit/agents": "1.7.0",
|
|
37
|
+
"@livekit/agents-plugin-openai": "^1.7.0",
|
|
38
|
+
"@livekit/agents-plugin-silero": "1.7.0",
|
|
39
|
+
"@livekit/agents-plugins-test": "1.7.0"
|
|
40
40
|
},
|
|
41
41
|
"dependencies": {
|
|
42
42
|
"ws": "^8.21.0"
|
|
43
43
|
},
|
|
44
44
|
"peerDependencies": {
|
|
45
|
-
"@livekit/rtc-node": "^0.13.
|
|
46
|
-
"@livekit/agents": "1.
|
|
45
|
+
"@livekit/rtc-node": "^0.13.34",
|
|
46
|
+
"@livekit/agents": "1.7.0"
|
|
47
47
|
},
|
|
48
48
|
"scripts": {
|
|
49
49
|
"build": "tsup --onSuccess \"pnpm build:types\"",
|
package/src/stt.ts
CHANGED
|
@@ -166,7 +166,11 @@ export class STT extends stt.STT {
|
|
|
166
166
|
);
|
|
167
167
|
}
|
|
168
168
|
|
|
169
|
-
return prerecordedTranscriptionToSpeechEvent(
|
|
169
|
+
return prerecordedTranscriptionToSpeechEvent(
|
|
170
|
+
this.#opts.language,
|
|
171
|
+
await resp.json(),
|
|
172
|
+
this.#opts.punctuate || this.#opts.smartFormat,
|
|
173
|
+
);
|
|
170
174
|
}
|
|
171
175
|
|
|
172
176
|
updateOptions(opts: Partial<STTOptions>) {
|
|
@@ -497,6 +501,7 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
497
501
|
this.#opts.language!,
|
|
498
502
|
json,
|
|
499
503
|
this.startTimeOffset,
|
|
504
|
+
this.#opts.punctuate || this.#opts.smartFormat,
|
|
500
505
|
);
|
|
501
506
|
|
|
502
507
|
// If, for some reason, we didn't get a SpeechStarted event but we got
|
|
@@ -546,7 +551,9 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
546
551
|
break;
|
|
547
552
|
}
|
|
548
553
|
default: {
|
|
549
|
-
this.#logger
|
|
554
|
+
this.#logger
|
|
555
|
+
.child({ 'lk.pii.message': json })
|
|
556
|
+
.warn('received unexpected message from Deepgram');
|
|
550
557
|
break;
|
|
551
558
|
}
|
|
552
559
|
}
|
|
@@ -555,7 +562,10 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
555
562
|
resolve();
|
|
556
563
|
}
|
|
557
564
|
} catch (err) {
|
|
558
|
-
this.#logger.error(
|
|
565
|
+
this.#logger.error(
|
|
566
|
+
{ error: err, 'lk.pii.message': msg.toString() },
|
|
567
|
+
'Deepgram STT failed to process message',
|
|
568
|
+
);
|
|
559
569
|
reject(err);
|
|
560
570
|
}
|
|
561
571
|
});
|
|
@@ -625,10 +635,22 @@ function createWav(frame: AudioFrame): Buffer {
|
|
|
625
635
|
return Buffer.concat([header, pcm]);
|
|
626
636
|
}
|
|
627
637
|
|
|
638
|
+
/**
|
|
639
|
+
* Return the word form matching the transcript requested from Deepgram.
|
|
640
|
+
*
|
|
641
|
+
* Both punctuate and smart_format produce a formatted transcript. When either is enabled,
|
|
642
|
+
* Deepgram provides punctuated_word alongside word. Fall back to word when it is absent or empty.
|
|
643
|
+
*/
|
|
644
|
+
function wordText(word: { [id: string]: any }, usePunctuatedWord: boolean): string {
|
|
645
|
+
const raw = word['word'] ?? '';
|
|
646
|
+
return usePunctuatedWord ? word['punctuated_word'] || raw : raw;
|
|
647
|
+
}
|
|
648
|
+
|
|
628
649
|
const liveTranscriptionToSpeechData = (
|
|
629
650
|
language: STTLanguages | string,
|
|
630
651
|
data: { [id: string]: any },
|
|
631
652
|
startTimeOffset: number = 0,
|
|
653
|
+
usePunctuatedWord: boolean = true,
|
|
632
654
|
): stt.SpeechData[] => {
|
|
633
655
|
const alts: any[] = data['channel']['alternatives'];
|
|
634
656
|
|
|
@@ -647,7 +669,7 @@ const liveTranscriptionToSpeechData = (
|
|
|
647
669
|
text: alt['transcript'],
|
|
648
670
|
words: wordsData.map((word) =>
|
|
649
671
|
createTimedString({
|
|
650
|
-
text: word
|
|
672
|
+
text: wordText(word, usePunctuatedWord),
|
|
651
673
|
startTime: (word['start'] ?? 0) + startTimeOffset,
|
|
652
674
|
endTime: (word['end'] ?? 0) + startTimeOffset,
|
|
653
675
|
confidence: word['confidence'] ?? 0.0,
|
|
@@ -661,6 +683,7 @@ const liveTranscriptionToSpeechData = (
|
|
|
661
683
|
function prerecordedTranscriptionToSpeechEvent(
|
|
662
684
|
language: STTLanguages | string | undefined,
|
|
663
685
|
data: { [id: string]: any },
|
|
686
|
+
usePunctuatedWord: boolean = true,
|
|
664
687
|
): stt.SpeechEvent {
|
|
665
688
|
const channel = data['results']['channels'][0];
|
|
666
689
|
const detectedLanguage = channel['detected_language'] ?? '';
|
|
@@ -675,7 +698,7 @@ function prerecordedTranscriptionToSpeechEvent(
|
|
|
675
698
|
text: alt['transcript'],
|
|
676
699
|
words: wordsData.map((word) =>
|
|
677
700
|
createTimedString({
|
|
678
|
-
text: word
|
|
701
|
+
text: wordText(word, usePunctuatedWord),
|
|
679
702
|
startTime: word['start'] ?? 0,
|
|
680
703
|
endTime: word['end'] ?? 0,
|
|
681
704
|
confidence: word['confidence'] ?? 0.0,
|
|
@@ -0,0 +1,78 @@
|
|
|
1
|
+
// SPDX-FileCopyrightText: 2026 LiveKit, Inc.
|
|
2
|
+
//
|
|
3
|
+
// SPDX-License-Identifier: Apache-2.0
|
|
4
|
+
import { log } from '@livekit/agents';
|
|
5
|
+
import { once } from 'node:events';
|
|
6
|
+
import { afterEach, describe, expect, it, vi } from 'vitest';
|
|
7
|
+
import { type WebSocket, WebSocketServer } from 'ws';
|
|
8
|
+
import { STT } from './stt.js';
|
|
9
|
+
import { STTv2 } from './stt_v2.js';
|
|
10
|
+
|
|
11
|
+
const servers: WebSocketServer[] = [];
|
|
12
|
+
const streams: Array<{ close(): void }> = [];
|
|
13
|
+
|
|
14
|
+
async function startServer(onConnection?: (socket: WebSocket) => void): Promise<string> {
|
|
15
|
+
const server = new WebSocketServer({ host: '127.0.0.1', port: 0 });
|
|
16
|
+
servers.push(server);
|
|
17
|
+
await once(server, 'listening');
|
|
18
|
+
if (onConnection) server.on('connection', onConnection);
|
|
19
|
+
|
|
20
|
+
const address = server.address();
|
|
21
|
+
if (address === null || typeof address === 'string') {
|
|
22
|
+
throw new Error('expected a TCP address');
|
|
23
|
+
}
|
|
24
|
+
return `ws://127.0.0.1:${address.port}`;
|
|
25
|
+
}
|
|
26
|
+
|
|
27
|
+
afterEach(async () => {
|
|
28
|
+
for (const stream of streams.splice(0)) stream.close();
|
|
29
|
+
await Promise.all(
|
|
30
|
+
servers.splice(0).map(async (server) => {
|
|
31
|
+
for (const client of server.clients) client.terminate();
|
|
32
|
+
await new Promise<void>((resolve) => server.close(() => resolve()));
|
|
33
|
+
}),
|
|
34
|
+
);
|
|
35
|
+
vi.restoreAllMocks();
|
|
36
|
+
});
|
|
37
|
+
|
|
38
|
+
describe('Deepgram STT logging', () => {
|
|
39
|
+
it('logs invalid messages as text with the parse error', async () => {
|
|
40
|
+
const errorSpy = vi.spyOn(log(), 'error');
|
|
41
|
+
const baseUrl = await startServer((socket) => {
|
|
42
|
+
setTimeout(() => socket.send('not-json'), 10);
|
|
43
|
+
});
|
|
44
|
+
const recognizer = new STT({ apiKey: 'test', baseUrl });
|
|
45
|
+
const stream = recognizer.stream();
|
|
46
|
+
streams.push(stream);
|
|
47
|
+
|
|
48
|
+
await vi.waitFor(() => {
|
|
49
|
+
expect(errorSpy).toHaveBeenCalledWith(
|
|
50
|
+
{
|
|
51
|
+
error: expect.any(SyntaxError),
|
|
52
|
+
'lk.pii.message': 'not-json',
|
|
53
|
+
},
|
|
54
|
+
'Deepgram STT failed to process message',
|
|
55
|
+
);
|
|
56
|
+
});
|
|
57
|
+
});
|
|
58
|
+
|
|
59
|
+
it('does not log the v2 connection URL', async () => {
|
|
60
|
+
const debugSpy = vi.spyOn(log(), 'debug');
|
|
61
|
+
const baseUrl = await startServer();
|
|
62
|
+
const recognizer = new STTv2({
|
|
63
|
+
apiKey: 'test',
|
|
64
|
+
endpointUrl: `${baseUrl}/v2/listen`,
|
|
65
|
+
keyterms: ['private-keyterm'],
|
|
66
|
+
});
|
|
67
|
+
const stream = recognizer.stream();
|
|
68
|
+
streams.push(stream);
|
|
69
|
+
|
|
70
|
+
await vi.waitFor(() => {
|
|
71
|
+
const connectionCalls = debugSpy.mock.calls.filter((call) =>
|
|
72
|
+
call.includes('connecting to Deepgram'),
|
|
73
|
+
);
|
|
74
|
+
expect(connectionCalls).toEqual([['connecting to Deepgram']]);
|
|
75
|
+
expect(JSON.stringify(connectionCalls)).not.toContain('private-keyterm');
|
|
76
|
+
});
|
|
77
|
+
});
|
|
78
|
+
});
|
package/src/stt_v2.ts
CHANGED
|
@@ -314,7 +314,7 @@ class SpeechStreamv2 extends stt.SpeechStream {
|
|
|
314
314
|
this.#reconnectEvent.clear();
|
|
315
315
|
|
|
316
316
|
const url = this.#getDeepgramUrl();
|
|
317
|
-
this.#logger.debug(
|
|
317
|
+
this.#logger.debug('connecting to Deepgram');
|
|
318
318
|
|
|
319
319
|
this.#ws = new WebSocket(url, {
|
|
320
320
|
headers: { Authorization: `Token ${this.#opts.apiKey}` },
|
|
@@ -506,7 +506,7 @@ class SpeechStreamv2 extends stt.SpeechStream {
|
|
|
506
506
|
this.#onEndOfSpeech();
|
|
507
507
|
}
|
|
508
508
|
} else if (data.type === 'Error') {
|
|
509
|
-
this.#logger.warn('deepgram sent an error'
|
|
509
|
+
this.#logger.warn({ 'lk.pii.data': data }, 'deepgram sent an error');
|
|
510
510
|
const desc = (data.description as string) || 'unknown error from deepgram';
|
|
511
511
|
throw new Error(`Deepgram API Error: ${desc}`);
|
|
512
512
|
}
|
package/src/tts.ts
CHANGED
|
@@ -422,7 +422,7 @@ export class SynthesizeStream extends tts.SynthesizeStream {
|
|
|
422
422
|
reject(new APIError('Deepgram TTS returned error', { body: message }));
|
|
423
423
|
return;
|
|
424
424
|
} else if (message.type !== 'Metadata') {
|
|
425
|
-
this.#logger.warn({ message }, 'Unknown Deepgram message type');
|
|
425
|
+
this.#logger.warn({ 'lk.pii.message': message }, 'Unknown Deepgram message type');
|
|
426
426
|
}
|
|
427
427
|
|
|
428
428
|
resetMessageTimeout(reject);
|