@livekit/agents-plugin-assemblyai 1.5.0 → 1.5.2
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/index.cjs +1 -1
- package/dist/index.js +1 -1
- package/dist/stt.cjs +44 -4
- package/dist/stt.cjs.map +1 -1
- package/dist/stt.d.cts +10 -1
- package/dist/stt.d.ts +10 -1
- package/dist/stt.d.ts.map +1 -1
- package/dist/stt.js +45 -4
- package/dist/stt.js.map +1 -1
- package/package.json +7 -7
- package/src/stt.ts +60 -3
package/dist/index.cjs
CHANGED
package/dist/index.js
CHANGED
package/dist/stt.cjs
CHANGED
|
@@ -47,6 +47,9 @@ const defaultSTTOptions = {
|
|
|
47
47
|
class STT extends import_agents.stt.STT {
|
|
48
48
|
#opts;
|
|
49
49
|
#streams = /* @__PURE__ */ new Set();
|
|
50
|
+
// set (user + session))
|
|
51
|
+
#userKeyterms;
|
|
52
|
+
#sessionKeyterms = [];
|
|
50
53
|
label = "assemblyai.STT";
|
|
51
54
|
get model() {
|
|
52
55
|
return this.#opts.speechModel;
|
|
@@ -55,10 +58,19 @@ class STT extends import_agents.stt.STT {
|
|
|
55
58
|
return "AssemblyAI";
|
|
56
59
|
}
|
|
57
60
|
constructor(opts = {}) {
|
|
61
|
+
const rawModel = opts.speechModel ?? defaultSTTOptions.speechModel;
|
|
62
|
+
const supportsCarryover = isU3ProModel(rawModel) || rawModel === "u3-pro";
|
|
63
|
+
if (opts.agentContextCarryover && !supportsCarryover) {
|
|
64
|
+
(0, import_agents.log)().warn(
|
|
65
|
+
`agentContextCarryover is enabled but model '${rawModel}' does not support it; ignoring`
|
|
66
|
+
);
|
|
67
|
+
}
|
|
58
68
|
super({
|
|
59
69
|
streaming: true,
|
|
60
70
|
interimResults: true,
|
|
61
|
-
alignedTranscript: "word"
|
|
71
|
+
alignedTranscript: "word",
|
|
72
|
+
keyterms: true,
|
|
73
|
+
chatContext: (opts.agentContextCarryover ?? false) && supportsCarryover
|
|
62
74
|
});
|
|
63
75
|
if (opts.speechModel === "u3-pro") {
|
|
64
76
|
(0, import_agents.log)().warn("'u3-pro' is deprecated, use 'universal-3-5-pro' instead.");
|
|
@@ -94,22 +106,50 @@ class STT extends import_agents.stt.STT {
|
|
|
94
106
|
apiKey,
|
|
95
107
|
minTurnSilence
|
|
96
108
|
};
|
|
109
|
+
this.#userKeyterms = [...this.#opts.keytermsPrompt ?? []];
|
|
97
110
|
}
|
|
98
111
|
// eslint-disable-next-line @typescript-eslint/no-unused-vars
|
|
99
112
|
async _recognize(_) {
|
|
100
113
|
throw new Error("Non-streaming recognize is not supported on AssemblyAI STT");
|
|
101
114
|
}
|
|
102
115
|
updateOptions(opts) {
|
|
103
|
-
|
|
116
|
+
const nextOpts = { ...opts };
|
|
117
|
+
if (nextOpts.keytermsPrompt !== void 0) {
|
|
118
|
+
this.#userKeyterms = [...nextOpts.keytermsPrompt];
|
|
119
|
+
nextOpts.keytermsPrompt = [.../* @__PURE__ */ new Set([...this.#userKeyterms, ...this.#sessionKeyterms])];
|
|
120
|
+
}
|
|
121
|
+
this.#opts = { ...this.#opts, ...nextOpts };
|
|
122
|
+
for (const ref of this.#streams) {
|
|
123
|
+
const stream = ref.deref();
|
|
124
|
+
if (stream) {
|
|
125
|
+
stream.updateOptions(nextOpts);
|
|
126
|
+
} else {
|
|
127
|
+
this.#streams.delete(ref);
|
|
128
|
+
}
|
|
129
|
+
}
|
|
130
|
+
}
|
|
131
|
+
_updateSessionKeyterms(keyterms) {
|
|
132
|
+
if (keyterms.length === this.#sessionKeyterms.length && keyterms.every((t, i) => t === this.#sessionKeyterms[i])) {
|
|
133
|
+
return;
|
|
134
|
+
}
|
|
135
|
+
this.#sessionKeyterms = [...keyterms];
|
|
136
|
+
const merged = [.../* @__PURE__ */ new Set([...this.#userKeyterms, ...keyterms])];
|
|
137
|
+
this.#opts.keytermsPrompt = merged;
|
|
104
138
|
for (const ref of this.#streams) {
|
|
105
139
|
const stream = ref.deref();
|
|
106
140
|
if (stream) {
|
|
107
|
-
stream.updateOptions(
|
|
141
|
+
stream.updateOptions({ keytermsPrompt: merged });
|
|
108
142
|
} else {
|
|
109
143
|
this.#streams.delete(ref);
|
|
110
144
|
}
|
|
111
145
|
}
|
|
112
146
|
}
|
|
147
|
+
_pushConversationItem(ev) {
|
|
148
|
+
const chatItem = ev.item;
|
|
149
|
+
if (chatItem instanceof import_agents.ChatMessage && chatItem.role === "assistant" && chatItem.textContent) {
|
|
150
|
+
this.updateOptions({ agentContext: chatItem.textContent });
|
|
151
|
+
}
|
|
152
|
+
}
|
|
113
153
|
stream(options) {
|
|
114
154
|
const stream = new SpeechStream(this, this.#opts, options == null ? void 0 : options.connOptions);
|
|
115
155
|
this.#streams.add(new WeakRef(stream));
|
|
@@ -217,7 +257,7 @@ class SpeechStream extends import_agents.stt.SpeechStream {
|
|
|
217
257
|
end_of_turn_confidence_threshold: this.#opts.endOfTurnConfidenceThreshold,
|
|
218
258
|
min_turn_silence: minSilence,
|
|
219
259
|
max_turn_silence: maxSilence,
|
|
220
|
-
keyterms_prompt: this.#opts.keytermsPrompt !== void 0 ? JSON.stringify(this.#opts.keytermsPrompt) : void 0,
|
|
260
|
+
keyterms_prompt: this.#opts.keytermsPrompt !== void 0 && this.#opts.keytermsPrompt.length > 0 ? JSON.stringify(this.#opts.keytermsPrompt) : void 0,
|
|
221
261
|
language_detection: languageDetection,
|
|
222
262
|
inactivity_timeout: this.#opts.inactivityTimeout,
|
|
223
263
|
prompt: this.#opts.prompt,
|
package/dist/stt.cjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2026 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n//\nimport {\n type APIConnectOptions,\n type AudioBuffer,\n AudioByteStream,\n Future,\n Task,\n createTimedString,\n delay,\n log,\n normalizeLanguage,\n stt,\n waitForAbort,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { RawData } from 'ws';\nimport { WebSocket } from 'ws';\nimport type { STTEncoding, STTModels, VoiceFocus } from './models.js';\n\n// Speech models in the Universal-3 Pro family, which share the same parameter support.\nconst U3_PRO_MODELS = ['u3-rt-pro', 'u3-rt-pro-beta-1', 'universal-3-5-pro'] as const;\n\nfunction isU3ProModel(model: STTModels): boolean {\n return U3_PRO_MODELS.includes(model as (typeof U3_PRO_MODELS)[number]);\n}\n\n// AssemblyAI Universal-Streaming (v3) message envelope. All fields are optional\n// since we narrow on `type` before reading anything else.\ninterface StreamEventMessage {\n type?: 'Begin' | 'SpeechStarted' | 'Turn' | 'Termination' | string;\n // Begin\n id?: string;\n expires_at?: number;\n // Turn\n transcript?: string;\n utterance?: string;\n end_of_turn?: boolean;\n end_of_turn_confidence?: number;\n turn_is_formatted?: boolean;\n language_code?: string;\n language_confidence?: number;\n speaker_label?: string;\n words?: Array<{\n text?: string;\n start?: number;\n end?: number;\n confidence?: number;\n speaker?: string;\n }>;\n // Termination\n audio_duration_seconds?: number;\n session_duration_seconds?: number;\n}\n\nfunction speechDataMetadata(data: StreamEventMessage): stt.SpeechData['metadata'] | undefined {\n const assemblyai: Record<string, number> = {};\n\n if (typeof data.language_confidence === 'number') {\n assemblyai.languageConfidence = data.language_confidence;\n }\n\n if (Object.keys(assemblyai).length === 0) return undefined;\n\n return { assemblyai };\n}\n\nexport interface STTOptions {\n apiKey?: string;\n sampleRate: number;\n /**\n * How large each chunk of audio is before being sent to AssemblyAI, in\n * milliseconds. Corresponds to Python's `buffer_size_seconds` (seconds there,\n * ms here per this repo's time-unit convention).\n */\n bufferSizeMs: number;\n encoding: STTEncoding;\n speechModel: STTModels;\n languageDetection?: boolean;\n /**\n * Session inactivity timeout in seconds. AssemblyAI accepts integer values\n * from 5 to 3600; when unset, no inactivity timeout is applied.\n */\n inactivityTimeout?: number;\n endOfTurnConfidenceThreshold?: number;\n /** Minimum silence (ms) before a confident end-of-turn is finalized. */\n minTurnSilence?: number;\n /** Maximum silence (ms) before end-of-turn is forced regardless of confidence. */\n maxTurnSilence?: number;\n formatTurns?: boolean;\n keytermsPrompt?: string[];\n /** Only supported with the Universal-3 Pro model family. */\n prompt?: string;\n /** Only supported with the Universal-3 Pro model family. */\n agentContext?: string;\n /** Only supported with the Universal-3 Pro model family. Set at connection time only. */\n previousContextNTurns?: number;\n vadThreshold?: number;\n /**\n * Enable speaker diarization. Note: AssemblyAI will return per-word speaker\n * labels, but the JS framework's `stt.SpeechData` type does not yet expose\n * a `speakerId` field (unlike the Python framework), so the labels are not\n * currently surfaced on emitted events. Setting this to `true` still has\n * effect server-side. Once the base `SpeechData` interface gains speaker\n * support, `#processStreamEvent` should forward `data.words[].speaker` too.\n */\n speakerLabels?: boolean;\n maxSpeakers?: number;\n domain?: string;\n /** Isolate the primary voice and suppress background noise. Connect-time only. */\n voiceFocus?: VoiceFocus;\n /** Background audio suppression aggressiveness, from 0.0 to 1.0. Connect-time only. */\n voiceFocusThreshold?: number;\n /**\n * Accuracy/latency preset for the Universal-3 Pro model family: `min_latency`, `balanced`,\n * or `max_accuracy`. Explicit turn-silence values still take precedence over mode defaults.\n */\n mode?: 'min_latency' | 'balanced' | 'max_accuracy';\n baseUrl: string;\n}\n\nconst defaultSTTOptions: STTOptions = {\n apiKey: process.env.ASSEMBLYAI_API_KEY,\n sampleRate: 16000,\n bufferSizeMs: 50,\n encoding: 'pcm_s16le',\n speechModel: 'universal-3-5-pro',\n baseUrl: 'wss://streaming.assemblyai.com',\n};\n\nexport class STT extends stt.STT {\n #opts: STTOptions;\n #streams = new Set<WeakRef<SpeechStream>>();\n label = 'assemblyai.STT';\n\n get model(): string {\n return this.#opts.speechModel;\n }\n\n get provider(): string {\n return 'AssemblyAI';\n }\n\n constructor(opts: Partial<STTOptions> = {}) {\n super({\n streaming: true,\n interimResults: true,\n alignedTranscript: 'word',\n });\n\n if (opts.speechModel === 'u3-pro') {\n log().warn(\"'u3-pro' is deprecated, use 'universal-3-5-pro' instead.\");\n opts.speechModel = 'universal-3-5-pro';\n }\n\n const speechModel = opts.speechModel ?? defaultSTTOptions.speechModel;\n if (!isU3ProModel(speechModel)) {\n for (const param of [\n 'prompt',\n 'agentContext',\n 'previousContextNTurns',\n 'voiceFocus',\n 'voiceFocusThreshold',\n 'mode',\n ] as const) {\n if (opts[param] !== undefined) {\n throw new Error(\n `The '${param}' parameter is only supported with the ${U3_PRO_MODELS.join(', ')} models.`,\n );\n }\n }\n }\n\n const apiKey = opts.apiKey ?? defaultSTTOptions.apiKey;\n if (!apiKey) {\n throw new Error(\n 'AssemblyAI API key is required. Pass one in via the `apiKey` parameter, or set it as the `ASSEMBLYAI_API_KEY` environment variable',\n );\n }\n\n // Minimize latency by default, but let AssemblyAI's mode preset control silence tuning.\n const minTurnSilence = opts.minTurnSilence ?? (opts.mode === undefined ? 100 : undefined);\n\n this.#opts = {\n ...defaultSTTOptions,\n ...opts,\n apiKey,\n minTurnSilence,\n };\n }\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n async _recognize(_: AudioBuffer): Promise<stt.SpeechEvent> {\n throw new Error('Non-streaming recognize is not supported on AssemblyAI STT');\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n for (const ref of this.#streams) {\n const stream = ref.deref();\n if (stream) {\n stream.updateOptions(opts);\n } else {\n this.#streams.delete(ref);\n }\n }\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): SpeechStream {\n const stream = new SpeechStream(this, this.#opts, options?.connOptions);\n this.#streams.add(new WeakRef(stream));\n return stream;\n }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n static readonly CLOSE_MSG = JSON.stringify({ type: 'Terminate' });\n\n #opts: STTOptions;\n #logger = log();\n #speechDurationInS = 0;\n #lastPreflightStartTime = 0;\n #pendingConfigMessages: Record<string, unknown>[] = [];\n #configMessagePending = new Future();\n #sessionId: string | null = null;\n #expiresAt: number | null = null;\n label = 'assemblyai.SpeechStream';\n\n constructor(stt: STT, opts: STTOptions, connOptions?: APIConnectOptions) {\n super(stt, opts.sampleRate, connOptions);\n this.#opts = opts;\n this.closed = false;\n }\n\n /**\n * The AssemblyAI session ID. Set when the WebSocket connection is established\n * (before any speech events). Null until the connection completes.\n * Share this with the AssemblyAI team when reporting issues.\n */\n get sessionId(): string | null {\n return this.#sessionId;\n }\n\n /**\n * Unix timestamp when the AssemblyAI session expires. Set alongside\n * {@link sessionId} when the WebSocket connection is established.\n */\n get expiresAt(): number | null {\n return this.#expiresAt;\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n\n const configMsg: Record<string, unknown> = { type: 'UpdateConfiguration' };\n if (opts.prompt !== undefined) configMsg.prompt = opts.prompt;\n if (opts.agentContext !== undefined) configMsg.agent_context = opts.agentContext;\n if (opts.keytermsPrompt !== undefined) configMsg.keyterms_prompt = opts.keytermsPrompt;\n if (opts.maxTurnSilence !== undefined) configMsg.max_turn_silence = opts.maxTurnSilence;\n if (opts.minTurnSilence !== undefined) configMsg.min_turn_silence = opts.minTurnSilence;\n if (opts.endOfTurnConfidenceThreshold !== undefined) {\n configMsg.end_of_turn_confidence_threshold = opts.endOfTurnConfidenceThreshold;\n }\n if (opts.vadThreshold !== undefined) configMsg.vad_threshold = opts.vadThreshold;\n\n // Only send if any actual fields (besides `type`) were specified.\n if (Object.keys(configMsg).length > 1) {\n this.#pendingConfigMessages.push(configMsg);\n if (!this.#configMessagePending.done) this.#configMessagePending.resolve();\n }\n }\n\n /**\n * Force-finalize the current turn immediately.\n */\n forceEndpoint() {\n this.#pendingConfigMessages.push({ type: 'ForceEndpoint' });\n if (!this.#configMessagePending.done) this.#configMessagePending.resolve();\n }\n\n // Deepgram-style reconnect loop around a single websocket lifetime.\n protected async run() {\n const maxRetry = 32;\n let retries = 0;\n\n while (!this.input.closed && !this.closed) {\n try {\n const ws = await this.#connectWS();\n await this.#runWS(ws);\n retries = 0;\n } catch (e) {\n if (!this.closed && !this.input.closed) {\n if (retries >= maxRetry) {\n throw new Error(`failed to connect to AssemblyAI after ${retries} attempts: ${e}`);\n }\n\n const retryDelaySeconds = Math.min(retries * 5, 10);\n retries++;\n\n this.#logger.warn(\n `failed to connect to AssemblyAI, retrying in ${retryDelaySeconds} seconds: ${e} (${retries}/${maxRetry})`,\n );\n await delay(retryDelaySeconds * 1000);\n } else {\n this.#logger.warn(\n `AssemblyAI disconnected, connection is closed: ${e} (inputClosed: ${this.input.closed}, isClosed: ${this.closed})`,\n );\n }\n }\n }\n\n this.closed = true;\n }\n\n async #connectWS(): Promise<WebSocket> {\n // Universal-3 Pro family models default both min and max silence to 100ms when unset.\n // When a mode preset is selected, leave them unset unless explicitly provided so the\n // server's per-mode silence tuning is not overridden by the latency-optimized default.\n let minSilence = this.#opts.minTurnSilence;\n let maxSilence = this.#opts.maxTurnSilence;\n if (isU3ProModel(this.#opts.speechModel)) {\n if (minSilence === undefined && this.#opts.mode === undefined) minSilence = 100;\n if (maxSilence === undefined) maxSilence = minSilence;\n }\n\n // Default language_detection to true for multilingual / u3-rt-pro-family models, false otherwise.\n const defaultLanguageDetection =\n this.#opts.speechModel.includes('multilingual') || isU3ProModel(this.#opts.speechModel);\n const languageDetection = this.#opts.languageDetection ?? defaultLanguageDetection;\n\n const liveConfig: Record<string, unknown> = {\n sample_rate: this.#opts.sampleRate,\n encoding: this.#opts.encoding,\n speech_model: this.#opts.speechModel,\n format_turns: this.#opts.formatTurns,\n end_of_turn_confidence_threshold: this.#opts.endOfTurnConfidenceThreshold,\n min_turn_silence: minSilence,\n max_turn_silence: maxSilence,\n keyterms_prompt:\n this.#opts.keytermsPrompt !== undefined\n ? JSON.stringify(this.#opts.keytermsPrompt)\n : undefined,\n language_detection: languageDetection,\n inactivity_timeout: this.#opts.inactivityTimeout,\n prompt: this.#opts.prompt,\n agent_context: this.#opts.agentContext,\n previous_context_n_turns: this.#opts.previousContextNTurns,\n vad_threshold: this.#opts.vadThreshold,\n speaker_labels: this.#opts.speakerLabels,\n max_speakers: this.#opts.maxSpeakers,\n domain: this.#opts.domain,\n voice_focus: this.#opts.voiceFocus,\n voice_focus_threshold: this.#opts.voiceFocusThreshold,\n mode: this.#opts.mode,\n };\n\n const url = new URL(`${this.#opts.baseUrl}/v3/ws`);\n // Python serializes booleans as the strings \"true\"/\"false\", so we mirror that.\n for (const [key, value] of Object.entries(liveConfig)) {\n if (value === undefined || value === null) continue;\n if (typeof value === 'boolean') {\n url.searchParams.append(key, value ? 'true' : 'false');\n } else {\n url.searchParams.append(key, String(value));\n }\n }\n\n const ws = new WebSocket(url, {\n headers: {\n Authorization: this.#opts.apiKey!,\n 'Content-Type': 'application/json',\n 'User-Agent': 'AssemblyAI/1.0 (integration=Livekit)',\n },\n });\n\n await new Promise<void>((resolve, reject) => {\n ws.on('open', () => resolve());\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(new Error(`WebSocket returned ${code}`)));\n });\n\n return ws;\n }\n\n async #runWS(ws: WebSocket) {\n let closing = false;\n const sessionController = new AbortController();\n\n // gets cancelled also when sendTask is complete\n const wsMonitor = Task.from(async (controller) => {\n const closed = new Promise<void>((_, reject) => {\n ws.once('close', (code, reason) => {\n if (!closing) {\n this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);\n reject(new Error('WebSocket closed'));\n }\n });\n });\n\n await Promise.race([closed, waitForAbort(controller.signal)]);\n });\n\n const sendTask = async () => {\n const samplesPerBuffer = Math.floor((this.#opts.sampleRate * this.#opts.bufferSizeMs) / 1000);\n const audioStream = new AudioByteStream(this.#opts.sampleRate, 1, samplesPerBuffer);\n\n const abortPromise = waitForAbort(this.abortSignal);\n const sessionAbort = waitForAbort(sessionController.signal);\n\n try {\n while (!this.closed) {\n const result = await Promise.race([this.input.next(), abortPromise, sessionAbort]);\n\n if (result === undefined) return; // aborted\n if (result.done) break;\n\n const data = result.value;\n\n let frames: AudioFrame[];\n if (data === SpeechStream.FLUSH_SENTINEL) {\n frames = audioStream.flush();\n } else if (data.sampleRate === this.#opts.sampleRate && data.channels === 1) {\n // AssemblyAI expects mono PCM. The base SpeechStream only resamples\n // sample rate, so reject any frame that is not already downmixed.\n frames = audioStream.write(data.data.buffer as ArrayBuffer);\n } else {\n throw new Error('sample rate or channel count of frame does not match');\n }\n\n for (const frame of frames) {\n this.#speechDurationInS += frame.samplesPerChannel / frame.sampleRate;\n ws.send(frame.data.buffer);\n }\n }\n } finally {\n closing = true;\n try {\n ws.send(SpeechStream.CLOSE_MSG);\n } catch {\n // ignore — socket may already be closing\n }\n wsMonitor.cancel();\n }\n };\n\n let messageHandler: ((msg: RawData, isBinary: boolean) => void) | null = null;\n const listenTask = Task.from(async (controller) => {\n const listenMessage = new Promise<void>((resolve, reject) => {\n messageHandler = (msg, isBinary) => {\n if (isBinary) {\n this.#logger.error('unexpected binary message from AssemblyAI');\n return;\n }\n try {\n const json = JSON.parse(msg.toString()) as StreamEventMessage;\n this.#processStreamEvent(json);\n if (this.closed || closing) {\n resolve();\n }\n } catch (err) {\n this.#logger.error(`AssemblyAI: error processing message: ${msg}`);\n reject(err);\n }\n };\n ws.on('message', messageHandler);\n });\n\n await Promise.race([listenMessage, waitForAbort(controller.signal)]);\n });\n\n const configTask = Task.from(async (controller) => {\n // Drain any messages queued while the socket was reconnecting.\n while (this.#pendingConfigMessages.length > 0) {\n const msg = this.#pendingConfigMessages.shift()!;\n ws.send(JSON.stringify(msg));\n }\n\n while (!controller.signal.aborted) {\n await Promise.race([this.#configMessagePending.await, waitForAbort(controller.signal)]);\n if (controller.signal.aborted) return;\n\n this.#configMessagePending = new Future();\n while (this.#pendingConfigMessages.length > 0) {\n const msg = this.#pendingConfigMessages.shift()!;\n ws.send(JSON.stringify(msg));\n }\n }\n });\n\n try {\n await Promise.all([sendTask(), listenTask.result, wsMonitor.result]);\n } finally {\n closing = true;\n sessionController.abort();\n listenTask.cancel();\n configTask.cancel();\n if (messageHandler) ws.off('message', messageHandler);\n try {\n ws.close();\n } catch {\n // ignore\n }\n }\n }\n\n #averageConfidence(words: Array<{ confidence?: number }>): number {\n if (words.length === 0) return 0;\n return words.reduce((sum, w) => sum + (w.confidence ?? 0), 0) / words.length;\n }\n\n #processStreamEvent(data: StreamEventMessage) {\n const messageType = data.type;\n\n if (messageType === 'Begin') {\n this.#sessionId = data.id ?? null;\n this.#expiresAt = data.expires_at ?? null;\n this.#logger.info(\n `AssemblyAI session started id=${this.#sessionId} expires_at=${this.#expiresAt}`,\n );\n return;\n }\n\n if (messageType === 'SpeechStarted') {\n this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n return;\n }\n\n if (messageType === 'Termination') {\n this.#logger.debug(\n `AssemblyAI session terminated audio_duration=${data.audio_duration_seconds}s session_duration=${data.session_duration_seconds}s`,\n );\n return;\n }\n\n if (messageType !== 'Turn') {\n return;\n }\n\n const words = data.words ?? [];\n const endOfTurn = Boolean(data.end_of_turn);\n const turnIsFormatted = Boolean(data.turn_is_formatted);\n const utterance = data.utterance ?? '';\n const transcript = data.transcript ?? '';\n const language = normalizeLanguage(data.language_code ?? 'en');\n const metadata = speechDataMetadata(data);\n\n // Word timestamps are in milliseconds:\n // https://www.assemblyai.com/docs/api-reference/streaming-api/streaming-api#receive.receiveTurn.words\n const timedWords = words.map((word) =>\n createTimedString({\n text: word.text ?? '',\n startTime: (word.start ?? 0) / 1000 + this.startTimeOffset,\n endTime: (word.end ?? 0) / 1000 + this.startTimeOffset,\n confidence: word.confidence ?? 0,\n startTimeOffset: this.startTimeOffset,\n }),\n );\n\n let startTime = 0;\n let endTime = 0;\n let confidence = 0;\n\n // `words` are cumulative for the turn — emit as an interim transcript.\n if (timedWords.length > 0) {\n const interimText = timedWords.map((w) => w.text).join(' ');\n startTime = timedWords[0]!.startTime ?? 0;\n endTime = timedWords[timedWords.length - 1]!.endTime ?? 0;\n confidence = this.#averageConfidence(timedWords);\n\n this.queue.put({\n type: stt.SpeechEventType.INTERIM_TRANSCRIPT,\n alternatives: [\n {\n language,\n text: interimText,\n startTime,\n endTime,\n confidence,\n words: timedWords,\n ...(metadata ? { metadata } : {}),\n },\n ],\n });\n }\n\n // `utterance` is chunk-based (not cumulative) — emit as a preflight transcript\n // covering only the words since the last preflight.\n if (utterance) {\n if (this.#lastPreflightStartTime === 0) {\n this.#lastPreflightStartTime = startTime;\n }\n\n const utteranceWords = timedWords.filter(\n (w) => w.startTime !== undefined && w.startTime >= this.#lastPreflightStartTime,\n );\n const utteranceConfidence = this.#averageConfidence(utteranceWords);\n\n this.queue.put({\n type: stt.SpeechEventType.PREFLIGHT_TRANSCRIPT,\n alternatives: [\n {\n language,\n text: utterance,\n startTime: this.#lastPreflightStartTime,\n endTime,\n confidence: utteranceConfidence,\n words: utteranceWords,\n ...(metadata ? { metadata } : {}),\n },\n ],\n });\n this.#lastPreflightStartTime = endTime;\n }\n\n // End-of-turn: emit FINAL_TRANSCRIPT + END_OF_SPEECH.\n // If the user asked for formatted turns, wait for a formatted final.\n const waitingForFormatted = this.#opts.formatTurns === true && !turnIsFormatted;\n if (endOfTurn && !waitingForFormatted) {\n this.queue.put({\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n alternatives: [\n {\n language,\n text: transcript,\n startTime,\n endTime,\n confidence,\n words: timedWords,\n ...(metadata ? { metadata } : {}),\n },\n ],\n });\n\n this.queue.put({ type: stt.SpeechEventType.END_OF_SPEECH });\n if (this.#speechDurationInS > 0) {\n this.queue.put({\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n // Propagate the AssemblyAI session id as the request id so metrics\n // can be correlated back to a specific connection, mirroring how\n // Deepgram surfaces its `request_id`.\n requestId: this.#sessionId ?? undefined,\n recognitionUsage: {\n audioDuration: this.#speechDurationInS,\n },\n });\n this.#speechDurationInS = 0;\n this.#lastPreflightStartTime = 0;\n }\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,oBAYO;AAGP,gBAA0B;AAI1B,MAAM,gBAAgB,CAAC,aAAa,oBAAoB,mBAAmB;AAE3E,SAAS,aAAa,OAA2B;AAC/C,SAAO,cAAc,SAAS,KAAuC;AACvE;AA8BA,SAAS,mBAAmB,MAAkE;AAC5F,QAAM,aAAqC,CAAC;AAE5C,MAAI,OAAO,KAAK,wBAAwB,UAAU;AAChD,eAAW,qBAAqB,KAAK;AAAA,EACvC;AAEA,MAAI,OAAO,KAAK,UAAU,EAAE,WAAW,EAAG,QAAO;AAEjD,SAAO,EAAE,WAAW;AACtB;AAwDA,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,YAAY;AAAA,EACZ,cAAc;AAAA,EACd,UAAU;AAAA,EACV,aAAa;AAAA,EACb,SAAS;AACX;AAEO,MAAM,YAAY,kBAAI,IAAI;AAAA,EAC/B;AAAA,EACA,WAAW,oBAAI,IAA2B;AAAA,EAC1C,QAAQ;AAAA,EAER,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,YAAY,OAA4B,CAAC,GAAG;AAC1C,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB;AAAA,MAChB,mBAAmB;AAAA,IACrB,CAAC;AAED,QAAI,KAAK,gBAAgB,UAAU;AACjC,6BAAI,EAAE,KAAK,0DAA0D;AACrE,WAAK,cAAc;AAAA,IACrB;AAEA,UAAM,cAAc,KAAK,eAAe,kBAAkB;AAC1D,QAAI,CAAC,aAAa,WAAW,GAAG;AAC9B,iBAAW,SAAS;AAAA,QAClB;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MACF,GAAY;AACV,YAAI,KAAK,KAAK,MAAM,QAAW;AAC7B,gBAAM,IAAI;AAAA,YACR,QAAQ,KAAK,0CAA0C,cAAc,KAAK,IAAI,CAAC;AAAA,UACjF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,SAAS,KAAK,UAAU,kBAAkB;AAChD,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAGA,UAAM,iBAAiB,KAAK,mBAAmB,KAAK,SAAS,SAAY,MAAM;AAE/E,SAAK,QAAQ;AAAA,MACX,GAAG;AAAA,MACH,GAAG;AAAA,MACH;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,WAAW,GAA0C;AACzD,UAAM,IAAI,MAAM,4DAA4D;AAAA,EAC9E;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,eAAW,OAAO,KAAK,UAAU;AAC/B,YAAM,SAAS,IAAI,MAAM;AACzB,UAAI,QAAQ;AACV,eAAO,cAAc,IAAI;AAAA,MAC3B,OAAO;AACL,aAAK,SAAS,OAAO,GAAG;AAAA,MAC1B;AAAA,IACF;AAAA,EACF;AAAA,EAEA,OAAO,SAA6D;AAClE,UAAM,SAAS,IAAI,aAAa,MAAM,KAAK,OAAO,mCAAS,WAAW;AACtE,SAAK,SAAS,IAAI,IAAI,QAAQ,MAAM,CAAC;AACrC,WAAO;AAAA,EACT;AACF;AAEO,MAAM,qBAAqB,kBAAI,aAAa;AAAA,EACjD,OAAgB,YAAY,KAAK,UAAU,EAAE,MAAM,YAAY,CAAC;AAAA,EAEhE;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,qBAAqB;AAAA,EACrB,0BAA0B;AAAA,EAC1B,yBAAoD,CAAC;AAAA,EACrD,wBAAwB,IAAI,qBAAO;AAAA,EACnC,aAA4B;AAAA,EAC5B,aAA4B;AAAA,EAC5B,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB,aAAiC;AACvE,UAAMA,MAAK,KAAK,YAAY,WAAW;AACvC,SAAK,QAAQ;AACb,SAAK,SAAS;AAAA,EAChB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,IAAI,YAA2B;AAC7B,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,IAAI,YAA2B;AAC7B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAEtC,UAAM,YAAqC,EAAE,MAAM,sBAAsB;AACzE,QAAI,KAAK,WAAW,OAAW,WAAU,SAAS,KAAK;AACvD,QAAI,KAAK,iBAAiB,OAAW,WAAU,gBAAgB,KAAK;AACpE,QAAI,KAAK,mBAAmB,OAAW,WAAU,kBAAkB,KAAK;AACxE,QAAI,KAAK,mBAAmB,OAAW,WAAU,mBAAmB,KAAK;AACzE,QAAI,KAAK,mBAAmB,OAAW,WAAU,mBAAmB,KAAK;AACzE,QAAI,KAAK,iCAAiC,QAAW;AACnD,gBAAU,mCAAmC,KAAK;AAAA,IACpD;AACA,QAAI,KAAK,iBAAiB,OAAW,WAAU,gBAAgB,KAAK;AAGpE,QAAI,OAAO,KAAK,SAAS,EAAE,SAAS,GAAG;AACrC,WAAK,uBAAuB,KAAK,SAAS;AAC1C,UAAI,CAAC,KAAK,sBAAsB,KAAM,MAAK,sBAAsB,QAAQ;AAAA,IAC3E;AAAA,EACF;AAAA;AAAA;AAAA;AAAA,EAKA,gBAAgB;AACd,SAAK,uBAAuB,KAAK,EAAE,MAAM,gBAAgB,CAAC;AAC1D,QAAI,CAAC,KAAK,sBAAsB,KAAM,MAAK,sBAAsB,QAAQ;AAAA,EAC3E;AAAA;AAAA,EAGA,MAAgB,MAAM;AACpB,UAAM,WAAW;AACjB,QAAI,UAAU;AAEd,WAAO,CAAC,KAAK,MAAM,UAAU,CAAC,KAAK,QAAQ;AACzC,UAAI;AACF,cAAM,KAAK,MAAM,KAAK,WAAW;AACjC,cAAM,KAAK,OAAO,EAAE;AACpB,kBAAU;AAAA,MACZ,SAAS,GAAG;AACV,YAAI,CAAC,KAAK,UAAU,CAAC,KAAK,MAAM,QAAQ;AACtC,cAAI,WAAW,UAAU;AACvB,kBAAM,IAAI,MAAM,yCAAyC,OAAO,cAAc,CAAC,EAAE;AAAA,UACnF;AAEA,gBAAM,oBAAoB,KAAK,IAAI,UAAU,GAAG,EAAE;AAClD;AAEA,eAAK,QAAQ;AAAA,YACX,gDAAgD,iBAAiB,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ;AAAA,UACzG;AACA,oBAAM,qBAAM,oBAAoB,GAAI;AAAA,QACtC,OAAO;AACL,eAAK,QAAQ;AAAA,YACX,kDAAkD,CAAC,kBAAkB,KAAK,MAAM,MAAM,eAAe,KAAK,MAAM;AAAA,UAClH;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAM,aAAiC;AAIrC,QAAI,aAAa,KAAK,MAAM;AAC5B,QAAI,aAAa,KAAK,MAAM;AAC5B,QAAI,aAAa,KAAK,MAAM,WAAW,GAAG;AACxC,UAAI,eAAe,UAAa,KAAK,MAAM,SAAS,OAAW,cAAa;AAC5E,UAAI,eAAe,OAAW,cAAa;AAAA,IAC7C;AAGA,UAAM,2BACJ,KAAK,MAAM,YAAY,SAAS,cAAc,KAAK,aAAa,KAAK,MAAM,WAAW;AACxF,UAAM,oBAAoB,KAAK,MAAM,qBAAqB;AAE1D,UAAM,aAAsC;AAAA,MAC1C,aAAa,KAAK,MAAM;AAAA,MACxB,UAAU,KAAK,MAAM;AAAA,MACrB,cAAc,KAAK,MAAM;AAAA,MACzB,cAAc,KAAK,MAAM;AAAA,MACzB,kCAAkC,KAAK,MAAM;AAAA,MAC7C,kBAAkB;AAAA,MAClB,kBAAkB;AAAA,MAClB,iBACE,KAAK,MAAM,mBAAmB,SAC1B,KAAK,UAAU,KAAK,MAAM,cAAc,IACxC;AAAA,MACN,oBAAoB;AAAA,MACpB,oBAAoB,KAAK,MAAM;AAAA,MAC/B,QAAQ,KAAK,MAAM;AAAA,MACnB,eAAe,KAAK,MAAM;AAAA,MAC1B,0BAA0B,KAAK,MAAM;AAAA,MACrC,eAAe,KAAK,MAAM;AAAA,MAC1B,gBAAgB,KAAK,MAAM;AAAA,MAC3B,cAAc,KAAK,MAAM;AAAA,MACzB,QAAQ,KAAK,MAAM;AAAA,MACnB,aAAa,KAAK,MAAM;AAAA,MACxB,uBAAuB,KAAK,MAAM;AAAA,MAClC,MAAM,KAAK,MAAM;AAAA,IACnB;AAEA,UAAM,MAAM,IAAI,IAAI,GAAG,KAAK,MAAM,OAAO,QAAQ;AAEjD,eAAW,CAAC,KAAK,KAAK,KAAK,OAAO,QAAQ,UAAU,GAAG;AACrD,UAAI,UAAU,UAAa,UAAU,KAAM;AAC3C,UAAI,OAAO,UAAU,WAAW;AAC9B,YAAI,aAAa,OAAO,KAAK,QAAQ,SAAS,OAAO;AAAA,MACvD,OAAO;AACL,YAAI,aAAa,OAAO,KAAK,OAAO,KAAK,CAAC;AAAA,MAC5C;AAAA,IACF;AAEA,UAAM,KAAK,IAAI,oBAAU,KAAK;AAAA,MAC5B,SAAS;AAAA,QACP,eAAe,KAAK,MAAM;AAAA,QAC1B,gBAAgB;AAAA,QAChB,cAAc;AAAA,MAChB;AAAA,IACF,CAAC;AAED,UAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,SAAG,GAAG,QAAQ,MAAM,QAAQ,CAAC;AAC7B,SAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,SAAG,GAAG,SAAS,CAAC,SAAS,OAAO,IAAI,MAAM,sBAAsB,IAAI,EAAE,CAAC,CAAC;AAAA,IAC1E,CAAC;AAED,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,OAAO,IAAe;AAC1B,QAAI,UAAU;AACd,UAAM,oBAAoB,IAAI,gBAAgB;AAG9C,UAAM,YAAY,mBAAK,KAAK,OAAO,eAAe;AAChD,YAAM,SAAS,IAAI,QAAc,CAAC,GAAG,WAAW;AAC9C,WAAG,KAAK,SAAS,CAAC,MAAM,WAAW;AACjC,cAAI,CAAC,SAAS;AACZ,iBAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,MAAM,EAAE;AAClE,mBAAO,IAAI,MAAM,kBAAkB,CAAC;AAAA,UACtC;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,YAAQ,4BAAa,WAAW,MAAM,CAAC,CAAC;AAAA,IAC9D,CAAC;AAED,UAAM,WAAW,YAAY;AAC3B,YAAM,mBAAmB,KAAK,MAAO,KAAK,MAAM,aAAa,KAAK,MAAM,eAAgB,GAAI;AAC5F,YAAM,cAAc,IAAI,8BAAgB,KAAK,MAAM,YAAY,GAAG,gBAAgB;AAElF,YAAM,mBAAe,4BAAa,KAAK,WAAW;AAClD,YAAM,mBAAe,4BAAa,kBAAkB,MAAM;AAE1D,UAAI;AACF,eAAO,CAAC,KAAK,QAAQ;AACnB,gBAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,KAAK,MAAM,KAAK,GAAG,cAAc,YAAY,CAAC;AAEjF,cAAI,WAAW,OAAW;AAC1B,cAAI,OAAO,KAAM;AAEjB,gBAAM,OAAO,OAAO;AAEpB,cAAI;AACJ,cAAI,SAAS,aAAa,gBAAgB;AACxC,qBAAS,YAAY,MAAM;AAAA,UAC7B,WAAW,KAAK,eAAe,KAAK,MAAM,cAAc,KAAK,aAAa,GAAG;AAG3E,qBAAS,YAAY,MAAM,KAAK,KAAK,MAAqB;AAAA,UAC5D,OAAO;AACL,kBAAM,IAAI,MAAM,sDAAsD;AAAA,UACxE;AAEA,qBAAW,SAAS,QAAQ;AAC1B,iBAAK,sBAAsB,MAAM,oBAAoB,MAAM;AAC3D,eAAG,KAAK,MAAM,KAAK,MAAM;AAAA,UAC3B;AAAA,QACF;AAAA,MACF,UAAE;AACA,kBAAU;AACV,YAAI;AACF,aAAG,KAAK,aAAa,SAAS;AAAA,QAChC,QAAQ;AAAA,QAER;AACA,kBAAU,OAAO;AAAA,MACnB;AAAA,IACF;AAEA,QAAI,iBAAqE;AACzE,UAAM,aAAa,mBAAK,KAAK,OAAO,eAAe;AACjD,YAAM,gBAAgB,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3D,yBAAiB,CAAC,KAAK,aAAa;AAClC,cAAI,UAAU;AACZ,iBAAK,QAAQ,MAAM,2CAA2C;AAC9D;AAAA,UACF;AACA,cAAI;AACF,kBAAM,OAAO,KAAK,MAAM,IAAI,SAAS,CAAC;AACtC,iBAAK,oBAAoB,IAAI;AAC7B,gBAAI,KAAK,UAAU,SAAS;AAC1B,sBAAQ;AAAA,YACV;AAAA,UACF,SAAS,KAAK;AACZ,iBAAK,QAAQ,MAAM,yCAAyC,GAAG,EAAE;AACjE,mBAAO,GAAG;AAAA,UACZ;AAAA,QACF;AACA,WAAG,GAAG,WAAW,cAAc;AAAA,MACjC,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,mBAAe,4BAAa,WAAW,MAAM,CAAC,CAAC;AAAA,IACrE,CAAC;AAED,UAAM,aAAa,mBAAK,KAAK,OAAO,eAAe;AAEjD,aAAO,KAAK,uBAAuB,SAAS,GAAG;AAC7C,cAAM,MAAM,KAAK,uBAAuB,MAAM;AAC9C,WAAG,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,MAC7B;AAEA,aAAO,CAAC,WAAW,OAAO,SAAS;AACjC,cAAM,QAAQ,KAAK,CAAC,KAAK,sBAAsB,WAAO,4BAAa,WAAW,MAAM,CAAC,CAAC;AACtF,YAAI,WAAW,OAAO,QAAS;AAE/B,aAAK,wBAAwB,IAAI,qBAAO;AACxC,eAAO,KAAK,uBAAuB,SAAS,GAAG;AAC7C,gBAAM,MAAM,KAAK,uBAAuB,MAAM;AAC9C,aAAG,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,QAC7B;AAAA,MACF;AAAA,IACF,CAAC;AAED,QAAI;AACF,YAAM,QAAQ,IAAI,CAAC,SAAS,GAAG,WAAW,QAAQ,UAAU,MAAM,CAAC;AAAA,IACrE,UAAE;AACA,gBAAU;AACV,wBAAkB,MAAM;AACxB,iBAAW,OAAO;AAClB,iBAAW,OAAO;AAClB,UAAI,eAAgB,IAAG,IAAI,WAAW,cAAc;AACpD,UAAI;AACF,WAAG,MAAM;AAAA,MACX,QAAQ;AAAA,MAER;AAAA,IACF;AAAA,EACF;AAAA,EAEA,mBAAmB,OAA+C;AAChE,QAAI,MAAM,WAAW,EAAG,QAAO;AAC/B,WAAO,MAAM,OAAO,CAAC,KAAK,MAAM,OAAO,EAAE,cAAc,IAAI,CAAC,IAAI,MAAM;AAAA,EACxE;AAAA,EAEA,oBAAoB,MAA0B;AAC5C,UAAM,cAAc,KAAK;AAEzB,QAAI,gBAAgB,SAAS;AAC3B,WAAK,aAAa,KAAK,MAAM;AAC7B,WAAK,aAAa,KAAK,cAAc;AACrC,WAAK,QAAQ;AAAA,QACX,iCAAiC,KAAK,UAAU,eAAe,KAAK,UAAU;AAAA,MAChF;AACA;AAAA,IACF;AAEA,QAAI,gBAAgB,iBAAiB;AACnC,WAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,gBAAgB,CAAC;AAC5D;AAAA,IACF;AAEA,QAAI,gBAAgB,eAAe;AACjC,WAAK,QAAQ;AAAA,QACX,gDAAgD,KAAK,sBAAsB,sBAAsB,KAAK,wBAAwB;AAAA,MAChI;AACA;AAAA,IACF;AAEA,QAAI,gBAAgB,QAAQ;AAC1B;AAAA,IACF;AAEA,UAAM,QAAQ,KAAK,SAAS,CAAC;AAC7B,UAAM,YAAY,QAAQ,KAAK,WAAW;AAC1C,UAAM,kBAAkB,QAAQ,KAAK,iBAAiB;AACtD,UAAM,YAAY,KAAK,aAAa;AACpC,UAAM,aAAa,KAAK,cAAc;AACtC,UAAM,eAAW,iCAAkB,KAAK,iBAAiB,IAAI;AAC7D,UAAM,WAAW,mBAAmB,IAAI;AAIxC,UAAM,aAAa,MAAM;AAAA,MAAI,CAAC,aAC5B,iCAAkB;AAAA,QAChB,MAAM,KAAK,QAAQ;AAAA,QACnB,YAAY,KAAK,SAAS,KAAK,MAAO,KAAK;AAAA,QAC3C,UAAU,KAAK,OAAO,KAAK,MAAO,KAAK;AAAA,QACvC,YAAY,KAAK,cAAc;AAAA,QAC/B,iBAAiB,KAAK;AAAA,MACxB,CAAC;AAAA,IACH;AAEA,QAAI,YAAY;AAChB,QAAI,UAAU;AACd,QAAI,aAAa;AAGjB,QAAI,WAAW,SAAS,GAAG;AACzB,YAAM,cAAc,WAAW,IAAI,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,GAAG;AAC1D,kBAAY,WAAW,CAAC,EAAG,aAAa;AACxC,gBAAU,WAAW,WAAW,SAAS,CAAC,EAAG,WAAW;AACxD,mBAAa,KAAK,mBAAmB,UAAU;AAE/C,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,kBAAI,gBAAgB;AAAA,QAC1B,cAAc;AAAA,UACZ;AAAA,YACE;AAAA,YACA,MAAM;AAAA,YACN;AAAA,YACA;AAAA,YACA;AAAA,YACA,OAAO;AAAA,YACP,GAAI,WAAW,EAAE,SAAS,IAAI,CAAC;AAAA,UACjC;AAAA,QACF;AAAA,MACF,CAAC;AAAA,IACH;AAIA,QAAI,WAAW;AACb,UAAI,KAAK,4BAA4B,GAAG;AACtC,aAAK,0BAA0B;AAAA,MACjC;AAEA,YAAM,iBAAiB,WAAW;AAAA,QAChC,CAAC,MAAM,EAAE,cAAc,UAAa,EAAE,aAAa,KAAK;AAAA,MAC1D;AACA,YAAM,sBAAsB,KAAK,mBAAmB,cAAc;AAElE,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,kBAAI,gBAAgB;AAAA,QAC1B,cAAc;AAAA,UACZ;AAAA,YACE;AAAA,YACA,MAAM;AAAA,YACN,WAAW,KAAK;AAAA,YAChB;AAAA,YACA,YAAY;AAAA,YACZ,OAAO;AAAA,YACP,GAAI,WAAW,EAAE,SAAS,IAAI,CAAC;AAAA,UACjC;AAAA,QACF;AAAA,MACF,CAAC;AACD,WAAK,0BAA0B;AAAA,IACjC;AAIA,UAAM,sBAAsB,KAAK,MAAM,gBAAgB,QAAQ,CAAC;AAChE,QAAI,aAAa,CAAC,qBAAqB;AACrC,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,kBAAI,gBAAgB;AAAA,QAC1B,cAAc;AAAA,UACZ;AAAA,YACE;AAAA,YACA,MAAM;AAAA,YACN;AAAA,YACA;AAAA,YACA;AAAA,YACA,OAAO;AAAA,YACP,GAAI,WAAW,EAAE,SAAS,IAAI,CAAC;AAAA,UACjC;AAAA,QACF;AAAA,MACF,CAAC;AAED,WAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,cAAc,CAAC;AAC1D,UAAI,KAAK,qBAAqB,GAAG;AAC/B,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,kBAAI,gBAAgB;AAAA;AAAA;AAAA;AAAA,UAI1B,WAAW,KAAK,cAAc;AAAA,UAC9B,kBAAkB;AAAA,YAChB,eAAe,KAAK;AAAA,UACtB;AAAA,QACF,CAAC;AACD,aAAK,qBAAqB;AAC1B,aAAK,0BAA0B;AAAA,MACjC;AAAA,IACF;AAAA,EACF;AACF;","names":["stt"]}
|
|
1
|
+
{"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2026 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n//\nimport {\n type APIConnectOptions,\n type AudioBuffer,\n AudioByteStream,\n ChatMessage,\n type ConversationItemAddedEvent,\n Future,\n Task,\n createTimedString,\n delay,\n log,\n normalizeLanguage,\n stt,\n waitForAbort,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { RawData } from 'ws';\nimport { WebSocket } from 'ws';\nimport type { STTEncoding, STTModels, VoiceFocus } from './models.js';\n\n// Speech models in the Universal-3 Pro family, which share the same parameter support.\nconst U3_PRO_MODELS = ['u3-rt-pro', 'u3-rt-pro-beta-1', 'universal-3-5-pro'] as const;\n\nfunction isU3ProModel(model: STTModels): boolean {\n return U3_PRO_MODELS.includes(model as (typeof U3_PRO_MODELS)[number]);\n}\n\n// AssemblyAI Universal-Streaming (v3) message envelope. All fields are optional\n// since we narrow on `type` before reading anything else.\ninterface StreamEventMessage {\n type?: 'Begin' | 'SpeechStarted' | 'Turn' | 'Termination' | string;\n // Begin\n id?: string;\n expires_at?: number;\n // Turn\n transcript?: string;\n utterance?: string;\n end_of_turn?: boolean;\n end_of_turn_confidence?: number;\n turn_is_formatted?: boolean;\n language_code?: string;\n language_confidence?: number;\n speaker_label?: string;\n words?: Array<{\n text?: string;\n start?: number;\n end?: number;\n confidence?: number;\n speaker?: string;\n }>;\n // Termination\n audio_duration_seconds?: number;\n session_duration_seconds?: number;\n}\n\nfunction speechDataMetadata(data: StreamEventMessage): stt.SpeechData['metadata'] | undefined {\n const assemblyai: Record<string, number> = {};\n\n if (typeof data.language_confidence === 'number') {\n assemblyai.languageConfidence = data.language_confidence;\n }\n\n if (Object.keys(assemblyai).length === 0) return undefined;\n\n return { assemblyai };\n}\n\nexport interface STTOptions {\n apiKey?: string;\n sampleRate: number;\n /**\n * How large each chunk of audio is before being sent to AssemblyAI, in\n * milliseconds. Corresponds to Python's `buffer_size_seconds` (seconds there,\n * ms here per this repo's time-unit convention).\n */\n bufferSizeMs: number;\n encoding: STTEncoding;\n speechModel: STTModels;\n languageDetection?: boolean;\n /**\n * Session inactivity timeout in seconds. AssemblyAI accepts integer values\n * from 5 to 3600; when unset, no inactivity timeout is applied.\n */\n inactivityTimeout?: number;\n endOfTurnConfidenceThreshold?: number;\n /** Minimum silence (ms) before a confident end-of-turn is finalized. */\n minTurnSilence?: number;\n /** Maximum silence (ms) before end-of-turn is forced regardless of confidence. */\n maxTurnSilence?: number;\n formatTurns?: boolean;\n keytermsPrompt?: string[];\n /** Only supported with the Universal-3 Pro model family. */\n prompt?: string;\n /** Only supported with the Universal-3 Pro model family. */\n agentContext?: string;\n /** Only supported with the Universal-3 Pro model family. Set at connection time only. */\n previousContextNTurns?: number;\n vadThreshold?: number;\n /**\n * Enable speaker diarization. Note: AssemblyAI will return per-word speaker\n * labels, but the JS framework's `stt.SpeechData` type does not yet expose\n * a `speakerId` field (unlike the Python framework), so the labels are not\n * currently surfaced on emitted events. Setting this to `true` still has\n * effect server-side. Once the base `SpeechData` interface gains speaker\n * support, `#processStreamEvent` should forward `data.words[].speaker` too.\n */\n speakerLabels?: boolean;\n maxSpeakers?: number;\n domain?: string;\n /** Isolate the primary voice and suppress background noise. Connect-time only. */\n voiceFocus?: VoiceFocus;\n /** Background audio suppression aggressiveness, from 0.0 to 1.0. Connect-time only. */\n voiceFocusThreshold?: number;\n /**\n * Accuracy/latency preset for the Universal-3 Pro model family: `min_latency`, `balanced`,\n * or `max_accuracy`. Explicit turn-silence values still take precedence over mode defaults.\n */\n mode?: 'min_latency' | 'balanced' | 'max_accuracy';\n /**\n * When the model supports it, let an `AgentSession` push each assistant reply into\n * `agentContext` so it is carried into the model's conversation context. Defaults to false;\n * set true to enable. Prior user turns are carried automatically by the model regardless of\n * this flag. Ignored on models without context support.\n */\n agentContextCarryover?: boolean;\n baseUrl: string;\n}\n\nconst defaultSTTOptions: STTOptions = {\n apiKey: process.env.ASSEMBLYAI_API_KEY,\n sampleRate: 16000,\n bufferSizeMs: 50,\n encoding: 'pcm_s16le',\n speechModel: 'universal-3-5-pro',\n baseUrl: 'wss://streaming.assemblyai.com',\n};\n\nexport class STT extends stt.STT {\n #opts: STTOptions;\n #streams = new Set<WeakRef<SpeechStream>>();\n // set (user + session))\n #userKeyterms: string[];\n #sessionKeyterms: string[] = [];\n label = 'assemblyai.STT';\n\n get model(): string {\n return this.#opts.speechModel;\n }\n\n get provider(): string {\n return 'AssemblyAI';\n }\n\n constructor(opts: Partial<STTOptions> = {}) {\n // u3-rt-pro family — \"u3-pro\" is normalized below — and is opt-in via the user)\n const rawModel = opts.speechModel ?? defaultSTTOptions.speechModel;\n const supportsCarryover = isU3ProModel(rawModel) || rawModel === 'u3-pro';\n if (opts.agentContextCarryover && !supportsCarryover) {\n log().warn(\n `agentContextCarryover is enabled but model '${rawModel}' does not support it; ignoring`,\n );\n }\n super({\n streaming: true,\n interimResults: true,\n alignedTranscript: 'word',\n keyterms: true,\n chatContext: (opts.agentContextCarryover ?? false) && supportsCarryover,\n });\n\n if (opts.speechModel === 'u3-pro') {\n log().warn(\"'u3-pro' is deprecated, use 'universal-3-5-pro' instead.\");\n opts.speechModel = 'universal-3-5-pro';\n }\n\n const speechModel = opts.speechModel ?? defaultSTTOptions.speechModel;\n if (!isU3ProModel(speechModel)) {\n for (const param of [\n 'prompt',\n 'agentContext',\n 'previousContextNTurns',\n 'voiceFocus',\n 'voiceFocusThreshold',\n 'mode',\n ] as const) {\n if (opts[param] !== undefined) {\n throw new Error(\n `The '${param}' parameter is only supported with the ${U3_PRO_MODELS.join(', ')} models.`,\n );\n }\n }\n }\n\n const apiKey = opts.apiKey ?? defaultSTTOptions.apiKey;\n if (!apiKey) {\n throw new Error(\n 'AssemblyAI API key is required. Pass one in via the `apiKey` parameter, or set it as the `ASSEMBLYAI_API_KEY` environment variable',\n );\n }\n\n // Minimize latency by default, but let AssemblyAI's mode preset control silence tuning.\n const minTurnSilence = opts.minTurnSilence ?? (opts.mode === undefined ? 100 : undefined);\n\n this.#opts = {\n ...defaultSTTOptions,\n ...opts,\n apiKey,\n minTurnSilence,\n };\n this.#userKeyterms = [...(this.#opts.keytermsPrompt ?? [])];\n }\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n async _recognize(_: AudioBuffer): Promise<stt.SpeechEvent> {\n throw new Error('Non-streaming recognize is not supported on AssemblyAI STT');\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n // session keyterms so a user update doesn't drop them)\n const nextOpts = { ...opts };\n if (nextOpts.keytermsPrompt !== undefined) {\n this.#userKeyterms = [...nextOpts.keytermsPrompt];\n nextOpts.keytermsPrompt = [...new Set([...this.#userKeyterms, ...this.#sessionKeyterms])];\n }\n this.#opts = { ...this.#opts, ...nextOpts };\n for (const ref of this.#streams) {\n const stream = ref.deref();\n if (stream) {\n stream.updateOptions(nextOpts);\n } else {\n this.#streams.delete(ref);\n }\n }\n }\n\n override _updateSessionKeyterms(keyterms: string[]): void {\n if (\n keyterms.length === this.#sessionKeyterms.length &&\n keyterms.every((t, i) => t === this.#sessionKeyterms[i])\n ) {\n return;\n }\n this.#sessionKeyterms = [...keyterms];\n const merged = [...new Set([...this.#userKeyterms, ...keyterms])];\n this.#opts.keytermsPrompt = merged;\n // applied live via the stream's UpdateConfiguration (no reconnect)\n for (const ref of this.#streams) {\n const stream = ref.deref();\n if (stream) {\n stream.updateOptions({ keytermsPrompt: merged });\n } else {\n this.#streams.delete(ref);\n }\n }\n }\n\n override _pushConversationItem(ev: ConversationItemAddedEvent): void {\n const chatItem = ev.item;\n if (chatItem instanceof ChatMessage && chatItem.role === 'assistant' && chatItem.textContent) {\n this.updateOptions({ agentContext: chatItem.textContent });\n }\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): SpeechStream {\n const stream = new SpeechStream(this, this.#opts, options?.connOptions);\n this.#streams.add(new WeakRef(stream));\n return stream;\n }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n static readonly CLOSE_MSG = JSON.stringify({ type: 'Terminate' });\n\n #opts: STTOptions;\n #logger = log();\n #speechDurationInS = 0;\n #lastPreflightStartTime = 0;\n #pendingConfigMessages: Record<string, unknown>[] = [];\n #configMessagePending = new Future();\n #sessionId: string | null = null;\n #expiresAt: number | null = null;\n label = 'assemblyai.SpeechStream';\n\n constructor(stt: STT, opts: STTOptions, connOptions?: APIConnectOptions) {\n super(stt, opts.sampleRate, connOptions);\n this.#opts = opts;\n this.closed = false;\n }\n\n /**\n * The AssemblyAI session ID. Set when the WebSocket connection is established\n * (before any speech events). Null until the connection completes.\n * Share this with the AssemblyAI team when reporting issues.\n */\n get sessionId(): string | null {\n return this.#sessionId;\n }\n\n /**\n * Unix timestamp when the AssemblyAI session expires. Set alongside\n * {@link sessionId} when the WebSocket connection is established.\n */\n get expiresAt(): number | null {\n return this.#expiresAt;\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n\n const configMsg: Record<string, unknown> = { type: 'UpdateConfiguration' };\n if (opts.prompt !== undefined) configMsg.prompt = opts.prompt;\n if (opts.agentContext !== undefined) configMsg.agent_context = opts.agentContext;\n if (opts.keytermsPrompt !== undefined) configMsg.keyterms_prompt = opts.keytermsPrompt;\n if (opts.maxTurnSilence !== undefined) configMsg.max_turn_silence = opts.maxTurnSilence;\n if (opts.minTurnSilence !== undefined) configMsg.min_turn_silence = opts.minTurnSilence;\n if (opts.endOfTurnConfidenceThreshold !== undefined) {\n configMsg.end_of_turn_confidence_threshold = opts.endOfTurnConfidenceThreshold;\n }\n if (opts.vadThreshold !== undefined) configMsg.vad_threshold = opts.vadThreshold;\n\n // Only send if any actual fields (besides `type`) were specified.\n if (Object.keys(configMsg).length > 1) {\n this.#pendingConfigMessages.push(configMsg);\n if (!this.#configMessagePending.done) this.#configMessagePending.resolve();\n }\n }\n\n /**\n * Force-finalize the current turn immediately.\n */\n forceEndpoint() {\n this.#pendingConfigMessages.push({ type: 'ForceEndpoint' });\n if (!this.#configMessagePending.done) this.#configMessagePending.resolve();\n }\n\n // Deepgram-style reconnect loop around a single websocket lifetime.\n protected async run() {\n const maxRetry = 32;\n let retries = 0;\n\n while (!this.input.closed && !this.closed) {\n try {\n const ws = await this.#connectWS();\n await this.#runWS(ws);\n retries = 0;\n } catch (e) {\n if (!this.closed && !this.input.closed) {\n if (retries >= maxRetry) {\n throw new Error(`failed to connect to AssemblyAI after ${retries} attempts: ${e}`);\n }\n\n const retryDelaySeconds = Math.min(retries * 5, 10);\n retries++;\n\n this.#logger.warn(\n `failed to connect to AssemblyAI, retrying in ${retryDelaySeconds} seconds: ${e} (${retries}/${maxRetry})`,\n );\n await delay(retryDelaySeconds * 1000);\n } else {\n this.#logger.warn(\n `AssemblyAI disconnected, connection is closed: ${e} (inputClosed: ${this.input.closed}, isClosed: ${this.closed})`,\n );\n }\n }\n }\n\n this.closed = true;\n }\n\n async #connectWS(): Promise<WebSocket> {\n // Universal-3 Pro family models default both min and max silence to 100ms when unset.\n // When a mode preset is selected, leave them unset unless explicitly provided so the\n // server's per-mode silence tuning is not overridden by the latency-optimized default.\n let minSilence = this.#opts.minTurnSilence;\n let maxSilence = this.#opts.maxTurnSilence;\n if (isU3ProModel(this.#opts.speechModel)) {\n if (minSilence === undefined && this.#opts.mode === undefined) minSilence = 100;\n if (maxSilence === undefined) maxSilence = minSilence;\n }\n\n // Default language_detection to true for multilingual / u3-rt-pro-family models, false otherwise.\n const defaultLanguageDetection =\n this.#opts.speechModel.includes('multilingual') || isU3ProModel(this.#opts.speechModel);\n const languageDetection = this.#opts.languageDetection ?? defaultLanguageDetection;\n\n const liveConfig: Record<string, unknown> = {\n sample_rate: this.#opts.sampleRate,\n encoding: this.#opts.encoding,\n speech_model: this.#opts.speechModel,\n format_turns: this.#opts.formatTurns,\n end_of_turn_confidence_threshold: this.#opts.endOfTurnConfidenceThreshold,\n min_turn_silence: minSilence,\n max_turn_silence: maxSilence,\n keyterms_prompt:\n this.#opts.keytermsPrompt !== undefined && this.#opts.keytermsPrompt.length > 0\n ? JSON.stringify(this.#opts.keytermsPrompt)\n : undefined,\n language_detection: languageDetection,\n inactivity_timeout: this.#opts.inactivityTimeout,\n prompt: this.#opts.prompt,\n agent_context: this.#opts.agentContext,\n previous_context_n_turns: this.#opts.previousContextNTurns,\n vad_threshold: this.#opts.vadThreshold,\n speaker_labels: this.#opts.speakerLabels,\n max_speakers: this.#opts.maxSpeakers,\n domain: this.#opts.domain,\n voice_focus: this.#opts.voiceFocus,\n voice_focus_threshold: this.#opts.voiceFocusThreshold,\n mode: this.#opts.mode,\n };\n\n const url = new URL(`${this.#opts.baseUrl}/v3/ws`);\n // Python serializes booleans as the strings \"true\"/\"false\", so we mirror that.\n for (const [key, value] of Object.entries(liveConfig)) {\n if (value === undefined || value === null) continue;\n if (typeof value === 'boolean') {\n url.searchParams.append(key, value ? 'true' : 'false');\n } else {\n url.searchParams.append(key, String(value));\n }\n }\n\n const ws = new WebSocket(url, {\n headers: {\n Authorization: this.#opts.apiKey!,\n 'Content-Type': 'application/json',\n 'User-Agent': 'AssemblyAI/1.0 (integration=Livekit)',\n },\n });\n\n await new Promise<void>((resolve, reject) => {\n ws.on('open', () => resolve());\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(new Error(`WebSocket returned ${code}`)));\n });\n\n return ws;\n }\n\n async #runWS(ws: WebSocket) {\n let closing = false;\n const sessionController = new AbortController();\n\n // gets cancelled also when sendTask is complete\n const wsMonitor = Task.from(async (controller) => {\n const closed = new Promise<void>((_, reject) => {\n ws.once('close', (code, reason) => {\n if (!closing) {\n this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);\n reject(new Error('WebSocket closed'));\n }\n });\n });\n\n await Promise.race([closed, waitForAbort(controller.signal)]);\n });\n\n const sendTask = async () => {\n const samplesPerBuffer = Math.floor((this.#opts.sampleRate * this.#opts.bufferSizeMs) / 1000);\n const audioStream = new AudioByteStream(this.#opts.sampleRate, 1, samplesPerBuffer);\n\n const abortPromise = waitForAbort(this.abortSignal);\n const sessionAbort = waitForAbort(sessionController.signal);\n\n try {\n while (!this.closed) {\n const result = await Promise.race([this.input.next(), abortPromise, sessionAbort]);\n\n if (result === undefined) return; // aborted\n if (result.done) break;\n\n const data = result.value;\n\n let frames: AudioFrame[];\n if (data === SpeechStream.FLUSH_SENTINEL) {\n frames = audioStream.flush();\n } else if (data.sampleRate === this.#opts.sampleRate && data.channels === 1) {\n // AssemblyAI expects mono PCM. The base SpeechStream only resamples\n // sample rate, so reject any frame that is not already downmixed.\n frames = audioStream.write(data.data.buffer as ArrayBuffer);\n } else {\n throw new Error('sample rate or channel count of frame does not match');\n }\n\n for (const frame of frames) {\n this.#speechDurationInS += frame.samplesPerChannel / frame.sampleRate;\n ws.send(frame.data.buffer);\n }\n }\n } finally {\n closing = true;\n try {\n ws.send(SpeechStream.CLOSE_MSG);\n } catch {\n // ignore — socket may already be closing\n }\n wsMonitor.cancel();\n }\n };\n\n let messageHandler: ((msg: RawData, isBinary: boolean) => void) | null = null;\n const listenTask = Task.from(async (controller) => {\n const listenMessage = new Promise<void>((resolve, reject) => {\n messageHandler = (msg, isBinary) => {\n if (isBinary) {\n this.#logger.error('unexpected binary message from AssemblyAI');\n return;\n }\n try {\n const json = JSON.parse(msg.toString()) as StreamEventMessage;\n this.#processStreamEvent(json);\n if (this.closed || closing) {\n resolve();\n }\n } catch (err) {\n this.#logger.error(`AssemblyAI: error processing message: ${msg}`);\n reject(err);\n }\n };\n ws.on('message', messageHandler);\n });\n\n await Promise.race([listenMessage, waitForAbort(controller.signal)]);\n });\n\n const configTask = Task.from(async (controller) => {\n // Drain any messages queued while the socket was reconnecting.\n while (this.#pendingConfigMessages.length > 0) {\n const msg = this.#pendingConfigMessages.shift()!;\n ws.send(JSON.stringify(msg));\n }\n\n while (!controller.signal.aborted) {\n await Promise.race([this.#configMessagePending.await, waitForAbort(controller.signal)]);\n if (controller.signal.aborted) return;\n\n this.#configMessagePending = new Future();\n while (this.#pendingConfigMessages.length > 0) {\n const msg = this.#pendingConfigMessages.shift()!;\n ws.send(JSON.stringify(msg));\n }\n }\n });\n\n try {\n await Promise.all([sendTask(), listenTask.result, wsMonitor.result]);\n } finally {\n closing = true;\n sessionController.abort();\n listenTask.cancel();\n configTask.cancel();\n if (messageHandler) ws.off('message', messageHandler);\n try {\n ws.close();\n } catch {\n // ignore\n }\n }\n }\n\n #averageConfidence(words: Array<{ confidence?: number }>): number {\n if (words.length === 0) return 0;\n return words.reduce((sum, w) => sum + (w.confidence ?? 0), 0) / words.length;\n }\n\n #processStreamEvent(data: StreamEventMessage) {\n const messageType = data.type;\n\n if (messageType === 'Begin') {\n this.#sessionId = data.id ?? null;\n this.#expiresAt = data.expires_at ?? null;\n this.#logger.info(\n `AssemblyAI session started id=${this.#sessionId} expires_at=${this.#expiresAt}`,\n );\n return;\n }\n\n if (messageType === 'SpeechStarted') {\n this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n return;\n }\n\n if (messageType === 'Termination') {\n this.#logger.debug(\n `AssemblyAI session terminated audio_duration=${data.audio_duration_seconds}s session_duration=${data.session_duration_seconds}s`,\n );\n return;\n }\n\n if (messageType !== 'Turn') {\n return;\n }\n\n const words = data.words ?? [];\n const endOfTurn = Boolean(data.end_of_turn);\n const turnIsFormatted = Boolean(data.turn_is_formatted);\n const utterance = data.utterance ?? '';\n const transcript = data.transcript ?? '';\n const language = normalizeLanguage(data.language_code ?? 'en');\n const metadata = speechDataMetadata(data);\n\n // Word timestamps are in milliseconds:\n // https://www.assemblyai.com/docs/api-reference/streaming-api/streaming-api#receive.receiveTurn.words\n const timedWords = words.map((word) =>\n createTimedString({\n text: word.text ?? '',\n startTime: (word.start ?? 0) / 1000 + this.startTimeOffset,\n endTime: (word.end ?? 0) / 1000 + this.startTimeOffset,\n confidence: word.confidence ?? 0,\n startTimeOffset: this.startTimeOffset,\n }),\n );\n\n let startTime = 0;\n let endTime = 0;\n let confidence = 0;\n\n // `words` are cumulative for the turn — emit as an interim transcript.\n if (timedWords.length > 0) {\n const interimText = timedWords.map((w) => w.text).join(' ');\n startTime = timedWords[0]!.startTime ?? 0;\n endTime = timedWords[timedWords.length - 1]!.endTime ?? 0;\n confidence = this.#averageConfidence(timedWords);\n\n this.queue.put({\n type: stt.SpeechEventType.INTERIM_TRANSCRIPT,\n alternatives: [\n {\n language,\n text: interimText,\n startTime,\n endTime,\n confidence,\n words: timedWords,\n ...(metadata ? { metadata } : {}),\n },\n ],\n });\n }\n\n // `utterance` is chunk-based (not cumulative) — emit as a preflight transcript\n // covering only the words since the last preflight.\n if (utterance) {\n if (this.#lastPreflightStartTime === 0) {\n this.#lastPreflightStartTime = startTime;\n }\n\n const utteranceWords = timedWords.filter(\n (w) => w.startTime !== undefined && w.startTime >= this.#lastPreflightStartTime,\n );\n const utteranceConfidence = this.#averageConfidence(utteranceWords);\n\n this.queue.put({\n type: stt.SpeechEventType.PREFLIGHT_TRANSCRIPT,\n alternatives: [\n {\n language,\n text: utterance,\n startTime: this.#lastPreflightStartTime,\n endTime,\n confidence: utteranceConfidence,\n words: utteranceWords,\n ...(metadata ? { metadata } : {}),\n },\n ],\n });\n this.#lastPreflightStartTime = endTime;\n }\n\n // End-of-turn: emit FINAL_TRANSCRIPT + END_OF_SPEECH.\n // If the user asked for formatted turns, wait for a formatted final.\n const waitingForFormatted = this.#opts.formatTurns === true && !turnIsFormatted;\n if (endOfTurn && !waitingForFormatted) {\n this.queue.put({\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n alternatives: [\n {\n language,\n text: transcript,\n startTime,\n endTime,\n confidence,\n words: timedWords,\n ...(metadata ? { metadata } : {}),\n },\n ],\n });\n\n this.queue.put({ type: stt.SpeechEventType.END_OF_SPEECH });\n if (this.#speechDurationInS > 0) {\n this.queue.put({\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n // Propagate the AssemblyAI session id as the request id so metrics\n // can be correlated back to a specific connection, mirroring how\n // Deepgram surfaces its `request_id`.\n requestId: this.#sessionId ?? undefined,\n recognitionUsage: {\n audioDuration: this.#speechDurationInS,\n },\n });\n this.#speechDurationInS = 0;\n this.#lastPreflightStartTime = 0;\n }\n }\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,oBAcO;AAGP,gBAA0B;AAI1B,MAAM,gBAAgB,CAAC,aAAa,oBAAoB,mBAAmB;AAE3E,SAAS,aAAa,OAA2B;AAC/C,SAAO,cAAc,SAAS,KAAuC;AACvE;AA8BA,SAAS,mBAAmB,MAAkE;AAC5F,QAAM,aAAqC,CAAC;AAE5C,MAAI,OAAO,KAAK,wBAAwB,UAAU;AAChD,eAAW,qBAAqB,KAAK;AAAA,EACvC;AAEA,MAAI,OAAO,KAAK,UAAU,EAAE,WAAW,EAAG,QAAO;AAEjD,SAAO,EAAE,WAAW;AACtB;AA+DA,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,YAAY;AAAA,EACZ,cAAc;AAAA,EACd,UAAU;AAAA,EACV,aAAa;AAAA,EACb,SAAS;AACX;AAEO,MAAM,YAAY,kBAAI,IAAI;AAAA,EAC/B;AAAA,EACA,WAAW,oBAAI,IAA2B;AAAA;AAAA,EAE1C;AAAA,EACA,mBAA6B,CAAC;AAAA,EAC9B,QAAQ;AAAA,EAER,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,YAAY,OAA4B,CAAC,GAAG;AAE1C,UAAM,WAAW,KAAK,eAAe,kBAAkB;AACvD,UAAM,oBAAoB,aAAa,QAAQ,KAAK,aAAa;AACjE,QAAI,KAAK,yBAAyB,CAAC,mBAAmB;AACpD,6BAAI,EAAE;AAAA,QACJ,+CAA+C,QAAQ;AAAA,MACzD;AAAA,IACF;AACA,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB;AAAA,MAChB,mBAAmB;AAAA,MACnB,UAAU;AAAA,MACV,cAAc,KAAK,yBAAyB,UAAU;AAAA,IACxD,CAAC;AAED,QAAI,KAAK,gBAAgB,UAAU;AACjC,6BAAI,EAAE,KAAK,0DAA0D;AACrE,WAAK,cAAc;AAAA,IACrB;AAEA,UAAM,cAAc,KAAK,eAAe,kBAAkB;AAC1D,QAAI,CAAC,aAAa,WAAW,GAAG;AAC9B,iBAAW,SAAS;AAAA,QAClB;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MACF,GAAY;AACV,YAAI,KAAK,KAAK,MAAM,QAAW;AAC7B,gBAAM,IAAI;AAAA,YACR,QAAQ,KAAK,0CAA0C,cAAc,KAAK,IAAI,CAAC;AAAA,UACjF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,SAAS,KAAK,UAAU,kBAAkB;AAChD,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAGA,UAAM,iBAAiB,KAAK,mBAAmB,KAAK,SAAS,SAAY,MAAM;AAE/E,SAAK,QAAQ;AAAA,MACX,GAAG;AAAA,MACH,GAAG;AAAA,MACH;AAAA,MACA;AAAA,IACF;AACA,SAAK,gBAAgB,CAAC,GAAI,KAAK,MAAM,kBAAkB,CAAC,CAAE;AAAA,EAC5D;AAAA;AAAA,EAGA,MAAM,WAAW,GAA0C;AACzD,UAAM,IAAI,MAAM,4DAA4D;AAAA,EAC9E;AAAA,EAEA,cAAc,MAA2B;AAEvC,UAAM,WAAW,EAAE,GAAG,KAAK;AAC3B,QAAI,SAAS,mBAAmB,QAAW;AACzC,WAAK,gBAAgB,CAAC,GAAG,SAAS,cAAc;AAChD,eAAS,iBAAiB,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAG,KAAK,eAAe,GAAG,KAAK,gBAAgB,CAAC,CAAC;AAAA,IAC1F;AACA,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,SAAS;AAC1C,eAAW,OAAO,KAAK,UAAU;AAC/B,YAAM,SAAS,IAAI,MAAM;AACzB,UAAI,QAAQ;AACV,eAAO,cAAc,QAAQ;AAAA,MAC/B,OAAO;AACL,aAAK,SAAS,OAAO,GAAG;AAAA,MAC1B;AAAA,IACF;AAAA,EACF;AAAA,EAES,uBAAuB,UAA0B;AACxD,QACE,SAAS,WAAW,KAAK,iBAAiB,UAC1C,SAAS,MAAM,CAAC,GAAG,MAAM,MAAM,KAAK,iBAAiB,CAAC,CAAC,GACvD;AACA;AAAA,IACF;AACA,SAAK,mBAAmB,CAAC,GAAG,QAAQ;AACpC,UAAM,SAAS,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAG,KAAK,eAAe,GAAG,QAAQ,CAAC,CAAC;AAChE,SAAK,MAAM,iBAAiB;AAE5B,eAAW,OAAO,KAAK,UAAU;AAC/B,YAAM,SAAS,IAAI,MAAM;AACzB,UAAI,QAAQ;AACV,eAAO,cAAc,EAAE,gBAAgB,OAAO,CAAC;AAAA,MACjD,OAAO;AACL,aAAK,SAAS,OAAO,GAAG;AAAA,MAC1B;AAAA,IACF;AAAA,EACF;AAAA,EAES,sBAAsB,IAAsC;AACnE,UAAM,WAAW,GAAG;AACpB,QAAI,oBAAoB,6BAAe,SAAS,SAAS,eAAe,SAAS,aAAa;AAC5F,WAAK,cAAc,EAAE,cAAc,SAAS,YAAY,CAAC;AAAA,IAC3D;AAAA,EACF;AAAA,EAEA,OAAO,SAA6D;AAClE,UAAM,SAAS,IAAI,aAAa,MAAM,KAAK,OAAO,mCAAS,WAAW;AACtE,SAAK,SAAS,IAAI,IAAI,QAAQ,MAAM,CAAC;AACrC,WAAO;AAAA,EACT;AACF;AAEO,MAAM,qBAAqB,kBAAI,aAAa;AAAA,EACjD,OAAgB,YAAY,KAAK,UAAU,EAAE,MAAM,YAAY,CAAC;AAAA,EAEhE;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,qBAAqB;AAAA,EACrB,0BAA0B;AAAA,EAC1B,yBAAoD,CAAC;AAAA,EACrD,wBAAwB,IAAI,qBAAO;AAAA,EACnC,aAA4B;AAAA,EAC5B,aAA4B;AAAA,EAC5B,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB,aAAiC;AACvE,UAAMA,MAAK,KAAK,YAAY,WAAW;AACvC,SAAK,QAAQ;AACb,SAAK,SAAS;AAAA,EAChB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,IAAI,YAA2B;AAC7B,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,IAAI,YAA2B;AAC7B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAEtC,UAAM,YAAqC,EAAE,MAAM,sBAAsB;AACzE,QAAI,KAAK,WAAW,OAAW,WAAU,SAAS,KAAK;AACvD,QAAI,KAAK,iBAAiB,OAAW,WAAU,gBAAgB,KAAK;AACpE,QAAI,KAAK,mBAAmB,OAAW,WAAU,kBAAkB,KAAK;AACxE,QAAI,KAAK,mBAAmB,OAAW,WAAU,mBAAmB,KAAK;AACzE,QAAI,KAAK,mBAAmB,OAAW,WAAU,mBAAmB,KAAK;AACzE,QAAI,KAAK,iCAAiC,QAAW;AACnD,gBAAU,mCAAmC,KAAK;AAAA,IACpD;AACA,QAAI,KAAK,iBAAiB,OAAW,WAAU,gBAAgB,KAAK;AAGpE,QAAI,OAAO,KAAK,SAAS,EAAE,SAAS,GAAG;AACrC,WAAK,uBAAuB,KAAK,SAAS;AAC1C,UAAI,CAAC,KAAK,sBAAsB,KAAM,MAAK,sBAAsB,QAAQ;AAAA,IAC3E;AAAA,EACF;AAAA;AAAA;AAAA;AAAA,EAKA,gBAAgB;AACd,SAAK,uBAAuB,KAAK,EAAE,MAAM,gBAAgB,CAAC;AAC1D,QAAI,CAAC,KAAK,sBAAsB,KAAM,MAAK,sBAAsB,QAAQ;AAAA,EAC3E;AAAA;AAAA,EAGA,MAAgB,MAAM;AACpB,UAAM,WAAW;AACjB,QAAI,UAAU;AAEd,WAAO,CAAC,KAAK,MAAM,UAAU,CAAC,KAAK,QAAQ;AACzC,UAAI;AACF,cAAM,KAAK,MAAM,KAAK,WAAW;AACjC,cAAM,KAAK,OAAO,EAAE;AACpB,kBAAU;AAAA,MACZ,SAAS,GAAG;AACV,YAAI,CAAC,KAAK,UAAU,CAAC,KAAK,MAAM,QAAQ;AACtC,cAAI,WAAW,UAAU;AACvB,kBAAM,IAAI,MAAM,yCAAyC,OAAO,cAAc,CAAC,EAAE;AAAA,UACnF;AAEA,gBAAM,oBAAoB,KAAK,IAAI,UAAU,GAAG,EAAE;AAClD;AAEA,eAAK,QAAQ;AAAA,YACX,gDAAgD,iBAAiB,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ;AAAA,UACzG;AACA,oBAAM,qBAAM,oBAAoB,GAAI;AAAA,QACtC,OAAO;AACL,eAAK,QAAQ;AAAA,YACX,kDAAkD,CAAC,kBAAkB,KAAK,MAAM,MAAM,eAAe,KAAK,MAAM;AAAA,UAClH;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAM,aAAiC;AAIrC,QAAI,aAAa,KAAK,MAAM;AAC5B,QAAI,aAAa,KAAK,MAAM;AAC5B,QAAI,aAAa,KAAK,MAAM,WAAW,GAAG;AACxC,UAAI,eAAe,UAAa,KAAK,MAAM,SAAS,OAAW,cAAa;AAC5E,UAAI,eAAe,OAAW,cAAa;AAAA,IAC7C;AAGA,UAAM,2BACJ,KAAK,MAAM,YAAY,SAAS,cAAc,KAAK,aAAa,KAAK,MAAM,WAAW;AACxF,UAAM,oBAAoB,KAAK,MAAM,qBAAqB;AAE1D,UAAM,aAAsC;AAAA,MAC1C,aAAa,KAAK,MAAM;AAAA,MACxB,UAAU,KAAK,MAAM;AAAA,MACrB,cAAc,KAAK,MAAM;AAAA,MACzB,cAAc,KAAK,MAAM;AAAA,MACzB,kCAAkC,KAAK,MAAM;AAAA,MAC7C,kBAAkB;AAAA,MAClB,kBAAkB;AAAA,MAClB,iBACE,KAAK,MAAM,mBAAmB,UAAa,KAAK,MAAM,eAAe,SAAS,IAC1E,KAAK,UAAU,KAAK,MAAM,cAAc,IACxC;AAAA,MACN,oBAAoB;AAAA,MACpB,oBAAoB,KAAK,MAAM;AAAA,MAC/B,QAAQ,KAAK,MAAM;AAAA,MACnB,eAAe,KAAK,MAAM;AAAA,MAC1B,0BAA0B,KAAK,MAAM;AAAA,MACrC,eAAe,KAAK,MAAM;AAAA,MAC1B,gBAAgB,KAAK,MAAM;AAAA,MAC3B,cAAc,KAAK,MAAM;AAAA,MACzB,QAAQ,KAAK,MAAM;AAAA,MACnB,aAAa,KAAK,MAAM;AAAA,MACxB,uBAAuB,KAAK,MAAM;AAAA,MAClC,MAAM,KAAK,MAAM;AAAA,IACnB;AAEA,UAAM,MAAM,IAAI,IAAI,GAAG,KAAK,MAAM,OAAO,QAAQ;AAEjD,eAAW,CAAC,KAAK,KAAK,KAAK,OAAO,QAAQ,UAAU,GAAG;AACrD,UAAI,UAAU,UAAa,UAAU,KAAM;AAC3C,UAAI,OAAO,UAAU,WAAW;AAC9B,YAAI,aAAa,OAAO,KAAK,QAAQ,SAAS,OAAO;AAAA,MACvD,OAAO;AACL,YAAI,aAAa,OAAO,KAAK,OAAO,KAAK,CAAC;AAAA,MAC5C;AAAA,IACF;AAEA,UAAM,KAAK,IAAI,oBAAU,KAAK;AAAA,MAC5B,SAAS;AAAA,QACP,eAAe,KAAK,MAAM;AAAA,QAC1B,gBAAgB;AAAA,QAChB,cAAc;AAAA,MAChB;AAAA,IACF,CAAC;AAED,UAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,SAAG,GAAG,QAAQ,MAAM,QAAQ,CAAC;AAC7B,SAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,SAAG,GAAG,SAAS,CAAC,SAAS,OAAO,IAAI,MAAM,sBAAsB,IAAI,EAAE,CAAC,CAAC;AAAA,IAC1E,CAAC;AAED,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,OAAO,IAAe;AAC1B,QAAI,UAAU;AACd,UAAM,oBAAoB,IAAI,gBAAgB;AAG9C,UAAM,YAAY,mBAAK,KAAK,OAAO,eAAe;AAChD,YAAM,SAAS,IAAI,QAAc,CAAC,GAAG,WAAW;AAC9C,WAAG,KAAK,SAAS,CAAC,MAAM,WAAW;AACjC,cAAI,CAAC,SAAS;AACZ,iBAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,MAAM,EAAE;AAClE,mBAAO,IAAI,MAAM,kBAAkB,CAAC;AAAA,UACtC;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,YAAQ,4BAAa,WAAW,MAAM,CAAC,CAAC;AAAA,IAC9D,CAAC;AAED,UAAM,WAAW,YAAY;AAC3B,YAAM,mBAAmB,KAAK,MAAO,KAAK,MAAM,aAAa,KAAK,MAAM,eAAgB,GAAI;AAC5F,YAAM,cAAc,IAAI,8BAAgB,KAAK,MAAM,YAAY,GAAG,gBAAgB;AAElF,YAAM,mBAAe,4BAAa,KAAK,WAAW;AAClD,YAAM,mBAAe,4BAAa,kBAAkB,MAAM;AAE1D,UAAI;AACF,eAAO,CAAC,KAAK,QAAQ;AACnB,gBAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,KAAK,MAAM,KAAK,GAAG,cAAc,YAAY,CAAC;AAEjF,cAAI,WAAW,OAAW;AAC1B,cAAI,OAAO,KAAM;AAEjB,gBAAM,OAAO,OAAO;AAEpB,cAAI;AACJ,cAAI,SAAS,aAAa,gBAAgB;AACxC,qBAAS,YAAY,MAAM;AAAA,UAC7B,WAAW,KAAK,eAAe,KAAK,MAAM,cAAc,KAAK,aAAa,GAAG;AAG3E,qBAAS,YAAY,MAAM,KAAK,KAAK,MAAqB;AAAA,UAC5D,OAAO;AACL,kBAAM,IAAI,MAAM,sDAAsD;AAAA,UACxE;AAEA,qBAAW,SAAS,QAAQ;AAC1B,iBAAK,sBAAsB,MAAM,oBAAoB,MAAM;AAC3D,eAAG,KAAK,MAAM,KAAK,MAAM;AAAA,UAC3B;AAAA,QACF;AAAA,MACF,UAAE;AACA,kBAAU;AACV,YAAI;AACF,aAAG,KAAK,aAAa,SAAS;AAAA,QAChC,QAAQ;AAAA,QAER;AACA,kBAAU,OAAO;AAAA,MACnB;AAAA,IACF;AAEA,QAAI,iBAAqE;AACzE,UAAM,aAAa,mBAAK,KAAK,OAAO,eAAe;AACjD,YAAM,gBAAgB,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3D,yBAAiB,CAAC,KAAK,aAAa;AAClC,cAAI,UAAU;AACZ,iBAAK,QAAQ,MAAM,2CAA2C;AAC9D;AAAA,UACF;AACA,cAAI;AACF,kBAAM,OAAO,KAAK,MAAM,IAAI,SAAS,CAAC;AACtC,iBAAK,oBAAoB,IAAI;AAC7B,gBAAI,KAAK,UAAU,SAAS;AAC1B,sBAAQ;AAAA,YACV;AAAA,UACF,SAAS,KAAK;AACZ,iBAAK,QAAQ,MAAM,yCAAyC,GAAG,EAAE;AACjE,mBAAO,GAAG;AAAA,UACZ;AAAA,QACF;AACA,WAAG,GAAG,WAAW,cAAc;AAAA,MACjC,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,mBAAe,4BAAa,WAAW,MAAM,CAAC,CAAC;AAAA,IACrE,CAAC;AAED,UAAM,aAAa,mBAAK,KAAK,OAAO,eAAe;AAEjD,aAAO,KAAK,uBAAuB,SAAS,GAAG;AAC7C,cAAM,MAAM,KAAK,uBAAuB,MAAM;AAC9C,WAAG,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,MAC7B;AAEA,aAAO,CAAC,WAAW,OAAO,SAAS;AACjC,cAAM,QAAQ,KAAK,CAAC,KAAK,sBAAsB,WAAO,4BAAa,WAAW,MAAM,CAAC,CAAC;AACtF,YAAI,WAAW,OAAO,QAAS;AAE/B,aAAK,wBAAwB,IAAI,qBAAO;AACxC,eAAO,KAAK,uBAAuB,SAAS,GAAG;AAC7C,gBAAM,MAAM,KAAK,uBAAuB,MAAM;AAC9C,aAAG,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,QAC7B;AAAA,MACF;AAAA,IACF,CAAC;AAED,QAAI;AACF,YAAM,QAAQ,IAAI,CAAC,SAAS,GAAG,WAAW,QAAQ,UAAU,MAAM,CAAC;AAAA,IACrE,UAAE;AACA,gBAAU;AACV,wBAAkB,MAAM;AACxB,iBAAW,OAAO;AAClB,iBAAW,OAAO;AAClB,UAAI,eAAgB,IAAG,IAAI,WAAW,cAAc;AACpD,UAAI;AACF,WAAG,MAAM;AAAA,MACX,QAAQ;AAAA,MAER;AAAA,IACF;AAAA,EACF;AAAA,EAEA,mBAAmB,OAA+C;AAChE,QAAI,MAAM,WAAW,EAAG,QAAO;AAC/B,WAAO,MAAM,OAAO,CAAC,KAAK,MAAM,OAAO,EAAE,cAAc,IAAI,CAAC,IAAI,MAAM;AAAA,EACxE;AAAA,EAEA,oBAAoB,MAA0B;AAC5C,UAAM,cAAc,KAAK;AAEzB,QAAI,gBAAgB,SAAS;AAC3B,WAAK,aAAa,KAAK,MAAM;AAC7B,WAAK,aAAa,KAAK,cAAc;AACrC,WAAK,QAAQ;AAAA,QACX,iCAAiC,KAAK,UAAU,eAAe,KAAK,UAAU;AAAA,MAChF;AACA;AAAA,IACF;AAEA,QAAI,gBAAgB,iBAAiB;AACnC,WAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,gBAAgB,CAAC;AAC5D;AAAA,IACF;AAEA,QAAI,gBAAgB,eAAe;AACjC,WAAK,QAAQ;AAAA,QACX,gDAAgD,KAAK,sBAAsB,sBAAsB,KAAK,wBAAwB;AAAA,MAChI;AACA;AAAA,IACF;AAEA,QAAI,gBAAgB,QAAQ;AAC1B;AAAA,IACF;AAEA,UAAM,QAAQ,KAAK,SAAS,CAAC;AAC7B,UAAM,YAAY,QAAQ,KAAK,WAAW;AAC1C,UAAM,kBAAkB,QAAQ,KAAK,iBAAiB;AACtD,UAAM,YAAY,KAAK,aAAa;AACpC,UAAM,aAAa,KAAK,cAAc;AACtC,UAAM,eAAW,iCAAkB,KAAK,iBAAiB,IAAI;AAC7D,UAAM,WAAW,mBAAmB,IAAI;AAIxC,UAAM,aAAa,MAAM;AAAA,MAAI,CAAC,aAC5B,iCAAkB;AAAA,QAChB,MAAM,KAAK,QAAQ;AAAA,QACnB,YAAY,KAAK,SAAS,KAAK,MAAO,KAAK;AAAA,QAC3C,UAAU,KAAK,OAAO,KAAK,MAAO,KAAK;AAAA,QACvC,YAAY,KAAK,cAAc;AAAA,QAC/B,iBAAiB,KAAK;AAAA,MACxB,CAAC;AAAA,IACH;AAEA,QAAI,YAAY;AAChB,QAAI,UAAU;AACd,QAAI,aAAa;AAGjB,QAAI,WAAW,SAAS,GAAG;AACzB,YAAM,cAAc,WAAW,IAAI,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,GAAG;AAC1D,kBAAY,WAAW,CAAC,EAAG,aAAa;AACxC,gBAAU,WAAW,WAAW,SAAS,CAAC,EAAG,WAAW;AACxD,mBAAa,KAAK,mBAAmB,UAAU;AAE/C,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,kBAAI,gBAAgB;AAAA,QAC1B,cAAc;AAAA,UACZ;AAAA,YACE;AAAA,YACA,MAAM;AAAA,YACN;AAAA,YACA;AAAA,YACA;AAAA,YACA,OAAO;AAAA,YACP,GAAI,WAAW,EAAE,SAAS,IAAI,CAAC;AAAA,UACjC;AAAA,QACF;AAAA,MACF,CAAC;AAAA,IACH;AAIA,QAAI,WAAW;AACb,UAAI,KAAK,4BAA4B,GAAG;AACtC,aAAK,0BAA0B;AAAA,MACjC;AAEA,YAAM,iBAAiB,WAAW;AAAA,QAChC,CAAC,MAAM,EAAE,cAAc,UAAa,EAAE,aAAa,KAAK;AAAA,MAC1D;AACA,YAAM,sBAAsB,KAAK,mBAAmB,cAAc;AAElE,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,kBAAI,gBAAgB;AAAA,QAC1B,cAAc;AAAA,UACZ;AAAA,YACE;AAAA,YACA,MAAM;AAAA,YACN,WAAW,KAAK;AAAA,YAChB;AAAA,YACA,YAAY;AAAA,YACZ,OAAO;AAAA,YACP,GAAI,WAAW,EAAE,SAAS,IAAI,CAAC;AAAA,UACjC;AAAA,QACF;AAAA,MACF,CAAC;AACD,WAAK,0BAA0B;AAAA,IACjC;AAIA,UAAM,sBAAsB,KAAK,MAAM,gBAAgB,QAAQ,CAAC;AAChE,QAAI,aAAa,CAAC,qBAAqB;AACrC,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,kBAAI,gBAAgB;AAAA,QAC1B,cAAc;AAAA,UACZ;AAAA,YACE;AAAA,YACA,MAAM;AAAA,YACN;AAAA,YACA;AAAA,YACA;AAAA,YACA,OAAO;AAAA,YACP,GAAI,WAAW,EAAE,SAAS,IAAI,CAAC;AAAA,UACjC;AAAA,QACF;AAAA,MACF,CAAC;AAED,WAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,cAAc,CAAC;AAC1D,UAAI,KAAK,qBAAqB,GAAG;AAC/B,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,kBAAI,gBAAgB;AAAA;AAAA;AAAA;AAAA,UAI1B,WAAW,KAAK,cAAc;AAAA,UAC9B,kBAAkB;AAAA,YAChB,eAAe,KAAK;AAAA,UACtB;AAAA,QACF,CAAC;AACD,aAAK,qBAAqB;AAC1B,aAAK,0BAA0B;AAAA,MACjC;AAAA,IACF;AAAA,EACF;AACF;","names":["stt"]}
|
package/dist/stt.d.cts
CHANGED
|
@@ -1,4 +1,4 @@
|
|
|
1
|
-
import { type APIConnectOptions, type AudioBuffer, stt } from '@livekit/agents';
|
|
1
|
+
import { type APIConnectOptions, type AudioBuffer, type ConversationItemAddedEvent, stt } from '@livekit/agents';
|
|
2
2
|
import type { STTEncoding, STTModels, VoiceFocus } from './models.js';
|
|
3
3
|
export interface STTOptions {
|
|
4
4
|
apiKey?: string;
|
|
@@ -51,6 +51,13 @@ export interface STTOptions {
|
|
|
51
51
|
* or `max_accuracy`. Explicit turn-silence values still take precedence over mode defaults.
|
|
52
52
|
*/
|
|
53
53
|
mode?: 'min_latency' | 'balanced' | 'max_accuracy';
|
|
54
|
+
/**
|
|
55
|
+
* When the model supports it, let an `AgentSession` push each assistant reply into
|
|
56
|
+
* `agentContext` so it is carried into the model's conversation context. Defaults to false;
|
|
57
|
+
* set true to enable. Prior user turns are carried automatically by the model regardless of
|
|
58
|
+
* this flag. Ignored on models without context support.
|
|
59
|
+
*/
|
|
60
|
+
agentContextCarryover?: boolean;
|
|
54
61
|
baseUrl: string;
|
|
55
62
|
}
|
|
56
63
|
export declare class STT extends stt.STT {
|
|
@@ -61,6 +68,8 @@ export declare class STT extends stt.STT {
|
|
|
61
68
|
constructor(opts?: Partial<STTOptions>);
|
|
62
69
|
_recognize(_: AudioBuffer): Promise<stt.SpeechEvent>;
|
|
63
70
|
updateOptions(opts: Partial<STTOptions>): void;
|
|
71
|
+
_updateSessionKeyterms(keyterms: string[]): void;
|
|
72
|
+
_pushConversationItem(ev: ConversationItemAddedEvent): void;
|
|
64
73
|
stream(options?: {
|
|
65
74
|
connOptions?: APIConnectOptions;
|
|
66
75
|
}): SpeechStream;
|
package/dist/stt.d.ts
CHANGED
|
@@ -1,4 +1,4 @@
|
|
|
1
|
-
import { type APIConnectOptions, type AudioBuffer, stt } from '@livekit/agents';
|
|
1
|
+
import { type APIConnectOptions, type AudioBuffer, type ConversationItemAddedEvent, stt } from '@livekit/agents';
|
|
2
2
|
import type { STTEncoding, STTModels, VoiceFocus } from './models.js';
|
|
3
3
|
export interface STTOptions {
|
|
4
4
|
apiKey?: string;
|
|
@@ -51,6 +51,13 @@ export interface STTOptions {
|
|
|
51
51
|
* or `max_accuracy`. Explicit turn-silence values still take precedence over mode defaults.
|
|
52
52
|
*/
|
|
53
53
|
mode?: 'min_latency' | 'balanced' | 'max_accuracy';
|
|
54
|
+
/**
|
|
55
|
+
* When the model supports it, let an `AgentSession` push each assistant reply into
|
|
56
|
+
* `agentContext` so it is carried into the model's conversation context. Defaults to false;
|
|
57
|
+
* set true to enable. Prior user turns are carried automatically by the model regardless of
|
|
58
|
+
* this flag. Ignored on models without context support.
|
|
59
|
+
*/
|
|
60
|
+
agentContextCarryover?: boolean;
|
|
54
61
|
baseUrl: string;
|
|
55
62
|
}
|
|
56
63
|
export declare class STT extends stt.STT {
|
|
@@ -61,6 +68,8 @@ export declare class STT extends stt.STT {
|
|
|
61
68
|
constructor(opts?: Partial<STTOptions>);
|
|
62
69
|
_recognize(_: AudioBuffer): Promise<stt.SpeechEvent>;
|
|
63
70
|
updateOptions(opts: Partial<STTOptions>): void;
|
|
71
|
+
_updateSessionKeyterms(keyterms: string[]): void;
|
|
72
|
+
_pushConversationItem(ev: ConversationItemAddedEvent): void;
|
|
64
73
|
stream(options?: {
|
|
65
74
|
connOptions?: APIConnectOptions;
|
|
66
75
|
}): SpeechStream;
|
package/dist/stt.d.ts.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"stt.d.ts","sourceRoot":"","sources":["../src/stt.ts"],"names":[],"mappings":"AAIA,OAAO,EACL,KAAK,iBAAiB,EACtB,KAAK,WAAW,
|
|
1
|
+
{"version":3,"file":"stt.d.ts","sourceRoot":"","sources":["../src/stt.ts"],"names":[],"mappings":"AAIA,OAAO,EACL,KAAK,iBAAiB,EACtB,KAAK,WAAW,EAGhB,KAAK,0BAA0B,EAO/B,GAAG,EAEJ,MAAM,iBAAiB,CAAC;AAIzB,OAAO,KAAK,EAAE,WAAW,EAAE,SAAS,EAAE,UAAU,EAAE,MAAM,aAAa,CAAC;AAiDtE,MAAM,WAAW,UAAU;IACzB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,UAAU,EAAE,MAAM,CAAC;IACnB;;;;OAIG;IACH,YAAY,EAAE,MAAM,CAAC;IACrB,QAAQ,EAAE,WAAW,CAAC;IACtB,WAAW,EAAE,SAAS,CAAC;IACvB,iBAAiB,CAAC,EAAE,OAAO,CAAC;IAC5B;;;OAGG;IACH,iBAAiB,CAAC,EAAE,MAAM,CAAC;IAC3B,4BAA4B,CAAC,EAAE,MAAM,CAAC;IACtC,wEAAwE;IACxE,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,kFAAkF;IAClF,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,WAAW,CAAC,EAAE,OAAO,CAAC;IACtB,cAAc,CAAC,EAAE,MAAM,EAAE,CAAC;IAC1B,4DAA4D;IAC5D,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,4DAA4D;IAC5D,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,yFAAyF;IACzF,qBAAqB,CAAC,EAAE,MAAM,CAAC;IAC/B,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB;;;;;;;OAOG;IACH,aAAa,CAAC,EAAE,OAAO,CAAC;IACxB,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,kFAAkF;IAClF,UAAU,CAAC,EAAE,UAAU,CAAC;IACxB,uFAAuF;IACvF,mBAAmB,CAAC,EAAE,MAAM,CAAC;IAC7B;;;OAGG;IACH,IAAI,CAAC,EAAE,aAAa,GAAG,UAAU,GAAG,cAAc,CAAC;IACnD;;;;;OAKG;IACH,qBAAqB,CAAC,EAAE,OAAO,CAAC;IAChC,OAAO,EAAE,MAAM,CAAC;CACjB;AAWD,qBAAa,GAAI,SAAQ,GAAG,CAAC,GAAG;;IAM9B,KAAK,SAAoB;IAEzB,IAAI,KAAK,IAAI,MAAM,CAElB;IAED,IAAI,QAAQ,IAAI,MAAM,CAErB;gBAEW,IAAI,GAAE,OAAO,CAAC,UAAU,CAAM;IA4DpC,UAAU,CAAC,CAAC,EAAE,WAAW,GAAG,OAAO,CAAC,GAAG,CAAC,WAAW,CAAC;IAI1D,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,UAAU,CAAC;IAkB9B,sBAAsB,CAAC,QAAQ,EAAE,MAAM,EAAE,GAAG,IAAI;IAqBhD,qBAAqB,CAAC,EAAE,EAAE,0BAA0B,GAAG,IAAI;IAOpE,MAAM,CAAC,OAAO,CAAC,EAAE;QAAE,WAAW,CAAC,EAAE,iBAAiB,CAAA;KAAE,GAAG,YAAY;CAKpE;AAED,qBAAa,YAAa,SAAQ,GAAG,CAAC,YAAY;;IAChD,MAAM,CAAC,QAAQ,CAAC,SAAS,SAAyC;IAUlE,KAAK,SAA6B;gBAEtB,GAAG,EAAE,GAAG,EAAE,IAAI,EAAE,UAAU,EAAE,WAAW,CAAC,EAAE,iBAAiB;IAMvE;;;;OAIG;IACH,IAAI,SAAS,IAAI,MAAM,GAAG,IAAI,CAE7B;IAED;;;OAGG;IACH,IAAI,SAAS,IAAI,MAAM,GAAG,IAAI,CAE7B;IAED,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,UAAU,CAAC;IAqBvC;;OAEG;IACH,aAAa;cAMG,GAAG;CAiXpB"}
|
package/dist/stt.js
CHANGED
|
@@ -1,5 +1,6 @@
|
|
|
1
1
|
import {
|
|
2
2
|
AudioByteStream,
|
|
3
|
+
ChatMessage,
|
|
3
4
|
Future,
|
|
4
5
|
Task,
|
|
5
6
|
createTimedString,
|
|
@@ -33,6 +34,9 @@ const defaultSTTOptions = {
|
|
|
33
34
|
class STT extends stt.STT {
|
|
34
35
|
#opts;
|
|
35
36
|
#streams = /* @__PURE__ */ new Set();
|
|
37
|
+
// set (user + session))
|
|
38
|
+
#userKeyterms;
|
|
39
|
+
#sessionKeyterms = [];
|
|
36
40
|
label = "assemblyai.STT";
|
|
37
41
|
get model() {
|
|
38
42
|
return this.#opts.speechModel;
|
|
@@ -41,10 +45,19 @@ class STT extends stt.STT {
|
|
|
41
45
|
return "AssemblyAI";
|
|
42
46
|
}
|
|
43
47
|
constructor(opts = {}) {
|
|
48
|
+
const rawModel = opts.speechModel ?? defaultSTTOptions.speechModel;
|
|
49
|
+
const supportsCarryover = isU3ProModel(rawModel) || rawModel === "u3-pro";
|
|
50
|
+
if (opts.agentContextCarryover && !supportsCarryover) {
|
|
51
|
+
log().warn(
|
|
52
|
+
`agentContextCarryover is enabled but model '${rawModel}' does not support it; ignoring`
|
|
53
|
+
);
|
|
54
|
+
}
|
|
44
55
|
super({
|
|
45
56
|
streaming: true,
|
|
46
57
|
interimResults: true,
|
|
47
|
-
alignedTranscript: "word"
|
|
58
|
+
alignedTranscript: "word",
|
|
59
|
+
keyterms: true,
|
|
60
|
+
chatContext: (opts.agentContextCarryover ?? false) && supportsCarryover
|
|
48
61
|
});
|
|
49
62
|
if (opts.speechModel === "u3-pro") {
|
|
50
63
|
log().warn("'u3-pro' is deprecated, use 'universal-3-5-pro' instead.");
|
|
@@ -80,22 +93,50 @@ class STT extends stt.STT {
|
|
|
80
93
|
apiKey,
|
|
81
94
|
minTurnSilence
|
|
82
95
|
};
|
|
96
|
+
this.#userKeyterms = [...this.#opts.keytermsPrompt ?? []];
|
|
83
97
|
}
|
|
84
98
|
// eslint-disable-next-line @typescript-eslint/no-unused-vars
|
|
85
99
|
async _recognize(_) {
|
|
86
100
|
throw new Error("Non-streaming recognize is not supported on AssemblyAI STT");
|
|
87
101
|
}
|
|
88
102
|
updateOptions(opts) {
|
|
89
|
-
|
|
103
|
+
const nextOpts = { ...opts };
|
|
104
|
+
if (nextOpts.keytermsPrompt !== void 0) {
|
|
105
|
+
this.#userKeyterms = [...nextOpts.keytermsPrompt];
|
|
106
|
+
nextOpts.keytermsPrompt = [.../* @__PURE__ */ new Set([...this.#userKeyterms, ...this.#sessionKeyterms])];
|
|
107
|
+
}
|
|
108
|
+
this.#opts = { ...this.#opts, ...nextOpts };
|
|
109
|
+
for (const ref of this.#streams) {
|
|
110
|
+
const stream = ref.deref();
|
|
111
|
+
if (stream) {
|
|
112
|
+
stream.updateOptions(nextOpts);
|
|
113
|
+
} else {
|
|
114
|
+
this.#streams.delete(ref);
|
|
115
|
+
}
|
|
116
|
+
}
|
|
117
|
+
}
|
|
118
|
+
_updateSessionKeyterms(keyterms) {
|
|
119
|
+
if (keyterms.length === this.#sessionKeyterms.length && keyterms.every((t, i) => t === this.#sessionKeyterms[i])) {
|
|
120
|
+
return;
|
|
121
|
+
}
|
|
122
|
+
this.#sessionKeyterms = [...keyterms];
|
|
123
|
+
const merged = [.../* @__PURE__ */ new Set([...this.#userKeyterms, ...keyterms])];
|
|
124
|
+
this.#opts.keytermsPrompt = merged;
|
|
90
125
|
for (const ref of this.#streams) {
|
|
91
126
|
const stream = ref.deref();
|
|
92
127
|
if (stream) {
|
|
93
|
-
stream.updateOptions(
|
|
128
|
+
stream.updateOptions({ keytermsPrompt: merged });
|
|
94
129
|
} else {
|
|
95
130
|
this.#streams.delete(ref);
|
|
96
131
|
}
|
|
97
132
|
}
|
|
98
133
|
}
|
|
134
|
+
_pushConversationItem(ev) {
|
|
135
|
+
const chatItem = ev.item;
|
|
136
|
+
if (chatItem instanceof ChatMessage && chatItem.role === "assistant" && chatItem.textContent) {
|
|
137
|
+
this.updateOptions({ agentContext: chatItem.textContent });
|
|
138
|
+
}
|
|
139
|
+
}
|
|
99
140
|
stream(options) {
|
|
100
141
|
const stream = new SpeechStream(this, this.#opts, options == null ? void 0 : options.connOptions);
|
|
101
142
|
this.#streams.add(new WeakRef(stream));
|
|
@@ -203,7 +244,7 @@ class SpeechStream extends stt.SpeechStream {
|
|
|
203
244
|
end_of_turn_confidence_threshold: this.#opts.endOfTurnConfidenceThreshold,
|
|
204
245
|
min_turn_silence: minSilence,
|
|
205
246
|
max_turn_silence: maxSilence,
|
|
206
|
-
keyterms_prompt: this.#opts.keytermsPrompt !== void 0 ? JSON.stringify(this.#opts.keytermsPrompt) : void 0,
|
|
247
|
+
keyterms_prompt: this.#opts.keytermsPrompt !== void 0 && this.#opts.keytermsPrompt.length > 0 ? JSON.stringify(this.#opts.keytermsPrompt) : void 0,
|
|
207
248
|
language_detection: languageDetection,
|
|
208
249
|
inactivity_timeout: this.#opts.inactivityTimeout,
|
|
209
250
|
prompt: this.#opts.prompt,
|
package/dist/stt.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2026 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n//\nimport {\n type APIConnectOptions,\n type AudioBuffer,\n AudioByteStream,\n Future,\n Task,\n createTimedString,\n delay,\n log,\n normalizeLanguage,\n stt,\n waitForAbort,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { RawData } from 'ws';\nimport { WebSocket } from 'ws';\nimport type { STTEncoding, STTModels, VoiceFocus } from './models.js';\n\n// Speech models in the Universal-3 Pro family, which share the same parameter support.\nconst U3_PRO_MODELS = ['u3-rt-pro', 'u3-rt-pro-beta-1', 'universal-3-5-pro'] as const;\n\nfunction isU3ProModel(model: STTModels): boolean {\n return U3_PRO_MODELS.includes(model as (typeof U3_PRO_MODELS)[number]);\n}\n\n// AssemblyAI Universal-Streaming (v3) message envelope. All fields are optional\n// since we narrow on `type` before reading anything else.\ninterface StreamEventMessage {\n type?: 'Begin' | 'SpeechStarted' | 'Turn' | 'Termination' | string;\n // Begin\n id?: string;\n expires_at?: number;\n // Turn\n transcript?: string;\n utterance?: string;\n end_of_turn?: boolean;\n end_of_turn_confidence?: number;\n turn_is_formatted?: boolean;\n language_code?: string;\n language_confidence?: number;\n speaker_label?: string;\n words?: Array<{\n text?: string;\n start?: number;\n end?: number;\n confidence?: number;\n speaker?: string;\n }>;\n // Termination\n audio_duration_seconds?: number;\n session_duration_seconds?: number;\n}\n\nfunction speechDataMetadata(data: StreamEventMessage): stt.SpeechData['metadata'] | undefined {\n const assemblyai: Record<string, number> = {};\n\n if (typeof data.language_confidence === 'number') {\n assemblyai.languageConfidence = data.language_confidence;\n }\n\n if (Object.keys(assemblyai).length === 0) return undefined;\n\n return { assemblyai };\n}\n\nexport interface STTOptions {\n apiKey?: string;\n sampleRate: number;\n /**\n * How large each chunk of audio is before being sent to AssemblyAI, in\n * milliseconds. Corresponds to Python's `buffer_size_seconds` (seconds there,\n * ms here per this repo's time-unit convention).\n */\n bufferSizeMs: number;\n encoding: STTEncoding;\n speechModel: STTModels;\n languageDetection?: boolean;\n /**\n * Session inactivity timeout in seconds. AssemblyAI accepts integer values\n * from 5 to 3600; when unset, no inactivity timeout is applied.\n */\n inactivityTimeout?: number;\n endOfTurnConfidenceThreshold?: number;\n /** Minimum silence (ms) before a confident end-of-turn is finalized. */\n minTurnSilence?: number;\n /** Maximum silence (ms) before end-of-turn is forced regardless of confidence. */\n maxTurnSilence?: number;\n formatTurns?: boolean;\n keytermsPrompt?: string[];\n /** Only supported with the Universal-3 Pro model family. */\n prompt?: string;\n /** Only supported with the Universal-3 Pro model family. */\n agentContext?: string;\n /** Only supported with the Universal-3 Pro model family. Set at connection time only. */\n previousContextNTurns?: number;\n vadThreshold?: number;\n /**\n * Enable speaker diarization. Note: AssemblyAI will return per-word speaker\n * labels, but the JS framework's `stt.SpeechData` type does not yet expose\n * a `speakerId` field (unlike the Python framework), so the labels are not\n * currently surfaced on emitted events. Setting this to `true` still has\n * effect server-side. Once the base `SpeechData` interface gains speaker\n * support, `#processStreamEvent` should forward `data.words[].speaker` too.\n */\n speakerLabels?: boolean;\n maxSpeakers?: number;\n domain?: string;\n /** Isolate the primary voice and suppress background noise. Connect-time only. */\n voiceFocus?: VoiceFocus;\n /** Background audio suppression aggressiveness, from 0.0 to 1.0. Connect-time only. */\n voiceFocusThreshold?: number;\n /**\n * Accuracy/latency preset for the Universal-3 Pro model family: `min_latency`, `balanced`,\n * or `max_accuracy`. Explicit turn-silence values still take precedence over mode defaults.\n */\n mode?: 'min_latency' | 'balanced' | 'max_accuracy';\n baseUrl: string;\n}\n\nconst defaultSTTOptions: STTOptions = {\n apiKey: process.env.ASSEMBLYAI_API_KEY,\n sampleRate: 16000,\n bufferSizeMs: 50,\n encoding: 'pcm_s16le',\n speechModel: 'universal-3-5-pro',\n baseUrl: 'wss://streaming.assemblyai.com',\n};\n\nexport class STT extends stt.STT {\n #opts: STTOptions;\n #streams = new Set<WeakRef<SpeechStream>>();\n label = 'assemblyai.STT';\n\n get model(): string {\n return this.#opts.speechModel;\n }\n\n get provider(): string {\n return 'AssemblyAI';\n }\n\n constructor(opts: Partial<STTOptions> = {}) {\n super({\n streaming: true,\n interimResults: true,\n alignedTranscript: 'word',\n });\n\n if (opts.speechModel === 'u3-pro') {\n log().warn(\"'u3-pro' is deprecated, use 'universal-3-5-pro' instead.\");\n opts.speechModel = 'universal-3-5-pro';\n }\n\n const speechModel = opts.speechModel ?? defaultSTTOptions.speechModel;\n if (!isU3ProModel(speechModel)) {\n for (const param of [\n 'prompt',\n 'agentContext',\n 'previousContextNTurns',\n 'voiceFocus',\n 'voiceFocusThreshold',\n 'mode',\n ] as const) {\n if (opts[param] !== undefined) {\n throw new Error(\n `The '${param}' parameter is only supported with the ${U3_PRO_MODELS.join(', ')} models.`,\n );\n }\n }\n }\n\n const apiKey = opts.apiKey ?? defaultSTTOptions.apiKey;\n if (!apiKey) {\n throw new Error(\n 'AssemblyAI API key is required. Pass one in via the `apiKey` parameter, or set it as the `ASSEMBLYAI_API_KEY` environment variable',\n );\n }\n\n // Minimize latency by default, but let AssemblyAI's mode preset control silence tuning.\n const minTurnSilence = opts.minTurnSilence ?? (opts.mode === undefined ? 100 : undefined);\n\n this.#opts = {\n ...defaultSTTOptions,\n ...opts,\n apiKey,\n minTurnSilence,\n };\n }\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n async _recognize(_: AudioBuffer): Promise<stt.SpeechEvent> {\n throw new Error('Non-streaming recognize is not supported on AssemblyAI STT');\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n for (const ref of this.#streams) {\n const stream = ref.deref();\n if (stream) {\n stream.updateOptions(opts);\n } else {\n this.#streams.delete(ref);\n }\n }\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): SpeechStream {\n const stream = new SpeechStream(this, this.#opts, options?.connOptions);\n this.#streams.add(new WeakRef(stream));\n return stream;\n }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n static readonly CLOSE_MSG = JSON.stringify({ type: 'Terminate' });\n\n #opts: STTOptions;\n #logger = log();\n #speechDurationInS = 0;\n #lastPreflightStartTime = 0;\n #pendingConfigMessages: Record<string, unknown>[] = [];\n #configMessagePending = new Future();\n #sessionId: string | null = null;\n #expiresAt: number | null = null;\n label = 'assemblyai.SpeechStream';\n\n constructor(stt: STT, opts: STTOptions, connOptions?: APIConnectOptions) {\n super(stt, opts.sampleRate, connOptions);\n this.#opts = opts;\n this.closed = false;\n }\n\n /**\n * The AssemblyAI session ID. Set when the WebSocket connection is established\n * (before any speech events). Null until the connection completes.\n * Share this with the AssemblyAI team when reporting issues.\n */\n get sessionId(): string | null {\n return this.#sessionId;\n }\n\n /**\n * Unix timestamp when the AssemblyAI session expires. Set alongside\n * {@link sessionId} when the WebSocket connection is established.\n */\n get expiresAt(): number | null {\n return this.#expiresAt;\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n\n const configMsg: Record<string, unknown> = { type: 'UpdateConfiguration' };\n if (opts.prompt !== undefined) configMsg.prompt = opts.prompt;\n if (opts.agentContext !== undefined) configMsg.agent_context = opts.agentContext;\n if (opts.keytermsPrompt !== undefined) configMsg.keyterms_prompt = opts.keytermsPrompt;\n if (opts.maxTurnSilence !== undefined) configMsg.max_turn_silence = opts.maxTurnSilence;\n if (opts.minTurnSilence !== undefined) configMsg.min_turn_silence = opts.minTurnSilence;\n if (opts.endOfTurnConfidenceThreshold !== undefined) {\n configMsg.end_of_turn_confidence_threshold = opts.endOfTurnConfidenceThreshold;\n }\n if (opts.vadThreshold !== undefined) configMsg.vad_threshold = opts.vadThreshold;\n\n // Only send if any actual fields (besides `type`) were specified.\n if (Object.keys(configMsg).length > 1) {\n this.#pendingConfigMessages.push(configMsg);\n if (!this.#configMessagePending.done) this.#configMessagePending.resolve();\n }\n }\n\n /**\n * Force-finalize the current turn immediately.\n */\n forceEndpoint() {\n this.#pendingConfigMessages.push({ type: 'ForceEndpoint' });\n if (!this.#configMessagePending.done) this.#configMessagePending.resolve();\n }\n\n // Deepgram-style reconnect loop around a single websocket lifetime.\n protected async run() {\n const maxRetry = 32;\n let retries = 0;\n\n while (!this.input.closed && !this.closed) {\n try {\n const ws = await this.#connectWS();\n await this.#runWS(ws);\n retries = 0;\n } catch (e) {\n if (!this.closed && !this.input.closed) {\n if (retries >= maxRetry) {\n throw new Error(`failed to connect to AssemblyAI after ${retries} attempts: ${e}`);\n }\n\n const retryDelaySeconds = Math.min(retries * 5, 10);\n retries++;\n\n this.#logger.warn(\n `failed to connect to AssemblyAI, retrying in ${retryDelaySeconds} seconds: ${e} (${retries}/${maxRetry})`,\n );\n await delay(retryDelaySeconds * 1000);\n } else {\n this.#logger.warn(\n `AssemblyAI disconnected, connection is closed: ${e} (inputClosed: ${this.input.closed}, isClosed: ${this.closed})`,\n );\n }\n }\n }\n\n this.closed = true;\n }\n\n async #connectWS(): Promise<WebSocket> {\n // Universal-3 Pro family models default both min and max silence to 100ms when unset.\n // When a mode preset is selected, leave them unset unless explicitly provided so the\n // server's per-mode silence tuning is not overridden by the latency-optimized default.\n let minSilence = this.#opts.minTurnSilence;\n let maxSilence = this.#opts.maxTurnSilence;\n if (isU3ProModel(this.#opts.speechModel)) {\n if (minSilence === undefined && this.#opts.mode === undefined) minSilence = 100;\n if (maxSilence === undefined) maxSilence = minSilence;\n }\n\n // Default language_detection to true for multilingual / u3-rt-pro-family models, false otherwise.\n const defaultLanguageDetection =\n this.#opts.speechModel.includes('multilingual') || isU3ProModel(this.#opts.speechModel);\n const languageDetection = this.#opts.languageDetection ?? defaultLanguageDetection;\n\n const liveConfig: Record<string, unknown> = {\n sample_rate: this.#opts.sampleRate,\n encoding: this.#opts.encoding,\n speech_model: this.#opts.speechModel,\n format_turns: this.#opts.formatTurns,\n end_of_turn_confidence_threshold: this.#opts.endOfTurnConfidenceThreshold,\n min_turn_silence: minSilence,\n max_turn_silence: maxSilence,\n keyterms_prompt:\n this.#opts.keytermsPrompt !== undefined\n ? JSON.stringify(this.#opts.keytermsPrompt)\n : undefined,\n language_detection: languageDetection,\n inactivity_timeout: this.#opts.inactivityTimeout,\n prompt: this.#opts.prompt,\n agent_context: this.#opts.agentContext,\n previous_context_n_turns: this.#opts.previousContextNTurns,\n vad_threshold: this.#opts.vadThreshold,\n speaker_labels: this.#opts.speakerLabels,\n max_speakers: this.#opts.maxSpeakers,\n domain: this.#opts.domain,\n voice_focus: this.#opts.voiceFocus,\n voice_focus_threshold: this.#opts.voiceFocusThreshold,\n mode: this.#opts.mode,\n };\n\n const url = new URL(`${this.#opts.baseUrl}/v3/ws`);\n // Python serializes booleans as the strings \"true\"/\"false\", so we mirror that.\n for (const [key, value] of Object.entries(liveConfig)) {\n if (value === undefined || value === null) continue;\n if (typeof value === 'boolean') {\n url.searchParams.append(key, value ? 'true' : 'false');\n } else {\n url.searchParams.append(key, String(value));\n }\n }\n\n const ws = new WebSocket(url, {\n headers: {\n Authorization: this.#opts.apiKey!,\n 'Content-Type': 'application/json',\n 'User-Agent': 'AssemblyAI/1.0 (integration=Livekit)',\n },\n });\n\n await new Promise<void>((resolve, reject) => {\n ws.on('open', () => resolve());\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(new Error(`WebSocket returned ${code}`)));\n });\n\n return ws;\n }\n\n async #runWS(ws: WebSocket) {\n let closing = false;\n const sessionController = new AbortController();\n\n // gets cancelled also when sendTask is complete\n const wsMonitor = Task.from(async (controller) => {\n const closed = new Promise<void>((_, reject) => {\n ws.once('close', (code, reason) => {\n if (!closing) {\n this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);\n reject(new Error('WebSocket closed'));\n }\n });\n });\n\n await Promise.race([closed, waitForAbort(controller.signal)]);\n });\n\n const sendTask = async () => {\n const samplesPerBuffer = Math.floor((this.#opts.sampleRate * this.#opts.bufferSizeMs) / 1000);\n const audioStream = new AudioByteStream(this.#opts.sampleRate, 1, samplesPerBuffer);\n\n const abortPromise = waitForAbort(this.abortSignal);\n const sessionAbort = waitForAbort(sessionController.signal);\n\n try {\n while (!this.closed) {\n const result = await Promise.race([this.input.next(), abortPromise, sessionAbort]);\n\n if (result === undefined) return; // aborted\n if (result.done) break;\n\n const data = result.value;\n\n let frames: AudioFrame[];\n if (data === SpeechStream.FLUSH_SENTINEL) {\n frames = audioStream.flush();\n } else if (data.sampleRate === this.#opts.sampleRate && data.channels === 1) {\n // AssemblyAI expects mono PCM. The base SpeechStream only resamples\n // sample rate, so reject any frame that is not already downmixed.\n frames = audioStream.write(data.data.buffer as ArrayBuffer);\n } else {\n throw new Error('sample rate or channel count of frame does not match');\n }\n\n for (const frame of frames) {\n this.#speechDurationInS += frame.samplesPerChannel / frame.sampleRate;\n ws.send(frame.data.buffer);\n }\n }\n } finally {\n closing = true;\n try {\n ws.send(SpeechStream.CLOSE_MSG);\n } catch {\n // ignore — socket may already be closing\n }\n wsMonitor.cancel();\n }\n };\n\n let messageHandler: ((msg: RawData, isBinary: boolean) => void) | null = null;\n const listenTask = Task.from(async (controller) => {\n const listenMessage = new Promise<void>((resolve, reject) => {\n messageHandler = (msg, isBinary) => {\n if (isBinary) {\n this.#logger.error('unexpected binary message from AssemblyAI');\n return;\n }\n try {\n const json = JSON.parse(msg.toString()) as StreamEventMessage;\n this.#processStreamEvent(json);\n if (this.closed || closing) {\n resolve();\n }\n } catch (err) {\n this.#logger.error(`AssemblyAI: error processing message: ${msg}`);\n reject(err);\n }\n };\n ws.on('message', messageHandler);\n });\n\n await Promise.race([listenMessage, waitForAbort(controller.signal)]);\n });\n\n const configTask = Task.from(async (controller) => {\n // Drain any messages queued while the socket was reconnecting.\n while (this.#pendingConfigMessages.length > 0) {\n const msg = this.#pendingConfigMessages.shift()!;\n ws.send(JSON.stringify(msg));\n }\n\n while (!controller.signal.aborted) {\n await Promise.race([this.#configMessagePending.await, waitForAbort(controller.signal)]);\n if (controller.signal.aborted) return;\n\n this.#configMessagePending = new Future();\n while (this.#pendingConfigMessages.length > 0) {\n const msg = this.#pendingConfigMessages.shift()!;\n ws.send(JSON.stringify(msg));\n }\n }\n });\n\n try {\n await Promise.all([sendTask(), listenTask.result, wsMonitor.result]);\n } finally {\n closing = true;\n sessionController.abort();\n listenTask.cancel();\n configTask.cancel();\n if (messageHandler) ws.off('message', messageHandler);\n try {\n ws.close();\n } catch {\n // ignore\n }\n }\n }\n\n #averageConfidence(words: Array<{ confidence?: number }>): number {\n if (words.length === 0) return 0;\n return words.reduce((sum, w) => sum + (w.confidence ?? 0), 0) / words.length;\n }\n\n #processStreamEvent(data: StreamEventMessage) {\n const messageType = data.type;\n\n if (messageType === 'Begin') {\n this.#sessionId = data.id ?? null;\n this.#expiresAt = data.expires_at ?? null;\n this.#logger.info(\n `AssemblyAI session started id=${this.#sessionId} expires_at=${this.#expiresAt}`,\n );\n return;\n }\n\n if (messageType === 'SpeechStarted') {\n this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n return;\n }\n\n if (messageType === 'Termination') {\n this.#logger.debug(\n `AssemblyAI session terminated audio_duration=${data.audio_duration_seconds}s session_duration=${data.session_duration_seconds}s`,\n );\n return;\n }\n\n if (messageType !== 'Turn') {\n return;\n }\n\n const words = data.words ?? [];\n const endOfTurn = Boolean(data.end_of_turn);\n const turnIsFormatted = Boolean(data.turn_is_formatted);\n const utterance = data.utterance ?? '';\n const transcript = data.transcript ?? '';\n const language = normalizeLanguage(data.language_code ?? 'en');\n const metadata = speechDataMetadata(data);\n\n // Word timestamps are in milliseconds:\n // https://www.assemblyai.com/docs/api-reference/streaming-api/streaming-api#receive.receiveTurn.words\n const timedWords = words.map((word) =>\n createTimedString({\n text: word.text ?? '',\n startTime: (word.start ?? 0) / 1000 + this.startTimeOffset,\n endTime: (word.end ?? 0) / 1000 + this.startTimeOffset,\n confidence: word.confidence ?? 0,\n startTimeOffset: this.startTimeOffset,\n }),\n );\n\n let startTime = 0;\n let endTime = 0;\n let confidence = 0;\n\n // `words` are cumulative for the turn — emit as an interim transcript.\n if (timedWords.length > 0) {\n const interimText = timedWords.map((w) => w.text).join(' ');\n startTime = timedWords[0]!.startTime ?? 0;\n endTime = timedWords[timedWords.length - 1]!.endTime ?? 0;\n confidence = this.#averageConfidence(timedWords);\n\n this.queue.put({\n type: stt.SpeechEventType.INTERIM_TRANSCRIPT,\n alternatives: [\n {\n language,\n text: interimText,\n startTime,\n endTime,\n confidence,\n words: timedWords,\n ...(metadata ? { metadata } : {}),\n },\n ],\n });\n }\n\n // `utterance` is chunk-based (not cumulative) — emit as a preflight transcript\n // covering only the words since the last preflight.\n if (utterance) {\n if (this.#lastPreflightStartTime === 0) {\n this.#lastPreflightStartTime = startTime;\n }\n\n const utteranceWords = timedWords.filter(\n (w) => w.startTime !== undefined && w.startTime >= this.#lastPreflightStartTime,\n );\n const utteranceConfidence = this.#averageConfidence(utteranceWords);\n\n this.queue.put({\n type: stt.SpeechEventType.PREFLIGHT_TRANSCRIPT,\n alternatives: [\n {\n language,\n text: utterance,\n startTime: this.#lastPreflightStartTime,\n endTime,\n confidence: utteranceConfidence,\n words: utteranceWords,\n ...(metadata ? { metadata } : {}),\n },\n ],\n });\n this.#lastPreflightStartTime = endTime;\n }\n\n // End-of-turn: emit FINAL_TRANSCRIPT + END_OF_SPEECH.\n // If the user asked for formatted turns, wait for a formatted final.\n const waitingForFormatted = this.#opts.formatTurns === true && !turnIsFormatted;\n if (endOfTurn && !waitingForFormatted) {\n this.queue.put({\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n alternatives: [\n {\n language,\n text: transcript,\n startTime,\n endTime,\n confidence,\n words: timedWords,\n ...(metadata ? { metadata } : {}),\n },\n ],\n });\n\n this.queue.put({ type: stt.SpeechEventType.END_OF_SPEECH });\n if (this.#speechDurationInS > 0) {\n this.queue.put({\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n // Propagate the AssemblyAI session id as the request id so metrics\n // can be correlated back to a specific connection, mirroring how\n // Deepgram surfaces its `request_id`.\n requestId: this.#sessionId ?? undefined,\n recognitionUsage: {\n audioDuration: this.#speechDurationInS,\n },\n });\n this.#speechDurationInS = 0;\n this.#lastPreflightStartTime = 0;\n }\n }\n }\n}\n"],"mappings":"AAIA;AAAA,EAGE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAGP,SAAS,iBAAiB;AAI1B,MAAM,gBAAgB,CAAC,aAAa,oBAAoB,mBAAmB;AAE3E,SAAS,aAAa,OAA2B;AAC/C,SAAO,cAAc,SAAS,KAAuC;AACvE;AA8BA,SAAS,mBAAmB,MAAkE;AAC5F,QAAM,aAAqC,CAAC;AAE5C,MAAI,OAAO,KAAK,wBAAwB,UAAU;AAChD,eAAW,qBAAqB,KAAK;AAAA,EACvC;AAEA,MAAI,OAAO,KAAK,UAAU,EAAE,WAAW,EAAG,QAAO;AAEjD,SAAO,EAAE,WAAW;AACtB;AAwDA,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,YAAY;AAAA,EACZ,cAAc;AAAA,EACd,UAAU;AAAA,EACV,aAAa;AAAA,EACb,SAAS;AACX;AAEO,MAAM,YAAY,IAAI,IAAI;AAAA,EAC/B;AAAA,EACA,WAAW,oBAAI,IAA2B;AAAA,EAC1C,QAAQ;AAAA,EAER,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,YAAY,OAA4B,CAAC,GAAG;AAC1C,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB;AAAA,MAChB,mBAAmB;AAAA,IACrB,CAAC;AAED,QAAI,KAAK,gBAAgB,UAAU;AACjC,UAAI,EAAE,KAAK,0DAA0D;AACrE,WAAK,cAAc;AAAA,IACrB;AAEA,UAAM,cAAc,KAAK,eAAe,kBAAkB;AAC1D,QAAI,CAAC,aAAa,WAAW,GAAG;AAC9B,iBAAW,SAAS;AAAA,QAClB;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MACF,GAAY;AACV,YAAI,KAAK,KAAK,MAAM,QAAW;AAC7B,gBAAM,IAAI;AAAA,YACR,QAAQ,KAAK,0CAA0C,cAAc,KAAK,IAAI,CAAC;AAAA,UACjF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,SAAS,KAAK,UAAU,kBAAkB;AAChD,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAGA,UAAM,iBAAiB,KAAK,mBAAmB,KAAK,SAAS,SAAY,MAAM;AAE/E,SAAK,QAAQ;AAAA,MACX,GAAG;AAAA,MACH,GAAG;AAAA,MACH;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,WAAW,GAA0C;AACzD,UAAM,IAAI,MAAM,4DAA4D;AAAA,EAC9E;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,eAAW,OAAO,KAAK,UAAU;AAC/B,YAAM,SAAS,IAAI,MAAM;AACzB,UAAI,QAAQ;AACV,eAAO,cAAc,IAAI;AAAA,MAC3B,OAAO;AACL,aAAK,SAAS,OAAO,GAAG;AAAA,MAC1B;AAAA,IACF;AAAA,EACF;AAAA,EAEA,OAAO,SAA6D;AAClE,UAAM,SAAS,IAAI,aAAa,MAAM,KAAK,OAAO,mCAAS,WAAW;AACtE,SAAK,SAAS,IAAI,IAAI,QAAQ,MAAM,CAAC;AACrC,WAAO;AAAA,EACT;AACF;AAEO,MAAM,qBAAqB,IAAI,aAAa;AAAA,EACjD,OAAgB,YAAY,KAAK,UAAU,EAAE,MAAM,YAAY,CAAC;AAAA,EAEhE;AAAA,EACA,UAAU,IAAI;AAAA,EACd,qBAAqB;AAAA,EACrB,0BAA0B;AAAA,EAC1B,yBAAoD,CAAC;AAAA,EACrD,wBAAwB,IAAI,OAAO;AAAA,EACnC,aAA4B;AAAA,EAC5B,aAA4B;AAAA,EAC5B,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB,aAAiC;AACvE,UAAMA,MAAK,KAAK,YAAY,WAAW;AACvC,SAAK,QAAQ;AACb,SAAK,SAAS;AAAA,EAChB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,IAAI,YAA2B;AAC7B,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,IAAI,YAA2B;AAC7B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAEtC,UAAM,YAAqC,EAAE,MAAM,sBAAsB;AACzE,QAAI,KAAK,WAAW,OAAW,WAAU,SAAS,KAAK;AACvD,QAAI,KAAK,iBAAiB,OAAW,WAAU,gBAAgB,KAAK;AACpE,QAAI,KAAK,mBAAmB,OAAW,WAAU,kBAAkB,KAAK;AACxE,QAAI,KAAK,mBAAmB,OAAW,WAAU,mBAAmB,KAAK;AACzE,QAAI,KAAK,mBAAmB,OAAW,WAAU,mBAAmB,KAAK;AACzE,QAAI,KAAK,iCAAiC,QAAW;AACnD,gBAAU,mCAAmC,KAAK;AAAA,IACpD;AACA,QAAI,KAAK,iBAAiB,OAAW,WAAU,gBAAgB,KAAK;AAGpE,QAAI,OAAO,KAAK,SAAS,EAAE,SAAS,GAAG;AACrC,WAAK,uBAAuB,KAAK,SAAS;AAC1C,UAAI,CAAC,KAAK,sBAAsB,KAAM,MAAK,sBAAsB,QAAQ;AAAA,IAC3E;AAAA,EACF;AAAA;AAAA;AAAA;AAAA,EAKA,gBAAgB;AACd,SAAK,uBAAuB,KAAK,EAAE,MAAM,gBAAgB,CAAC;AAC1D,QAAI,CAAC,KAAK,sBAAsB,KAAM,MAAK,sBAAsB,QAAQ;AAAA,EAC3E;AAAA;AAAA,EAGA,MAAgB,MAAM;AACpB,UAAM,WAAW;AACjB,QAAI,UAAU;AAEd,WAAO,CAAC,KAAK,MAAM,UAAU,CAAC,KAAK,QAAQ;AACzC,UAAI;AACF,cAAM,KAAK,MAAM,KAAK,WAAW;AACjC,cAAM,KAAK,OAAO,EAAE;AACpB,kBAAU;AAAA,MACZ,SAAS,GAAG;AACV,YAAI,CAAC,KAAK,UAAU,CAAC,KAAK,MAAM,QAAQ;AACtC,cAAI,WAAW,UAAU;AACvB,kBAAM,IAAI,MAAM,yCAAyC,OAAO,cAAc,CAAC,EAAE;AAAA,UACnF;AAEA,gBAAM,oBAAoB,KAAK,IAAI,UAAU,GAAG,EAAE;AAClD;AAEA,eAAK,QAAQ;AAAA,YACX,gDAAgD,iBAAiB,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ;AAAA,UACzG;AACA,gBAAM,MAAM,oBAAoB,GAAI;AAAA,QACtC,OAAO;AACL,eAAK,QAAQ;AAAA,YACX,kDAAkD,CAAC,kBAAkB,KAAK,MAAM,MAAM,eAAe,KAAK,MAAM;AAAA,UAClH;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAM,aAAiC;AAIrC,QAAI,aAAa,KAAK,MAAM;AAC5B,QAAI,aAAa,KAAK,MAAM;AAC5B,QAAI,aAAa,KAAK,MAAM,WAAW,GAAG;AACxC,UAAI,eAAe,UAAa,KAAK,MAAM,SAAS,OAAW,cAAa;AAC5E,UAAI,eAAe,OAAW,cAAa;AAAA,IAC7C;AAGA,UAAM,2BACJ,KAAK,MAAM,YAAY,SAAS,cAAc,KAAK,aAAa,KAAK,MAAM,WAAW;AACxF,UAAM,oBAAoB,KAAK,MAAM,qBAAqB;AAE1D,UAAM,aAAsC;AAAA,MAC1C,aAAa,KAAK,MAAM;AAAA,MACxB,UAAU,KAAK,MAAM;AAAA,MACrB,cAAc,KAAK,MAAM;AAAA,MACzB,cAAc,KAAK,MAAM;AAAA,MACzB,kCAAkC,KAAK,MAAM;AAAA,MAC7C,kBAAkB;AAAA,MAClB,kBAAkB;AAAA,MAClB,iBACE,KAAK,MAAM,mBAAmB,SAC1B,KAAK,UAAU,KAAK,MAAM,cAAc,IACxC;AAAA,MACN,oBAAoB;AAAA,MACpB,oBAAoB,KAAK,MAAM;AAAA,MAC/B,QAAQ,KAAK,MAAM;AAAA,MACnB,eAAe,KAAK,MAAM;AAAA,MAC1B,0BAA0B,KAAK,MAAM;AAAA,MACrC,eAAe,KAAK,MAAM;AAAA,MAC1B,gBAAgB,KAAK,MAAM;AAAA,MAC3B,cAAc,KAAK,MAAM;AAAA,MACzB,QAAQ,KAAK,MAAM;AAAA,MACnB,aAAa,KAAK,MAAM;AAAA,MACxB,uBAAuB,KAAK,MAAM;AAAA,MAClC,MAAM,KAAK,MAAM;AAAA,IACnB;AAEA,UAAM,MAAM,IAAI,IAAI,GAAG,KAAK,MAAM,OAAO,QAAQ;AAEjD,eAAW,CAAC,KAAK,KAAK,KAAK,OAAO,QAAQ,UAAU,GAAG;AACrD,UAAI,UAAU,UAAa,UAAU,KAAM;AAC3C,UAAI,OAAO,UAAU,WAAW;AAC9B,YAAI,aAAa,OAAO,KAAK,QAAQ,SAAS,OAAO;AAAA,MACvD,OAAO;AACL,YAAI,aAAa,OAAO,KAAK,OAAO,KAAK,CAAC;AAAA,MAC5C;AAAA,IACF;AAEA,UAAM,KAAK,IAAI,UAAU,KAAK;AAAA,MAC5B,SAAS;AAAA,QACP,eAAe,KAAK,MAAM;AAAA,QAC1B,gBAAgB;AAAA,QAChB,cAAc;AAAA,MAChB;AAAA,IACF,CAAC;AAED,UAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,SAAG,GAAG,QAAQ,MAAM,QAAQ,CAAC;AAC7B,SAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,SAAG,GAAG,SAAS,CAAC,SAAS,OAAO,IAAI,MAAM,sBAAsB,IAAI,EAAE,CAAC,CAAC;AAAA,IAC1E,CAAC;AAED,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,OAAO,IAAe;AAC1B,QAAI,UAAU;AACd,UAAM,oBAAoB,IAAI,gBAAgB;AAG9C,UAAM,YAAY,KAAK,KAAK,OAAO,eAAe;AAChD,YAAM,SAAS,IAAI,QAAc,CAAC,GAAG,WAAW;AAC9C,WAAG,KAAK,SAAS,CAAC,MAAM,WAAW;AACjC,cAAI,CAAC,SAAS;AACZ,iBAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,MAAM,EAAE;AAClE,mBAAO,IAAI,MAAM,kBAAkB,CAAC;AAAA,UACtC;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,QAAQ,aAAa,WAAW,MAAM,CAAC,CAAC;AAAA,IAC9D,CAAC;AAED,UAAM,WAAW,YAAY;AAC3B,YAAM,mBAAmB,KAAK,MAAO,KAAK,MAAM,aAAa,KAAK,MAAM,eAAgB,GAAI;AAC5F,YAAM,cAAc,IAAI,gBAAgB,KAAK,MAAM,YAAY,GAAG,gBAAgB;AAElF,YAAM,eAAe,aAAa,KAAK,WAAW;AAClD,YAAM,eAAe,aAAa,kBAAkB,MAAM;AAE1D,UAAI;AACF,eAAO,CAAC,KAAK,QAAQ;AACnB,gBAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,KAAK,MAAM,KAAK,GAAG,cAAc,YAAY,CAAC;AAEjF,cAAI,WAAW,OAAW;AAC1B,cAAI,OAAO,KAAM;AAEjB,gBAAM,OAAO,OAAO;AAEpB,cAAI;AACJ,cAAI,SAAS,aAAa,gBAAgB;AACxC,qBAAS,YAAY,MAAM;AAAA,UAC7B,WAAW,KAAK,eAAe,KAAK,MAAM,cAAc,KAAK,aAAa,GAAG;AAG3E,qBAAS,YAAY,MAAM,KAAK,KAAK,MAAqB;AAAA,UAC5D,OAAO;AACL,kBAAM,IAAI,MAAM,sDAAsD;AAAA,UACxE;AAEA,qBAAW,SAAS,QAAQ;AAC1B,iBAAK,sBAAsB,MAAM,oBAAoB,MAAM;AAC3D,eAAG,KAAK,MAAM,KAAK,MAAM;AAAA,UAC3B;AAAA,QACF;AAAA,MACF,UAAE;AACA,kBAAU;AACV,YAAI;AACF,aAAG,KAAK,aAAa,SAAS;AAAA,QAChC,QAAQ;AAAA,QAER;AACA,kBAAU,OAAO;AAAA,MACnB;AAAA,IACF;AAEA,QAAI,iBAAqE;AACzE,UAAM,aAAa,KAAK,KAAK,OAAO,eAAe;AACjD,YAAM,gBAAgB,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3D,yBAAiB,CAAC,KAAK,aAAa;AAClC,cAAI,UAAU;AACZ,iBAAK,QAAQ,MAAM,2CAA2C;AAC9D;AAAA,UACF;AACA,cAAI;AACF,kBAAM,OAAO,KAAK,MAAM,IAAI,SAAS,CAAC;AACtC,iBAAK,oBAAoB,IAAI;AAC7B,gBAAI,KAAK,UAAU,SAAS;AAC1B,sBAAQ;AAAA,YACV;AAAA,UACF,SAAS,KAAK;AACZ,iBAAK,QAAQ,MAAM,yCAAyC,GAAG,EAAE;AACjE,mBAAO,GAAG;AAAA,UACZ;AAAA,QACF;AACA,WAAG,GAAG,WAAW,cAAc;AAAA,MACjC,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,eAAe,aAAa,WAAW,MAAM,CAAC,CAAC;AAAA,IACrE,CAAC;AAED,UAAM,aAAa,KAAK,KAAK,OAAO,eAAe;AAEjD,aAAO,KAAK,uBAAuB,SAAS,GAAG;AAC7C,cAAM,MAAM,KAAK,uBAAuB,MAAM;AAC9C,WAAG,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,MAC7B;AAEA,aAAO,CAAC,WAAW,OAAO,SAAS;AACjC,cAAM,QAAQ,KAAK,CAAC,KAAK,sBAAsB,OAAO,aAAa,WAAW,MAAM,CAAC,CAAC;AACtF,YAAI,WAAW,OAAO,QAAS;AAE/B,aAAK,wBAAwB,IAAI,OAAO;AACxC,eAAO,KAAK,uBAAuB,SAAS,GAAG;AAC7C,gBAAM,MAAM,KAAK,uBAAuB,MAAM;AAC9C,aAAG,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,QAC7B;AAAA,MACF;AAAA,IACF,CAAC;AAED,QAAI;AACF,YAAM,QAAQ,IAAI,CAAC,SAAS,GAAG,WAAW,QAAQ,UAAU,MAAM,CAAC;AAAA,IACrE,UAAE;AACA,gBAAU;AACV,wBAAkB,MAAM;AACxB,iBAAW,OAAO;AAClB,iBAAW,OAAO;AAClB,UAAI,eAAgB,IAAG,IAAI,WAAW,cAAc;AACpD,UAAI;AACF,WAAG,MAAM;AAAA,MACX,QAAQ;AAAA,MAER;AAAA,IACF;AAAA,EACF;AAAA,EAEA,mBAAmB,OAA+C;AAChE,QAAI,MAAM,WAAW,EAAG,QAAO;AAC/B,WAAO,MAAM,OAAO,CAAC,KAAK,MAAM,OAAO,EAAE,cAAc,IAAI,CAAC,IAAI,MAAM;AAAA,EACxE;AAAA,EAEA,oBAAoB,MAA0B;AAC5C,UAAM,cAAc,KAAK;AAEzB,QAAI,gBAAgB,SAAS;AAC3B,WAAK,aAAa,KAAK,MAAM;AAC7B,WAAK,aAAa,KAAK,cAAc;AACrC,WAAK,QAAQ;AAAA,QACX,iCAAiC,KAAK,UAAU,eAAe,KAAK,UAAU;AAAA,MAChF;AACA;AAAA,IACF;AAEA,QAAI,gBAAgB,iBAAiB;AACnC,WAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,gBAAgB,CAAC;AAC5D;AAAA,IACF;AAEA,QAAI,gBAAgB,eAAe;AACjC,WAAK,QAAQ;AAAA,QACX,gDAAgD,KAAK,sBAAsB,sBAAsB,KAAK,wBAAwB;AAAA,MAChI;AACA;AAAA,IACF;AAEA,QAAI,gBAAgB,QAAQ;AAC1B;AAAA,IACF;AAEA,UAAM,QAAQ,KAAK,SAAS,CAAC;AAC7B,UAAM,YAAY,QAAQ,KAAK,WAAW;AAC1C,UAAM,kBAAkB,QAAQ,KAAK,iBAAiB;AACtD,UAAM,YAAY,KAAK,aAAa;AACpC,UAAM,aAAa,KAAK,cAAc;AACtC,UAAM,WAAW,kBAAkB,KAAK,iBAAiB,IAAI;AAC7D,UAAM,WAAW,mBAAmB,IAAI;AAIxC,UAAM,aAAa,MAAM;AAAA,MAAI,CAAC,SAC5B,kBAAkB;AAAA,QAChB,MAAM,KAAK,QAAQ;AAAA,QACnB,YAAY,KAAK,SAAS,KAAK,MAAO,KAAK;AAAA,QAC3C,UAAU,KAAK,OAAO,KAAK,MAAO,KAAK;AAAA,QACvC,YAAY,KAAK,cAAc;AAAA,QAC/B,iBAAiB,KAAK;AAAA,MACxB,CAAC;AAAA,IACH;AAEA,QAAI,YAAY;AAChB,QAAI,UAAU;AACd,QAAI,aAAa;AAGjB,QAAI,WAAW,SAAS,GAAG;AACzB,YAAM,cAAc,WAAW,IAAI,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,GAAG;AAC1D,kBAAY,WAAW,CAAC,EAAG,aAAa;AACxC,gBAAU,WAAW,WAAW,SAAS,CAAC,EAAG,WAAW;AACxD,mBAAa,KAAK,mBAAmB,UAAU;AAE/C,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,IAAI,gBAAgB;AAAA,QAC1B,cAAc;AAAA,UACZ;AAAA,YACE;AAAA,YACA,MAAM;AAAA,YACN;AAAA,YACA;AAAA,YACA;AAAA,YACA,OAAO;AAAA,YACP,GAAI,WAAW,EAAE,SAAS,IAAI,CAAC;AAAA,UACjC;AAAA,QACF;AAAA,MACF,CAAC;AAAA,IACH;AAIA,QAAI,WAAW;AACb,UAAI,KAAK,4BAA4B,GAAG;AACtC,aAAK,0BAA0B;AAAA,MACjC;AAEA,YAAM,iBAAiB,WAAW;AAAA,QAChC,CAAC,MAAM,EAAE,cAAc,UAAa,EAAE,aAAa,KAAK;AAAA,MAC1D;AACA,YAAM,sBAAsB,KAAK,mBAAmB,cAAc;AAElE,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,IAAI,gBAAgB;AAAA,QAC1B,cAAc;AAAA,UACZ;AAAA,YACE;AAAA,YACA,MAAM;AAAA,YACN,WAAW,KAAK;AAAA,YAChB;AAAA,YACA,YAAY;AAAA,YACZ,OAAO;AAAA,YACP,GAAI,WAAW,EAAE,SAAS,IAAI,CAAC;AAAA,UACjC;AAAA,QACF;AAAA,MACF,CAAC;AACD,WAAK,0BAA0B;AAAA,IACjC;AAIA,UAAM,sBAAsB,KAAK,MAAM,gBAAgB,QAAQ,CAAC;AAChE,QAAI,aAAa,CAAC,qBAAqB;AACrC,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,IAAI,gBAAgB;AAAA,QAC1B,cAAc;AAAA,UACZ;AAAA,YACE;AAAA,YACA,MAAM;AAAA,YACN;AAAA,YACA;AAAA,YACA;AAAA,YACA,OAAO;AAAA,YACP,GAAI,WAAW,EAAE,SAAS,IAAI,CAAC;AAAA,UACjC;AAAA,QACF;AAAA,MACF,CAAC;AAED,WAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,cAAc,CAAC;AAC1D,UAAI,KAAK,qBAAqB,GAAG;AAC/B,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,IAAI,gBAAgB;AAAA;AAAA;AAAA;AAAA,UAI1B,WAAW,KAAK,cAAc;AAAA,UAC9B,kBAAkB;AAAA,YAChB,eAAe,KAAK;AAAA,UACtB;AAAA,QACF,CAAC;AACD,aAAK,qBAAqB;AAC1B,aAAK,0BAA0B;AAAA,MACjC;AAAA,IACF;AAAA,EACF;AACF;","names":["stt"]}
|
|
1
|
+
{"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2026 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n//\nimport {\n type APIConnectOptions,\n type AudioBuffer,\n AudioByteStream,\n ChatMessage,\n type ConversationItemAddedEvent,\n Future,\n Task,\n createTimedString,\n delay,\n log,\n normalizeLanguage,\n stt,\n waitForAbort,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { RawData } from 'ws';\nimport { WebSocket } from 'ws';\nimport type { STTEncoding, STTModels, VoiceFocus } from './models.js';\n\n// Speech models in the Universal-3 Pro family, which share the same parameter support.\nconst U3_PRO_MODELS = ['u3-rt-pro', 'u3-rt-pro-beta-1', 'universal-3-5-pro'] as const;\n\nfunction isU3ProModel(model: STTModels): boolean {\n return U3_PRO_MODELS.includes(model as (typeof U3_PRO_MODELS)[number]);\n}\n\n// AssemblyAI Universal-Streaming (v3) message envelope. All fields are optional\n// since we narrow on `type` before reading anything else.\ninterface StreamEventMessage {\n type?: 'Begin' | 'SpeechStarted' | 'Turn' | 'Termination' | string;\n // Begin\n id?: string;\n expires_at?: number;\n // Turn\n transcript?: string;\n utterance?: string;\n end_of_turn?: boolean;\n end_of_turn_confidence?: number;\n turn_is_formatted?: boolean;\n language_code?: string;\n language_confidence?: number;\n speaker_label?: string;\n words?: Array<{\n text?: string;\n start?: number;\n end?: number;\n confidence?: number;\n speaker?: string;\n }>;\n // Termination\n audio_duration_seconds?: number;\n session_duration_seconds?: number;\n}\n\nfunction speechDataMetadata(data: StreamEventMessage): stt.SpeechData['metadata'] | undefined {\n const assemblyai: Record<string, number> = {};\n\n if (typeof data.language_confidence === 'number') {\n assemblyai.languageConfidence = data.language_confidence;\n }\n\n if (Object.keys(assemblyai).length === 0) return undefined;\n\n return { assemblyai };\n}\n\nexport interface STTOptions {\n apiKey?: string;\n sampleRate: number;\n /**\n * How large each chunk of audio is before being sent to AssemblyAI, in\n * milliseconds. Corresponds to Python's `buffer_size_seconds` (seconds there,\n * ms here per this repo's time-unit convention).\n */\n bufferSizeMs: number;\n encoding: STTEncoding;\n speechModel: STTModels;\n languageDetection?: boolean;\n /**\n * Session inactivity timeout in seconds. AssemblyAI accepts integer values\n * from 5 to 3600; when unset, no inactivity timeout is applied.\n */\n inactivityTimeout?: number;\n endOfTurnConfidenceThreshold?: number;\n /** Minimum silence (ms) before a confident end-of-turn is finalized. */\n minTurnSilence?: number;\n /** Maximum silence (ms) before end-of-turn is forced regardless of confidence. */\n maxTurnSilence?: number;\n formatTurns?: boolean;\n keytermsPrompt?: string[];\n /** Only supported with the Universal-3 Pro model family. */\n prompt?: string;\n /** Only supported with the Universal-3 Pro model family. */\n agentContext?: string;\n /** Only supported with the Universal-3 Pro model family. Set at connection time only. */\n previousContextNTurns?: number;\n vadThreshold?: number;\n /**\n * Enable speaker diarization. Note: AssemblyAI will return per-word speaker\n * labels, but the JS framework's `stt.SpeechData` type does not yet expose\n * a `speakerId` field (unlike the Python framework), so the labels are not\n * currently surfaced on emitted events. Setting this to `true` still has\n * effect server-side. Once the base `SpeechData` interface gains speaker\n * support, `#processStreamEvent` should forward `data.words[].speaker` too.\n */\n speakerLabels?: boolean;\n maxSpeakers?: number;\n domain?: string;\n /** Isolate the primary voice and suppress background noise. Connect-time only. */\n voiceFocus?: VoiceFocus;\n /** Background audio suppression aggressiveness, from 0.0 to 1.0. Connect-time only. */\n voiceFocusThreshold?: number;\n /**\n * Accuracy/latency preset for the Universal-3 Pro model family: `min_latency`, `balanced`,\n * or `max_accuracy`. Explicit turn-silence values still take precedence over mode defaults.\n */\n mode?: 'min_latency' | 'balanced' | 'max_accuracy';\n /**\n * When the model supports it, let an `AgentSession` push each assistant reply into\n * `agentContext` so it is carried into the model's conversation context. Defaults to false;\n * set true to enable. Prior user turns are carried automatically by the model regardless of\n * this flag. Ignored on models without context support.\n */\n agentContextCarryover?: boolean;\n baseUrl: string;\n}\n\nconst defaultSTTOptions: STTOptions = {\n apiKey: process.env.ASSEMBLYAI_API_KEY,\n sampleRate: 16000,\n bufferSizeMs: 50,\n encoding: 'pcm_s16le',\n speechModel: 'universal-3-5-pro',\n baseUrl: 'wss://streaming.assemblyai.com',\n};\n\nexport class STT extends stt.STT {\n #opts: STTOptions;\n #streams = new Set<WeakRef<SpeechStream>>();\n // set (user + session))\n #userKeyterms: string[];\n #sessionKeyterms: string[] = [];\n label = 'assemblyai.STT';\n\n get model(): string {\n return this.#opts.speechModel;\n }\n\n get provider(): string {\n return 'AssemblyAI';\n }\n\n constructor(opts: Partial<STTOptions> = {}) {\n // u3-rt-pro family — \"u3-pro\" is normalized below — and is opt-in via the user)\n const rawModel = opts.speechModel ?? defaultSTTOptions.speechModel;\n const supportsCarryover = isU3ProModel(rawModel) || rawModel === 'u3-pro';\n if (opts.agentContextCarryover && !supportsCarryover) {\n log().warn(\n `agentContextCarryover is enabled but model '${rawModel}' does not support it; ignoring`,\n );\n }\n super({\n streaming: true,\n interimResults: true,\n alignedTranscript: 'word',\n keyterms: true,\n chatContext: (opts.agentContextCarryover ?? false) && supportsCarryover,\n });\n\n if (opts.speechModel === 'u3-pro') {\n log().warn(\"'u3-pro' is deprecated, use 'universal-3-5-pro' instead.\");\n opts.speechModel = 'universal-3-5-pro';\n }\n\n const speechModel = opts.speechModel ?? defaultSTTOptions.speechModel;\n if (!isU3ProModel(speechModel)) {\n for (const param of [\n 'prompt',\n 'agentContext',\n 'previousContextNTurns',\n 'voiceFocus',\n 'voiceFocusThreshold',\n 'mode',\n ] as const) {\n if (opts[param] !== undefined) {\n throw new Error(\n `The '${param}' parameter is only supported with the ${U3_PRO_MODELS.join(', ')} models.`,\n );\n }\n }\n }\n\n const apiKey = opts.apiKey ?? defaultSTTOptions.apiKey;\n if (!apiKey) {\n throw new Error(\n 'AssemblyAI API key is required. Pass one in via the `apiKey` parameter, or set it as the `ASSEMBLYAI_API_KEY` environment variable',\n );\n }\n\n // Minimize latency by default, but let AssemblyAI's mode preset control silence tuning.\n const minTurnSilence = opts.minTurnSilence ?? (opts.mode === undefined ? 100 : undefined);\n\n this.#opts = {\n ...defaultSTTOptions,\n ...opts,\n apiKey,\n minTurnSilence,\n };\n this.#userKeyterms = [...(this.#opts.keytermsPrompt ?? [])];\n }\n\n // eslint-disable-next-line @typescript-eslint/no-unused-vars\n async _recognize(_: AudioBuffer): Promise<stt.SpeechEvent> {\n throw new Error('Non-streaming recognize is not supported on AssemblyAI STT');\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n // session keyterms so a user update doesn't drop them)\n const nextOpts = { ...opts };\n if (nextOpts.keytermsPrompt !== undefined) {\n this.#userKeyterms = [...nextOpts.keytermsPrompt];\n nextOpts.keytermsPrompt = [...new Set([...this.#userKeyterms, ...this.#sessionKeyterms])];\n }\n this.#opts = { ...this.#opts, ...nextOpts };\n for (const ref of this.#streams) {\n const stream = ref.deref();\n if (stream) {\n stream.updateOptions(nextOpts);\n } else {\n this.#streams.delete(ref);\n }\n }\n }\n\n override _updateSessionKeyterms(keyterms: string[]): void {\n if (\n keyterms.length === this.#sessionKeyterms.length &&\n keyterms.every((t, i) => t === this.#sessionKeyterms[i])\n ) {\n return;\n }\n this.#sessionKeyterms = [...keyterms];\n const merged = [...new Set([...this.#userKeyterms, ...keyterms])];\n this.#opts.keytermsPrompt = merged;\n // applied live via the stream's UpdateConfiguration (no reconnect)\n for (const ref of this.#streams) {\n const stream = ref.deref();\n if (stream) {\n stream.updateOptions({ keytermsPrompt: merged });\n } else {\n this.#streams.delete(ref);\n }\n }\n }\n\n override _pushConversationItem(ev: ConversationItemAddedEvent): void {\n const chatItem = ev.item;\n if (chatItem instanceof ChatMessage && chatItem.role === 'assistant' && chatItem.textContent) {\n this.updateOptions({ agentContext: chatItem.textContent });\n }\n }\n\n stream(options?: { connOptions?: APIConnectOptions }): SpeechStream {\n const stream = new SpeechStream(this, this.#opts, options?.connOptions);\n this.#streams.add(new WeakRef(stream));\n return stream;\n }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n static readonly CLOSE_MSG = JSON.stringify({ type: 'Terminate' });\n\n #opts: STTOptions;\n #logger = log();\n #speechDurationInS = 0;\n #lastPreflightStartTime = 0;\n #pendingConfigMessages: Record<string, unknown>[] = [];\n #configMessagePending = new Future();\n #sessionId: string | null = null;\n #expiresAt: number | null = null;\n label = 'assemblyai.SpeechStream';\n\n constructor(stt: STT, opts: STTOptions, connOptions?: APIConnectOptions) {\n super(stt, opts.sampleRate, connOptions);\n this.#opts = opts;\n this.closed = false;\n }\n\n /**\n * The AssemblyAI session ID. Set when the WebSocket connection is established\n * (before any speech events). Null until the connection completes.\n * Share this with the AssemblyAI team when reporting issues.\n */\n get sessionId(): string | null {\n return this.#sessionId;\n }\n\n /**\n * Unix timestamp when the AssemblyAI session expires. Set alongside\n * {@link sessionId} when the WebSocket connection is established.\n */\n get expiresAt(): number | null {\n return this.#expiresAt;\n }\n\n updateOptions(opts: Partial<STTOptions>) {\n this.#opts = { ...this.#opts, ...opts };\n\n const configMsg: Record<string, unknown> = { type: 'UpdateConfiguration' };\n if (opts.prompt !== undefined) configMsg.prompt = opts.prompt;\n if (opts.agentContext !== undefined) configMsg.agent_context = opts.agentContext;\n if (opts.keytermsPrompt !== undefined) configMsg.keyterms_prompt = opts.keytermsPrompt;\n if (opts.maxTurnSilence !== undefined) configMsg.max_turn_silence = opts.maxTurnSilence;\n if (opts.minTurnSilence !== undefined) configMsg.min_turn_silence = opts.minTurnSilence;\n if (opts.endOfTurnConfidenceThreshold !== undefined) {\n configMsg.end_of_turn_confidence_threshold = opts.endOfTurnConfidenceThreshold;\n }\n if (opts.vadThreshold !== undefined) configMsg.vad_threshold = opts.vadThreshold;\n\n // Only send if any actual fields (besides `type`) were specified.\n if (Object.keys(configMsg).length > 1) {\n this.#pendingConfigMessages.push(configMsg);\n if (!this.#configMessagePending.done) this.#configMessagePending.resolve();\n }\n }\n\n /**\n * Force-finalize the current turn immediately.\n */\n forceEndpoint() {\n this.#pendingConfigMessages.push({ type: 'ForceEndpoint' });\n if (!this.#configMessagePending.done) this.#configMessagePending.resolve();\n }\n\n // Deepgram-style reconnect loop around a single websocket lifetime.\n protected async run() {\n const maxRetry = 32;\n let retries = 0;\n\n while (!this.input.closed && !this.closed) {\n try {\n const ws = await this.#connectWS();\n await this.#runWS(ws);\n retries = 0;\n } catch (e) {\n if (!this.closed && !this.input.closed) {\n if (retries >= maxRetry) {\n throw new Error(`failed to connect to AssemblyAI after ${retries} attempts: ${e}`);\n }\n\n const retryDelaySeconds = Math.min(retries * 5, 10);\n retries++;\n\n this.#logger.warn(\n `failed to connect to AssemblyAI, retrying in ${retryDelaySeconds} seconds: ${e} (${retries}/${maxRetry})`,\n );\n await delay(retryDelaySeconds * 1000);\n } else {\n this.#logger.warn(\n `AssemblyAI disconnected, connection is closed: ${e} (inputClosed: ${this.input.closed}, isClosed: ${this.closed})`,\n );\n }\n }\n }\n\n this.closed = true;\n }\n\n async #connectWS(): Promise<WebSocket> {\n // Universal-3 Pro family models default both min and max silence to 100ms when unset.\n // When a mode preset is selected, leave them unset unless explicitly provided so the\n // server's per-mode silence tuning is not overridden by the latency-optimized default.\n let minSilence = this.#opts.minTurnSilence;\n let maxSilence = this.#opts.maxTurnSilence;\n if (isU3ProModel(this.#opts.speechModel)) {\n if (minSilence === undefined && this.#opts.mode === undefined) minSilence = 100;\n if (maxSilence === undefined) maxSilence = minSilence;\n }\n\n // Default language_detection to true for multilingual / u3-rt-pro-family models, false otherwise.\n const defaultLanguageDetection =\n this.#opts.speechModel.includes('multilingual') || isU3ProModel(this.#opts.speechModel);\n const languageDetection = this.#opts.languageDetection ?? defaultLanguageDetection;\n\n const liveConfig: Record<string, unknown> = {\n sample_rate: this.#opts.sampleRate,\n encoding: this.#opts.encoding,\n speech_model: this.#opts.speechModel,\n format_turns: this.#opts.formatTurns,\n end_of_turn_confidence_threshold: this.#opts.endOfTurnConfidenceThreshold,\n min_turn_silence: minSilence,\n max_turn_silence: maxSilence,\n keyterms_prompt:\n this.#opts.keytermsPrompt !== undefined && this.#opts.keytermsPrompt.length > 0\n ? JSON.stringify(this.#opts.keytermsPrompt)\n : undefined,\n language_detection: languageDetection,\n inactivity_timeout: this.#opts.inactivityTimeout,\n prompt: this.#opts.prompt,\n agent_context: this.#opts.agentContext,\n previous_context_n_turns: this.#opts.previousContextNTurns,\n vad_threshold: this.#opts.vadThreshold,\n speaker_labels: this.#opts.speakerLabels,\n max_speakers: this.#opts.maxSpeakers,\n domain: this.#opts.domain,\n voice_focus: this.#opts.voiceFocus,\n voice_focus_threshold: this.#opts.voiceFocusThreshold,\n mode: this.#opts.mode,\n };\n\n const url = new URL(`${this.#opts.baseUrl}/v3/ws`);\n // Python serializes booleans as the strings \"true\"/\"false\", so we mirror that.\n for (const [key, value] of Object.entries(liveConfig)) {\n if (value === undefined || value === null) continue;\n if (typeof value === 'boolean') {\n url.searchParams.append(key, value ? 'true' : 'false');\n } else {\n url.searchParams.append(key, String(value));\n }\n }\n\n const ws = new WebSocket(url, {\n headers: {\n Authorization: this.#opts.apiKey!,\n 'Content-Type': 'application/json',\n 'User-Agent': 'AssemblyAI/1.0 (integration=Livekit)',\n },\n });\n\n await new Promise<void>((resolve, reject) => {\n ws.on('open', () => resolve());\n ws.on('error', (error) => reject(error));\n ws.on('close', (code) => reject(new Error(`WebSocket returned ${code}`)));\n });\n\n return ws;\n }\n\n async #runWS(ws: WebSocket) {\n let closing = false;\n const sessionController = new AbortController();\n\n // gets cancelled also when sendTask is complete\n const wsMonitor = Task.from(async (controller) => {\n const closed = new Promise<void>((_, reject) => {\n ws.once('close', (code, reason) => {\n if (!closing) {\n this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);\n reject(new Error('WebSocket closed'));\n }\n });\n });\n\n await Promise.race([closed, waitForAbort(controller.signal)]);\n });\n\n const sendTask = async () => {\n const samplesPerBuffer = Math.floor((this.#opts.sampleRate * this.#opts.bufferSizeMs) / 1000);\n const audioStream = new AudioByteStream(this.#opts.sampleRate, 1, samplesPerBuffer);\n\n const abortPromise = waitForAbort(this.abortSignal);\n const sessionAbort = waitForAbort(sessionController.signal);\n\n try {\n while (!this.closed) {\n const result = await Promise.race([this.input.next(), abortPromise, sessionAbort]);\n\n if (result === undefined) return; // aborted\n if (result.done) break;\n\n const data = result.value;\n\n let frames: AudioFrame[];\n if (data === SpeechStream.FLUSH_SENTINEL) {\n frames = audioStream.flush();\n } else if (data.sampleRate === this.#opts.sampleRate && data.channels === 1) {\n // AssemblyAI expects mono PCM. The base SpeechStream only resamples\n // sample rate, so reject any frame that is not already downmixed.\n frames = audioStream.write(data.data.buffer as ArrayBuffer);\n } else {\n throw new Error('sample rate or channel count of frame does not match');\n }\n\n for (const frame of frames) {\n this.#speechDurationInS += frame.samplesPerChannel / frame.sampleRate;\n ws.send(frame.data.buffer);\n }\n }\n } finally {\n closing = true;\n try {\n ws.send(SpeechStream.CLOSE_MSG);\n } catch {\n // ignore — socket may already be closing\n }\n wsMonitor.cancel();\n }\n };\n\n let messageHandler: ((msg: RawData, isBinary: boolean) => void) | null = null;\n const listenTask = Task.from(async (controller) => {\n const listenMessage = new Promise<void>((resolve, reject) => {\n messageHandler = (msg, isBinary) => {\n if (isBinary) {\n this.#logger.error('unexpected binary message from AssemblyAI');\n return;\n }\n try {\n const json = JSON.parse(msg.toString()) as StreamEventMessage;\n this.#processStreamEvent(json);\n if (this.closed || closing) {\n resolve();\n }\n } catch (err) {\n this.#logger.error(`AssemblyAI: error processing message: ${msg}`);\n reject(err);\n }\n };\n ws.on('message', messageHandler);\n });\n\n await Promise.race([listenMessage, waitForAbort(controller.signal)]);\n });\n\n const configTask = Task.from(async (controller) => {\n // Drain any messages queued while the socket was reconnecting.\n while (this.#pendingConfigMessages.length > 0) {\n const msg = this.#pendingConfigMessages.shift()!;\n ws.send(JSON.stringify(msg));\n }\n\n while (!controller.signal.aborted) {\n await Promise.race([this.#configMessagePending.await, waitForAbort(controller.signal)]);\n if (controller.signal.aborted) return;\n\n this.#configMessagePending = new Future();\n while (this.#pendingConfigMessages.length > 0) {\n const msg = this.#pendingConfigMessages.shift()!;\n ws.send(JSON.stringify(msg));\n }\n }\n });\n\n try {\n await Promise.all([sendTask(), listenTask.result, wsMonitor.result]);\n } finally {\n closing = true;\n sessionController.abort();\n listenTask.cancel();\n configTask.cancel();\n if (messageHandler) ws.off('message', messageHandler);\n try {\n ws.close();\n } catch {\n // ignore\n }\n }\n }\n\n #averageConfidence(words: Array<{ confidence?: number }>): number {\n if (words.length === 0) return 0;\n return words.reduce((sum, w) => sum + (w.confidence ?? 0), 0) / words.length;\n }\n\n #processStreamEvent(data: StreamEventMessage) {\n const messageType = data.type;\n\n if (messageType === 'Begin') {\n this.#sessionId = data.id ?? null;\n this.#expiresAt = data.expires_at ?? null;\n this.#logger.info(\n `AssemblyAI session started id=${this.#sessionId} expires_at=${this.#expiresAt}`,\n );\n return;\n }\n\n if (messageType === 'SpeechStarted') {\n this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n return;\n }\n\n if (messageType === 'Termination') {\n this.#logger.debug(\n `AssemblyAI session terminated audio_duration=${data.audio_duration_seconds}s session_duration=${data.session_duration_seconds}s`,\n );\n return;\n }\n\n if (messageType !== 'Turn') {\n return;\n }\n\n const words = data.words ?? [];\n const endOfTurn = Boolean(data.end_of_turn);\n const turnIsFormatted = Boolean(data.turn_is_formatted);\n const utterance = data.utterance ?? '';\n const transcript = data.transcript ?? '';\n const language = normalizeLanguage(data.language_code ?? 'en');\n const metadata = speechDataMetadata(data);\n\n // Word timestamps are in milliseconds:\n // https://www.assemblyai.com/docs/api-reference/streaming-api/streaming-api#receive.receiveTurn.words\n const timedWords = words.map((word) =>\n createTimedString({\n text: word.text ?? '',\n startTime: (word.start ?? 0) / 1000 + this.startTimeOffset,\n endTime: (word.end ?? 0) / 1000 + this.startTimeOffset,\n confidence: word.confidence ?? 0,\n startTimeOffset: this.startTimeOffset,\n }),\n );\n\n let startTime = 0;\n let endTime = 0;\n let confidence = 0;\n\n // `words` are cumulative for the turn — emit as an interim transcript.\n if (timedWords.length > 0) {\n const interimText = timedWords.map((w) => w.text).join(' ');\n startTime = timedWords[0]!.startTime ?? 0;\n endTime = timedWords[timedWords.length - 1]!.endTime ?? 0;\n confidence = this.#averageConfidence(timedWords);\n\n this.queue.put({\n type: stt.SpeechEventType.INTERIM_TRANSCRIPT,\n alternatives: [\n {\n language,\n text: interimText,\n startTime,\n endTime,\n confidence,\n words: timedWords,\n ...(metadata ? { metadata } : {}),\n },\n ],\n });\n }\n\n // `utterance` is chunk-based (not cumulative) — emit as a preflight transcript\n // covering only the words since the last preflight.\n if (utterance) {\n if (this.#lastPreflightStartTime === 0) {\n this.#lastPreflightStartTime = startTime;\n }\n\n const utteranceWords = timedWords.filter(\n (w) => w.startTime !== undefined && w.startTime >= this.#lastPreflightStartTime,\n );\n const utteranceConfidence = this.#averageConfidence(utteranceWords);\n\n this.queue.put({\n type: stt.SpeechEventType.PREFLIGHT_TRANSCRIPT,\n alternatives: [\n {\n language,\n text: utterance,\n startTime: this.#lastPreflightStartTime,\n endTime,\n confidence: utteranceConfidence,\n words: utteranceWords,\n ...(metadata ? { metadata } : {}),\n },\n ],\n });\n this.#lastPreflightStartTime = endTime;\n }\n\n // End-of-turn: emit FINAL_TRANSCRIPT + END_OF_SPEECH.\n // If the user asked for formatted turns, wait for a formatted final.\n const waitingForFormatted = this.#opts.formatTurns === true && !turnIsFormatted;\n if (endOfTurn && !waitingForFormatted) {\n this.queue.put({\n type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n alternatives: [\n {\n language,\n text: transcript,\n startTime,\n endTime,\n confidence,\n words: timedWords,\n ...(metadata ? { metadata } : {}),\n },\n ],\n });\n\n this.queue.put({ type: stt.SpeechEventType.END_OF_SPEECH });\n if (this.#speechDurationInS > 0) {\n this.queue.put({\n type: stt.SpeechEventType.RECOGNITION_USAGE,\n // Propagate the AssemblyAI session id as the request id so metrics\n // can be correlated back to a specific connection, mirroring how\n // Deepgram surfaces its `request_id`.\n requestId: this.#sessionId ?? undefined,\n recognitionUsage: {\n audioDuration: this.#speechDurationInS,\n },\n });\n this.#speechDurationInS = 0;\n this.#lastPreflightStartTime = 0;\n }\n }\n }\n}\n"],"mappings":"AAIA;AAAA,EAGE;AAAA,EACA;AAAA,EAEA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAGP,SAAS,iBAAiB;AAI1B,MAAM,gBAAgB,CAAC,aAAa,oBAAoB,mBAAmB;AAE3E,SAAS,aAAa,OAA2B;AAC/C,SAAO,cAAc,SAAS,KAAuC;AACvE;AA8BA,SAAS,mBAAmB,MAAkE;AAC5F,QAAM,aAAqC,CAAC;AAE5C,MAAI,OAAO,KAAK,wBAAwB,UAAU;AAChD,eAAW,qBAAqB,KAAK;AAAA,EACvC;AAEA,MAAI,OAAO,KAAK,UAAU,EAAE,WAAW,EAAG,QAAO;AAEjD,SAAO,EAAE,WAAW;AACtB;AA+DA,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,YAAY;AAAA,EACZ,cAAc;AAAA,EACd,UAAU;AAAA,EACV,aAAa;AAAA,EACb,SAAS;AACX;AAEO,MAAM,YAAY,IAAI,IAAI;AAAA,EAC/B;AAAA,EACA,WAAW,oBAAI,IAA2B;AAAA;AAAA,EAE1C;AAAA,EACA,mBAA6B,CAAC;AAAA,EAC9B,QAAQ;AAAA,EAER,IAAI,QAAgB;AAClB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,YAAY,OAA4B,CAAC,GAAG;AAE1C,UAAM,WAAW,KAAK,eAAe,kBAAkB;AACvD,UAAM,oBAAoB,aAAa,QAAQ,KAAK,aAAa;AACjE,QAAI,KAAK,yBAAyB,CAAC,mBAAmB;AACpD,UAAI,EAAE;AAAA,QACJ,+CAA+C,QAAQ;AAAA,MACzD;AAAA,IACF;AACA,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB;AAAA,MAChB,mBAAmB;AAAA,MACnB,UAAU;AAAA,MACV,cAAc,KAAK,yBAAyB,UAAU;AAAA,IACxD,CAAC;AAED,QAAI,KAAK,gBAAgB,UAAU;AACjC,UAAI,EAAE,KAAK,0DAA0D;AACrE,WAAK,cAAc;AAAA,IACrB;AAEA,UAAM,cAAc,KAAK,eAAe,kBAAkB;AAC1D,QAAI,CAAC,aAAa,WAAW,GAAG;AAC9B,iBAAW,SAAS;AAAA,QAClB;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MACF,GAAY;AACV,YAAI,KAAK,KAAK,MAAM,QAAW;AAC7B,gBAAM,IAAI;AAAA,YACR,QAAQ,KAAK,0CAA0C,cAAc,KAAK,IAAI,CAAC;AAAA,UACjF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,SAAS,KAAK,UAAU,kBAAkB;AAChD,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAGA,UAAM,iBAAiB,KAAK,mBAAmB,KAAK,SAAS,SAAY,MAAM;AAE/E,SAAK,QAAQ;AAAA,MACX,GAAG;AAAA,MACH,GAAG;AAAA,MACH;AAAA,MACA;AAAA,IACF;AACA,SAAK,gBAAgB,CAAC,GAAI,KAAK,MAAM,kBAAkB,CAAC,CAAE;AAAA,EAC5D;AAAA;AAAA,EAGA,MAAM,WAAW,GAA0C;AACzD,UAAM,IAAI,MAAM,4DAA4D;AAAA,EAC9E;AAAA,EAEA,cAAc,MAA2B;AAEvC,UAAM,WAAW,EAAE,GAAG,KAAK;AAC3B,QAAI,SAAS,mBAAmB,QAAW;AACzC,WAAK,gBAAgB,CAAC,GAAG,SAAS,cAAc;AAChD,eAAS,iBAAiB,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAG,KAAK,eAAe,GAAG,KAAK,gBAAgB,CAAC,CAAC;AAAA,IAC1F;AACA,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,SAAS;AAC1C,eAAW,OAAO,KAAK,UAAU;AAC/B,YAAM,SAAS,IAAI,MAAM;AACzB,UAAI,QAAQ;AACV,eAAO,cAAc,QAAQ;AAAA,MAC/B,OAAO;AACL,aAAK,SAAS,OAAO,GAAG;AAAA,MAC1B;AAAA,IACF;AAAA,EACF;AAAA,EAES,uBAAuB,UAA0B;AACxD,QACE,SAAS,WAAW,KAAK,iBAAiB,UAC1C,SAAS,MAAM,CAAC,GAAG,MAAM,MAAM,KAAK,iBAAiB,CAAC,CAAC,GACvD;AACA;AAAA,IACF;AACA,SAAK,mBAAmB,CAAC,GAAG,QAAQ;AACpC,UAAM,SAAS,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAG,KAAK,eAAe,GAAG,QAAQ,CAAC,CAAC;AAChE,SAAK,MAAM,iBAAiB;AAE5B,eAAW,OAAO,KAAK,UAAU;AAC/B,YAAM,SAAS,IAAI,MAAM;AACzB,UAAI,QAAQ;AACV,eAAO,cAAc,EAAE,gBAAgB,OAAO,CAAC;AAAA,MACjD,OAAO;AACL,aAAK,SAAS,OAAO,GAAG;AAAA,MAC1B;AAAA,IACF;AAAA,EACF;AAAA,EAES,sBAAsB,IAAsC;AACnE,UAAM,WAAW,GAAG;AACpB,QAAI,oBAAoB,eAAe,SAAS,SAAS,eAAe,SAAS,aAAa;AAC5F,WAAK,cAAc,EAAE,cAAc,SAAS,YAAY,CAAC;AAAA,IAC3D;AAAA,EACF;AAAA,EAEA,OAAO,SAA6D;AAClE,UAAM,SAAS,IAAI,aAAa,MAAM,KAAK,OAAO,mCAAS,WAAW;AACtE,SAAK,SAAS,IAAI,IAAI,QAAQ,MAAM,CAAC;AACrC,WAAO;AAAA,EACT;AACF;AAEO,MAAM,qBAAqB,IAAI,aAAa;AAAA,EACjD,OAAgB,YAAY,KAAK,UAAU,EAAE,MAAM,YAAY,CAAC;AAAA,EAEhE;AAAA,EACA,UAAU,IAAI;AAAA,EACd,qBAAqB;AAAA,EACrB,0BAA0B;AAAA,EAC1B,yBAAoD,CAAC;AAAA,EACrD,wBAAwB,IAAI,OAAO;AAAA,EACnC,aAA4B;AAAA,EAC5B,aAA4B;AAAA,EAC5B,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB,aAAiC;AACvE,UAAMA,MAAK,KAAK,YAAY,WAAW;AACvC,SAAK,QAAQ;AACb,SAAK,SAAS;AAAA,EAChB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,IAAI,YAA2B;AAC7B,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,IAAI,YAA2B;AAC7B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAEtC,UAAM,YAAqC,EAAE,MAAM,sBAAsB;AACzE,QAAI,KAAK,WAAW,OAAW,WAAU,SAAS,KAAK;AACvD,QAAI,KAAK,iBAAiB,OAAW,WAAU,gBAAgB,KAAK;AACpE,QAAI,KAAK,mBAAmB,OAAW,WAAU,kBAAkB,KAAK;AACxE,QAAI,KAAK,mBAAmB,OAAW,WAAU,mBAAmB,KAAK;AACzE,QAAI,KAAK,mBAAmB,OAAW,WAAU,mBAAmB,KAAK;AACzE,QAAI,KAAK,iCAAiC,QAAW;AACnD,gBAAU,mCAAmC,KAAK;AAAA,IACpD;AACA,QAAI,KAAK,iBAAiB,OAAW,WAAU,gBAAgB,KAAK;AAGpE,QAAI,OAAO,KAAK,SAAS,EAAE,SAAS,GAAG;AACrC,WAAK,uBAAuB,KAAK,SAAS;AAC1C,UAAI,CAAC,KAAK,sBAAsB,KAAM,MAAK,sBAAsB,QAAQ;AAAA,IAC3E;AAAA,EACF;AAAA;AAAA;AAAA;AAAA,EAKA,gBAAgB;AACd,SAAK,uBAAuB,KAAK,EAAE,MAAM,gBAAgB,CAAC;AAC1D,QAAI,CAAC,KAAK,sBAAsB,KAAM,MAAK,sBAAsB,QAAQ;AAAA,EAC3E;AAAA;AAAA,EAGA,MAAgB,MAAM;AACpB,UAAM,WAAW;AACjB,QAAI,UAAU;AAEd,WAAO,CAAC,KAAK,MAAM,UAAU,CAAC,KAAK,QAAQ;AACzC,UAAI;AACF,cAAM,KAAK,MAAM,KAAK,WAAW;AACjC,cAAM,KAAK,OAAO,EAAE;AACpB,kBAAU;AAAA,MACZ,SAAS,GAAG;AACV,YAAI,CAAC,KAAK,UAAU,CAAC,KAAK,MAAM,QAAQ;AACtC,cAAI,WAAW,UAAU;AACvB,kBAAM,IAAI,MAAM,yCAAyC,OAAO,cAAc,CAAC,EAAE;AAAA,UACnF;AAEA,gBAAM,oBAAoB,KAAK,IAAI,UAAU,GAAG,EAAE;AAClD;AAEA,eAAK,QAAQ;AAAA,YACX,gDAAgD,iBAAiB,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ;AAAA,UACzG;AACA,gBAAM,MAAM,oBAAoB,GAAI;AAAA,QACtC,OAAO;AACL,eAAK,QAAQ;AAAA,YACX,kDAAkD,CAAC,kBAAkB,KAAK,MAAM,MAAM,eAAe,KAAK,MAAM;AAAA,UAClH;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAM,aAAiC;AAIrC,QAAI,aAAa,KAAK,MAAM;AAC5B,QAAI,aAAa,KAAK,MAAM;AAC5B,QAAI,aAAa,KAAK,MAAM,WAAW,GAAG;AACxC,UAAI,eAAe,UAAa,KAAK,MAAM,SAAS,OAAW,cAAa;AAC5E,UAAI,eAAe,OAAW,cAAa;AAAA,IAC7C;AAGA,UAAM,2BACJ,KAAK,MAAM,YAAY,SAAS,cAAc,KAAK,aAAa,KAAK,MAAM,WAAW;AACxF,UAAM,oBAAoB,KAAK,MAAM,qBAAqB;AAE1D,UAAM,aAAsC;AAAA,MAC1C,aAAa,KAAK,MAAM;AAAA,MACxB,UAAU,KAAK,MAAM;AAAA,MACrB,cAAc,KAAK,MAAM;AAAA,MACzB,cAAc,KAAK,MAAM;AAAA,MACzB,kCAAkC,KAAK,MAAM;AAAA,MAC7C,kBAAkB;AAAA,MAClB,kBAAkB;AAAA,MAClB,iBACE,KAAK,MAAM,mBAAmB,UAAa,KAAK,MAAM,eAAe,SAAS,IAC1E,KAAK,UAAU,KAAK,MAAM,cAAc,IACxC;AAAA,MACN,oBAAoB;AAAA,MACpB,oBAAoB,KAAK,MAAM;AAAA,MAC/B,QAAQ,KAAK,MAAM;AAAA,MACnB,eAAe,KAAK,MAAM;AAAA,MAC1B,0BAA0B,KAAK,MAAM;AAAA,MACrC,eAAe,KAAK,MAAM;AAAA,MAC1B,gBAAgB,KAAK,MAAM;AAAA,MAC3B,cAAc,KAAK,MAAM;AAAA,MACzB,QAAQ,KAAK,MAAM;AAAA,MACnB,aAAa,KAAK,MAAM;AAAA,MACxB,uBAAuB,KAAK,MAAM;AAAA,MAClC,MAAM,KAAK,MAAM;AAAA,IACnB;AAEA,UAAM,MAAM,IAAI,IAAI,GAAG,KAAK,MAAM,OAAO,QAAQ;AAEjD,eAAW,CAAC,KAAK,KAAK,KAAK,OAAO,QAAQ,UAAU,GAAG;AACrD,UAAI,UAAU,UAAa,UAAU,KAAM;AAC3C,UAAI,OAAO,UAAU,WAAW;AAC9B,YAAI,aAAa,OAAO,KAAK,QAAQ,SAAS,OAAO;AAAA,MACvD,OAAO;AACL,YAAI,aAAa,OAAO,KAAK,OAAO,KAAK,CAAC;AAAA,MAC5C;AAAA,IACF;AAEA,UAAM,KAAK,IAAI,UAAU,KAAK;AAAA,MAC5B,SAAS;AAAA,QACP,eAAe,KAAK,MAAM;AAAA,QAC1B,gBAAgB;AAAA,QAChB,cAAc;AAAA,MAChB;AAAA,IACF,CAAC;AAED,UAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,SAAG,GAAG,QAAQ,MAAM,QAAQ,CAAC;AAC7B,SAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,SAAG,GAAG,SAAS,CAAC,SAAS,OAAO,IAAI,MAAM,sBAAsB,IAAI,EAAE,CAAC,CAAC;AAAA,IAC1E,CAAC;AAED,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,OAAO,IAAe;AAC1B,QAAI,UAAU;AACd,UAAM,oBAAoB,IAAI,gBAAgB;AAG9C,UAAM,YAAY,KAAK,KAAK,OAAO,eAAe;AAChD,YAAM,SAAS,IAAI,QAAc,CAAC,GAAG,WAAW;AAC9C,WAAG,KAAK,SAAS,CAAC,MAAM,WAAW;AACjC,cAAI,CAAC,SAAS;AACZ,iBAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,MAAM,EAAE;AAClE,mBAAO,IAAI,MAAM,kBAAkB,CAAC;AAAA,UACtC;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,QAAQ,aAAa,WAAW,MAAM,CAAC,CAAC;AAAA,IAC9D,CAAC;AAED,UAAM,WAAW,YAAY;AAC3B,YAAM,mBAAmB,KAAK,MAAO,KAAK,MAAM,aAAa,KAAK,MAAM,eAAgB,GAAI;AAC5F,YAAM,cAAc,IAAI,gBAAgB,KAAK,MAAM,YAAY,GAAG,gBAAgB;AAElF,YAAM,eAAe,aAAa,KAAK,WAAW;AAClD,YAAM,eAAe,aAAa,kBAAkB,MAAM;AAE1D,UAAI;AACF,eAAO,CAAC,KAAK,QAAQ;AACnB,gBAAM,SAAS,MAAM,QAAQ,KAAK,CAAC,KAAK,MAAM,KAAK,GAAG,cAAc,YAAY,CAAC;AAEjF,cAAI,WAAW,OAAW;AAC1B,cAAI,OAAO,KAAM;AAEjB,gBAAM,OAAO,OAAO;AAEpB,cAAI;AACJ,cAAI,SAAS,aAAa,gBAAgB;AACxC,qBAAS,YAAY,MAAM;AAAA,UAC7B,WAAW,KAAK,eAAe,KAAK,MAAM,cAAc,KAAK,aAAa,GAAG;AAG3E,qBAAS,YAAY,MAAM,KAAK,KAAK,MAAqB;AAAA,UAC5D,OAAO;AACL,kBAAM,IAAI,MAAM,sDAAsD;AAAA,UACxE;AAEA,qBAAW,SAAS,QAAQ;AAC1B,iBAAK,sBAAsB,MAAM,oBAAoB,MAAM;AAC3D,eAAG,KAAK,MAAM,KAAK,MAAM;AAAA,UAC3B;AAAA,QACF;AAAA,MACF,UAAE;AACA,kBAAU;AACV,YAAI;AACF,aAAG,KAAK,aAAa,SAAS;AAAA,QAChC,QAAQ;AAAA,QAER;AACA,kBAAU,OAAO;AAAA,MACnB;AAAA,IACF;AAEA,QAAI,iBAAqE;AACzE,UAAM,aAAa,KAAK,KAAK,OAAO,eAAe;AACjD,YAAM,gBAAgB,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3D,yBAAiB,CAAC,KAAK,aAAa;AAClC,cAAI,UAAU;AACZ,iBAAK,QAAQ,MAAM,2CAA2C;AAC9D;AAAA,UACF;AACA,cAAI;AACF,kBAAM,OAAO,KAAK,MAAM,IAAI,SAAS,CAAC;AACtC,iBAAK,oBAAoB,IAAI;AAC7B,gBAAI,KAAK,UAAU,SAAS;AAC1B,sBAAQ;AAAA,YACV;AAAA,UACF,SAAS,KAAK;AACZ,iBAAK,QAAQ,MAAM,yCAAyC,GAAG,EAAE;AACjE,mBAAO,GAAG;AAAA,UACZ;AAAA,QACF;AACA,WAAG,GAAG,WAAW,cAAc;AAAA,MACjC,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,eAAe,aAAa,WAAW,MAAM,CAAC,CAAC;AAAA,IACrE,CAAC;AAED,UAAM,aAAa,KAAK,KAAK,OAAO,eAAe;AAEjD,aAAO,KAAK,uBAAuB,SAAS,GAAG;AAC7C,cAAM,MAAM,KAAK,uBAAuB,MAAM;AAC9C,WAAG,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,MAC7B;AAEA,aAAO,CAAC,WAAW,OAAO,SAAS;AACjC,cAAM,QAAQ,KAAK,CAAC,KAAK,sBAAsB,OAAO,aAAa,WAAW,MAAM,CAAC,CAAC;AACtF,YAAI,WAAW,OAAO,QAAS;AAE/B,aAAK,wBAAwB,IAAI,OAAO;AACxC,eAAO,KAAK,uBAAuB,SAAS,GAAG;AAC7C,gBAAM,MAAM,KAAK,uBAAuB,MAAM;AAC9C,aAAG,KAAK,KAAK,UAAU,GAAG,CAAC;AAAA,QAC7B;AAAA,MACF;AAAA,IACF,CAAC;AAED,QAAI;AACF,YAAM,QAAQ,IAAI,CAAC,SAAS,GAAG,WAAW,QAAQ,UAAU,MAAM,CAAC;AAAA,IACrE,UAAE;AACA,gBAAU;AACV,wBAAkB,MAAM;AACxB,iBAAW,OAAO;AAClB,iBAAW,OAAO;AAClB,UAAI,eAAgB,IAAG,IAAI,WAAW,cAAc;AACpD,UAAI;AACF,WAAG,MAAM;AAAA,MACX,QAAQ;AAAA,MAER;AAAA,IACF;AAAA,EACF;AAAA,EAEA,mBAAmB,OAA+C;AAChE,QAAI,MAAM,WAAW,EAAG,QAAO;AAC/B,WAAO,MAAM,OAAO,CAAC,KAAK,MAAM,OAAO,EAAE,cAAc,IAAI,CAAC,IAAI,MAAM;AAAA,EACxE;AAAA,EAEA,oBAAoB,MAA0B;AAC5C,UAAM,cAAc,KAAK;AAEzB,QAAI,gBAAgB,SAAS;AAC3B,WAAK,aAAa,KAAK,MAAM;AAC7B,WAAK,aAAa,KAAK,cAAc;AACrC,WAAK,QAAQ;AAAA,QACX,iCAAiC,KAAK,UAAU,eAAe,KAAK,UAAU;AAAA,MAChF;AACA;AAAA,IACF;AAEA,QAAI,gBAAgB,iBAAiB;AACnC,WAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,gBAAgB,CAAC;AAC5D;AAAA,IACF;AAEA,QAAI,gBAAgB,eAAe;AACjC,WAAK,QAAQ;AAAA,QACX,gDAAgD,KAAK,sBAAsB,sBAAsB,KAAK,wBAAwB;AAAA,MAChI;AACA;AAAA,IACF;AAEA,QAAI,gBAAgB,QAAQ;AAC1B;AAAA,IACF;AAEA,UAAM,QAAQ,KAAK,SAAS,CAAC;AAC7B,UAAM,YAAY,QAAQ,KAAK,WAAW;AAC1C,UAAM,kBAAkB,QAAQ,KAAK,iBAAiB;AACtD,UAAM,YAAY,KAAK,aAAa;AACpC,UAAM,aAAa,KAAK,cAAc;AACtC,UAAM,WAAW,kBAAkB,KAAK,iBAAiB,IAAI;AAC7D,UAAM,WAAW,mBAAmB,IAAI;AAIxC,UAAM,aAAa,MAAM;AAAA,MAAI,CAAC,SAC5B,kBAAkB;AAAA,QAChB,MAAM,KAAK,QAAQ;AAAA,QACnB,YAAY,KAAK,SAAS,KAAK,MAAO,KAAK;AAAA,QAC3C,UAAU,KAAK,OAAO,KAAK,MAAO,KAAK;AAAA,QACvC,YAAY,KAAK,cAAc;AAAA,QAC/B,iBAAiB,KAAK;AAAA,MACxB,CAAC;AAAA,IACH;AAEA,QAAI,YAAY;AAChB,QAAI,UAAU;AACd,QAAI,aAAa;AAGjB,QAAI,WAAW,SAAS,GAAG;AACzB,YAAM,cAAc,WAAW,IAAI,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,GAAG;AAC1D,kBAAY,WAAW,CAAC,EAAG,aAAa;AACxC,gBAAU,WAAW,WAAW,SAAS,CAAC,EAAG,WAAW;AACxD,mBAAa,KAAK,mBAAmB,UAAU;AAE/C,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,IAAI,gBAAgB;AAAA,QAC1B,cAAc;AAAA,UACZ;AAAA,YACE;AAAA,YACA,MAAM;AAAA,YACN;AAAA,YACA;AAAA,YACA;AAAA,YACA,OAAO;AAAA,YACP,GAAI,WAAW,EAAE,SAAS,IAAI,CAAC;AAAA,UACjC;AAAA,QACF;AAAA,MACF,CAAC;AAAA,IACH;AAIA,QAAI,WAAW;AACb,UAAI,KAAK,4BAA4B,GAAG;AACtC,aAAK,0BAA0B;AAAA,MACjC;AAEA,YAAM,iBAAiB,WAAW;AAAA,QAChC,CAAC,MAAM,EAAE,cAAc,UAAa,EAAE,aAAa,KAAK;AAAA,MAC1D;AACA,YAAM,sBAAsB,KAAK,mBAAmB,cAAc;AAElE,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,IAAI,gBAAgB;AAAA,QAC1B,cAAc;AAAA,UACZ;AAAA,YACE;AAAA,YACA,MAAM;AAAA,YACN,WAAW,KAAK;AAAA,YAChB;AAAA,YACA,YAAY;AAAA,YACZ,OAAO;AAAA,YACP,GAAI,WAAW,EAAE,SAAS,IAAI,CAAC;AAAA,UACjC;AAAA,QACF;AAAA,MACF,CAAC;AACD,WAAK,0BAA0B;AAAA,IACjC;AAIA,UAAM,sBAAsB,KAAK,MAAM,gBAAgB,QAAQ,CAAC;AAChE,QAAI,aAAa,CAAC,qBAAqB;AACrC,WAAK,MAAM,IAAI;AAAA,QACb,MAAM,IAAI,gBAAgB;AAAA,QAC1B,cAAc;AAAA,UACZ;AAAA,YACE;AAAA,YACA,MAAM;AAAA,YACN;AAAA,YACA;AAAA,YACA;AAAA,YACA,OAAO;AAAA,YACP,GAAI,WAAW,EAAE,SAAS,IAAI,CAAC;AAAA,UACjC;AAAA,QACF;AAAA,MACF,CAAC;AAED,WAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,cAAc,CAAC;AAC1D,UAAI,KAAK,qBAAqB,GAAG;AAC/B,aAAK,MAAM,IAAI;AAAA,UACb,MAAM,IAAI,gBAAgB;AAAA;AAAA;AAAA;AAAA,UAI1B,WAAW,KAAK,cAAc;AAAA,UAC9B,kBAAkB;AAAA,YAChB,eAAe,KAAK;AAAA,UACtB;AAAA,QACF,CAAC;AACD,aAAK,qBAAqB;AAC1B,aAAK,0BAA0B;AAAA,MACjC;AAAA,IACF;AAAA,EACF;AACF;","names":["stt"]}
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@livekit/agents-plugin-assemblyai",
|
|
3
|
-
"version": "1.5.
|
|
3
|
+
"version": "1.5.2",
|
|
4
4
|
"description": "AssemblyAI plugin for LiveKit Agents for Node.js",
|
|
5
5
|
"main": "dist/index.js",
|
|
6
6
|
"require": "dist/index.cjs",
|
|
@@ -25,21 +25,21 @@
|
|
|
25
25
|
"README.md"
|
|
26
26
|
],
|
|
27
27
|
"devDependencies": {
|
|
28
|
-
"@livekit/rtc-node": "^0.13.
|
|
28
|
+
"@livekit/rtc-node": "^0.13.31",
|
|
29
29
|
"@microsoft/api-extractor": "^7.35.0",
|
|
30
30
|
"@types/ws": "^8.5.10",
|
|
31
31
|
"tsup": "^8.3.5",
|
|
32
32
|
"typescript": "^5.0.0",
|
|
33
|
-
"@livekit/agents": "1.5.
|
|
34
|
-
"@livekit/agents-
|
|
35
|
-
"@livekit/agents-
|
|
33
|
+
"@livekit/agents": "1.5.2",
|
|
34
|
+
"@livekit/agents-plugins-test": "1.5.2",
|
|
35
|
+
"@livekit/agents-plugin-silero": "1.5.2"
|
|
36
36
|
},
|
|
37
37
|
"dependencies": {
|
|
38
38
|
"ws": "^8.16.0"
|
|
39
39
|
},
|
|
40
40
|
"peerDependencies": {
|
|
41
|
-
"@livekit/rtc-node": "^0.13.
|
|
42
|
-
"@livekit/agents": "1.5.
|
|
41
|
+
"@livekit/rtc-node": "^0.13.31",
|
|
42
|
+
"@livekit/agents": "1.5.2"
|
|
43
43
|
},
|
|
44
44
|
"scripts": {
|
|
45
45
|
"build": "tsup --onSuccess \"pnpm build:types\"",
|
package/src/stt.ts
CHANGED
|
@@ -6,6 +6,8 @@ import {
|
|
|
6
6
|
type APIConnectOptions,
|
|
7
7
|
type AudioBuffer,
|
|
8
8
|
AudioByteStream,
|
|
9
|
+
ChatMessage,
|
|
10
|
+
type ConversationItemAddedEvent,
|
|
9
11
|
Future,
|
|
10
12
|
Task,
|
|
11
13
|
createTimedString,
|
|
@@ -118,6 +120,13 @@ export interface STTOptions {
|
|
|
118
120
|
* or `max_accuracy`. Explicit turn-silence values still take precedence over mode defaults.
|
|
119
121
|
*/
|
|
120
122
|
mode?: 'min_latency' | 'balanced' | 'max_accuracy';
|
|
123
|
+
/**
|
|
124
|
+
* When the model supports it, let an `AgentSession` push each assistant reply into
|
|
125
|
+
* `agentContext` so it is carried into the model's conversation context. Defaults to false;
|
|
126
|
+
* set true to enable. Prior user turns are carried automatically by the model regardless of
|
|
127
|
+
* this flag. Ignored on models without context support.
|
|
128
|
+
*/
|
|
129
|
+
agentContextCarryover?: boolean;
|
|
121
130
|
baseUrl: string;
|
|
122
131
|
}
|
|
123
132
|
|
|
@@ -133,6 +142,9 @@ const defaultSTTOptions: STTOptions = {
|
|
|
133
142
|
export class STT extends stt.STT {
|
|
134
143
|
#opts: STTOptions;
|
|
135
144
|
#streams = new Set<WeakRef<SpeechStream>>();
|
|
145
|
+
// set (user + session))
|
|
146
|
+
#userKeyterms: string[];
|
|
147
|
+
#sessionKeyterms: string[] = [];
|
|
136
148
|
label = 'assemblyai.STT';
|
|
137
149
|
|
|
138
150
|
get model(): string {
|
|
@@ -144,10 +156,20 @@ export class STT extends stt.STT {
|
|
|
144
156
|
}
|
|
145
157
|
|
|
146
158
|
constructor(opts: Partial<STTOptions> = {}) {
|
|
159
|
+
// u3-rt-pro family — "u3-pro" is normalized below — and is opt-in via the user)
|
|
160
|
+
const rawModel = opts.speechModel ?? defaultSTTOptions.speechModel;
|
|
161
|
+
const supportsCarryover = isU3ProModel(rawModel) || rawModel === 'u3-pro';
|
|
162
|
+
if (opts.agentContextCarryover && !supportsCarryover) {
|
|
163
|
+
log().warn(
|
|
164
|
+
`agentContextCarryover is enabled but model '${rawModel}' does not support it; ignoring`,
|
|
165
|
+
);
|
|
166
|
+
}
|
|
147
167
|
super({
|
|
148
168
|
streaming: true,
|
|
149
169
|
interimResults: true,
|
|
150
170
|
alignedTranscript: 'word',
|
|
171
|
+
keyterms: true,
|
|
172
|
+
chatContext: (opts.agentContextCarryover ?? false) && supportsCarryover,
|
|
151
173
|
});
|
|
152
174
|
|
|
153
175
|
if (opts.speechModel === 'u3-pro') {
|
|
@@ -189,6 +211,7 @@ export class STT extends stt.STT {
|
|
|
189
211
|
apiKey,
|
|
190
212
|
minTurnSilence,
|
|
191
213
|
};
|
|
214
|
+
this.#userKeyterms = [...(this.#opts.keytermsPrompt ?? [])];
|
|
192
215
|
}
|
|
193
216
|
|
|
194
217
|
// eslint-disable-next-line @typescript-eslint/no-unused-vars
|
|
@@ -197,17 +220,51 @@ export class STT extends stt.STT {
|
|
|
197
220
|
}
|
|
198
221
|
|
|
199
222
|
updateOptions(opts: Partial<STTOptions>) {
|
|
200
|
-
|
|
223
|
+
// session keyterms so a user update doesn't drop them)
|
|
224
|
+
const nextOpts = { ...opts };
|
|
225
|
+
if (nextOpts.keytermsPrompt !== undefined) {
|
|
226
|
+
this.#userKeyterms = [...nextOpts.keytermsPrompt];
|
|
227
|
+
nextOpts.keytermsPrompt = [...new Set([...this.#userKeyterms, ...this.#sessionKeyterms])];
|
|
228
|
+
}
|
|
229
|
+
this.#opts = { ...this.#opts, ...nextOpts };
|
|
230
|
+
for (const ref of this.#streams) {
|
|
231
|
+
const stream = ref.deref();
|
|
232
|
+
if (stream) {
|
|
233
|
+
stream.updateOptions(nextOpts);
|
|
234
|
+
} else {
|
|
235
|
+
this.#streams.delete(ref);
|
|
236
|
+
}
|
|
237
|
+
}
|
|
238
|
+
}
|
|
239
|
+
|
|
240
|
+
override _updateSessionKeyterms(keyterms: string[]): void {
|
|
241
|
+
if (
|
|
242
|
+
keyterms.length === this.#sessionKeyterms.length &&
|
|
243
|
+
keyterms.every((t, i) => t === this.#sessionKeyterms[i])
|
|
244
|
+
) {
|
|
245
|
+
return;
|
|
246
|
+
}
|
|
247
|
+
this.#sessionKeyterms = [...keyterms];
|
|
248
|
+
const merged = [...new Set([...this.#userKeyterms, ...keyterms])];
|
|
249
|
+
this.#opts.keytermsPrompt = merged;
|
|
250
|
+
// applied live via the stream's UpdateConfiguration (no reconnect)
|
|
201
251
|
for (const ref of this.#streams) {
|
|
202
252
|
const stream = ref.deref();
|
|
203
253
|
if (stream) {
|
|
204
|
-
stream.updateOptions(
|
|
254
|
+
stream.updateOptions({ keytermsPrompt: merged });
|
|
205
255
|
} else {
|
|
206
256
|
this.#streams.delete(ref);
|
|
207
257
|
}
|
|
208
258
|
}
|
|
209
259
|
}
|
|
210
260
|
|
|
261
|
+
override _pushConversationItem(ev: ConversationItemAddedEvent): void {
|
|
262
|
+
const chatItem = ev.item;
|
|
263
|
+
if (chatItem instanceof ChatMessage && chatItem.role === 'assistant' && chatItem.textContent) {
|
|
264
|
+
this.updateOptions({ agentContext: chatItem.textContent });
|
|
265
|
+
}
|
|
266
|
+
}
|
|
267
|
+
|
|
211
268
|
stream(options?: { connOptions?: APIConnectOptions }): SpeechStream {
|
|
212
269
|
const stream = new SpeechStream(this, this.#opts, options?.connOptions);
|
|
213
270
|
this.#streams.add(new WeakRef(stream));
|
|
@@ -339,7 +396,7 @@ export class SpeechStream extends stt.SpeechStream {
|
|
|
339
396
|
min_turn_silence: minSilence,
|
|
340
397
|
max_turn_silence: maxSilence,
|
|
341
398
|
keyterms_prompt:
|
|
342
|
-
this.#opts.keytermsPrompt !== undefined
|
|
399
|
+
this.#opts.keytermsPrompt !== undefined && this.#opts.keytermsPrompt.length > 0
|
|
343
400
|
? JSON.stringify(this.#opts.keytermsPrompt)
|
|
344
401
|
: undefined,
|
|
345
402
|
language_detection: languageDetection,
|