@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.
Files changed (226) hide show
  1. package/README.md +7 -7
  2. package/dist/cognitive_substrate/IGMI.d.ts +2 -0
  3. package/dist/cognitive_substrate/IGMI.d.ts.map +1 -1
  4. package/dist/cognitive_substrate/IGMI.js.map +1 -1
  5. package/dist/cognitive_substrate/personas/IPersonaDefinition.d.ts +2 -0
  6. package/dist/cognitive_substrate/personas/IPersonaDefinition.d.ts.map +1 -1
  7. package/dist/core/audio/AdaptiveVAD.d.ts +147 -0
  8. package/dist/core/audio/AdaptiveVAD.d.ts.map +1 -0
  9. package/dist/core/audio/AdaptiveVAD.js +239 -0
  10. package/dist/core/audio/AdaptiveVAD.js.map +1 -0
  11. package/dist/core/audio/AudioProcessor.d.ts +121 -0
  12. package/dist/core/audio/AudioProcessor.d.ts.map +1 -0
  13. package/dist/core/audio/AudioProcessor.js +299 -0
  14. package/dist/core/audio/AudioProcessor.js.map +1 -0
  15. package/dist/core/audio/EnvironmentalCalibrator.d.ts +187 -0
  16. package/dist/core/audio/EnvironmentalCalibrator.d.ts.map +1 -0
  17. package/dist/core/audio/EnvironmentalCalibrator.js +360 -0
  18. package/dist/core/audio/EnvironmentalCalibrator.js.map +1 -0
  19. package/dist/core/audio/SilenceDetector.d.ts +111 -0
  20. package/dist/core/audio/SilenceDetector.d.ts.map +1 -0
  21. package/dist/core/audio/SilenceDetector.js +178 -0
  22. package/dist/core/audio/SilenceDetector.js.map +1 -0
  23. package/dist/core/llm/IPromptEngine.d.ts +2 -0
  24. package/dist/core/llm/IPromptEngine.d.ts.map +1 -1
  25. package/dist/core/llm/IPromptEngine.js.map +1 -1
  26. package/dist/core/llm/providers/implementations/OllamaProvider.js +2 -2
  27. package/dist/core/llm/providers/implementations/OllamaProvider.js.map +1 -1
  28. package/dist/core/llm/providers/implementations/OpenAIProvider.d.ts.map +1 -1
  29. package/dist/core/llm/providers/implementations/OpenAIProvider.js +8 -1
  30. package/dist/core/llm/providers/implementations/OpenAIProvider.js.map +1 -1
  31. package/dist/discovery/CapabilityIndex.d.ts.map +1 -1
  32. package/dist/discovery/CapabilityIndex.js +18 -4
  33. package/dist/discovery/CapabilityIndex.js.map +1 -1
  34. package/dist/extensions/ExtensionManager.d.ts.map +1 -1
  35. package/dist/extensions/ExtensionManager.js +9 -1
  36. package/dist/extensions/ExtensionManager.js.map +1 -1
  37. package/dist/extensions/types.d.ts +20 -2
  38. package/dist/extensions/types.d.ts.map +1 -1
  39. package/dist/extensions/types.js +4 -0
  40. package/dist/extensions/types.js.map +1 -1
  41. package/dist/index.d.ts +2 -0
  42. package/dist/index.d.ts.map +1 -1
  43. package/dist/index.js +4 -0
  44. package/dist/index.js.map +1 -1
  45. package/dist/memory/CognitiveMemoryManager.d.ts +133 -0
  46. package/dist/memory/CognitiveMemoryManager.d.ts.map +1 -0
  47. package/dist/memory/CognitiveMemoryManager.js +519 -0
  48. package/dist/memory/CognitiveMemoryManager.js.map +1 -0
  49. package/dist/memory/config.d.ts +135 -0
  50. package/dist/memory/config.d.ts.map +1 -0
  51. package/dist/memory/config.js +28 -0
  52. package/dist/memory/config.js.map +1 -0
  53. package/dist/memory/consolidation/ConsolidationPipeline.d.ts +70 -0
  54. package/dist/memory/consolidation/ConsolidationPipeline.d.ts.map +1 -0
  55. package/dist/memory/consolidation/ConsolidationPipeline.js +301 -0
  56. package/dist/memory/consolidation/ConsolidationPipeline.js.map +1 -0
  57. package/dist/memory/context/CompactionEngine.d.ts +21 -0
  58. package/dist/memory/context/CompactionEngine.d.ts.map +1 -0
  59. package/dist/memory/context/CompactionEngine.js +35 -0
  60. package/dist/memory/context/CompactionEngine.js.map +1 -0
  61. package/dist/memory/context/CompactionLog.d.ts +50 -0
  62. package/dist/memory/context/CompactionLog.d.ts.map +1 -0
  63. package/dist/memory/context/CompactionLog.js +139 -0
  64. package/dist/memory/context/CompactionLog.js.map +1 -0
  65. package/dist/memory/context/ContextWindowManager.d.ts +114 -0
  66. package/dist/memory/context/ContextWindowManager.d.ts.map +1 -0
  67. package/dist/memory/context/ContextWindowManager.js +225 -0
  68. package/dist/memory/context/ContextWindowManager.js.map +1 -0
  69. package/dist/memory/context/RollingSummaryChain.d.ts +53 -0
  70. package/dist/memory/context/RollingSummaryChain.d.ts.map +1 -0
  71. package/dist/memory/context/RollingSummaryChain.js +193 -0
  72. package/dist/memory/context/RollingSummaryChain.js.map +1 -0
  73. package/dist/memory/context/index.d.ts +15 -0
  74. package/dist/memory/context/index.d.ts.map +1 -0
  75. package/dist/memory/context/index.js +14 -0
  76. package/dist/memory/context/index.js.map +1 -0
  77. package/dist/memory/context/strategies/HierarchicalStrategy.d.ts +25 -0
  78. package/dist/memory/context/strategies/HierarchicalStrategy.d.ts.map +1 -0
  79. package/dist/memory/context/strategies/HierarchicalStrategy.js +240 -0
  80. package/dist/memory/context/strategies/HierarchicalStrategy.js.map +1 -0
  81. package/dist/memory/context/strategies/HybridStrategy.d.ts +26 -0
  82. package/dist/memory/context/strategies/HybridStrategy.d.ts.map +1 -0
  83. package/dist/memory/context/strategies/HybridStrategy.js +245 -0
  84. package/dist/memory/context/strategies/HybridStrategy.js.map +1 -0
  85. package/dist/memory/context/strategies/SlidingSummaryStrategy.d.ts +22 -0
  86. package/dist/memory/context/strategies/SlidingSummaryStrategy.d.ts.map +1 -0
  87. package/dist/memory/context/strategies/SlidingSummaryStrategy.js +203 -0
  88. package/dist/memory/context/strategies/SlidingSummaryStrategy.js.map +1 -0
  89. package/dist/memory/context/types.d.ts +120 -0
  90. package/dist/memory/context/types.d.ts.map +1 -0
  91. package/dist/memory/context/types.js +17 -0
  92. package/dist/memory/context/types.js.map +1 -0
  93. package/dist/memory/decay/DecayModel.d.ts +87 -0
  94. package/dist/memory/decay/DecayModel.d.ts.map +1 -0
  95. package/dist/memory/decay/DecayModel.js +117 -0
  96. package/dist/memory/decay/DecayModel.js.map +1 -0
  97. package/dist/memory/decay/RetrievalPriorityScorer.d.ts +63 -0
  98. package/dist/memory/decay/RetrievalPriorityScorer.d.ts.map +1 -0
  99. package/dist/memory/decay/RetrievalPriorityScorer.js +110 -0
  100. package/dist/memory/decay/RetrievalPriorityScorer.js.map +1 -0
  101. package/dist/memory/encoding/ContentFeatureDetector.d.ts +37 -0
  102. package/dist/memory/encoding/ContentFeatureDetector.d.ts.map +1 -0
  103. package/dist/memory/encoding/ContentFeatureDetector.js +176 -0
  104. package/dist/memory/encoding/ContentFeatureDetector.js.map +1 -0
  105. package/dist/memory/encoding/EncodingModel.d.ts +67 -0
  106. package/dist/memory/encoding/EncodingModel.d.ts.map +1 -0
  107. package/dist/memory/encoding/EncodingModel.js +171 -0
  108. package/dist/memory/encoding/EncodingModel.js.map +1 -0
  109. package/dist/memory/extension/CognitiveMemoryExtension.d.ts +18 -0
  110. package/dist/memory/extension/CognitiveMemoryExtension.d.ts.map +1 -0
  111. package/dist/memory/extension/CognitiveMemoryExtension.js +131 -0
  112. package/dist/memory/extension/CognitiveMemoryExtension.js.map +1 -0
  113. package/dist/memory/graph/GraphologyMemoryGraph.d.ts +29 -0
  114. package/dist/memory/graph/GraphologyMemoryGraph.d.ts.map +1 -0
  115. package/dist/memory/graph/GraphologyMemoryGraph.js +224 -0
  116. package/dist/memory/graph/GraphologyMemoryGraph.js.map +1 -0
  117. package/dist/memory/graph/IMemoryGraph.d.ts +70 -0
  118. package/dist/memory/graph/IMemoryGraph.d.ts.map +1 -0
  119. package/dist/memory/graph/IMemoryGraph.js +15 -0
  120. package/dist/memory/graph/IMemoryGraph.js.map +1 -0
  121. package/dist/memory/graph/KnowledgeGraphMemoryGraph.d.ts +38 -0
  122. package/dist/memory/graph/KnowledgeGraphMemoryGraph.d.ts.map +1 -0
  123. package/dist/memory/graph/KnowledgeGraphMemoryGraph.js +264 -0
  124. package/dist/memory/graph/KnowledgeGraphMemoryGraph.js.map +1 -0
  125. package/dist/memory/graph/SpreadingActivation.d.ts +36 -0
  126. package/dist/memory/graph/SpreadingActivation.d.ts.map +1 -0
  127. package/dist/memory/graph/SpreadingActivation.js +100 -0
  128. package/dist/memory/graph/SpreadingActivation.js.map +1 -0
  129. package/dist/memory/index.d.ts +52 -0
  130. package/dist/memory/index.d.ts.map +1 -0
  131. package/dist/memory/index.js +43 -0
  132. package/dist/memory/index.js.map +1 -0
  133. package/dist/memory/observation/MemoryObserver.d.ts +59 -0
  134. package/dist/memory/observation/MemoryObserver.d.ts.map +1 -0
  135. package/dist/memory/observation/MemoryObserver.js +136 -0
  136. package/dist/memory/observation/MemoryObserver.js.map +1 -0
  137. package/dist/memory/observation/MemoryReflector.d.ts +51 -0
  138. package/dist/memory/observation/MemoryReflector.d.ts.map +1 -0
  139. package/dist/memory/observation/MemoryReflector.js +184 -0
  140. package/dist/memory/observation/MemoryReflector.js.map +1 -0
  141. package/dist/memory/observation/ObservationBuffer.d.ts +51 -0
  142. package/dist/memory/observation/ObservationBuffer.d.ts.map +1 -0
  143. package/dist/memory/observation/ObservationBuffer.js +81 -0
  144. package/dist/memory/observation/ObservationBuffer.js.map +1 -0
  145. package/dist/memory/prompt/MemoryFormatters.d.ts +22 -0
  146. package/dist/memory/prompt/MemoryFormatters.d.ts.map +1 -0
  147. package/dist/memory/prompt/MemoryFormatters.js +113 -0
  148. package/dist/memory/prompt/MemoryFormatters.js.map +1 -0
  149. package/dist/memory/prompt/MemoryPromptAssembler.d.ts +40 -0
  150. package/dist/memory/prompt/MemoryPromptAssembler.d.ts.map +1 -0
  151. package/dist/memory/prompt/MemoryPromptAssembler.js +180 -0
  152. package/dist/memory/prompt/MemoryPromptAssembler.js.map +1 -0
  153. package/dist/memory/prospective/ProspectiveMemoryManager.d.ts +83 -0
  154. package/dist/memory/prospective/ProspectiveMemoryManager.d.ts.map +1 -0
  155. package/dist/memory/prospective/ProspectiveMemoryManager.js +128 -0
  156. package/dist/memory/prospective/ProspectiveMemoryManager.js.map +1 -0
  157. package/dist/memory/store/MemoryStore.d.ts +76 -0
  158. package/dist/memory/store/MemoryStore.d.ts.map +1 -0
  159. package/dist/memory/store/MemoryStore.js +320 -0
  160. package/dist/memory/store/MemoryStore.js.map +1 -0
  161. package/dist/memory/types.d.ts +184 -0
  162. package/dist/memory/types.d.ts.map +1 -0
  163. package/dist/memory/types.js +14 -0
  164. package/dist/memory/types.js.map +1 -0
  165. package/dist/memory/working/CognitiveWorkingMemory.d.ts +87 -0
  166. package/dist/memory/working/CognitiveWorkingMemory.d.ts.map +1 -0
  167. package/dist/memory/working/CognitiveWorkingMemory.js +230 -0
  168. package/dist/memory/working/CognitiveWorkingMemory.js.map +1 -0
  169. package/dist/rag/EmbeddingManager.d.ts.map +1 -1
  170. package/dist/rag/EmbeddingManager.js +8 -1
  171. package/dist/rag/EmbeddingManager.js.map +1 -1
  172. package/dist/rag/HydeRetriever.d.ts +111 -0
  173. package/dist/rag/HydeRetriever.d.ts.map +1 -0
  174. package/dist/rag/HydeRetriever.js +143 -0
  175. package/dist/rag/HydeRetriever.js.map +1 -0
  176. package/dist/rag/IRetrievalAugmentor.d.ts +15 -0
  177. package/dist/rag/IRetrievalAugmentor.d.ts.map +1 -1
  178. package/dist/rag/index.d.ts +1 -0
  179. package/dist/rag/index.d.ts.map +1 -1
  180. package/dist/rag/index.js +2 -0
  181. package/dist/rag/index.js.map +1 -1
  182. package/dist/speech/SpeechProviderRegistry.d.ts +17 -0
  183. package/dist/speech/SpeechProviderRegistry.d.ts.map +1 -0
  184. package/dist/speech/SpeechProviderRegistry.js +47 -0
  185. package/dist/speech/SpeechProviderRegistry.js.map +1 -0
  186. package/dist/speech/SpeechRuntime.d.ts +30 -0
  187. package/dist/speech/SpeechRuntime.d.ts.map +1 -0
  188. package/dist/speech/SpeechRuntime.js +156 -0
  189. package/dist/speech/SpeechRuntime.js.map +1 -0
  190. package/dist/speech/SpeechSession.d.ts +37 -0
  191. package/dist/speech/SpeechSession.d.ts.map +1 -0
  192. package/dist/speech/SpeechSession.js +285 -0
  193. package/dist/speech/SpeechSession.js.map +1 -0
  194. package/dist/speech/audio.d.ts +3 -0
  195. package/dist/speech/audio.d.ts.map +1 -0
  196. package/dist/speech/audio.js +39 -0
  197. package/dist/speech/audio.js.map +1 -0
  198. package/dist/speech/index.d.ts +11 -0
  199. package/dist/speech/index.d.ts.map +1 -0
  200. package/dist/speech/index.js +11 -0
  201. package/dist/speech/index.js.map +1 -0
  202. package/dist/speech/providerCatalog.d.ts +7 -0
  203. package/dist/speech/providerCatalog.d.ts.map +1 -0
  204. package/dist/speech/providerCatalog.js +243 -0
  205. package/dist/speech/providerCatalog.js.map +1 -0
  206. package/dist/speech/providers/BuiltInAdaptiveVadProvider.d.ts +20 -0
  207. package/dist/speech/providers/BuiltInAdaptiveVadProvider.d.ts.map +1 -0
  208. package/dist/speech/providers/BuiltInAdaptiveVadProvider.js +29 -0
  209. package/dist/speech/providers/BuiltInAdaptiveVadProvider.js.map +1 -0
  210. package/dist/speech/providers/ElevenLabsTextToSpeechProvider.d.ts +20 -0
  211. package/dist/speech/providers/ElevenLabsTextToSpeechProvider.d.ts.map +1 -0
  212. package/dist/speech/providers/ElevenLabsTextToSpeechProvider.js +95 -0
  213. package/dist/speech/providers/ElevenLabsTextToSpeechProvider.js.map +1 -0
  214. package/dist/speech/providers/OpenAITextToSpeechProvider.d.ts +20 -0
  215. package/dist/speech/providers/OpenAITextToSpeechProvider.d.ts.map +1 -0
  216. package/dist/speech/providers/OpenAITextToSpeechProvider.js +75 -0
  217. package/dist/speech/providers/OpenAITextToSpeechProvider.js.map +1 -0
  218. package/dist/speech/providers/OpenAIWhisperSpeechToTextProvider.d.ts +18 -0
  219. package/dist/speech/providers/OpenAIWhisperSpeechToTextProvider.d.ts.map +1 -0
  220. package/dist/speech/providers/OpenAIWhisperSpeechToTextProvider.js +109 -0
  221. package/dist/speech/providers/OpenAIWhisperSpeechToTextProvider.js.map +1 -0
  222. package/dist/speech/types.d.ts +272 -0
  223. package/dist/speech/types.d.ts.map +1 -0
  224. package/dist/speech/types.js +2 -0
  225. package/dist/speech/types.js.map +1 -0
  226. package/package.json +22 -2
@@ -0,0 +1,20 @@
1
+ import type { AdaptiveVADConfig } from '../../core/audio/AdaptiveVAD.js';
2
+ import type { CalibrationConfig } from '../../core/audio/EnvironmentalCalibrator.js';
3
+ import type { SpeechVadDecision, SpeechVadProvider } from '../types.js';
4
+ export interface BuiltInAdaptiveVadProviderConfig {
5
+ sampleRate?: number;
6
+ frameDurationMs?: number;
7
+ calibration?: CalibrationConfig;
8
+ vad?: AdaptiveVADConfig;
9
+ }
10
+ export declare class BuiltInAdaptiveVadProvider implements SpeechVadProvider {
11
+ readonly id = "agentos-adaptive-vad";
12
+ readonly displayName = "AgentOS Adaptive VAD";
13
+ private readonly calibrator;
14
+ private readonly vad;
15
+ constructor(config?: BuiltInAdaptiveVadProviderConfig);
16
+ processFrame(frame: Float32Array): SpeechVadDecision;
17
+ reset(): void;
18
+ getNoiseProfile(): import("../../core/audio/EnvironmentalCalibrator.js").NoiseProfile | null;
19
+ }
20
+ //# sourceMappingURL=BuiltInAdaptiveVadProvider.d.ts.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"BuiltInAdaptiveVadProvider.d.ts","sourceRoot":"","sources":["../../../src/speech/providers/BuiltInAdaptiveVadProvider.ts"],"names":[],"mappings":"AAEA,OAAO,KAAK,EACV,iBAAiB,EAClB,MAAM,iCAAiC,CAAC;AACzC,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,6CAA6C,CAAC;AACrF,OAAO,KAAK,EAAE,iBAAiB,EAAE,iBAAiB,EAAE,MAAM,aAAa,CAAC;AAExE,MAAM,WAAW,gCAAgC;IAC/C,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,WAAW,CAAC,EAAE,iBAAiB,CAAC;IAChC,GAAG,CAAC,EAAE,iBAAiB,CAAC;CACzB;AAED,qBAAa,0BAA2B,YAAW,iBAAiB;IAClE,SAAgB,EAAE,0BAA0B;IAC5C,SAAgB,WAAW,0BAA0B;IACrD,OAAO,CAAC,QAAQ,CAAC,UAAU,CAA0B;IACrD,OAAO,CAAC,QAAQ,CAAC,GAAG,CAAc;gBAEtB,MAAM,GAAE,gCAAqC;IAYzD,YAAY,CAAC,KAAK,EAAE,YAAY,GAAG,iBAAiB;IAUpD,KAAK,IAAI,IAAI;IAIb,eAAe;CAGhB"}
@@ -0,0 +1,29 @@
1
+ import { AdaptiveVAD } from '../../core/audio/AdaptiveVAD.js';
2
+ import { EnvironmentalCalibrator } from '../../core/audio/EnvironmentalCalibrator.js';
3
+ export class BuiltInAdaptiveVadProvider {
4
+ constructor(config = {}) {
5
+ this.id = 'agentos-adaptive-vad';
6
+ this.displayName = 'AgentOS Adaptive VAD';
7
+ this.calibrator = new EnvironmentalCalibrator({
8
+ sampleRate: config.sampleRate ?? 16000,
9
+ ...(config.calibration ?? {}),
10
+ });
11
+ this.vad = new AdaptiveVAD(config.vad ?? {}, this.calibrator, config.frameDurationMs ?? 20);
12
+ }
13
+ processFrame(frame) {
14
+ const result = this.vad.processFrame(frame);
15
+ return {
16
+ isSpeech: result.isSpeech,
17
+ confidence: result.confidence,
18
+ result,
19
+ profile: this.calibrator.getCurrentProfile(),
20
+ };
21
+ }
22
+ reset() {
23
+ this.vad.resetState();
24
+ }
25
+ getNoiseProfile() {
26
+ return this.calibrator.getCurrentProfile();
27
+ }
28
+ }
29
+ //# sourceMappingURL=BuiltInAdaptiveVadProvider.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"BuiltInAdaptiveVadProvider.js","sourceRoot":"","sources":["../../../src/speech/providers/BuiltInAdaptiveVadProvider.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,WAAW,EAAE,MAAM,iCAAiC,CAAC;AAC9D,OAAO,EAAE,uBAAuB,EAAE,MAAM,6CAA6C,CAAC;AActF,MAAM,OAAO,0BAA0B;IAMrC,YAAY,SAA2C,EAAE;QALzC,OAAE,GAAG,sBAAsB,CAAC;QAC5B,gBAAW,GAAG,sBAAsB,CAAC;QAKnD,IAAI,CAAC,UAAU,GAAG,IAAI,uBAAuB,CAAC;YAC5C,UAAU,EAAE,MAAM,CAAC,UAAU,IAAI,KAAM;YACvC,GAAG,CAAC,MAAM,CAAC,WAAW,IAAI,EAAE,CAAC;SAC9B,CAAC,CAAC;QACH,IAAI,CAAC,GAAG,GAAG,IAAI,WAAW,CACxB,MAAM,CAAC,GAAG,IAAI,EAAE,EAChB,IAAI,CAAC,UAAU,EACf,MAAM,CAAC,eAAe,IAAI,EAAE,CAC7B,CAAC;IACJ,CAAC;IAED,YAAY,CAAC,KAAmB;QAC9B,MAAM,MAAM,GAAG,IAAI,CAAC,GAAG,CAAC,YAAY,CAAC,KAAK,CAAC,CAAC;QAC5C,OAAO;YACL,QAAQ,EAAE,MAAM,CAAC,QAAQ;YACzB,UAAU,EAAE,MAAM,CAAC,UAAU;YAC7B,MAAM;YACN,OAAO,EAAE,IAAI,CAAC,UAAU,CAAC,iBAAiB,EAAE;SAC7C,CAAC;IACJ,CAAC;IAED,KAAK;QACH,IAAI,CAAC,GAAG,CAAC,UAAU,EAAE,CAAC;IACxB,CAAC;IAED,eAAe;QACb,OAAO,IAAI,CAAC,UAAU,CAAC,iBAAiB,EAAE,CAAC;IAC7C,CAAC;CACF"}
@@ -0,0 +1,20 @@
1
+ import type { SpeechSynthesisOptions, SpeechSynthesisResult, SpeechVoice, TextToSpeechProvider } from '../types.js';
2
+ export interface ElevenLabsTextToSpeechProviderConfig {
3
+ apiKey: string;
4
+ baseUrl?: string;
5
+ voiceId?: string;
6
+ model?: string;
7
+ fetchImpl?: typeof fetch;
8
+ }
9
+ export declare class ElevenLabsTextToSpeechProvider implements TextToSpeechProvider {
10
+ private readonly config;
11
+ readonly id = "elevenlabs";
12
+ readonly displayName = "ElevenLabs";
13
+ readonly supportsStreaming = true;
14
+ private readonly fetchImpl;
15
+ constructor(config: ElevenLabsTextToSpeechProviderConfig);
16
+ getProviderName(): string;
17
+ synthesize(text: string, options?: SpeechSynthesisOptions): Promise<SpeechSynthesisResult>;
18
+ listAvailableVoices(): Promise<SpeechVoice[]>;
19
+ }
20
+ //# sourceMappingURL=ElevenLabsTextToSpeechProvider.d.ts.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"ElevenLabsTextToSpeechProvider.d.ts","sourceRoot":"","sources":["../../../src/speech/providers/ElevenLabsTextToSpeechProvider.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EACV,sBAAsB,EACtB,qBAAqB,EACrB,WAAW,EACX,oBAAoB,EACrB,MAAM,aAAa,CAAC;AAErB,MAAM,WAAW,oCAAoC;IACnD,MAAM,EAAE,MAAM,CAAC;IACf,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,SAAS,CAAC,EAAE,OAAO,KAAK,CAAC;CAC1B;AAED,qBAAa,8BAA+B,YAAW,oBAAoB;IAM7D,OAAO,CAAC,QAAQ,CAAC,MAAM;IALnC,SAAgB,EAAE,gBAAgB;IAClC,SAAgB,WAAW,gBAAgB;IAC3C,SAAgB,iBAAiB,QAAQ;IACzC,OAAO,CAAC,QAAQ,CAAC,SAAS,CAAe;gBAEZ,MAAM,EAAE,oCAAoC;IAIzE,eAAe,IAAI,MAAM;IAInB,UAAU,CACd,IAAI,EAAE,MAAM,EACZ,OAAO,GAAE,sBAA2B,GACnC,OAAO,CAAC,qBAAqB,CAAC;IA8D3B,mBAAmB,IAAI,OAAO,CAAC,WAAW,EAAE,CAAC;CAwCpD"}
@@ -0,0 +1,95 @@
1
+ export class ElevenLabsTextToSpeechProvider {
2
+ constructor(config) {
3
+ this.config = config;
4
+ this.id = 'elevenlabs';
5
+ this.displayName = 'ElevenLabs';
6
+ this.supportsStreaming = true;
7
+ this.fetchImpl = config.fetchImpl ?? fetch;
8
+ }
9
+ getProviderName() {
10
+ return this.displayName;
11
+ }
12
+ async synthesize(text, options = {}) {
13
+ const voiceId = options.voice ??
14
+ this.config.voiceId ??
15
+ (typeof options.providerSpecificOptions?.voiceId === 'string'
16
+ ? options.providerSpecificOptions.voiceId
17
+ : undefined) ??
18
+ 'EXAVITQu4vr4xnSDxMaL';
19
+ const model = options.model ?? this.config.model ?? 'eleven_multilingual_v2';
20
+ const response = await this.fetchImpl(`${this.config.baseUrl ?? 'https://api.elevenlabs.io/v1'}/text-to-speech/${voiceId}`, {
21
+ method: 'POST',
22
+ headers: {
23
+ 'xi-api-key': this.config.apiKey,
24
+ 'Content-Type': 'application/json',
25
+ Accept: 'audio/mpeg',
26
+ },
27
+ body: JSON.stringify({
28
+ text,
29
+ model_id: model,
30
+ voice_settings: {
31
+ stability: typeof options.providerSpecificOptions?.stability === 'number'
32
+ ? options.providerSpecificOptions.stability
33
+ : 0.5,
34
+ similarity_boost: typeof options.providerSpecificOptions?.similarityBoost === 'number'
35
+ ? options.providerSpecificOptions.similarityBoost
36
+ : 0.75,
37
+ style: typeof options.providerSpecificOptions?.style === 'number'
38
+ ? options.providerSpecificOptions.style
39
+ : undefined,
40
+ use_speaker_boost: typeof options.providerSpecificOptions?.useSpeakerBoost === 'boolean'
41
+ ? options.providerSpecificOptions.useSpeakerBoost
42
+ : true,
43
+ },
44
+ }),
45
+ });
46
+ if (!response.ok) {
47
+ const message = await response.text();
48
+ throw new Error(`ElevenLabs synthesis failed (${response.status}): ${message}`);
49
+ }
50
+ const audioBuffer = Buffer.from(await response.arrayBuffer());
51
+ return {
52
+ audioBuffer,
53
+ mimeType: 'audio/mpeg',
54
+ cost: 0,
55
+ voiceUsed: voiceId,
56
+ providerName: this.displayName,
57
+ usage: {
58
+ characters: text.length,
59
+ modelUsed: model,
60
+ },
61
+ };
62
+ }
63
+ async listAvailableVoices() {
64
+ const response = await this.fetchImpl(`${this.config.baseUrl ?? 'https://api.elevenlabs.io/v1'}/voices`, {
65
+ method: 'GET',
66
+ headers: {
67
+ 'xi-api-key': this.config.apiKey,
68
+ },
69
+ });
70
+ if (!response.ok) {
71
+ return [];
72
+ }
73
+ const payload = (await response.json());
74
+ return (payload.voices ?? [])
75
+ .filter((voice) => typeof voice === 'object' && voice !== null)
76
+ .map((voice) => {
77
+ const labels = typeof voice.labels === 'object' && voice.labels !== null
78
+ ? voice.labels
79
+ : {};
80
+ return {
81
+ id: typeof voice.voice_id === 'string' ? voice.voice_id : '',
82
+ name: typeof voice.name === 'string' ? voice.name : 'Unknown',
83
+ lang: typeof labels.accent === 'string'
84
+ ? labels.accent
85
+ : typeof labels.language === 'string'
86
+ ? labels.language
87
+ : 'various',
88
+ description: typeof voice.description === 'string' ? voice.description : undefined,
89
+ provider: this.id,
90
+ };
91
+ })
92
+ .filter((voice) => voice.id);
93
+ }
94
+ }
95
+ //# sourceMappingURL=ElevenLabsTextToSpeechProvider.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"ElevenLabsTextToSpeechProvider.js","sourceRoot":"","sources":["../../../src/speech/providers/ElevenLabsTextToSpeechProvider.ts"],"names":[],"mappings":"AAeA,MAAM,OAAO,8BAA8B;IAMzC,YAA6B,MAA4C;QAA5C,WAAM,GAAN,MAAM,CAAsC;QALzD,OAAE,GAAG,YAAY,CAAC;QAClB,gBAAW,GAAG,YAAY,CAAC;QAC3B,sBAAiB,GAAG,IAAI,CAAC;QAIvC,IAAI,CAAC,SAAS,GAAG,MAAM,CAAC,SAAS,IAAI,KAAK,CAAC;IAC7C,CAAC;IAED,eAAe;QACb,OAAO,IAAI,CAAC,WAAW,CAAC;IAC1B,CAAC;IAED,KAAK,CAAC,UAAU,CACd,IAAY,EACZ,UAAkC,EAAE;QAEpC,MAAM,OAAO,GACX,OAAO,CAAC,KAAK;YACb,IAAI,CAAC,MAAM,CAAC,OAAO;YACnB,CAAC,OAAO,OAAO,CAAC,uBAAuB,EAAE,OAAO,KAAK,QAAQ;gBAC3D,CAAC,CAAC,OAAO,CAAC,uBAAuB,CAAC,OAAO;gBACzC,CAAC,CAAC,SAAS,CAAC;YACd,sBAAsB,CAAC;QACzB,MAAM,KAAK,GAAG,OAAO,CAAC,KAAK,IAAI,IAAI,CAAC,MAAM,CAAC,KAAK,IAAI,wBAAwB,CAAC;QAC7E,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,SAAS,CACnC,GAAG,IAAI,CAAC,MAAM,CAAC,OAAO,IAAI,8BAA8B,mBAAmB,OAAO,EAAE,EACpF;YACE,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,YAAY,EAAE,IAAI,CAAC,MAAM,CAAC,MAAM;gBAChC,cAAc,EAAE,kBAAkB;gBAClC,MAAM,EAAE,YAAY;aACrB;YACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC;gBACnB,IAAI;gBACJ,QAAQ,EAAE,KAAK;gBACf,cAAc,EAAE;oBACd,SAAS,EACP,OAAO,OAAO,CAAC,uBAAuB,EAAE,SAAS,KAAK,QAAQ;wBAC5D,CAAC,CAAC,OAAO,CAAC,uBAAuB,CAAC,SAAS;wBAC3C,CAAC,CAAC,GAAG;oBACT,gBAAgB,EACd,OAAO,OAAO,CAAC,uBAAuB,EAAE,eAAe,KAAK,QAAQ;wBAClE,CAAC,CAAC,OAAO,CAAC,uBAAuB,CAAC,eAAe;wBACjD,CAAC,CAAC,IAAI;oBACV,KAAK,EACH,OAAO,OAAO,CAAC,uBAAuB,EAAE,KAAK,KAAK,QAAQ;wBACxD,CAAC,CAAC,OAAO,CAAC,uBAAuB,CAAC,KAAK;wBACvC,CAAC,CAAC,SAAS;oBACf,iBAAiB,EACf,OAAO,OAAO,CAAC,uBAAuB,EAAE,eAAe,KAAK,SAAS;wBACnE,CAAC,CAAC,OAAO,CAAC,uBAAuB,CAAC,eAAe;wBACjD,CAAC,CAAC,IAAI;iBACX;aACF,CAAC;SACH,CACF,CAAC;QAEF,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;YACjB,MAAM,OAAO,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;YACtC,MAAM,IAAI,KAAK,CAAC,gCAAgC,QAAQ,CAAC,MAAM,MAAM,OAAO,EAAE,CAAC,CAAC;QAClF,CAAC;QAED,MAAM,WAAW,GAAG,MAAM,CAAC,IAAI,CAAC,MAAM,QAAQ,CAAC,WAAW,EAAE,CAAC,CAAC;QAC9D,OAAO;YACL,WAAW;YACX,QAAQ,EAAE,YAAY;YACtB,IAAI,EAAE,CAAC;YACP,SAAS,EAAE,OAAO;YAClB,YAAY,EAAE,IAAI,CAAC,WAAW;YAC9B,KAAK,EAAE;gBACL,UAAU,EAAE,IAAI,CAAC,MAAM;gBACvB,SAAS,EAAE,KAAK;aACjB;SACF,CAAC;IACJ,CAAC;IAED,KAAK,CAAC,mBAAmB;QACvB,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,SAAS,CACnC,GAAG,IAAI,CAAC,MAAM,CAAC,OAAO,IAAI,8BAA8B,SAAS,EACjE;YACE,MAAM,EAAE,KAAK;YACb,OAAO,EAAE;gBACP,YAAY,EAAE,IAAI,CAAC,MAAM,CAAC,MAAM;aACjC;SACF,CACF,CAAC;QAEF,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;YACjB,OAAO,EAAE,CAAC;QACZ,CAAC;QAED,MAAM,OAAO,GAAG,CAAC,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAgD,CAAC;QACvF,OAAO,CAAC,OAAO,CAAC,MAAM,IAAI,EAAE,CAAC;aAC1B,MAAM,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,OAAO,KAAK,KAAK,QAAQ,IAAI,KAAK,KAAK,IAAI,CAAC;aAC9D,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE;YACb,MAAM,MAAM,GACV,OAAO,KAAK,CAAC,MAAM,KAAK,QAAQ,IAAI,KAAK,CAAC,MAAM,KAAK,IAAI;gBACvD,CAAC,CAAE,KAAK,CAAC,MAAkC;gBAC3C,CAAC,CAAC,EAAE,CAAC;YAET,OAAO;gBACL,EAAE,EAAE,OAAO,KAAK,CAAC,QAAQ,KAAK,QAAQ,CAAC,CAAC,CAAC,KAAK,CAAC,QAAQ,CAAC,CAAC,CAAC,EAAE;gBAC5D,IAAI,EAAE,OAAO,KAAK,CAAC,IAAI,KAAK,QAAQ,CAAC,CAAC,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,CAAC,SAAS;gBAC7D,IAAI,EACF,OAAO,MAAM,CAAC,MAAM,KAAK,QAAQ;oBAC/B,CAAC,CAAC,MAAM,CAAC,MAAM;oBACf,CAAC,CAAC,OAAO,MAAM,CAAC,QAAQ,KAAK,QAAQ;wBACrC,CAAC,CAAC,MAAM,CAAC,QAAQ;wBACjB,CAAC,CAAC,SAAS;gBACf,WAAW,EACT,OAAO,KAAK,CAAC,WAAW,KAAK,QAAQ,CAAC,CAAC,CAAC,KAAK,CAAC,WAAW,CAAC,CAAC,CAAC,SAAS;gBACvE,QAAQ,EAAE,IAAI,CAAC,EAAE;aAClB,CAAC;QACJ,CAAC,CAAC;aACD,MAAM,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC;IACjC,CAAC;CACF"}
@@ -0,0 +1,20 @@
1
+ import type { SpeechSynthesisOptions, SpeechSynthesisResult, SpeechVoice, TextToSpeechProvider } from '../types.js';
2
+ export interface OpenAITextToSpeechProviderConfig {
3
+ apiKey: string;
4
+ baseUrl?: string;
5
+ model?: string;
6
+ voice?: string;
7
+ fetchImpl?: typeof fetch;
8
+ }
9
+ export declare class OpenAITextToSpeechProvider implements TextToSpeechProvider {
10
+ private readonly config;
11
+ readonly id = "openai-tts";
12
+ readonly displayName = "OpenAI TTS";
13
+ readonly supportsStreaming = true;
14
+ private readonly fetchImpl;
15
+ constructor(config: OpenAITextToSpeechProviderConfig);
16
+ getProviderName(): string;
17
+ synthesize(text: string, options?: SpeechSynthesisOptions): Promise<SpeechSynthesisResult>;
18
+ listAvailableVoices(): Promise<SpeechVoice[]>;
19
+ }
20
+ //# sourceMappingURL=OpenAITextToSpeechProvider.d.ts.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"OpenAITextToSpeechProvider.d.ts","sourceRoot":"","sources":["../../../src/speech/providers/OpenAITextToSpeechProvider.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EACV,sBAAsB,EACtB,qBAAqB,EACrB,WAAW,EACX,oBAAoB,EACrB,MAAM,aAAa,CAAC;AAErB,MAAM,WAAW,gCAAgC;IAC/C,MAAM,EAAE,MAAM,CAAC;IACf,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,SAAS,CAAC,EAAE,OAAO,KAAK,CAAC;CAC1B;AA4BD,qBAAa,0BAA2B,YAAW,oBAAoB;IAMzD,OAAO,CAAC,QAAQ,CAAC,MAAM;IALnC,SAAgB,EAAE,gBAAgB;IAClC,SAAgB,WAAW,gBAAgB;IAC3C,SAAgB,iBAAiB,QAAQ;IACzC,OAAO,CAAC,QAAQ,CAAC,SAAS,CAAe;gBAEZ,MAAM,EAAE,gCAAgC;IAIrE,eAAe,IAAI,MAAM;IAInB,UAAU,CACd,IAAI,EAAE,MAAM,EACZ,OAAO,GAAE,sBAA2B,GACnC,OAAO,CAAC,qBAAqB,CAAC;IAyC3B,mBAAmB,IAAI,OAAO,CAAC,WAAW,EAAE,CAAC;CAGpD"}
@@ -0,0 +1,75 @@
1
+ const OPENAI_VOICES = [
2
+ { id: 'alloy', name: 'Alloy', provider: 'openai-tts', lang: 'various', isDefault: false },
3
+ { id: 'echo', name: 'Echo', provider: 'openai-tts', lang: 'various', isDefault: false },
4
+ { id: 'fable', name: 'Fable', provider: 'openai-tts', lang: 'various', isDefault: false },
5
+ { id: 'onyx', name: 'Onyx', provider: 'openai-tts', lang: 'various', isDefault: false },
6
+ { id: 'nova', name: 'Nova', provider: 'openai-tts', lang: 'various', isDefault: true },
7
+ { id: 'shimmer', name: 'Shimmer', provider: 'openai-tts', lang: 'various', isDefault: false },
8
+ ];
9
+ function mimeTypeForOutput(format) {
10
+ switch (format) {
11
+ case 'opus':
12
+ return 'audio/opus';
13
+ case 'aac':
14
+ return 'audio/aac';
15
+ case 'flac':
16
+ return 'audio/flac';
17
+ case 'wav':
18
+ return 'audio/wav';
19
+ case 'pcm':
20
+ return 'audio/L16';
21
+ default:
22
+ return 'audio/mpeg';
23
+ }
24
+ }
25
+ export class OpenAITextToSpeechProvider {
26
+ constructor(config) {
27
+ this.config = config;
28
+ this.id = 'openai-tts';
29
+ this.displayName = 'OpenAI TTS';
30
+ this.supportsStreaming = true;
31
+ this.fetchImpl = config.fetchImpl ?? fetch;
32
+ }
33
+ getProviderName() {
34
+ return this.displayName;
35
+ }
36
+ async synthesize(text, options = {}) {
37
+ const model = options.model ?? this.config.model ?? 'tts-1';
38
+ const voice = options.voice ?? this.config.voice ?? 'nova';
39
+ const outputFormat = options.outputFormat ?? 'mp3';
40
+ const response = await this.fetchImpl(`${this.config.baseUrl ?? 'https://api.openai.com/v1'}/audio/speech`, {
41
+ method: 'POST',
42
+ headers: {
43
+ Authorization: `Bearer ${this.config.apiKey}`,
44
+ 'Content-Type': 'application/json',
45
+ },
46
+ body: JSON.stringify({
47
+ model,
48
+ voice,
49
+ input: text,
50
+ response_format: outputFormat,
51
+ speed: options.speed,
52
+ }),
53
+ });
54
+ if (!response.ok) {
55
+ const message = await response.text();
56
+ throw new Error(`OpenAI TTS synthesis failed (${response.status}): ${message}`);
57
+ }
58
+ const audioBuffer = Buffer.from(await response.arrayBuffer());
59
+ return {
60
+ audioBuffer,
61
+ mimeType: mimeTypeForOutput(outputFormat),
62
+ cost: 0,
63
+ voiceUsed: voice,
64
+ providerName: this.displayName,
65
+ usage: {
66
+ characters: text.length,
67
+ modelUsed: model,
68
+ },
69
+ };
70
+ }
71
+ async listAvailableVoices() {
72
+ return [...OPENAI_VOICES];
73
+ }
74
+ }
75
+ //# sourceMappingURL=OpenAITextToSpeechProvider.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"OpenAITextToSpeechProvider.js","sourceRoot":"","sources":["../../../src/speech/providers/OpenAITextToSpeechProvider.ts"],"names":[],"mappings":"AAeA,MAAM,aAAa,GAA2B;IAC5C,EAAE,EAAE,EAAE,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,QAAQ,EAAE,YAAY,EAAE,IAAI,EAAE,SAAS,EAAE,SAAS,EAAE,KAAK,EAAE;IACzF,EAAE,EAAE,EAAE,MAAM,EAAE,IAAI,EAAE,MAAM,EAAE,QAAQ,EAAE,YAAY,EAAE,IAAI,EAAE,SAAS,EAAE,SAAS,EAAE,KAAK,EAAE;IACvF,EAAE,EAAE,EAAE,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,QAAQ,EAAE,YAAY,EAAE,IAAI,EAAE,SAAS,EAAE,SAAS,EAAE,KAAK,EAAE;IACzF,EAAE,EAAE,EAAE,MAAM,EAAE,IAAI,EAAE,MAAM,EAAE,QAAQ,EAAE,YAAY,EAAE,IAAI,EAAE,SAAS,EAAE,SAAS,EAAE,KAAK,EAAE;IACvF,EAAE,EAAE,EAAE,MAAM,EAAE,IAAI,EAAE,MAAM,EAAE,QAAQ,EAAE,YAAY,EAAE,IAAI,EAAE,SAAS,EAAE,SAAS,EAAE,IAAI,EAAE;IACtF,EAAE,EAAE,EAAE,SAAS,EAAE,IAAI,EAAE,SAAS,EAAE,QAAQ,EAAE,YAAY,EAAE,IAAI,EAAE,SAAS,EAAE,SAAS,EAAE,KAAK,EAAE;CAC9F,CAAC;AAEF,SAAS,iBAAiB,CAAC,MAA0B;IACnD,QAAQ,MAAM,EAAE,CAAC;QACf,KAAK,MAAM;YACT,OAAO,YAAY,CAAC;QACtB,KAAK,KAAK;YACR,OAAO,WAAW,CAAC;QACrB,KAAK,MAAM;YACT,OAAO,YAAY,CAAC;QACtB,KAAK,KAAK;YACR,OAAO,WAAW,CAAC;QACrB,KAAK,KAAK;YACR,OAAO,WAAW,CAAC;QACrB;YACE,OAAO,YAAY,CAAC;IACxB,CAAC;AACH,CAAC;AAED,MAAM,OAAO,0BAA0B;IAMrC,YAA6B,MAAwC;QAAxC,WAAM,GAAN,MAAM,CAAkC;QALrD,OAAE,GAAG,YAAY,CAAC;QAClB,gBAAW,GAAG,YAAY,CAAC;QAC3B,sBAAiB,GAAG,IAAI,CAAC;QAIvC,IAAI,CAAC,SAAS,GAAG,MAAM,CAAC,SAAS,IAAI,KAAK,CAAC;IAC7C,CAAC;IAED,eAAe;QACb,OAAO,IAAI,CAAC,WAAW,CAAC;IAC1B,CAAC;IAED,KAAK,CAAC,UAAU,CACd,IAAY,EACZ,UAAkC,EAAE;QAEpC,MAAM,KAAK,GAAG,OAAO,CAAC,KAAK,IAAI,IAAI,CAAC,MAAM,CAAC,KAAK,IAAI,OAAO,CAAC;QAC5D,MAAM,KAAK,GAAG,OAAO,CAAC,KAAK,IAAI,IAAI,CAAC,MAAM,CAAC,KAAK,IAAI,MAAM,CAAC;QAC3D,MAAM,YAAY,GAAG,OAAO,CAAC,YAAY,IAAI,KAAK,CAAC;QACnD,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,SAAS,CACnC,GAAG,IAAI,CAAC,MAAM,CAAC,OAAO,IAAI,2BAA2B,eAAe,EACpE;YACE,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,aAAa,EAAE,UAAU,IAAI,CAAC,MAAM,CAAC,MAAM,EAAE;gBAC7C,cAAc,EAAE,kBAAkB;aACnC;YACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC;gBACnB,KAAK;gBACL,KAAK;gBACL,KAAK,EAAE,IAAI;gBACX,eAAe,EAAE,YAAY;gBAC7B,KAAK,EAAE,OAAO,CAAC,KAAK;aACrB,CAAC;SACH,CACF,CAAC;QAEF,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;YACjB,MAAM,OAAO,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;YACtC,MAAM,IAAI,KAAK,CAAC,gCAAgC,QAAQ,CAAC,MAAM,MAAM,OAAO,EAAE,CAAC,CAAC;QAClF,CAAC;QAED,MAAM,WAAW,GAAG,MAAM,CAAC,IAAI,CAAC,MAAM,QAAQ,CAAC,WAAW,EAAE,CAAC,CAAC;QAC9D,OAAO;YACL,WAAW;YACX,QAAQ,EAAE,iBAAiB,CAAC,YAAY,CAAC;YACzC,IAAI,EAAE,CAAC;YACP,SAAS,EAAE,KAAK;YAChB,YAAY,EAAE,IAAI,CAAC,WAAW;YAC9B,KAAK,EAAE;gBACL,UAAU,EAAE,IAAI,CAAC,MAAM;gBACvB,SAAS,EAAE,KAAK;aACjB;SACF,CAAC;IACJ,CAAC;IAED,KAAK,CAAC,mBAAmB;QACvB,OAAO,CAAC,GAAG,aAAa,CAAC,CAAC;IAC5B,CAAC;CACF"}
@@ -0,0 +1,18 @@
1
+ import type { SpeechAudioInput, SpeechToTextProvider, SpeechTranscriptionOptions, SpeechTranscriptionResult } from '../types.js';
2
+ export interface OpenAIWhisperSpeechToTextProviderConfig {
3
+ apiKey: string;
4
+ baseUrl?: string;
5
+ model?: string;
6
+ fetchImpl?: typeof fetch;
7
+ }
8
+ export declare class OpenAIWhisperSpeechToTextProvider implements SpeechToTextProvider {
9
+ private readonly config;
10
+ readonly id = "openai-whisper";
11
+ readonly displayName = "OpenAI Whisper";
12
+ readonly supportsStreaming = false;
13
+ private readonly fetchImpl;
14
+ constructor(config: OpenAIWhisperSpeechToTextProviderConfig);
15
+ getProviderName(): string;
16
+ transcribe(audio: SpeechAudioInput, options?: SpeechTranscriptionOptions): Promise<SpeechTranscriptionResult>;
17
+ }
18
+ //# sourceMappingURL=OpenAIWhisperSpeechToTextProvider.d.ts.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"OpenAIWhisperSpeechToTextProvider.d.ts","sourceRoot":"","sources":["../../../src/speech/providers/OpenAIWhisperSpeechToTextProvider.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EACV,gBAAgB,EAEhB,oBAAoB,EACpB,0BAA0B,EAC1B,yBAAyB,EAE1B,MAAM,aAAa,CAAC;AAErB,MAAM,WAAW,uCAAuC;IACtD,MAAM,EAAE,MAAM,CAAC;IACf,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,SAAS,CAAC,EAAE,OAAO,KAAK,CAAC;CAC1B;AA8CD,qBAAa,iCAAkC,YAAW,oBAAoB;IAMhE,OAAO,CAAC,QAAQ,CAAC,MAAM;IALnC,SAAgB,EAAE,oBAAoB;IACtC,SAAgB,WAAW,oBAAoB;IAC/C,SAAgB,iBAAiB,SAAS;IAC1C,OAAO,CAAC,QAAQ,CAAC,SAAS,CAAe;gBAEZ,MAAM,EAAE,uCAAuC;IAI5E,eAAe,IAAI,MAAM;IAInB,UAAU,CACd,KAAK,EAAE,gBAAgB,EACvB,OAAO,GAAE,0BAA+B,GACvC,OAAO,CAAC,yBAAyB,CAAC;CAoEtC"}
@@ -0,0 +1,109 @@
1
+ function normalizeSegments(input) {
2
+ if (!Array.isArray(input))
3
+ return undefined;
4
+ return input
5
+ .filter((segment) => typeof segment === 'object' && segment !== null)
6
+ .map((segment) => {
7
+ const item = segment;
8
+ return {
9
+ text: typeof item.text === 'string' ? item.text : '',
10
+ startTime: typeof item.start === 'number' ? item.start : 0,
11
+ endTime: typeof item.end === 'number' ? item.end : 0,
12
+ confidence: typeof item.confidence === 'number' ? item.confidence : undefined,
13
+ speaker: typeof item.speaker === 'string' || typeof item.speaker === 'number'
14
+ ? item.speaker
15
+ : undefined,
16
+ words: Array.isArray(item.words)
17
+ ? item.words
18
+ .filter((word) => typeof word === 'object' && word !== null)
19
+ .map((word) => {
20
+ const value = word;
21
+ return {
22
+ word: typeof value.word === 'string' ? value.word : '',
23
+ start: typeof value.start === 'number' ? value.start : 0,
24
+ end: typeof value.end === 'number' ? value.end : 0,
25
+ confidence: typeof value.confidence === 'number' ? value.confidence : undefined,
26
+ };
27
+ })
28
+ : undefined,
29
+ id: typeof item.id === 'number' ? item.id : undefined,
30
+ seek: typeof item.seek === 'number' ? item.seek : undefined,
31
+ tokens: Array.isArray(item.tokens)
32
+ ? item.tokens.filter((token) => typeof token === 'number')
33
+ : undefined,
34
+ temperature: typeof item.temperature === 'number' ? item.temperature : undefined,
35
+ avg_logprob: typeof item.avg_logprob === 'number' ? item.avg_logprob : undefined,
36
+ compression_ratio: typeof item.compression_ratio === 'number' ? item.compression_ratio : undefined,
37
+ no_speech_prob: typeof item.no_speech_prob === 'number' ? item.no_speech_prob : undefined,
38
+ };
39
+ });
40
+ }
41
+ export class OpenAIWhisperSpeechToTextProvider {
42
+ constructor(config) {
43
+ this.config = config;
44
+ this.id = 'openai-whisper';
45
+ this.displayName = 'OpenAI Whisper';
46
+ this.supportsStreaming = false;
47
+ this.fetchImpl = config.fetchImpl ?? fetch;
48
+ }
49
+ getProviderName() {
50
+ return this.displayName;
51
+ }
52
+ async transcribe(audio, options = {}) {
53
+ const form = new FormData();
54
+ const responseFormat = (options.responseFormat ?? 'verbose_json');
55
+ const model = options.model ?? this.config.model ?? 'whisper-1';
56
+ const fileName = audio.fileName ?? `speech.${audio.format ?? 'wav'}`;
57
+ form.append('file', new Blob([Uint8Array.from(audio.data)], { type: audio.mimeType ?? 'audio/wav' }), fileName);
58
+ form.append('model', model);
59
+ form.append('response_format', responseFormat);
60
+ if (options.language)
61
+ form.append('language', options.language);
62
+ if (options.prompt)
63
+ form.append('prompt', options.prompt);
64
+ if (typeof options.temperature === 'number') {
65
+ form.append('temperature', String(options.temperature));
66
+ }
67
+ const response = await this.fetchImpl(`${this.config.baseUrl ?? 'https://api.openai.com/v1'}/audio/transcriptions`, {
68
+ method: 'POST',
69
+ headers: {
70
+ Authorization: `Bearer ${this.config.apiKey}`,
71
+ },
72
+ body: form,
73
+ });
74
+ if (!response.ok) {
75
+ const message = await response.text();
76
+ throw new Error(`OpenAI Whisper transcription failed (${response.status}): ${message}`);
77
+ }
78
+ if (responseFormat === 'text' || response.headers.get('content-type')?.includes('text/plain')) {
79
+ const text = await response.text();
80
+ return {
81
+ text,
82
+ language: options.language,
83
+ durationSeconds: audio.durationSeconds,
84
+ cost: 0,
85
+ isFinal: true,
86
+ usage: {
87
+ durationMinutes: (audio.durationSeconds ?? 0) / 60,
88
+ modelUsed: model,
89
+ },
90
+ };
91
+ }
92
+ const payload = (await response.json());
93
+ const durationSeconds = typeof payload.duration === 'number' ? payload.duration : audio.durationSeconds;
94
+ return {
95
+ text: typeof payload.text === 'string' ? payload.text : '',
96
+ language: typeof payload.language === 'string' ? payload.language : options.language,
97
+ durationSeconds,
98
+ cost: 0,
99
+ segments: normalizeSegments(payload.segments),
100
+ providerResponse: payload,
101
+ isFinal: true,
102
+ usage: {
103
+ durationMinutes: (durationSeconds ?? 0) / 60,
104
+ modelUsed: model,
105
+ },
106
+ };
107
+ }
108
+ }
109
+ //# sourceMappingURL=OpenAIWhisperSpeechToTextProvider.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"OpenAIWhisperSpeechToTextProvider.js","sourceRoot":"","sources":["../../../src/speech/providers/OpenAIWhisperSpeechToTextProvider.ts"],"names":[],"mappings":"AAgBA,SAAS,iBAAiB,CAAC,KAAc;IACvC,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC;QAAE,OAAO,SAAS,CAAC;IAC5C,OAAO,KAAK;SACT,MAAM,CAAC,CAAC,OAAO,EAAE,EAAE,CAAC,OAAO,OAAO,KAAK,QAAQ,IAAI,OAAO,KAAK,IAAI,CAAC;SACpE,GAAG,CAAC,CAAC,OAAO,EAAE,EAAE;QACf,MAAM,IAAI,GAAG,OAAkC,CAAC;QAChD,OAAO;YACL,IAAI,EAAE,OAAO,IAAI,CAAC,IAAI,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE;YACpD,SAAS,EAAE,OAAO,IAAI,CAAC,KAAK,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;YAC1D,OAAO,EAAE,OAAO,IAAI,CAAC,GAAG,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;YACpD,UAAU,EAAE,OAAO,IAAI,CAAC,UAAU,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC,CAAC,SAAS;YAC7E,OAAO,EACL,OAAO,IAAI,CAAC,OAAO,KAAK,QAAQ,IAAI,OAAO,IAAI,CAAC,OAAO,KAAK,QAAQ;gBAClE,CAAC,CAAC,IAAI,CAAC,OAAO;gBACd,CAAC,CAAC,SAAS;YACf,KAAK,EAAE,KAAK,CAAC,OAAO,CAAC,IAAI,CAAC,KAAK,CAAC;gBAC9B,CAAC,CAAC,IAAI,CAAC,KAAK;qBACP,MAAM,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,OAAO,IAAI,KAAK,QAAQ,IAAI,IAAI,KAAK,IAAI,CAAC;qBAC3D,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE;oBACZ,MAAM,KAAK,GAAG,IAA+B,CAAC;oBAC9C,OAAO;wBACL,IAAI,EAAE,OAAO,KAAK,CAAC,IAAI,KAAK,QAAQ,CAAC,CAAC,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE;wBACtD,KAAK,EAAE,OAAO,KAAK,CAAC,KAAK,KAAK,QAAQ,CAAC,CAAC,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;wBACxD,GAAG,EAAE,OAAO,KAAK,CAAC,GAAG,KAAK,QAAQ,CAAC,CAAC,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;wBAClD,UAAU,EACR,OAAO,KAAK,CAAC,UAAU,KAAK,QAAQ,CAAC,CAAC,CAAC,KAAK,CAAC,UAAU,CAAC,CAAC,CAAC,SAAS;qBACtE,CAAC;gBACJ,CAAC,CAAC;gBACN,CAAC,CAAC,SAAS;YACb,EAAE,EAAE,OAAO,IAAI,CAAC,EAAE,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,CAAC,SAAS;YACrD,IAAI,EAAE,OAAO,IAAI,CAAC,IAAI,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC,SAAS;YAC3D,MAAM,EAAE,KAAK,CAAC,OAAO,CAAC,IAAI,CAAC,MAAM,CAAC;gBAChC,CAAC,CAAC,IAAI,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC,KAAK,EAAmB,EAAE,CAAC,OAAO,KAAK,KAAK,QAAQ,CAAC;gBAC3E,CAAC,CAAC,SAAS;YACb,WAAW,EAAE,OAAO,IAAI,CAAC,WAAW,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,WAAW,CAAC,CAAC,CAAC,SAAS;YAChF,WAAW,EAAE,OAAO,IAAI,CAAC,WAAW,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,WAAW,CAAC,CAAC,CAAC,SAAS;YAChF,iBAAiB,EACf,OAAO,IAAI,CAAC,iBAAiB,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,iBAAiB,CAAC,CAAC,CAAC,SAAS;YACjF,cAAc,EACZ,OAAO,IAAI,CAAC,cAAc,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,cAAc,CAAC,CAAC,CAAC,SAAS;SAC5E,CAAC;IACJ,CAAC,CAAC,CAAC;AACP,CAAC;AAED,MAAM,OAAO,iCAAiC;IAM5C,YAA6B,MAA+C;QAA/C,WAAM,GAAN,MAAM,CAAyC;QAL5D,OAAE,GAAG,gBAAgB,CAAC;QACtB,gBAAW,GAAG,gBAAgB,CAAC;QAC/B,sBAAiB,GAAG,KAAK,CAAC;QAIxC,IAAI,CAAC,SAAS,GAAG,MAAM,CAAC,SAAS,IAAI,KAAK,CAAC;IAC7C,CAAC;IAED,eAAe;QACb,OAAO,IAAI,CAAC,WAAW,CAAC;IAC1B,CAAC;IAED,KAAK,CAAC,UAAU,CACd,KAAuB,EACvB,UAAsC,EAAE;QAExC,MAAM,IAAI,GAAG,IAAI,QAAQ,EAAE,CAAC;QAC5B,MAAM,cAAc,GAAG,CAAC,OAAO,CAAC,cAAc,IAAI,cAAc,CAAyB,CAAC;QAC1F,MAAM,KAAK,GAAG,OAAO,CAAC,KAAK,IAAI,IAAI,CAAC,MAAM,CAAC,KAAK,IAAI,WAAW,CAAC;QAChE,MAAM,QAAQ,GAAG,KAAK,CAAC,QAAQ,IAAI,UAAU,KAAK,CAAC,MAAM,IAAI,KAAK,EAAE,CAAC;QAErE,IAAI,CAAC,MAAM,CACT,MAAM,EACN,IAAI,IAAI,CAAC,CAAC,UAAU,CAAC,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,EAAE,EAAE,IAAI,EAAE,KAAK,CAAC,QAAQ,IAAI,WAAW,EAAE,CAAC,EAChF,QAAQ,CACT,CAAC;QACF,IAAI,CAAC,MAAM,CAAC,OAAO,EAAE,KAAK,CAAC,CAAC;QAC5B,IAAI,CAAC,MAAM,CAAC,iBAAiB,EAAE,cAAc,CAAC,CAAC;QAC/C,IAAI,OAAO,CAAC,QAAQ;YAAE,IAAI,CAAC,MAAM,CAAC,UAAU,EAAE,OAAO,CAAC,QAAQ,CAAC,CAAC;QAChE,IAAI,OAAO,CAAC,MAAM;YAAE,IAAI,CAAC,MAAM,CAAC,QAAQ,EAAE,OAAO,CAAC,MAAM,CAAC,CAAC;QAC1D,IAAI,OAAO,OAAO,CAAC,WAAW,KAAK,QAAQ,EAAE,CAAC;YAC5C,IAAI,CAAC,MAAM,CAAC,aAAa,EAAE,MAAM,CAAC,OAAO,CAAC,WAAW,CAAC,CAAC,CAAC;QAC1D,CAAC;QAED,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,SAAS,CACnC,GAAG,IAAI,CAAC,MAAM,CAAC,OAAO,IAAI,2BAA2B,uBAAuB,EAC5E;YACE,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,aAAa,EAAE,UAAU,IAAI,CAAC,MAAM,CAAC,MAAM,EAAE;aAC9C;YACD,IAAI,EAAE,IAAI;SACX,CACF,CAAC;QAEF,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;YACjB,MAAM,OAAO,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;YACtC,MAAM,IAAI,KAAK,CAAC,wCAAwC,QAAQ,CAAC,MAAM,MAAM,OAAO,EAAE,CAAC,CAAC;QAC1F,CAAC;QAED,IAAI,cAAc,KAAK,MAAM,IAAI,QAAQ,CAAC,OAAO,CAAC,GAAG,CAAC,cAAc,CAAC,EAAE,QAAQ,CAAC,YAAY,CAAC,EAAE,CAAC;YAC9F,MAAM,IAAI,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;YACnC,OAAO;gBACL,IAAI;gBACJ,QAAQ,EAAE,OAAO,CAAC,QAAQ;gBAC1B,eAAe,EAAE,KAAK,CAAC,eAAe;gBACtC,IAAI,EAAE,CAAC;gBACP,OAAO,EAAE,IAAI;gBACb,KAAK,EAAE;oBACL,eAAe,EAAE,CAAC,KAAK,CAAC,eAAe,IAAI,CAAC,CAAC,GAAG,EAAE;oBAClD,SAAS,EAAE,KAAK;iBACjB;aACF,CAAC;QACJ,CAAC;QAED,MAAM,OAAO,GAAG,CAAC,MAAM,QAAQ,CAAC,IAAI,EAAE,CAA4B,CAAC;QACnE,MAAM,eAAe,GACnB,OAAO,OAAO,CAAC,QAAQ,KAAK,QAAQ,CAAC,CAAC,CAAC,OAAO,CAAC,QAAQ,CAAC,CAAC,CAAC,KAAK,CAAC,eAAe,CAAC;QAElF,OAAO;YACL,IAAI,EAAE,OAAO,OAAO,CAAC,IAAI,KAAK,QAAQ,CAAC,CAAC,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE;YAC1D,QAAQ,EAAE,OAAO,OAAO,CAAC,QAAQ,KAAK,QAAQ,CAAC,CAAC,CAAC,OAAO,CAAC,QAAQ,CAAC,CAAC,CAAC,OAAO,CAAC,QAAQ;YACpF,eAAe;YACf,IAAI,EAAE,CAAC;YACP,QAAQ,EAAE,iBAAiB,CAAC,OAAO,CAAC,QAAQ,CAAC;YAC7C,gBAAgB,EAAE,OAAO;YACzB,OAAO,EAAE,IAAI;YACb,KAAK,EAAE;gBACL,eAAe,EAAE,CAAC,eAAe,IAAI,CAAC,CAAC,GAAG,EAAE;gBAC5C,SAAS,EAAE,KAAK;aACjB;SACF,CAAC;IACJ,CAAC;CACF"}