@framers/agentos 0.1.46 → 0.1.48
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/README.md +7 -7
- package/dist/cognitive_substrate/IGMI.d.ts +2 -0
- package/dist/cognitive_substrate/IGMI.d.ts.map +1 -1
- package/dist/cognitive_substrate/IGMI.js.map +1 -1
- package/dist/cognitive_substrate/personas/IPersonaDefinition.d.ts +2 -0
- package/dist/cognitive_substrate/personas/IPersonaDefinition.d.ts.map +1 -1
- package/dist/core/audio/AdaptiveVAD.d.ts +147 -0
- package/dist/core/audio/AdaptiveVAD.d.ts.map +1 -0
- package/dist/core/audio/AdaptiveVAD.js +239 -0
- package/dist/core/audio/AdaptiveVAD.js.map +1 -0
- package/dist/core/audio/AudioProcessor.d.ts +121 -0
- package/dist/core/audio/AudioProcessor.d.ts.map +1 -0
- package/dist/core/audio/AudioProcessor.js +299 -0
- package/dist/core/audio/AudioProcessor.js.map +1 -0
- package/dist/core/audio/EnvironmentalCalibrator.d.ts +187 -0
- package/dist/core/audio/EnvironmentalCalibrator.d.ts.map +1 -0
- package/dist/core/audio/EnvironmentalCalibrator.js +360 -0
- package/dist/core/audio/EnvironmentalCalibrator.js.map +1 -0
- package/dist/core/audio/SilenceDetector.d.ts +111 -0
- package/dist/core/audio/SilenceDetector.d.ts.map +1 -0
- package/dist/core/audio/SilenceDetector.js +178 -0
- package/dist/core/audio/SilenceDetector.js.map +1 -0
- package/dist/core/llm/IPromptEngine.d.ts +2 -0
- package/dist/core/llm/IPromptEngine.d.ts.map +1 -1
- package/dist/core/llm/IPromptEngine.js.map +1 -1
- package/dist/core/llm/providers/implementations/OllamaProvider.js +2 -2
- package/dist/core/llm/providers/implementations/OllamaProvider.js.map +1 -1
- package/dist/core/llm/providers/implementations/OpenAIProvider.d.ts.map +1 -1
- package/dist/core/llm/providers/implementations/OpenAIProvider.js +8 -1
- package/dist/core/llm/providers/implementations/OpenAIProvider.js.map +1 -1
- package/dist/discovery/CapabilityIndex.d.ts.map +1 -1
- package/dist/discovery/CapabilityIndex.js +18 -4
- package/dist/discovery/CapabilityIndex.js.map +1 -1
- package/dist/extensions/ExtensionManager.d.ts.map +1 -1
- package/dist/extensions/ExtensionManager.js +9 -1
- package/dist/extensions/ExtensionManager.js.map +1 -1
- package/dist/extensions/types.d.ts +20 -2
- package/dist/extensions/types.d.ts.map +1 -1
- package/dist/extensions/types.js +4 -0
- package/dist/extensions/types.js.map +1 -1
- package/dist/index.d.ts +2 -0
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +4 -0
- package/dist/index.js.map +1 -1
- package/dist/memory/CognitiveMemoryManager.d.ts +133 -0
- package/dist/memory/CognitiveMemoryManager.d.ts.map +1 -0
- package/dist/memory/CognitiveMemoryManager.js +519 -0
- package/dist/memory/CognitiveMemoryManager.js.map +1 -0
- package/dist/memory/config.d.ts +135 -0
- package/dist/memory/config.d.ts.map +1 -0
- package/dist/memory/config.js +28 -0
- package/dist/memory/config.js.map +1 -0
- package/dist/memory/consolidation/ConsolidationPipeline.d.ts +70 -0
- package/dist/memory/consolidation/ConsolidationPipeline.d.ts.map +1 -0
- package/dist/memory/consolidation/ConsolidationPipeline.js +301 -0
- package/dist/memory/consolidation/ConsolidationPipeline.js.map +1 -0
- package/dist/memory/context/CompactionEngine.d.ts +21 -0
- package/dist/memory/context/CompactionEngine.d.ts.map +1 -0
- package/dist/memory/context/CompactionEngine.js +35 -0
- package/dist/memory/context/CompactionEngine.js.map +1 -0
- package/dist/memory/context/CompactionLog.d.ts +50 -0
- package/dist/memory/context/CompactionLog.d.ts.map +1 -0
- package/dist/memory/context/CompactionLog.js +139 -0
- package/dist/memory/context/CompactionLog.js.map +1 -0
- package/dist/memory/context/ContextWindowManager.d.ts +114 -0
- package/dist/memory/context/ContextWindowManager.d.ts.map +1 -0
- package/dist/memory/context/ContextWindowManager.js +225 -0
- package/dist/memory/context/ContextWindowManager.js.map +1 -0
- package/dist/memory/context/RollingSummaryChain.d.ts +53 -0
- package/dist/memory/context/RollingSummaryChain.d.ts.map +1 -0
- package/dist/memory/context/RollingSummaryChain.js +193 -0
- package/dist/memory/context/RollingSummaryChain.js.map +1 -0
- package/dist/memory/context/index.d.ts +15 -0
- package/dist/memory/context/index.d.ts.map +1 -0
- package/dist/memory/context/index.js +14 -0
- package/dist/memory/context/index.js.map +1 -0
- package/dist/memory/context/strategies/HierarchicalStrategy.d.ts +25 -0
- package/dist/memory/context/strategies/HierarchicalStrategy.d.ts.map +1 -0
- package/dist/memory/context/strategies/HierarchicalStrategy.js +240 -0
- package/dist/memory/context/strategies/HierarchicalStrategy.js.map +1 -0
- package/dist/memory/context/strategies/HybridStrategy.d.ts +26 -0
- package/dist/memory/context/strategies/HybridStrategy.d.ts.map +1 -0
- package/dist/memory/context/strategies/HybridStrategy.js +245 -0
- package/dist/memory/context/strategies/HybridStrategy.js.map +1 -0
- package/dist/memory/context/strategies/SlidingSummaryStrategy.d.ts +22 -0
- package/dist/memory/context/strategies/SlidingSummaryStrategy.d.ts.map +1 -0
- package/dist/memory/context/strategies/SlidingSummaryStrategy.js +203 -0
- package/dist/memory/context/strategies/SlidingSummaryStrategy.js.map +1 -0
- package/dist/memory/context/types.d.ts +120 -0
- package/dist/memory/context/types.d.ts.map +1 -0
- package/dist/memory/context/types.js +17 -0
- package/dist/memory/context/types.js.map +1 -0
- package/dist/memory/decay/DecayModel.d.ts +87 -0
- package/dist/memory/decay/DecayModel.d.ts.map +1 -0
- package/dist/memory/decay/DecayModel.js +117 -0
- package/dist/memory/decay/DecayModel.js.map +1 -0
- package/dist/memory/decay/RetrievalPriorityScorer.d.ts +63 -0
- package/dist/memory/decay/RetrievalPriorityScorer.d.ts.map +1 -0
- package/dist/memory/decay/RetrievalPriorityScorer.js +110 -0
- package/dist/memory/decay/RetrievalPriorityScorer.js.map +1 -0
- package/dist/memory/encoding/ContentFeatureDetector.d.ts +37 -0
- package/dist/memory/encoding/ContentFeatureDetector.d.ts.map +1 -0
- package/dist/memory/encoding/ContentFeatureDetector.js +176 -0
- package/dist/memory/encoding/ContentFeatureDetector.js.map +1 -0
- package/dist/memory/encoding/EncodingModel.d.ts +67 -0
- package/dist/memory/encoding/EncodingModel.d.ts.map +1 -0
- package/dist/memory/encoding/EncodingModel.js +171 -0
- package/dist/memory/encoding/EncodingModel.js.map +1 -0
- package/dist/memory/extension/CognitiveMemoryExtension.d.ts +18 -0
- package/dist/memory/extension/CognitiveMemoryExtension.d.ts.map +1 -0
- package/dist/memory/extension/CognitiveMemoryExtension.js +131 -0
- package/dist/memory/extension/CognitiveMemoryExtension.js.map +1 -0
- package/dist/memory/graph/GraphologyMemoryGraph.d.ts +29 -0
- package/dist/memory/graph/GraphologyMemoryGraph.d.ts.map +1 -0
- package/dist/memory/graph/GraphologyMemoryGraph.js +224 -0
- package/dist/memory/graph/GraphologyMemoryGraph.js.map +1 -0
- package/dist/memory/graph/IMemoryGraph.d.ts +70 -0
- package/dist/memory/graph/IMemoryGraph.d.ts.map +1 -0
- package/dist/memory/graph/IMemoryGraph.js +15 -0
- package/dist/memory/graph/IMemoryGraph.js.map +1 -0
- package/dist/memory/graph/KnowledgeGraphMemoryGraph.d.ts +38 -0
- package/dist/memory/graph/KnowledgeGraphMemoryGraph.d.ts.map +1 -0
- package/dist/memory/graph/KnowledgeGraphMemoryGraph.js +264 -0
- package/dist/memory/graph/KnowledgeGraphMemoryGraph.js.map +1 -0
- package/dist/memory/graph/SpreadingActivation.d.ts +36 -0
- package/dist/memory/graph/SpreadingActivation.d.ts.map +1 -0
- package/dist/memory/graph/SpreadingActivation.js +100 -0
- package/dist/memory/graph/SpreadingActivation.js.map +1 -0
- package/dist/memory/index.d.ts +52 -0
- package/dist/memory/index.d.ts.map +1 -0
- package/dist/memory/index.js +43 -0
- package/dist/memory/index.js.map +1 -0
- package/dist/memory/observation/MemoryObserver.d.ts +59 -0
- package/dist/memory/observation/MemoryObserver.d.ts.map +1 -0
- package/dist/memory/observation/MemoryObserver.js +136 -0
- package/dist/memory/observation/MemoryObserver.js.map +1 -0
- package/dist/memory/observation/MemoryReflector.d.ts +51 -0
- package/dist/memory/observation/MemoryReflector.d.ts.map +1 -0
- package/dist/memory/observation/MemoryReflector.js +184 -0
- package/dist/memory/observation/MemoryReflector.js.map +1 -0
- package/dist/memory/observation/ObservationBuffer.d.ts +51 -0
- package/dist/memory/observation/ObservationBuffer.d.ts.map +1 -0
- package/dist/memory/observation/ObservationBuffer.js +81 -0
- package/dist/memory/observation/ObservationBuffer.js.map +1 -0
- package/dist/memory/prompt/MemoryFormatters.d.ts +22 -0
- package/dist/memory/prompt/MemoryFormatters.d.ts.map +1 -0
- package/dist/memory/prompt/MemoryFormatters.js +113 -0
- package/dist/memory/prompt/MemoryFormatters.js.map +1 -0
- package/dist/memory/prompt/MemoryPromptAssembler.d.ts +40 -0
- package/dist/memory/prompt/MemoryPromptAssembler.d.ts.map +1 -0
- package/dist/memory/prompt/MemoryPromptAssembler.js +180 -0
- package/dist/memory/prompt/MemoryPromptAssembler.js.map +1 -0
- package/dist/memory/prospective/ProspectiveMemoryManager.d.ts +83 -0
- package/dist/memory/prospective/ProspectiveMemoryManager.d.ts.map +1 -0
- package/dist/memory/prospective/ProspectiveMemoryManager.js +128 -0
- package/dist/memory/prospective/ProspectiveMemoryManager.js.map +1 -0
- package/dist/memory/store/MemoryStore.d.ts +76 -0
- package/dist/memory/store/MemoryStore.d.ts.map +1 -0
- package/dist/memory/store/MemoryStore.js +320 -0
- package/dist/memory/store/MemoryStore.js.map +1 -0
- package/dist/memory/types.d.ts +184 -0
- package/dist/memory/types.d.ts.map +1 -0
- package/dist/memory/types.js +14 -0
- package/dist/memory/types.js.map +1 -0
- package/dist/memory/working/CognitiveWorkingMemory.d.ts +87 -0
- package/dist/memory/working/CognitiveWorkingMemory.d.ts.map +1 -0
- package/dist/memory/working/CognitiveWorkingMemory.js +230 -0
- package/dist/memory/working/CognitiveWorkingMemory.js.map +1 -0
- package/dist/rag/EmbeddingManager.d.ts.map +1 -1
- package/dist/rag/EmbeddingManager.js +8 -1
- package/dist/rag/EmbeddingManager.js.map +1 -1
- package/dist/rag/HydeRetriever.d.ts +111 -0
- package/dist/rag/HydeRetriever.d.ts.map +1 -0
- package/dist/rag/HydeRetriever.js +143 -0
- package/dist/rag/HydeRetriever.js.map +1 -0
- package/dist/rag/IRetrievalAugmentor.d.ts +15 -0
- package/dist/rag/IRetrievalAugmentor.d.ts.map +1 -1
- package/dist/rag/index.d.ts +1 -0
- package/dist/rag/index.d.ts.map +1 -1
- package/dist/rag/index.js +2 -0
- package/dist/rag/index.js.map +1 -1
- package/dist/speech/SpeechProviderRegistry.d.ts +17 -0
- package/dist/speech/SpeechProviderRegistry.d.ts.map +1 -0
- package/dist/speech/SpeechProviderRegistry.js +47 -0
- package/dist/speech/SpeechProviderRegistry.js.map +1 -0
- package/dist/speech/SpeechRuntime.d.ts +30 -0
- package/dist/speech/SpeechRuntime.d.ts.map +1 -0
- package/dist/speech/SpeechRuntime.js +156 -0
- package/dist/speech/SpeechRuntime.js.map +1 -0
- package/dist/speech/SpeechSession.d.ts +37 -0
- package/dist/speech/SpeechSession.d.ts.map +1 -0
- package/dist/speech/SpeechSession.js +285 -0
- package/dist/speech/SpeechSession.js.map +1 -0
- package/dist/speech/audio.d.ts +3 -0
- package/dist/speech/audio.d.ts.map +1 -0
- package/dist/speech/audio.js +39 -0
- package/dist/speech/audio.js.map +1 -0
- package/dist/speech/index.d.ts +11 -0
- package/dist/speech/index.d.ts.map +1 -0
- package/dist/speech/index.js +11 -0
- package/dist/speech/index.js.map +1 -0
- package/dist/speech/providerCatalog.d.ts +7 -0
- package/dist/speech/providerCatalog.d.ts.map +1 -0
- package/dist/speech/providerCatalog.js +243 -0
- package/dist/speech/providerCatalog.js.map +1 -0
- package/dist/speech/providers/BuiltInAdaptiveVadProvider.d.ts +20 -0
- package/dist/speech/providers/BuiltInAdaptiveVadProvider.d.ts.map +1 -0
- package/dist/speech/providers/BuiltInAdaptiveVadProvider.js +29 -0
- package/dist/speech/providers/BuiltInAdaptiveVadProvider.js.map +1 -0
- package/dist/speech/providers/ElevenLabsTextToSpeechProvider.d.ts +20 -0
- package/dist/speech/providers/ElevenLabsTextToSpeechProvider.d.ts.map +1 -0
- package/dist/speech/providers/ElevenLabsTextToSpeechProvider.js +95 -0
- package/dist/speech/providers/ElevenLabsTextToSpeechProvider.js.map +1 -0
- package/dist/speech/providers/OpenAITextToSpeechProvider.d.ts +20 -0
- package/dist/speech/providers/OpenAITextToSpeechProvider.d.ts.map +1 -0
- package/dist/speech/providers/OpenAITextToSpeechProvider.js +75 -0
- package/dist/speech/providers/OpenAITextToSpeechProvider.js.map +1 -0
- package/dist/speech/providers/OpenAIWhisperSpeechToTextProvider.d.ts +18 -0
- package/dist/speech/providers/OpenAIWhisperSpeechToTextProvider.d.ts.map +1 -0
- package/dist/speech/providers/OpenAIWhisperSpeechToTextProvider.js +109 -0
- package/dist/speech/providers/OpenAIWhisperSpeechToTextProvider.js.map +1 -0
- package/dist/speech/types.d.ts +272 -0
- package/dist/speech/types.d.ts.map +1 -0
- package/dist/speech/types.js +2 -0
- package/dist/speech/types.js.map +1 -0
- package/package.json +22 -2
|
@@ -0,0 +1,285 @@
|
|
|
1
|
+
import { EventEmitter } from 'node:events';
|
|
2
|
+
import { AdaptiveVAD } from '../core/audio/AdaptiveVAD.js';
|
|
3
|
+
import { EnvironmentalCalibrator } from '../core/audio/EnvironmentalCalibrator.js';
|
|
4
|
+
import { SilenceDetector } from '../core/audio/SilenceDetector.js';
|
|
5
|
+
import { encodeFloat32ToWav } from './audio.js';
|
|
6
|
+
export class SpeechSession extends EventEmitter {
|
|
7
|
+
on(event, listener) {
|
|
8
|
+
return super.on(event, listener);
|
|
9
|
+
}
|
|
10
|
+
emit(event, ...args) {
|
|
11
|
+
return super.emit(event, ...args);
|
|
12
|
+
}
|
|
13
|
+
constructor(config = {}, providers = {}) {
|
|
14
|
+
super();
|
|
15
|
+
this.state = 'idle';
|
|
16
|
+
this.capturedFrames = [];
|
|
17
|
+
this.currentSpeechStartedAt = null;
|
|
18
|
+
this.wakeWordDetected = false;
|
|
19
|
+
this.transcriptionPromise = null;
|
|
20
|
+
this.lastExternalVadSpeech = false;
|
|
21
|
+
this.config = {
|
|
22
|
+
mode: config.mode ?? 'vad',
|
|
23
|
+
sampleRate: config.sampleRate ?? 16000,
|
|
24
|
+
frameDurationMs: config.frameDurationMs ?? 20,
|
|
25
|
+
audioFileName: config.audioFileName ?? 'speech-session.wav',
|
|
26
|
+
autoTranscribeOnSpeechEnd: config.autoTranscribeOnSpeechEnd ?? true,
|
|
27
|
+
...config,
|
|
28
|
+
};
|
|
29
|
+
this.providers = providers;
|
|
30
|
+
this.calibrator = new EnvironmentalCalibrator({
|
|
31
|
+
sampleRate: this.config.sampleRate,
|
|
32
|
+
});
|
|
33
|
+
this.vad = new AdaptiveVAD(this.config.vad ?? {}, this.calibrator, this.config.frameDurationMs);
|
|
34
|
+
this.silenceDetector = new SilenceDetector(this.config.silence ?? {});
|
|
35
|
+
this.bindVadEvents();
|
|
36
|
+
this.bindSilenceEvents();
|
|
37
|
+
}
|
|
38
|
+
getState() {
|
|
39
|
+
return this.state;
|
|
40
|
+
}
|
|
41
|
+
async start() {
|
|
42
|
+
if (this.state === 'closed') {
|
|
43
|
+
throw new Error('SpeechSession is closed.');
|
|
44
|
+
}
|
|
45
|
+
if (this.config.mode === 'wake-word') {
|
|
46
|
+
if (!this.providers.wakeWord) {
|
|
47
|
+
throw new Error('Wake-word mode requires a wake-word provider.');
|
|
48
|
+
}
|
|
49
|
+
this.wakeWordDetected = false;
|
|
50
|
+
this.changeState('wake-listening');
|
|
51
|
+
return;
|
|
52
|
+
}
|
|
53
|
+
this.changeState('listening');
|
|
54
|
+
}
|
|
55
|
+
async stop() {
|
|
56
|
+
this.resetBuffers();
|
|
57
|
+
this.changeState('idle');
|
|
58
|
+
}
|
|
59
|
+
async flush(reason = 'manual') {
|
|
60
|
+
await this.finalizeUtterance(reason);
|
|
61
|
+
}
|
|
62
|
+
async close() {
|
|
63
|
+
this.resetBuffers();
|
|
64
|
+
this.silenceDetector.dispose();
|
|
65
|
+
this.providers.wakeWord?.dispose?.();
|
|
66
|
+
this.changeState('closed');
|
|
67
|
+
}
|
|
68
|
+
async ingestFrame(frame) {
|
|
69
|
+
if (this.state === 'closed' || this.state === 'idle')
|
|
70
|
+
return;
|
|
71
|
+
if (this.config.mode === 'wake-word' && !this.wakeWordDetected) {
|
|
72
|
+
const detection = await this.providers.wakeWord?.detect(frame, this.config.sampleRate);
|
|
73
|
+
if (detection) {
|
|
74
|
+
this.wakeWordDetected = true;
|
|
75
|
+
this.emit('wake_word_detected', { detection });
|
|
76
|
+
this.changeState('listening');
|
|
77
|
+
}
|
|
78
|
+
return;
|
|
79
|
+
}
|
|
80
|
+
if (this.providers.vad) {
|
|
81
|
+
const decision = this.providers.vad.processFrame(frame);
|
|
82
|
+
this.emit('vad_result', decision);
|
|
83
|
+
this.handleExternalVadDecision(decision);
|
|
84
|
+
const shouldCapture = this.capturedFrames.length > 0 || decision.isSpeech || this.lastExternalVadSpeech;
|
|
85
|
+
if (shouldCapture) {
|
|
86
|
+
this.capturedFrames.push(new Float32Array(frame));
|
|
87
|
+
}
|
|
88
|
+
this.lastExternalVadSpeech = decision.isSpeech;
|
|
89
|
+
return;
|
|
90
|
+
}
|
|
91
|
+
const result = this.vad.processFrame(frame);
|
|
92
|
+
const decision = {
|
|
93
|
+
isSpeech: result.isSpeech,
|
|
94
|
+
confidence: result.confidence,
|
|
95
|
+
result,
|
|
96
|
+
profile: this.calibrator.getCurrentProfile(),
|
|
97
|
+
};
|
|
98
|
+
this.emit('vad_result', decision);
|
|
99
|
+
const shouldCapture = this.capturedFrames.length > 0 || result.isSpeech || this.vad.getCurrentState().isSpeaking;
|
|
100
|
+
if (shouldCapture) {
|
|
101
|
+
this.capturedFrames.push(new Float32Array(frame));
|
|
102
|
+
}
|
|
103
|
+
}
|
|
104
|
+
async transcribeAudio(audioBuffer, options = {}, captureOverride) {
|
|
105
|
+
if (!this.providers.stt) {
|
|
106
|
+
throw new Error('No speech-to-text provider configured.');
|
|
107
|
+
}
|
|
108
|
+
this.changeState('transcribing');
|
|
109
|
+
try {
|
|
110
|
+
const capture = captureOverride ?? this.createCapture(audioBuffer);
|
|
111
|
+
const result = await this.providers.stt.transcribe({
|
|
112
|
+
data: audioBuffer,
|
|
113
|
+
fileName: capture.fileName,
|
|
114
|
+
mimeType: capture.mimeType,
|
|
115
|
+
format: 'wav',
|
|
116
|
+
sampleRate: capture.sampleRate,
|
|
117
|
+
durationSeconds: capture.durationMs / 1000,
|
|
118
|
+
}, {
|
|
119
|
+
...this.config.sttOptions,
|
|
120
|
+
...options,
|
|
121
|
+
});
|
|
122
|
+
this.emit('transcript_final', { result, capture });
|
|
123
|
+
this.changeState(this.config.mode === 'wake-word' ? 'wake-listening' : 'listening');
|
|
124
|
+
this.wakeWordDetected = false;
|
|
125
|
+
}
|
|
126
|
+
catch (error) {
|
|
127
|
+
this.handleError(error);
|
|
128
|
+
}
|
|
129
|
+
}
|
|
130
|
+
async speak(text, options = {}) {
|
|
131
|
+
if (!this.providers.tts) {
|
|
132
|
+
throw new Error('No text-to-speech provider configured.');
|
|
133
|
+
}
|
|
134
|
+
this.changeState('playing');
|
|
135
|
+
this.emit('synthesis_started', { text });
|
|
136
|
+
try {
|
|
137
|
+
const result = await this.providers.tts.synthesize(text, {
|
|
138
|
+
...this.config.ttsOptions,
|
|
139
|
+
...options,
|
|
140
|
+
});
|
|
141
|
+
this.emit('synthesis_completed', { text, result });
|
|
142
|
+
this.changeState(this.config.mode === 'wake-word' ? 'wake-listening' : 'listening');
|
|
143
|
+
this.wakeWordDetected = false;
|
|
144
|
+
return result;
|
|
145
|
+
}
|
|
146
|
+
catch (error) {
|
|
147
|
+
this.handleError(error);
|
|
148
|
+
throw error;
|
|
149
|
+
}
|
|
150
|
+
}
|
|
151
|
+
interrupt() {
|
|
152
|
+
this.resetBuffers();
|
|
153
|
+
this.changeState('interrupted');
|
|
154
|
+
this.changeState(this.config.mode === 'wake-word' ? 'wake-listening' : 'listening');
|
|
155
|
+
}
|
|
156
|
+
bindVadEvents() {
|
|
157
|
+
this.vad.on('speech_start', (result) => {
|
|
158
|
+
this.currentSpeechStartedAt = Date.now();
|
|
159
|
+
this.changeState('capturing');
|
|
160
|
+
this.emit('speech_started', {
|
|
161
|
+
timestamp: this.currentSpeechStartedAt,
|
|
162
|
+
vad: result,
|
|
163
|
+
});
|
|
164
|
+
this.silenceDetector.handleSpeechStart(result);
|
|
165
|
+
});
|
|
166
|
+
this.vad.on('voice_activity', (result) => {
|
|
167
|
+
this.silenceDetector.handleVoiceActivity(result);
|
|
168
|
+
});
|
|
169
|
+
this.vad.on('no_voice_activity', (result) => {
|
|
170
|
+
this.silenceDetector.handleNoVoiceActivity(result);
|
|
171
|
+
});
|
|
172
|
+
this.vad.on('speech_end', (result, durationMs) => {
|
|
173
|
+
this.emit('speech_ended', {
|
|
174
|
+
timestamp: Date.now(),
|
|
175
|
+
vad: result,
|
|
176
|
+
durationMs,
|
|
177
|
+
});
|
|
178
|
+
this.silenceDetector.handleSpeechEnd(result, durationMs);
|
|
179
|
+
});
|
|
180
|
+
}
|
|
181
|
+
bindSilenceEvents() {
|
|
182
|
+
this.silenceDetector.on('significant_pause_detected', (pauseDurationMs) => {
|
|
183
|
+
this.emit('significant_pause', pauseDurationMs);
|
|
184
|
+
});
|
|
185
|
+
this.silenceDetector.on('utterance_end_detected', () => {
|
|
186
|
+
void this.finalizeUtterance('silence-timeout');
|
|
187
|
+
});
|
|
188
|
+
}
|
|
189
|
+
handleExternalVadDecision(decision) {
|
|
190
|
+
const result = decision.result ?? this.createSyntheticVadResult(decision);
|
|
191
|
+
if (decision.isSpeech) {
|
|
192
|
+
if (!this.lastExternalVadSpeech) {
|
|
193
|
+
this.currentSpeechStartedAt = Date.now();
|
|
194
|
+
this.changeState('capturing');
|
|
195
|
+
this.emit('speech_started', {
|
|
196
|
+
timestamp: this.currentSpeechStartedAt,
|
|
197
|
+
vad: result,
|
|
198
|
+
});
|
|
199
|
+
this.silenceDetector.handleSpeechStart(result);
|
|
200
|
+
}
|
|
201
|
+
this.silenceDetector.handleVoiceActivity(result);
|
|
202
|
+
return;
|
|
203
|
+
}
|
|
204
|
+
if (this.lastExternalVadSpeech) {
|
|
205
|
+
const durationMs = this.currentSpeechStartedAt
|
|
206
|
+
? Date.now() - this.currentSpeechStartedAt
|
|
207
|
+
: this.config.frameDurationMs;
|
|
208
|
+
this.emit('speech_ended', {
|
|
209
|
+
timestamp: Date.now(),
|
|
210
|
+
vad: result,
|
|
211
|
+
durationMs,
|
|
212
|
+
});
|
|
213
|
+
this.silenceDetector.handleSpeechEnd(result, durationMs);
|
|
214
|
+
return;
|
|
215
|
+
}
|
|
216
|
+
this.silenceDetector.handleNoVoiceActivity(result);
|
|
217
|
+
}
|
|
218
|
+
async finalizeUtterance(reason) {
|
|
219
|
+
if (this.transcriptionPromise) {
|
|
220
|
+
await this.transcriptionPromise;
|
|
221
|
+
return;
|
|
222
|
+
}
|
|
223
|
+
if (this.capturedFrames.length === 0) {
|
|
224
|
+
this.changeState(this.config.mode === 'wake-word' ? 'wake-listening' : 'listening');
|
|
225
|
+
return;
|
|
226
|
+
}
|
|
227
|
+
const audioBuffer = encodeFloat32ToWav(this.capturedFrames, this.config.sampleRate);
|
|
228
|
+
const capture = this.createCapture(audioBuffer);
|
|
229
|
+
this.emit('utterance_captured', { reason, capture });
|
|
230
|
+
this.resetBuffers();
|
|
231
|
+
if (!this.config.autoTranscribeOnSpeechEnd || !this.providers.stt) {
|
|
232
|
+
this.changeState(this.config.mode === 'wake-word' ? 'wake-listening' : 'listening');
|
|
233
|
+
return;
|
|
234
|
+
}
|
|
235
|
+
this.transcriptionPromise = this.transcribeAudio(audioBuffer, {}, capture).finally(() => {
|
|
236
|
+
this.transcriptionPromise = null;
|
|
237
|
+
});
|
|
238
|
+
await this.transcriptionPromise;
|
|
239
|
+
}
|
|
240
|
+
createCapture(audioBuffer) {
|
|
241
|
+
const bufferedSampleCount = this.capturedFrames.reduce((sum, frame) => sum + frame.length, 0);
|
|
242
|
+
const wavPayloadBytes = Math.max(0, audioBuffer.length - 44);
|
|
243
|
+
const derivedSampleCount = bufferedSampleCount > 0 ? bufferedSampleCount : wavPayloadBytes / 2;
|
|
244
|
+
const durationMs = (derivedSampleCount / this.config.sampleRate) * 1000;
|
|
245
|
+
return {
|
|
246
|
+
audioBuffer,
|
|
247
|
+
mimeType: 'audio/wav',
|
|
248
|
+
fileName: this.config.audioFileName,
|
|
249
|
+
sampleRate: this.config.sampleRate,
|
|
250
|
+
durationMs,
|
|
251
|
+
frameCount: this.capturedFrames.length,
|
|
252
|
+
};
|
|
253
|
+
}
|
|
254
|
+
resetBuffers() {
|
|
255
|
+
this.capturedFrames.length = 0;
|
|
256
|
+
this.currentSpeechStartedAt = null;
|
|
257
|
+
this.lastExternalVadSpeech = false;
|
|
258
|
+
this.vad.resetState();
|
|
259
|
+
this.silenceDetector.reset();
|
|
260
|
+
this.providers.vad?.reset();
|
|
261
|
+
this.providers.wakeWord?.reset?.();
|
|
262
|
+
}
|
|
263
|
+
createSyntheticVadResult(decision) {
|
|
264
|
+
return {
|
|
265
|
+
isSpeech: decision.isSpeech,
|
|
266
|
+
frameEnergy: decision.isSpeech ? 1 : 0,
|
|
267
|
+
currentSpeechThreshold: 0,
|
|
268
|
+
currentSilenceThreshold: 0,
|
|
269
|
+
confidence: decision.confidence,
|
|
270
|
+
};
|
|
271
|
+
}
|
|
272
|
+
changeState(next) {
|
|
273
|
+
if (this.state === next)
|
|
274
|
+
return;
|
|
275
|
+
const previous = this.state;
|
|
276
|
+
this.state = next;
|
|
277
|
+
this.emit('state_changed', { previous, current: next });
|
|
278
|
+
}
|
|
279
|
+
handleError(error) {
|
|
280
|
+
const normalized = error instanceof Error ? error : new Error(String(error));
|
|
281
|
+
this.changeState('error');
|
|
282
|
+
this.emit('error', { error: normalized });
|
|
283
|
+
}
|
|
284
|
+
}
|
|
285
|
+
//# sourceMappingURL=SpeechSession.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"SpeechSession.js","sourceRoot":"","sources":["../../src/speech/SpeechSession.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,EAAE,WAAW,EAAE,MAAM,8BAA8B,CAAC;AAE3D,OAAO,EAAE,uBAAuB,EAAE,MAAM,0CAA0C,CAAC;AACnF,OAAO,EAAE,eAAe,EAAE,MAAM,kCAAkC,CAAC;AACnE,OAAO,EAAE,kBAAkB,EAAE,MAAM,YAAY,CAAC;AAahD,MAAM,OAAO,aAAc,SAAQ,YAAY;IAmB7B,EAAE,CAChB,KAAQ,EACR,QAAkC;QAElC,OAAO,KAAK,CAAC,EAAE,CAAC,KAAK,EAAE,QAAQ,CAAC,CAAC;IACnC,CAAC;IAEe,IAAI,CAClB,KAAQ,EACR,GAAG,IAA0C;QAE7C,OAAO,KAAK,CAAC,IAAI,CAAC,KAAK,EAAE,GAAG,IAAI,CAAC,CAAC;IACpC,CAAC;IAED,YAAY,SAA8B,EAAE,EAAE,YAAoC,EAAE;QAClF,KAAK,EAAE,CAAC;QAtBF,UAAK,GAAuB,MAAM,CAAC;QAC1B,mBAAc,GAAmB,EAAE,CAAC;QAC7C,2BAAsB,GAAkB,IAAI,CAAC;QAC7C,qBAAgB,GAAG,KAAK,CAAC;QACzB,yBAAoB,GAAyB,IAAI,CAAC;QAClD,0BAAqB,GAAG,KAAK,CAAC;QAkBpC,IAAI,CAAC,MAAM,GAAG;YACZ,IAAI,EAAE,MAAM,CAAC,IAAI,IAAI,KAAK;YAC1B,UAAU,EAAE,MAAM,CAAC,UAAU,IAAI,KAAM;YACvC,eAAe,EAAE,MAAM,CAAC,eAAe,IAAI,EAAE;YAC7C,aAAa,EAAE,MAAM,CAAC,aAAa,IAAI,oBAAoB;YAC3D,yBAAyB,EAAE,MAAM,CAAC,yBAAyB,IAAI,IAAI;YACnE,GAAG,MAAM;SACV,CAAC;QACF,IAAI,CAAC,SAAS,GAAG,SAAS,CAAC;QAC3B,IAAI,CAAC,UAAU,GAAG,IAAI,uBAAuB,CAAC;YAC5C,UAAU,EAAE,IAAI,CAAC,MAAM,CAAC,UAAU;SACnC,CAAC,CAAC;QACH,IAAI,CAAC,GAAG,GAAG,IAAI,WAAW,CACxB,IAAI,CAAC,MAAM,CAAC,GAAG,IAAI,EAAE,EACrB,IAAI,CAAC,UAAU,EACf,IAAI,CAAC,MAAM,CAAC,eAAe,CAC5B,CAAC;QACF,IAAI,CAAC,eAAe,GAAG,IAAI,eAAe,CAAC,IAAI,CAAC,MAAM,CAAC,OAAO,IAAI,EAAE,CAAC,CAAC;QACtE,IAAI,CAAC,aAAa,EAAE,CAAC;QACrB,IAAI,CAAC,iBAAiB,EAAE,CAAC;IAC3B,CAAC;IAED,QAAQ;QACN,OAAO,IAAI,CAAC,KAAK,CAAC;IACpB,CAAC;IAED,KAAK,CAAC,KAAK;QACT,IAAI,IAAI,CAAC,KAAK,KAAK,QAAQ,EAAE,CAAC;YAC5B,MAAM,IAAI,KAAK,CAAC,0BAA0B,CAAC,CAAC;QAC9C,CAAC;QACD,IAAI,IAAI,CAAC,MAAM,CAAC,IAAI,KAAK,WAAW,EAAE,CAAC;YACrC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,QAAQ,EAAE,CAAC;gBAC7B,MAAM,IAAI,KAAK,CAAC,+CAA+C,CAAC,CAAC;YACnE,CAAC;YACD,IAAI,CAAC,gBAAgB,GAAG,KAAK,CAAC;YAC9B,IAAI,CAAC,WAAW,CAAC,gBAAgB,CAAC,CAAC;YACnC,OAAO;QACT,CAAC;QACD,IAAI,CAAC,WAAW,CAAC,WAAW,CAAC,CAAC;IAChC,CAAC;IAED,KAAK,CAAC,IAAI;QACR,IAAI,CAAC,YAAY,EAAE,CAAC;QACpB,IAAI,CAAC,WAAW,CAAC,MAAM,CAAC,CAAC;IAC3B,CAAC;IAED,KAAK,CAAC,KAAK,CAAC,SAAsC,QAAQ;QACxD,MAAM,IAAI,CAAC,iBAAiB,CAAC,MAAM,CAAC,CAAC;IACvC,CAAC;IAED,KAAK,CAAC,KAAK;QACT,IAAI,CAAC,YAAY,EAAE,CAAC;QACpB,IAAI,CAAC,eAAe,CAAC,OAAO,EAAE,CAAC;QAC/B,IAAI,CAAC,SAAS,CAAC,QAAQ,EAAE,OAAO,EAAE,EAAE,CAAC;QACrC,IAAI,CAAC,WAAW,CAAC,QAAQ,CAAC,CAAC;IAC7B,CAAC;IAED,KAAK,CAAC,WAAW,CAAC,KAAmB;QACnC,IAAI,IAAI,CAAC,KAAK,KAAK,QAAQ,IAAI,IAAI,CAAC,KAAK,KAAK,MAAM;YAAE,OAAO;QAE7D,IAAI,IAAI,CAAC,MAAM,CAAC,IAAI,KAAK,WAAW,IAAI,CAAC,IAAI,CAAC,gBAAgB,EAAE,CAAC;YAC/D,MAAM,SAAS,GAAG,MAAM,IAAI,CAAC,SAAS,CAAC,QAAQ,EAAE,MAAM,CAAC,KAAK,EAAE,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,CAAC;YACvF,IAAI,SAAS,EAAE,CAAC;gBACd,IAAI,CAAC,gBAAgB,GAAG,IAAI,CAAC;gBAC7B,IAAI,CAAC,IAAI,CAAC,oBAAoB,EAAE,EAAE,SAAS,EAAE,CAAC,CAAC;gBAC/C,IAAI,CAAC,WAAW,CAAC,WAAW,CAAC,CAAC;YAChC,CAAC;YACD,OAAO;QACT,CAAC;QAED,IAAI,IAAI,CAAC,SAAS,CAAC,GAAG,EAAE,CAAC;YACvB,MAAM,QAAQ,GAAG,IAAI,CAAC,SAAS,CAAC,GAAG,CAAC,YAAY,CAAC,KAAK,CAAC,CAAC;YACxD,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE,QAAQ,CAAC,CAAC;YAClC,IAAI,CAAC,yBAAyB,CAAC,QAAQ,CAAC,CAAC;YACzC,MAAM,aAAa,GACjB,IAAI,CAAC,cAAc,CAAC,MAAM,GAAG,CAAC,IAAI,QAAQ,CAAC,QAAQ,IAAI,IAAI,CAAC,qBAAqB,CAAC;YACpF,IAAI,aAAa,EAAE,CAAC;gBAClB,IAAI,CAAC,cAAc,CAAC,IAAI,CAAC,IAAI,YAAY,CAAC,KAAK,CAAC,CAAC,CAAC;YACpD,CAAC;YACD,IAAI,CAAC,qBAAqB,GAAG,QAAQ,CAAC,QAAQ,CAAC;YAC/C,OAAO;QACT,CAAC;QAED,MAAM,MAAM,GAAG,IAAI,CAAC,GAAG,CAAC,YAAY,CAAC,KAAK,CAAC,CAAC;QAC5C,MAAM,QAAQ,GAAsB;YAClC,QAAQ,EAAE,MAAM,CAAC,QAAQ;YACzB,UAAU,EAAE,MAAM,CAAC,UAAU;YAC7B,MAAM;YACN,OAAO,EAAE,IAAI,CAAC,UAAU,CAAC,iBAAiB,EAAE;SAC7C,CAAC;QACF,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE,QAAQ,CAAC,CAAC;QAElC,MAAM,aAAa,GACjB,IAAI,CAAC,cAAc,CAAC,MAAM,GAAG,CAAC,IAAI,MAAM,CAAC,QAAQ,IAAI,IAAI,CAAC,GAAG,CAAC,eAAe,EAAE,CAAC,UAAU,CAAC;QAC7F,IAAI,aAAa,EAAE,CAAC;YAClB,IAAI,CAAC,cAAc,CAAC,IAAI,CAAC,IAAI,YAAY,CAAC,KAAK,CAAC,CAAC,CAAC;QACpD,CAAC;IACH,CAAC;IAED,KAAK,CAAC,eAAe,CACnB,WAAmB,EACnB,UAAsC,EAAE,EACxC,eAA2C;QAE3C,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,GAAG,EAAE,CAAC;YACxB,MAAM,IAAI,KAAK,CAAC,wCAAwC,CAAC,CAAC;QAC5D,CAAC;QACD,IAAI,CAAC,WAAW,CAAC,cAAc,CAAC,CAAC;QACjC,IAAI,CAAC;YACH,MAAM,OAAO,GAAG,eAAe,IAAI,IAAI,CAAC,aAAa,CAAC,WAAW,CAAC,CAAC;YACnE,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,SAAS,CAAC,GAAG,CAAC,UAAU,CAChD;gBACE,IAAI,EAAE,WAAW;gBACjB,QAAQ,EAAE,OAAO,CAAC,QAAQ;gBAC1B,QAAQ,EAAE,OAAO,CAAC,QAAQ;gBAC1B,MAAM,EAAE,KAAK;gBACb,UAAU,EAAE,OAAO,CAAC,UAAU;gBAC9B,eAAe,EAAE,OAAO,CAAC,UAAU,GAAG,IAAI;aAC3C,EACD;gBACE,GAAG,IAAI,CAAC,MAAM,CAAC,UAAU;gBACzB,GAAG,OAAO;aACX,CACF,CAAC;YACF,IAAI,CAAC,IAAI,CAAC,kBAAkB,EAAE,EAAE,MAAM,EAAE,OAAO,EAAE,CAAC,CAAC;YACnD,IAAI,CAAC,WAAW,CAAC,IAAI,CAAC,MAAM,CAAC,IAAI,KAAK,WAAW,CAAC,CAAC,CAAC,gBAAgB,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC;YACpF,IAAI,CAAC,gBAAgB,GAAG,KAAK,CAAC;QAChC,CAAC;QAAC,OAAO,KAAK,EAAE,CAAC;YACf,IAAI,CAAC,WAAW,CAAC,KAAK,CAAC,CAAC;QAC1B,CAAC;IACH,CAAC;IAED,KAAK,CAAC,KAAK,CAAC,IAAY,EAAE,UAAkC,EAAE;QAC5D,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,GAAG,EAAE,CAAC;YACxB,MAAM,IAAI,KAAK,CAAC,wCAAwC,CAAC,CAAC;QAC5D,CAAC;QACD,IAAI,CAAC,WAAW,CAAC,SAAS,CAAC,CAAC;QAC5B,IAAI,CAAC,IAAI,CAAC,mBAAmB,EAAE,EAAE,IAAI,EAAE,CAAC,CAAC;QACzC,IAAI,CAAC;YACH,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,SAAS,CAAC,GAAG,CAAC,UAAU,CAAC,IAAI,EAAE;gBACvD,GAAG,IAAI,CAAC,MAAM,CAAC,UAAU;gBACzB,GAAG,OAAO;aACX,CAAC,CAAC;YACH,IAAI,CAAC,IAAI,CAAC,qBAAqB,EAAE,EAAE,IAAI,EAAE,MAAM,EAAE,CAAC,CAAC;YACnD,IAAI,CAAC,WAAW,CAAC,IAAI,CAAC,MAAM,CAAC,IAAI,KAAK,WAAW,CAAC,CAAC,CAAC,gBAAgB,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC;YACpF,IAAI,CAAC,gBAAgB,GAAG,KAAK,CAAC;YAC9B,OAAO,MAAM,CAAC;QAChB,CAAC;QAAC,OAAO,KAAK,EAAE,CAAC;YACf,IAAI,CAAC,WAAW,CAAC,KAAK,CAAC,CAAC;YACxB,MAAM,KAAK,CAAC;QACd,CAAC;IACH,CAAC;IAED,SAAS;QACP,IAAI,CAAC,YAAY,EAAE,CAAC;QACpB,IAAI,CAAC,WAAW,CAAC,aAAa,CAAC,CAAC;QAChC,IAAI,CAAC,WAAW,CAAC,IAAI,CAAC,MAAM,CAAC,IAAI,KAAK,WAAW,CAAC,CAAC,CAAC,gBAAgB,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC;IACtF,CAAC;IAEO,aAAa;QACnB,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,cAAc,EAAE,CAAC,MAAM,EAAE,EAAE;YACrC,IAAI,CAAC,sBAAsB,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;YACzC,IAAI,CAAC,WAAW,CAAC,WAAW,CAAC,CAAC;YAC9B,IAAI,CAAC,IAAI,CAAC,gBAAgB,EAAE;gBAC1B,SAAS,EAAE,IAAI,CAAC,sBAAsB;gBACtC,GAAG,EAAE,MAAM;aACZ,CAAC,CAAC;YACH,IAAI,CAAC,eAAe,CAAC,iBAAiB,CAAC,MAAM,CAAC,CAAC;QACjD,CAAC,CAAC,CAAC;QAEH,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,gBAAgB,EAAE,CAAC,MAAM,EAAE,EAAE;YACvC,IAAI,CAAC,eAAe,CAAC,mBAAmB,CAAC,MAAM,CAAC,CAAC;QACnD,CAAC,CAAC,CAAC;QAEH,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,mBAAmB,EAAE,CAAC,MAAM,EAAE,EAAE;YAC1C,IAAI,CAAC,eAAe,CAAC,qBAAqB,CAAC,MAAM,CAAC,CAAC;QACrD,CAAC,CAAC,CAAC;QAEH,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,YAAY,EAAE,CAAC,MAAM,EAAE,UAAU,EAAE,EAAE;YAC/C,IAAI,CAAC,IAAI,CAAC,cAAc,EAAE;gBACxB,SAAS,EAAE,IAAI,CAAC,GAAG,EAAE;gBACrB,GAAG,EAAE,MAAM;gBACX,UAAU;aACX,CAAC,CAAC;YACH,IAAI,CAAC,eAAe,CAAC,eAAe,CAAC,MAAM,EAAE,UAAU,CAAC,CAAC;QAC3D,CAAC,CAAC,CAAC;IACL,CAAC;IAEO,iBAAiB;QACvB,IAAI,CAAC,eAAe,CAAC,EAAE,CAAC,4BAA4B,EAAE,CAAC,eAAe,EAAE,EAAE;YACxE,IAAI,CAAC,IAAI,CAAC,mBAAmB,EAAE,eAAe,CAAC,CAAC;QAClD,CAAC,CAAC,CAAC;QAEH,IAAI,CAAC,eAAe,CAAC,EAAE,CAAC,wBAAwB,EAAE,GAAG,EAAE;YACrD,KAAK,IAAI,CAAC,iBAAiB,CAAC,iBAAiB,CAAC,CAAC;QACjD,CAAC,CAAC,CAAC;IACL,CAAC;IAEO,yBAAyB,CAAC,QAA2B;QAC3D,MAAM,MAAM,GAAG,QAAQ,CAAC,MAAM,IAAI,IAAI,CAAC,wBAAwB,CAAC,QAAQ,CAAC,CAAC;QAC1E,IAAI,QAAQ,CAAC,QAAQ,EAAE,CAAC;YACtB,IAAI,CAAC,IAAI,CAAC,qBAAqB,EAAE,CAAC;gBAChC,IAAI,CAAC,sBAAsB,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;gBACzC,IAAI,CAAC,WAAW,CAAC,WAAW,CAAC,CAAC;gBAC9B,IAAI,CAAC,IAAI,CAAC,gBAAgB,EAAE;oBAC1B,SAAS,EAAE,IAAI,CAAC,sBAAsB;oBACtC,GAAG,EAAE,MAAM;iBACZ,CAAC,CAAC;gBACH,IAAI,CAAC,eAAe,CAAC,iBAAiB,CAAC,MAAM,CAAC,CAAC;YACjD,CAAC;YACD,IAAI,CAAC,eAAe,CAAC,mBAAmB,CAAC,MAAM,CAAC,CAAC;YACjD,OAAO;QACT,CAAC;QAED,IAAI,IAAI,CAAC,qBAAqB,EAAE,CAAC;YAC/B,MAAM,UAAU,GAAG,IAAI,CAAC,sBAAsB;gBAC5C,CAAC,CAAC,IAAI,CAAC,GAAG,EAAE,GAAG,IAAI,CAAC,sBAAsB;gBAC1C,CAAC,CAAC,IAAI,CAAC,MAAM,CAAC,eAAe,CAAC;YAChC,IAAI,CAAC,IAAI,CAAC,cAAc,EAAE;gBACxB,SAAS,EAAE,IAAI,CAAC,GAAG,EAAE;gBACrB,GAAG,EAAE,MAAM;gBACX,UAAU;aACX,CAAC,CAAC;YACH,IAAI,CAAC,eAAe,CAAC,eAAe,CAAC,MAAM,EAAE,UAAU,CAAC,CAAC;YACzD,OAAO;QACT,CAAC;QAED,IAAI,CAAC,eAAe,CAAC,qBAAqB,CAAC,MAAM,CAAC,CAAC;IACrD,CAAC;IAEO,KAAK,CAAC,iBAAiB,CAAC,MAAmC;QACjE,IAAI,IAAI,CAAC,oBAAoB,EAAE,CAAC;YAC9B,MAAM,IAAI,CAAC,oBAAoB,CAAC;YAChC,OAAO;QACT,CAAC;QACD,IAAI,IAAI,CAAC,cAAc,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YACrC,IAAI,CAAC,WAAW,CAAC,IAAI,CAAC,MAAM,CAAC,IAAI,KAAK,WAAW,CAAC,CAAC,CAAC,gBAAgB,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC;YACpF,OAAO;QACT,CAAC;QAED,MAAM,WAAW,GAAG,kBAAkB,CAAC,IAAI,CAAC,cAAc,EAAE,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,CAAC;QACpF,MAAM,OAAO,GAAG,IAAI,CAAC,aAAa,CAAC,WAAW,CAAC,CAAC;QAChD,IAAI,CAAC,IAAI,CAAC,oBAAoB,EAAE,EAAE,MAAM,EAAE,OAAO,EAAE,CAAC,CAAC;QACrD,IAAI,CAAC,YAAY,EAAE,CAAC;QAEpB,IAAI,CAAC,IAAI,CAAC,MAAM,CAAC,yBAAyB,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,GAAG,EAAE,CAAC;YAClE,IAAI,CAAC,WAAW,CAAC,IAAI,CAAC,MAAM,CAAC,IAAI,KAAK,WAAW,CAAC,CAAC,CAAC,gBAAgB,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC;YACpF,OAAO;QACT,CAAC;QAED,IAAI,CAAC,oBAAoB,GAAG,IAAI,CAAC,eAAe,CAAC,WAAW,EAAE,EAAE,EAAE,OAAO,CAAC,CAAC,OAAO,CAAC,GAAG,EAAE;YACtF,IAAI,CAAC,oBAAoB,GAAG,IAAI,CAAC;QACnC,CAAC,CAAC,CAAC;QACH,MAAM,IAAI,CAAC,oBAAoB,CAAC;IAClC,CAAC;IAEO,aAAa,CAAC,WAAmB;QACvC,MAAM,mBAAmB,GAAG,IAAI,CAAC,cAAc,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,KAAK,EAAE,EAAE,CAAC,GAAG,GAAG,KAAK,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC;QAC9F,MAAM,eAAe,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,WAAW,CAAC,MAAM,GAAG,EAAE,CAAC,CAAC;QAC7D,MAAM,kBAAkB,GAAG,mBAAmB,GAAG,CAAC,CAAC,CAAC,CAAC,mBAAmB,CAAC,CAAC,CAAC,eAAe,GAAG,CAAC,CAAC;QAC/F,MAAM,UAAU,GAAG,CAAC,kBAAkB,GAAG,IAAI,CAAC,MAAM,CAAC,UAAU,CAAC,GAAG,IAAI,CAAC;QAExE,OAAO;YACL,WAAW;YACX,QAAQ,EAAE,WAAW;YACrB,QAAQ,EAAE,IAAI,CAAC,MAAM,CAAC,aAAa;YACnC,UAAU,EAAE,IAAI,CAAC,MAAM,CAAC,UAAU;YAClC,UAAU;YACV,UAAU,EAAE,IAAI,CAAC,cAAc,CAAC,MAAM;SACvC,CAAC;IACJ,CAAC;IAEO,YAAY;QAClB,IAAI,CAAC,cAAc,CAAC,MAAM,GAAG,CAAC,CAAC;QAC/B,IAAI,CAAC,sBAAsB,GAAG,IAAI,CAAC;QACnC,IAAI,CAAC,qBAAqB,GAAG,KAAK,CAAC;QACnC,IAAI,CAAC,GAAG,CAAC,UAAU,EAAE,CAAC;QACtB,IAAI,CAAC,eAAe,CAAC,KAAK,EAAE,CAAC;QAC7B,IAAI,CAAC,SAAS,CAAC,GAAG,EAAE,KAAK,EAAE,CAAC;QAC5B,IAAI,CAAC,SAAS,CAAC,QAAQ,EAAE,KAAK,EAAE,EAAE,CAAC;IACrC,CAAC;IAEO,wBAAwB,CAAC,QAA2B;QAC1D,OAAO;YACL,QAAQ,EAAE,QAAQ,CAAC,QAAQ;YAC3B,WAAW,EAAE,QAAQ,CAAC,QAAQ,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;YACtC,sBAAsB,EAAE,CAAC;YACzB,uBAAuB,EAAE,CAAC;YAC1B,UAAU,EAAE,QAAQ,CAAC,UAAU;SAChC,CAAC;IACJ,CAAC;IAEO,WAAW,CAAC,IAAwB;QAC1C,IAAI,IAAI,CAAC,KAAK,KAAK,IAAI;YAAE,OAAO;QAChC,MAAM,QAAQ,GAAG,IAAI,CAAC,KAAK,CAAC;QAC5B,IAAI,CAAC,KAAK,GAAG,IAAI,CAAC;QAClB,IAAI,CAAC,IAAI,CAAC,eAAe,EAAE,EAAE,QAAQ,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC,CAAC;IAC1D,CAAC;IAEO,WAAW,CAAC,KAAc;QAChC,MAAM,UAAU,GAAG,KAAK,YAAY,KAAK,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC;QAC7E,IAAI,CAAC,WAAW,CAAC,OAAO,CAAC,CAAC;QAC1B,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,EAAE,KAAK,EAAE,UAAU,EAAE,CAAC,CAAC;IAC5C,CAAC;CACF"}
|
|
@@ -0,0 +1,3 @@
|
|
|
1
|
+
export declare function concatFloat32AudioFrames(frames: readonly Float32Array[]): Float32Array;
|
|
2
|
+
export declare function encodeFloat32ToWav(input: Float32Array | readonly Float32Array[], sampleRate: number, channelCount?: number): Buffer;
|
|
3
|
+
//# sourceMappingURL=audio.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"audio.d.ts","sourceRoot":"","sources":["../../src/speech/audio.ts"],"names":[],"mappings":"AAAA,wBAAgB,wBAAwB,CAAC,MAAM,EAAE,SAAS,YAAY,EAAE,GAAG,YAAY,CAWtF;AAOD,wBAAgB,kBAAkB,CAChC,KAAK,EAAE,YAAY,GAAG,SAAS,YAAY,EAAE,EAC7C,UAAU,EAAE,MAAM,EAClB,YAAY,SAAI,GACf,MAAM,CA4BR"}
|
|
@@ -0,0 +1,39 @@
|
|
|
1
|
+
export function concatFloat32AudioFrames(frames) {
|
|
2
|
+
const totalLength = frames.reduce((sum, frame) => sum + frame.length, 0);
|
|
3
|
+
const output = new Float32Array(totalLength);
|
|
4
|
+
let offset = 0;
|
|
5
|
+
for (const frame of frames) {
|
|
6
|
+
output.set(frame, offset);
|
|
7
|
+
offset += frame.length;
|
|
8
|
+
}
|
|
9
|
+
return output;
|
|
10
|
+
}
|
|
11
|
+
function clampPcm16(sample) {
|
|
12
|
+
const value = Math.max(-1, Math.min(1, sample));
|
|
13
|
+
return value < 0 ? Math.round(value * 0x8000) : Math.round(value * 0x7fff);
|
|
14
|
+
}
|
|
15
|
+
export function encodeFloat32ToWav(input, sampleRate, channelCount = 1) {
|
|
16
|
+
const audio = input instanceof Float32Array ? input : concatFloat32AudioFrames(input);
|
|
17
|
+
const pcmData = Buffer.alloc(audio.length * 2);
|
|
18
|
+
for (let index = 0; index < audio.length; index += 1) {
|
|
19
|
+
pcmData.writeInt16LE(clampPcm16(audio[index] ?? 0), index * 2);
|
|
20
|
+
}
|
|
21
|
+
const header = Buffer.alloc(44);
|
|
22
|
+
const byteRate = sampleRate * channelCount * 2;
|
|
23
|
+
const blockAlign = channelCount * 2;
|
|
24
|
+
header.write('RIFF', 0);
|
|
25
|
+
header.writeUInt32LE(36 + pcmData.length, 4);
|
|
26
|
+
header.write('WAVE', 8);
|
|
27
|
+
header.write('fmt ', 12);
|
|
28
|
+
header.writeUInt32LE(16, 16);
|
|
29
|
+
header.writeUInt16LE(1, 20);
|
|
30
|
+
header.writeUInt16LE(channelCount, 22);
|
|
31
|
+
header.writeUInt32LE(sampleRate, 24);
|
|
32
|
+
header.writeUInt32LE(byteRate, 28);
|
|
33
|
+
header.writeUInt16LE(blockAlign, 32);
|
|
34
|
+
header.writeUInt16LE(16, 34);
|
|
35
|
+
header.write('data', 36);
|
|
36
|
+
header.writeUInt32LE(pcmData.length, 40);
|
|
37
|
+
return Buffer.concat([header, pcmData]);
|
|
38
|
+
}
|
|
39
|
+
//# sourceMappingURL=audio.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"audio.js","sourceRoot":"","sources":["../../src/speech/audio.ts"],"names":[],"mappings":"AAAA,MAAM,UAAU,wBAAwB,CAAC,MAA+B;IACtE,MAAM,WAAW,GAAG,MAAM,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,KAAK,EAAE,EAAE,CAAC,GAAG,GAAG,KAAK,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC;IACzE,MAAM,MAAM,GAAG,IAAI,YAAY,CAAC,WAAW,CAAC,CAAC;IAE7C,IAAI,MAAM,GAAG,CAAC,CAAC;IACf,KAAK,MAAM,KAAK,IAAI,MAAM,EAAE,CAAC;QAC3B,MAAM,CAAC,GAAG,CAAC,KAAK,EAAE,MAAM,CAAC,CAAC;QAC1B,MAAM,IAAI,KAAK,CAAC,MAAM,CAAC;IACzB,CAAC;IAED,OAAO,MAAM,CAAC;AAChB,CAAC;AAED,SAAS,UAAU,CAAC,MAAc;IAChC,MAAM,KAAK,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,MAAM,CAAC,CAAC,CAAC;IAChD,OAAO,KAAK,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,KAAK,GAAG,MAAM,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,KAAK,GAAG,MAAM,CAAC,CAAC;AAC7E,CAAC;AAED,MAAM,UAAU,kBAAkB,CAChC,KAA6C,EAC7C,UAAkB,EAClB,YAAY,GAAG,CAAC;IAEhB,MAAM,KAAK,GACT,KAAK,YAAY,YAAY,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,wBAAwB,CAAC,KAAK,CAAC,CAAC;IAC1E,MAAM,OAAO,GAAG,MAAM,CAAC,KAAK,CAAC,KAAK,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;IAE/C,KAAK,IAAI,KAAK,GAAG,CAAC,EAAE,KAAK,GAAG,KAAK,CAAC,MAAM,EAAE,KAAK,IAAI,CAAC,EAAE,CAAC;QACrD,OAAO,CAAC,YAAY,CAAC,UAAU,CAAC,KAAK,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,EAAE,KAAK,GAAG,CAAC,CAAC,CAAC;IACjE,CAAC;IAED,MAAM,MAAM,GAAG,MAAM,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC;IAChC,MAAM,QAAQ,GAAG,UAAU,GAAG,YAAY,GAAG,CAAC,CAAC;IAC/C,MAAM,UAAU,GAAG,YAAY,GAAG,CAAC,CAAC;IAEpC,MAAM,CAAC,KAAK,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC;IACxB,MAAM,CAAC,aAAa,CAAC,EAAE,GAAG,OAAO,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC;IAC7C,MAAM,CAAC,KAAK,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC;IACxB,MAAM,CAAC,KAAK,CAAC,MAAM,EAAE,EAAE,CAAC,CAAC;IACzB,MAAM,CAAC,aAAa,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC;IAC7B,MAAM,CAAC,aAAa,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;IAC5B,MAAM,CAAC,aAAa,CAAC,YAAY,EAAE,EAAE,CAAC,CAAC;IACvC,MAAM,CAAC,aAAa,CAAC,UAAU,EAAE,EAAE,CAAC,CAAC;IACrC,MAAM,CAAC,aAAa,CAAC,QAAQ,EAAE,EAAE,CAAC,CAAC;IACnC,MAAM,CAAC,aAAa,CAAC,UAAU,EAAE,EAAE,CAAC,CAAC;IACrC,MAAM,CAAC,aAAa,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC;IAC7B,MAAM,CAAC,KAAK,CAAC,MAAM,EAAE,EAAE,CAAC,CAAC;IACzB,MAAM,CAAC,aAAa,CAAC,OAAO,CAAC,MAAM,EAAE,EAAE,CAAC,CAAC;IAEzC,OAAO,MAAM,CAAC,MAAM,CAAC,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC,CAAC;AAC1C,CAAC"}
|
|
@@ -0,0 +1,11 @@
|
|
|
1
|
+
export * from './types.js';
|
|
2
|
+
export * from './audio.js';
|
|
3
|
+
export * from './providerCatalog.js';
|
|
4
|
+
export * from './SpeechProviderRegistry.js';
|
|
5
|
+
export * from './SpeechSession.js';
|
|
6
|
+
export * from './SpeechRuntime.js';
|
|
7
|
+
export * from './providers/OpenAIWhisperSpeechToTextProvider.js';
|
|
8
|
+
export * from './providers/OpenAITextToSpeechProvider.js';
|
|
9
|
+
export * from './providers/ElevenLabsTextToSpeechProvider.js';
|
|
10
|
+
export * from './providers/BuiltInAdaptiveVadProvider.js';
|
|
11
|
+
//# sourceMappingURL=index.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/speech/index.ts"],"names":[],"mappings":"AAAA,cAAc,YAAY,CAAC;AAC3B,cAAc,YAAY,CAAC;AAC3B,cAAc,sBAAsB,CAAC;AACrC,cAAc,6BAA6B,CAAC;AAC5C,cAAc,oBAAoB,CAAC;AACnC,cAAc,oBAAoB,CAAC;AACnC,cAAc,kDAAkD,CAAC;AACjE,cAAc,2CAA2C,CAAC;AAC1D,cAAc,+CAA+C,CAAC;AAC9D,cAAc,2CAA2C,CAAC"}
|
|
@@ -0,0 +1,11 @@
|
|
|
1
|
+
export * from './types.js';
|
|
2
|
+
export * from './audio.js';
|
|
3
|
+
export * from './providerCatalog.js';
|
|
4
|
+
export * from './SpeechProviderRegistry.js';
|
|
5
|
+
export * from './SpeechSession.js';
|
|
6
|
+
export * from './SpeechRuntime.js';
|
|
7
|
+
export * from './providers/OpenAIWhisperSpeechToTextProvider.js';
|
|
8
|
+
export * from './providers/OpenAITextToSpeechProvider.js';
|
|
9
|
+
export * from './providers/ElevenLabsTextToSpeechProvider.js';
|
|
10
|
+
export * from './providers/BuiltInAdaptiveVadProvider.js';
|
|
11
|
+
//# sourceMappingURL=index.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"index.js","sourceRoot":"","sources":["../../src/speech/index.ts"],"names":[],"mappings":"AAAA,cAAc,YAAY,CAAC;AAC3B,cAAc,YAAY,CAAC;AAC3B,cAAc,sBAAsB,CAAC;AACrC,cAAc,6BAA6B,CAAC;AAC5C,cAAc,oBAAoB,CAAC;AACnC,cAAc,oBAAoB,CAAC;AACnC,cAAc,kDAAkD,CAAC;AACjE,cAAc,2CAA2C,CAAC;AAC1D,cAAc,+CAA+C,CAAC;AAC9D,cAAc,2CAA2C,CAAC"}
|
|
@@ -0,0 +1,7 @@
|
|
|
1
|
+
import type { SpeechProviderCatalogEntry, SpeechProviderKind } from './types.js';
|
|
2
|
+
export declare const SPEECH_PROVIDER_CATALOG: readonly SpeechProviderCatalogEntry[];
|
|
3
|
+
export declare function getSpeechProviderCatalog(kind?: SpeechProviderKind): SpeechProviderCatalogEntry[];
|
|
4
|
+
export declare function getSpeechProviderKinds(): SpeechProviderKind[];
|
|
5
|
+
export declare function findSpeechProviderCatalogEntry(id: string): SpeechProviderCatalogEntry | undefined;
|
|
6
|
+
export declare function isSpeechProviderConfigured(entry: SpeechProviderCatalogEntry, env?: Record<string, string | undefined>): boolean;
|
|
7
|
+
//# sourceMappingURL=providerCatalog.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"providerCatalog.d.ts","sourceRoot":"","sources":["../../src/speech/providerCatalog.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,0BAA0B,EAAE,kBAAkB,EAAE,MAAM,YAAY,CAAC;AAEjF,eAAO,MAAM,uBAAuB,EAAE,SAAS,0BAA0B,EAmO/D,CAAC;AAEX,wBAAgB,wBAAwB,CACtC,IAAI,CAAC,EAAE,kBAAkB,GACxB,0BAA0B,EAAE,CAG9B;AAED,wBAAgB,sBAAsB,IAAI,kBAAkB,EAAE,CAE7D;AAED,wBAAgB,8BAA8B,CAC5C,EAAE,EAAE,MAAM,GACT,0BAA0B,GAAG,SAAS,CAExC;AAED,wBAAgB,0BAA0B,CACxC,KAAK,EAAE,0BAA0B,EACjC,GAAG,GAAE,MAAM,CAAC,MAAM,EAAE,MAAM,GAAG,SAAS,CAAe,GACpD,OAAO,CAET"}
|
|
@@ -0,0 +1,243 @@
|
|
|
1
|
+
export const SPEECH_PROVIDER_CATALOG = [
|
|
2
|
+
{
|
|
3
|
+
id: 'twilio',
|
|
4
|
+
kind: 'telephony',
|
|
5
|
+
label: 'Twilio',
|
|
6
|
+
envVars: ['TWILIO_ACCOUNT_SID', 'TWILIO_AUTH_TOKEN'],
|
|
7
|
+
local: false,
|
|
8
|
+
description: 'Telephony call control via Twilio programmable voice.',
|
|
9
|
+
extensionName: 'voice-twilio',
|
|
10
|
+
features: ['telephony', 'webhooks', 'media-streams'],
|
|
11
|
+
},
|
|
12
|
+
{
|
|
13
|
+
id: 'telnyx',
|
|
14
|
+
kind: 'telephony',
|
|
15
|
+
label: 'Telnyx',
|
|
16
|
+
envVars: ['TELNYX_API_KEY', 'TELNYX_CONNECTION_ID'],
|
|
17
|
+
local: false,
|
|
18
|
+
description: 'Telephony call control via Telnyx Call Control.',
|
|
19
|
+
extensionName: 'voice-telnyx',
|
|
20
|
+
features: ['telephony', 'sip', 'media-streams'],
|
|
21
|
+
},
|
|
22
|
+
{
|
|
23
|
+
id: 'plivo',
|
|
24
|
+
kind: 'telephony',
|
|
25
|
+
label: 'Plivo',
|
|
26
|
+
envVars: ['PLIVO_AUTH_ID', 'PLIVO_AUTH_TOKEN'],
|
|
27
|
+
local: false,
|
|
28
|
+
description: 'Telephony call control via the Plivo Voice API.',
|
|
29
|
+
extensionName: 'voice-plivo',
|
|
30
|
+
features: ['telephony', 'xml'],
|
|
31
|
+
},
|
|
32
|
+
{
|
|
33
|
+
id: 'openai-whisper',
|
|
34
|
+
kind: 'stt',
|
|
35
|
+
label: 'OpenAI Whisper',
|
|
36
|
+
envVars: ['OPENAI_API_KEY'],
|
|
37
|
+
local: false,
|
|
38
|
+
description: 'Hosted Whisper speech-to-text for file and utterance transcription.',
|
|
39
|
+
defaultModel: 'whisper-1',
|
|
40
|
+
streaming: false,
|
|
41
|
+
features: ['cloud', 'timestamps', 'transcription'],
|
|
42
|
+
},
|
|
43
|
+
{
|
|
44
|
+
id: 'deepgram',
|
|
45
|
+
kind: 'stt',
|
|
46
|
+
label: 'Deepgram',
|
|
47
|
+
envVars: ['DEEPGRAM_API_KEY'],
|
|
48
|
+
local: false,
|
|
49
|
+
description: 'Streaming-first speech-to-text for live conversations.',
|
|
50
|
+
streaming: true,
|
|
51
|
+
features: ['cloud', 'streaming'],
|
|
52
|
+
},
|
|
53
|
+
{
|
|
54
|
+
id: 'assemblyai',
|
|
55
|
+
kind: 'stt',
|
|
56
|
+
label: 'AssemblyAI',
|
|
57
|
+
envVars: ['ASSEMBLYAI_API_KEY'],
|
|
58
|
+
local: false,
|
|
59
|
+
description: 'Hosted speech-to-text with diarization and analytics.',
|
|
60
|
+
streaming: true,
|
|
61
|
+
features: ['cloud', 'streaming', 'diarization'],
|
|
62
|
+
},
|
|
63
|
+
{
|
|
64
|
+
id: 'google-cloud-stt',
|
|
65
|
+
kind: 'stt',
|
|
66
|
+
label: 'Google Cloud STT',
|
|
67
|
+
envVars: ['GOOGLE_STT_CREDENTIALS'],
|
|
68
|
+
local: false,
|
|
69
|
+
description: 'Google Cloud speech recognition.',
|
|
70
|
+
streaming: true,
|
|
71
|
+
features: ['cloud', 'streaming'],
|
|
72
|
+
},
|
|
73
|
+
{
|
|
74
|
+
id: 'azure-speech-stt',
|
|
75
|
+
kind: 'stt',
|
|
76
|
+
label: 'Azure Speech STT',
|
|
77
|
+
envVars: ['AZURE_SPEECH_KEY', 'AZURE_SPEECH_REGION'],
|
|
78
|
+
local: false,
|
|
79
|
+
description: 'Azure speech recognition for batch and live transcription.',
|
|
80
|
+
streaming: true,
|
|
81
|
+
features: ['cloud', 'streaming'],
|
|
82
|
+
},
|
|
83
|
+
{
|
|
84
|
+
id: 'whisper-local',
|
|
85
|
+
kind: 'stt',
|
|
86
|
+
label: 'Whisper.cpp',
|
|
87
|
+
envVars: [],
|
|
88
|
+
local: true,
|
|
89
|
+
description: 'Local Whisper.cpp offline transcription.',
|
|
90
|
+
features: ['local', 'offline'],
|
|
91
|
+
},
|
|
92
|
+
{
|
|
93
|
+
id: 'vosk',
|
|
94
|
+
kind: 'stt',
|
|
95
|
+
label: 'Vosk',
|
|
96
|
+
envVars: [],
|
|
97
|
+
local: true,
|
|
98
|
+
description: 'Local streaming speech recognition via Vosk.',
|
|
99
|
+
streaming: true,
|
|
100
|
+
features: ['local', 'offline', 'streaming'],
|
|
101
|
+
},
|
|
102
|
+
{
|
|
103
|
+
id: 'nvidia-nemo',
|
|
104
|
+
kind: 'stt',
|
|
105
|
+
label: 'NVIDIA NeMo',
|
|
106
|
+
envVars: [],
|
|
107
|
+
local: true,
|
|
108
|
+
description: 'Local/offline ASR via NVIDIA NeMo.',
|
|
109
|
+
features: ['local', 'offline'],
|
|
110
|
+
},
|
|
111
|
+
{
|
|
112
|
+
id: 'openai-tts',
|
|
113
|
+
kind: 'tts',
|
|
114
|
+
label: 'OpenAI TTS',
|
|
115
|
+
envVars: ['OPENAI_API_KEY'],
|
|
116
|
+
local: false,
|
|
117
|
+
streaming: true,
|
|
118
|
+
description: 'Hosted speech synthesis via OpenAI speech models.',
|
|
119
|
+
defaultModel: 'tts-1',
|
|
120
|
+
defaultVoice: 'nova',
|
|
121
|
+
features: ['cloud', 'tts'],
|
|
122
|
+
},
|
|
123
|
+
{
|
|
124
|
+
id: 'elevenlabs',
|
|
125
|
+
kind: 'tts',
|
|
126
|
+
label: 'ElevenLabs',
|
|
127
|
+
envVars: ['ELEVENLABS_API_KEY'],
|
|
128
|
+
local: false,
|
|
129
|
+
streaming: true,
|
|
130
|
+
description: 'High-quality neural TTS with voice cloning.',
|
|
131
|
+
extensionName: 'voice-synthesis',
|
|
132
|
+
defaultModel: 'eleven_multilingual_v2',
|
|
133
|
+
features: ['cloud', 'tts', 'voice-cloning'],
|
|
134
|
+
},
|
|
135
|
+
{
|
|
136
|
+
id: 'google-cloud-tts',
|
|
137
|
+
kind: 'tts',
|
|
138
|
+
label: 'Google Cloud TTS',
|
|
139
|
+
envVars: ['GOOGLE_TTS_CREDENTIALS'],
|
|
140
|
+
local: false,
|
|
141
|
+
description: 'Google Cloud neural text-to-speech.',
|
|
142
|
+
features: ['cloud', 'tts'],
|
|
143
|
+
},
|
|
144
|
+
{
|
|
145
|
+
id: 'amazon-polly',
|
|
146
|
+
kind: 'tts',
|
|
147
|
+
label: 'Amazon Polly',
|
|
148
|
+
envVars: ['AWS_ACCESS_KEY_ID', 'AWS_SECRET_ACCESS_KEY'],
|
|
149
|
+
local: false,
|
|
150
|
+
streaming: true,
|
|
151
|
+
description: 'Amazon Polly neural text-to-speech.',
|
|
152
|
+
features: ['cloud', 'tts'],
|
|
153
|
+
},
|
|
154
|
+
{
|
|
155
|
+
id: 'azure-speech-tts',
|
|
156
|
+
kind: 'tts',
|
|
157
|
+
label: 'Azure Speech TTS',
|
|
158
|
+
envVars: ['AZURE_SPEECH_KEY', 'AZURE_SPEECH_REGION'],
|
|
159
|
+
local: false,
|
|
160
|
+
streaming: true,
|
|
161
|
+
description: 'Azure neural text-to-speech.',
|
|
162
|
+
features: ['cloud', 'tts'],
|
|
163
|
+
},
|
|
164
|
+
{
|
|
165
|
+
id: 'piper',
|
|
166
|
+
kind: 'tts',
|
|
167
|
+
label: 'Piper',
|
|
168
|
+
envVars: [],
|
|
169
|
+
local: true,
|
|
170
|
+
description: 'Offline local TTS via Piper.',
|
|
171
|
+
features: ['local', 'offline', 'tts'],
|
|
172
|
+
},
|
|
173
|
+
{
|
|
174
|
+
id: 'coqui',
|
|
175
|
+
kind: 'tts',
|
|
176
|
+
label: 'Coqui XTTS',
|
|
177
|
+
envVars: [],
|
|
178
|
+
local: true,
|
|
179
|
+
streaming: true,
|
|
180
|
+
description: 'Local XTTS / Coqui speech synthesis.',
|
|
181
|
+
features: ['local', 'tts', 'voice-cloning'],
|
|
182
|
+
},
|
|
183
|
+
{
|
|
184
|
+
id: 'bark',
|
|
185
|
+
kind: 'tts',
|
|
186
|
+
label: 'Bark',
|
|
187
|
+
envVars: [],
|
|
188
|
+
local: true,
|
|
189
|
+
description: 'Local Bark expressive speech synthesis.',
|
|
190
|
+
features: ['local', 'tts'],
|
|
191
|
+
},
|
|
192
|
+
{
|
|
193
|
+
id: 'styletts2',
|
|
194
|
+
kind: 'tts',
|
|
195
|
+
label: 'StyleTTS2',
|
|
196
|
+
envVars: [],
|
|
197
|
+
local: true,
|
|
198
|
+
description: 'Local StyleTTS2 expressive speech synthesis.',
|
|
199
|
+
features: ['local', 'tts'],
|
|
200
|
+
},
|
|
201
|
+
{
|
|
202
|
+
id: 'agentos-adaptive-vad',
|
|
203
|
+
kind: 'vad',
|
|
204
|
+
label: 'AgentOS Adaptive VAD',
|
|
205
|
+
envVars: [],
|
|
206
|
+
local: true,
|
|
207
|
+
description: 'Built-in adaptive VAD backed by environmental calibration.',
|
|
208
|
+
features: ['local', 'vad', 'adaptive'],
|
|
209
|
+
},
|
|
210
|
+
{
|
|
211
|
+
id: 'porcupine',
|
|
212
|
+
kind: 'wake-word',
|
|
213
|
+
label: 'Porcupine',
|
|
214
|
+
envVars: ['PICOVOICE_ACCESS_KEY'],
|
|
215
|
+
local: true,
|
|
216
|
+
description: 'Wake-word detection via Picovoice Porcupine.',
|
|
217
|
+
features: ['local', 'wake-word'],
|
|
218
|
+
},
|
|
219
|
+
{
|
|
220
|
+
id: 'openwakeword',
|
|
221
|
+
kind: 'wake-word',
|
|
222
|
+
label: 'OpenWakeWord',
|
|
223
|
+
envVars: [],
|
|
224
|
+
local: true,
|
|
225
|
+
description: 'Wake-word detection via OpenWakeWord.',
|
|
226
|
+
features: ['local', 'wake-word'],
|
|
227
|
+
},
|
|
228
|
+
];
|
|
229
|
+
export function getSpeechProviderCatalog(kind) {
|
|
230
|
+
if (!kind)
|
|
231
|
+
return [...SPEECH_PROVIDER_CATALOG];
|
|
232
|
+
return SPEECH_PROVIDER_CATALOG.filter((entry) => entry.kind === kind);
|
|
233
|
+
}
|
|
234
|
+
export function getSpeechProviderKinds() {
|
|
235
|
+
return ['telephony', 'stt', 'tts', 'vad', 'wake-word'];
|
|
236
|
+
}
|
|
237
|
+
export function findSpeechProviderCatalogEntry(id) {
|
|
238
|
+
return SPEECH_PROVIDER_CATALOG.find((entry) => entry.id === id);
|
|
239
|
+
}
|
|
240
|
+
export function isSpeechProviderConfigured(entry, env = process.env) {
|
|
241
|
+
return entry.local || entry.envVars.every((envVar) => Boolean(env[envVar]));
|
|
242
|
+
}
|
|
243
|
+
//# sourceMappingURL=providerCatalog.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"providerCatalog.js","sourceRoot":"","sources":["../../src/speech/providerCatalog.ts"],"names":[],"mappings":"AAEA,MAAM,CAAC,MAAM,uBAAuB,GAA0C;IAC5E;QACE,EAAE,EAAE,QAAQ;QACZ,IAAI,EAAE,WAAW;QACjB,KAAK,EAAE,QAAQ;QACf,OAAO,EAAE,CAAC,oBAAoB,EAAE,mBAAmB,CAAC;QACpD,KAAK,EAAE,KAAK;QACZ,WAAW,EAAE,uDAAuD;QACpE,aAAa,EAAE,cAAc;QAC7B,QAAQ,EAAE,CAAC,WAAW,EAAE,UAAU,EAAE,eAAe,CAAC;KACrD;IACD;QACE,EAAE,EAAE,QAAQ;QACZ,IAAI,EAAE,WAAW;QACjB,KAAK,EAAE,QAAQ;QACf,OAAO,EAAE,CAAC,gBAAgB,EAAE,sBAAsB,CAAC;QACnD,KAAK,EAAE,KAAK;QACZ,WAAW,EAAE,iDAAiD;QAC9D,aAAa,EAAE,cAAc;QAC7B,QAAQ,EAAE,CAAC,WAAW,EAAE,KAAK,EAAE,eAAe,CAAC;KAChD;IACD;QACE,EAAE,EAAE,OAAO;QACX,IAAI,EAAE,WAAW;QACjB,KAAK,EAAE,OAAO;QACd,OAAO,EAAE,CAAC,eAAe,EAAE,kBAAkB,CAAC;QAC9C,KAAK,EAAE,KAAK;QACZ,WAAW,EAAE,iDAAiD;QAC9D,aAAa,EAAE,aAAa;QAC5B,QAAQ,EAAE,CAAC,WAAW,EAAE,KAAK,CAAC;KAC/B;IACD;QACE,EAAE,EAAE,gBAAgB;QACpB,IAAI,EAAE,KAAK;QACX,KAAK,EAAE,gBAAgB;QACvB,OAAO,EAAE,CAAC,gBAAgB,CAAC;QAC3B,KAAK,EAAE,KAAK;QACZ,WAAW,EAAE,qEAAqE;QAClF,YAAY,EAAE,WAAW;QACzB,SAAS,EAAE,KAAK;QAChB,QAAQ,EAAE,CAAC,OAAO,EAAE,YAAY,EAAE,eAAe,CAAC;KACnD;IACD;QACE,EAAE,EAAE,UAAU;QACd,IAAI,EAAE,KAAK;QACX,KAAK,EAAE,UAAU;QACjB,OAAO,EAAE,CAAC,kBAAkB,CAAC;QAC7B,KAAK,EAAE,KAAK;QACZ,WAAW,EAAE,wDAAwD;QACrE,SAAS,EAAE,IAAI;QACf,QAAQ,EAAE,CAAC,OAAO,EAAE,WAAW,CAAC;KACjC;IACD;QACE,EAAE,EAAE,YAAY;QAChB,IAAI,EAAE,KAAK;QACX,KAAK,EAAE,YAAY;QACnB,OAAO,EAAE,CAAC,oBAAoB,CAAC;QAC/B,KAAK,EAAE,KAAK;QACZ,WAAW,EAAE,uDAAuD;QACpE,SAAS,EAAE,IAAI;QACf,QAAQ,EAAE,CAAC,OAAO,EAAE,WAAW,EAAE,aAAa,CAAC;KAChD;IACD;QACE,EAAE,EAAE,kBAAkB;QACtB,IAAI,EAAE,KAAK;QACX,KAAK,EAAE,kBAAkB;QACzB,OAAO,EAAE,CAAC,wBAAwB,CAAC;QACnC,KAAK,EAAE,KAAK;QACZ,WAAW,EAAE,kCAAkC;QAC/C,SAAS,EAAE,IAAI;QACf,QAAQ,EAAE,CAAC,OAAO,EAAE,WAAW,CAAC;KACjC;IACD;QACE,EAAE,EAAE,kBAAkB;QACtB,IAAI,EAAE,KAAK;QACX,KAAK,EAAE,kBAAkB;QACzB,OAAO,EAAE,CAAC,kBAAkB,EAAE,qBAAqB,CAAC;QACpD,KAAK,EAAE,KAAK;QACZ,WAAW,EAAE,4DAA4D;QACzE,SAAS,EAAE,IAAI;QACf,QAAQ,EAAE,CAAC,OAAO,EAAE,WAAW,CAAC;KACjC;IACD;QACE,EAAE,EAAE,eAAe;QACnB,IAAI,EAAE,KAAK;QACX,KAAK,EAAE,aAAa;QACpB,OAAO,EAAE,EAAE;QACX,KAAK,EAAE,IAAI;QACX,WAAW,EAAE,0CAA0C;QACvD,QAAQ,EAAE,CAAC,OAAO,EAAE,SAAS,CAAC;KAC/B;IACD;QACE,EAAE,EAAE,MAAM;QACV,IAAI,EAAE,KAAK;QACX,KAAK,EAAE,MAAM;QACb,OAAO,EAAE,EAAE;QACX,KAAK,EAAE,IAAI;QACX,WAAW,EAAE,8CAA8C;QAC3D,SAAS,EAAE,IAAI;QACf,QAAQ,EAAE,CAAC,OAAO,EAAE,SAAS,EAAE,WAAW,CAAC;KAC5C;IACD;QACE,EAAE,EAAE,aAAa;QACjB,IAAI,EAAE,KAAK;QACX,KAAK,EAAE,aAAa;QACpB,OAAO,EAAE,EAAE;QACX,KAAK,EAAE,IAAI;QACX,WAAW,EAAE,oCAAoC;QACjD,QAAQ,EAAE,CAAC,OAAO,EAAE,SAAS,CAAC;KAC/B;IACD;QACE,EAAE,EAAE,YAAY;QAChB,IAAI,EAAE,KAAK;QACX,KAAK,EAAE,YAAY;QACnB,OAAO,EAAE,CAAC,gBAAgB,CAAC;QAC3B,KAAK,EAAE,KAAK;QACZ,SAAS,EAAE,IAAI;QACf,WAAW,EAAE,mDAAmD;QAChE,YAAY,EAAE,OAAO;QACrB,YAAY,EAAE,MAAM;QACpB,QAAQ,EAAE,CAAC,OAAO,EAAE,KAAK,CAAC;KAC3B;IACD;QACE,EAAE,EAAE,YAAY;QAChB,IAAI,EAAE,KAAK;QACX,KAAK,EAAE,YAAY;QACnB,OAAO,EAAE,CAAC,oBAAoB,CAAC;QAC/B,KAAK,EAAE,KAAK;QACZ,SAAS,EAAE,IAAI;QACf,WAAW,EAAE,6CAA6C;QAC1D,aAAa,EAAE,iBAAiB;QAChC,YAAY,EAAE,wBAAwB;QACtC,QAAQ,EAAE,CAAC,OAAO,EAAE,KAAK,EAAE,eAAe,CAAC;KAC5C;IACD;QACE,EAAE,EAAE,kBAAkB;QACtB,IAAI,EAAE,KAAK;QACX,KAAK,EAAE,kBAAkB;QACzB,OAAO,EAAE,CAAC,wBAAwB,CAAC;QACnC,KAAK,EAAE,KAAK;QACZ,WAAW,EAAE,qCAAqC;QAClD,QAAQ,EAAE,CAAC,OAAO,EAAE,KAAK,CAAC;KAC3B;IACD;QACE,EAAE,EAAE,cAAc;QAClB,IAAI,EAAE,KAAK;QACX,KAAK,EAAE,cAAc;QACrB,OAAO,EAAE,CAAC,mBAAmB,EAAE,uBAAuB,CAAC;QACvD,KAAK,EAAE,KAAK;QACZ,SAAS,EAAE,IAAI;QACf,WAAW,EAAE,qCAAqC;QAClD,QAAQ,EAAE,CAAC,OAAO,EAAE,KAAK,CAAC;KAC3B;IACD;QACE,EAAE,EAAE,kBAAkB;QACtB,IAAI,EAAE,KAAK;QACX,KAAK,EAAE,kBAAkB;QACzB,OAAO,EAAE,CAAC,kBAAkB,EAAE,qBAAqB,CAAC;QACpD,KAAK,EAAE,KAAK;QACZ,SAAS,EAAE,IAAI;QACf,WAAW,EAAE,8BAA8B;QAC3C,QAAQ,EAAE,CAAC,OAAO,EAAE,KAAK,CAAC;KAC3B;IACD;QACE,EAAE,EAAE,OAAO;QACX,IAAI,EAAE,KAAK;QACX,KAAK,EAAE,OAAO;QACd,OAAO,EAAE,EAAE;QACX,KAAK,EAAE,IAAI;QACX,WAAW,EAAE,8BAA8B;QAC3C,QAAQ,EAAE,CAAC,OAAO,EAAE,SAAS,EAAE,KAAK,CAAC;KACtC;IACD;QACE,EAAE,EAAE,OAAO;QACX,IAAI,EAAE,KAAK;QACX,KAAK,EAAE,YAAY;QACnB,OAAO,EAAE,EAAE;QACX,KAAK,EAAE,IAAI;QACX,SAAS,EAAE,IAAI;QACf,WAAW,EAAE,sCAAsC;QACnD,QAAQ,EAAE,CAAC,OAAO,EAAE,KAAK,EAAE,eAAe,CAAC;KAC5C;IACD;QACE,EAAE,EAAE,MAAM;QACV,IAAI,EAAE,KAAK;QACX,KAAK,EAAE,MAAM;QACb,OAAO,EAAE,EAAE;QACX,KAAK,EAAE,IAAI;QACX,WAAW,EAAE,yCAAyC;QACtD,QAAQ,EAAE,CAAC,OAAO,EAAE,KAAK,CAAC;KAC3B;IACD;QACE,EAAE,EAAE,WAAW;QACf,IAAI,EAAE,KAAK;QACX,KAAK,EAAE,WAAW;QAClB,OAAO,EAAE,EAAE;QACX,KAAK,EAAE,IAAI;QACX,WAAW,EAAE,8CAA8C;QAC3D,QAAQ,EAAE,CAAC,OAAO,EAAE,KAAK,CAAC;KAC3B;IACD;QACE,EAAE,EAAE,sBAAsB;QAC1B,IAAI,EAAE,KAAK;QACX,KAAK,EAAE,sBAAsB;QAC7B,OAAO,EAAE,EAAE;QACX,KAAK,EAAE,IAAI;QACX,WAAW,EAAE,4DAA4D;QACzE,QAAQ,EAAE,CAAC,OAAO,EAAE,KAAK,EAAE,UAAU,CAAC;KACvC;IACD;QACE,EAAE,EAAE,WAAW;QACf,IAAI,EAAE,WAAW;QACjB,KAAK,EAAE,WAAW;QAClB,OAAO,EAAE,CAAC,sBAAsB,CAAC;QACjC,KAAK,EAAE,IAAI;QACX,WAAW,EAAE,8CAA8C;QAC3D,QAAQ,EAAE,CAAC,OAAO,EAAE,WAAW,CAAC;KACjC;IACD;QACE,EAAE,EAAE,cAAc;QAClB,IAAI,EAAE,WAAW;QACjB,KAAK,EAAE,cAAc;QACrB,OAAO,EAAE,EAAE;QACX,KAAK,EAAE,IAAI;QACX,WAAW,EAAE,uCAAuC;QACpD,QAAQ,EAAE,CAAC,OAAO,EAAE,WAAW,CAAC;KACjC;CACO,CAAC;AAEX,MAAM,UAAU,wBAAwB,CACtC,IAAyB;IAEzB,IAAI,CAAC,IAAI;QAAE,OAAO,CAAC,GAAG,uBAAuB,CAAC,CAAC;IAC/C,OAAO,uBAAuB,CAAC,MAAM,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,IAAI,KAAK,IAAI,CAAC,CAAC;AACxE,CAAC;AAED,MAAM,UAAU,sBAAsB;IACpC,OAAO,CAAC,WAAW,EAAE,KAAK,EAAE,KAAK,EAAE,KAAK,EAAE,WAAW,CAAC,CAAC;AACzD,CAAC;AAED,MAAM,UAAU,8BAA8B,CAC5C,EAAU;IAEV,OAAO,uBAAuB,CAAC,IAAI,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,EAAE,KAAK,EAAE,CAAC,CAAC;AAClE,CAAC;AAED,MAAM,UAAU,0BAA0B,CACxC,KAAiC,EACjC,MAA0C,OAAO,CAAC,GAAG;IAErD,OAAO,KAAK,CAAC,KAAK,IAAI,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC,CAAC,MAAM,EAAE,EAAE,CAAC,OAAO,CAAC,GAAG,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC;AAC9E,CAAC"}
|