@qvac/sdk 0.8.3 → 0.9.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (216) hide show
  1. package/NOTICE +267 -63
  2. package/dist/client/api/diffusion.d.ts +32 -0
  3. package/dist/client/api/diffusion.js +103 -0
  4. package/dist/client/api/embed.d.ts +9 -3
  5. package/dist/client/api/embed.js +4 -1
  6. package/dist/client/api/finetune.d.ts +69 -0
  7. package/dist/client/api/finetune.js +139 -0
  8. package/dist/client/api/get-model-info.d.ts +1 -1
  9. package/dist/client/api/index.d.ts +4 -0
  10. package/dist/client/api/index.js +4 -0
  11. package/dist/client/api/rag.d.ts +1 -1
  12. package/dist/client/api/rag.js +1 -1
  13. package/dist/client/api/resume.d.ts +14 -0
  14. package/dist/client/api/resume.js +20 -0
  15. package/dist/client/api/suspend.d.ts +13 -0
  16. package/dist/client/api/suspend.js +19 -0
  17. package/dist/client/api/transcribe.d.ts +25 -17
  18. package/dist/client/api/transcribe.js +115 -35
  19. package/dist/client/rpc/bare-client.d.ts +5 -1
  20. package/dist/client/rpc/bare-client.js +70 -57
  21. package/dist/client/rpc/expo-rpc-client.d.ts +4 -0
  22. package/dist/client/rpc/expo-rpc-client.js +8 -0
  23. package/dist/client/rpc/node-rpc-client.d.ts +4 -0
  24. package/dist/client/rpc/node-rpc-client.js +8 -0
  25. package/dist/client/rpc/rpc-client.d.ts +13 -0
  26. package/dist/client/rpc/rpc-client.js +131 -1
  27. package/dist/examples/delegated-inference/composite.d.ts +2 -0
  28. package/dist/examples/delegated-inference/composite.js +99 -0
  29. package/dist/examples/diffusion-flux2-klein.d.ts +2 -0
  30. package/dist/examples/diffusion-flux2-klein.js +46 -0
  31. package/dist/examples/diffusion-simple.d.ts +2 -0
  32. package/dist/examples/diffusion-simple.js +17 -0
  33. package/dist/examples/diffusion-txt2img.d.ts +2 -0
  34. package/dist/examples/diffusion-txt2img.js +40 -0
  35. package/dist/examples/embed-p2p.js +2 -2
  36. package/dist/examples/finetune/llamacpp-finetune.d.ts +2 -0
  37. package/dist/examples/finetune/llamacpp-finetune.js +78 -0
  38. package/dist/examples/logging-streaming.js +1 -1
  39. package/dist/examples/mcp-websearch.js +2 -0
  40. package/dist/examples/ocr-fasttext.js +11 -0
  41. package/dist/examples/profiling/per-call.js +3 -3
  42. package/dist/examples/rag/rag-chromadb.js +2 -2
  43. package/dist/examples/rag/rag-hyperdb/pipeline.js +1 -1
  44. package/dist/examples/rag/rag-lancedb.js +2 -2
  45. package/dist/examples/rag/rag-sqlite.js +2 -2
  46. package/dist/examples/suspend-resume.d.ts +2 -0
  47. package/dist/examples/suspend-resume.js +52 -0
  48. package/dist/examples/transcription/parakeet-ctc-filesystem.js +3 -5
  49. package/dist/examples/transcription/parakeet-microphone-record.js +42 -33
  50. package/dist/examples/transcription/parakeet-sortformer.js +13 -3
  51. package/dist/examples/transcription/parakeet-tdt-filesystem.js +5 -7
  52. package/dist/examples/transcription/whispercpp-filesystem-streaming.d.ts +2 -0
  53. package/dist/examples/transcription/whispercpp-filesystem-streaming.js +99 -0
  54. package/dist/examples/transcription/whispercpp-microphone-record.js +78 -299
  55. package/dist/examples/transcription/whispercpp-prompt.js +11 -0
  56. package/dist/examples/translation/translation-llm.js +2 -2
  57. package/dist/examples/tts/supertonic-multilingual.d.ts +2 -0
  58. package/dist/examples/tts/supertonic-multilingual.js +56 -0
  59. package/dist/examples/tts/supertonic.js +12 -7
  60. package/dist/index.d.ts +3 -2
  61. package/dist/index.js +3 -2
  62. package/dist/logging/utils.js +1 -1
  63. package/dist/models/registry/models.d.ts +15176 -4264
  64. package/dist/models/registry/models.js +18657 -7404
  65. package/dist/models/update-models/codegen.js +2 -2
  66. package/dist/models/update-models/naming.js +26 -0
  67. package/dist/pear/pre.js +2 -0
  68. package/dist/polyfills/bare-globals.d.ts +2 -0
  69. package/dist/polyfills/bare-globals.js +4 -0
  70. package/dist/polyfills/bare-globals.noop.d.ts +2 -0
  71. package/dist/polyfills/bare-globals.noop.js +1 -0
  72. package/dist/profiling/exporters.js +16 -11
  73. package/dist/profiling/model-execution.d.ts +16 -0
  74. package/dist/profiling/model-execution.js +33 -0
  75. package/dist/schemas/common.d.ts +967 -198
  76. package/dist/schemas/common.js +16 -1
  77. package/dist/schemas/completion-stream.d.ts +14 -6
  78. package/dist/schemas/completion-stream.js +4 -3
  79. package/dist/schemas/download-asset.d.ts +7 -3
  80. package/dist/schemas/embed.d.ts +19 -0
  81. package/dist/schemas/embed.js +7 -0
  82. package/dist/schemas/engine-addon-map.js +4 -1
  83. package/dist/schemas/finetune.d.ts +620 -0
  84. package/dist/schemas/finetune.js +161 -0
  85. package/dist/schemas/get-model-info.d.ts +2 -0
  86. package/dist/schemas/get-model-info.js +1 -0
  87. package/dist/schemas/index.d.ts +5 -1
  88. package/dist/schemas/index.js +5 -1
  89. package/dist/schemas/llamacpp-config.d.ts +26 -6
  90. package/dist/schemas/llamacpp-config.js +12 -0
  91. package/dist/schemas/load-model.d.ts +4005 -1400
  92. package/dist/schemas/load-model.js +34 -1
  93. package/dist/schemas/model-src-utils.d.ts +10 -2
  94. package/dist/schemas/model-types.d.ts +19 -1
  95. package/dist/schemas/model-types.js +11 -0
  96. package/dist/schemas/ocr.d.ts +2 -0
  97. package/dist/schemas/plugin.d.ts +24 -3
  98. package/dist/schemas/plugin.js +14 -0
  99. package/dist/schemas/profiling.d.ts +4 -0
  100. package/dist/schemas/profiling.js +4 -0
  101. package/dist/schemas/registry.d.ts +18 -0
  102. package/dist/schemas/registry.js +2 -0
  103. package/dist/schemas/reload-config.d.ts +9 -1
  104. package/dist/schemas/resume.d.ts +10 -0
  105. package/dist/schemas/resume.js +7 -0
  106. package/dist/schemas/sdcpp-config.d.ts +327 -0
  107. package/dist/schemas/sdcpp-config.js +102 -0
  108. package/dist/schemas/sdk-config.d.ts +1225 -235
  109. package/dist/schemas/sdk-config.js +4 -0
  110. package/dist/schemas/sdk-errors-server.d.ts +2 -0
  111. package/dist/schemas/sdk-errors-server.js +12 -0
  112. package/dist/schemas/suspend.d.ts +10 -0
  113. package/dist/schemas/suspend.js +7 -0
  114. package/dist/schemas/text-to-speech.d.ts +187 -10
  115. package/dist/schemas/text-to-speech.js +12 -4
  116. package/dist/schemas/transcription-config.d.ts +16 -56
  117. package/dist/schemas/transcription-config.js +0 -2
  118. package/dist/schemas/transcription.d.ts +66 -6
  119. package/dist/schemas/transcription.js +29 -7
  120. package/dist/schemas/translate.d.ts +14 -4
  121. package/dist/schemas/translate.js +10 -2
  122. package/dist/schemas/translation-config.d.ts +30 -10
  123. package/dist/server/bare/delegate-rpc-client.js +3 -3
  124. package/dist/server/bare/hyperswarm.js +15 -3
  125. package/dist/server/bare/ops/embed.d.ts +8 -4
  126. package/dist/server/bare/ops/embed.js +15 -4
  127. package/dist/server/bare/ops/transcribe.d.ts +6 -2
  128. package/dist/server/bare/ops/transcribe.js +53 -1
  129. package/dist/server/bare/ops/translate.d.ts +4 -1
  130. package/dist/server/bare/ops/translate.js +28 -65
  131. package/dist/server/bare/ops/unload-model.js +2 -0
  132. package/dist/server/bare/plugins/index.d.ts +1 -0
  133. package/dist/server/bare/plugins/index.js +1 -0
  134. package/dist/server/bare/plugins/llamacpp-completion/ops/completion-stream.d.ts +3 -1
  135. package/dist/server/bare/plugins/llamacpp-completion/ops/completion-stream.js +45 -9
  136. package/dist/server/bare/plugins/llamacpp-completion/ops/finetune.d.ts +11 -0
  137. package/dist/server/bare/plugins/llamacpp-completion/ops/finetune.js +112 -0
  138. package/dist/server/bare/plugins/llamacpp-completion/plugin.d.ts +175 -5
  139. package/dist/server/bare/plugins/llamacpp-completion/plugin.js +45 -35
  140. package/dist/server/bare/plugins/llamacpp-embedding/plugin.d.ts +11 -1
  141. package/dist/server/bare/plugins/llamacpp-embedding/plugin.js +9 -4
  142. package/dist/server/bare/plugins/nmtcpp-translation/plugin.d.ts +17 -2
  143. package/dist/server/bare/plugins/nmtcpp-translation/plugin.js +16 -15
  144. package/dist/server/bare/plugins/onnx-ocr/ops/ocr-stream.d.ts +5 -7
  145. package/dist/server/bare/plugins/onnx-ocr/ops/ocr-stream.js +11 -3
  146. package/dist/server/bare/plugins/onnx-ocr/plugin.d.ts +2 -0
  147. package/dist/server/bare/plugins/onnx-ocr/plugin.js +24 -19
  148. package/dist/server/bare/plugins/onnx-tts/ops/text-to-speech.d.ts +7 -2
  149. package/dist/server/bare/plugins/onnx-tts/ops/text-to-speech.js +22 -97
  150. package/dist/server/bare/plugins/onnx-tts/plugin.d.ts +98 -9
  151. package/dist/server/bare/plugins/onnx-tts/plugin.js +85 -46
  152. package/dist/server/bare/plugins/parakeet-transcription/plugin.d.ts +30 -60
  153. package/dist/server/bare/plugins/parakeet-transcription/plugin.js +49 -37
  154. package/dist/server/bare/plugins/sdcpp-generation/ops/diffusion.d.ts +3 -0
  155. package/dist/server/bare/plugins/sdcpp-generation/ops/diffusion.js +51 -0
  156. package/dist/server/bare/plugins/sdcpp-generation/plugin.d.ts +278 -0
  157. package/dist/server/bare/plugins/sdcpp-generation/plugin.js +70 -0
  158. package/dist/server/bare/plugins/whispercpp-transcription/plugin.d.ts +38 -2
  159. package/dist/server/bare/plugins/whispercpp-transcription/plugin.js +43 -22
  160. package/dist/server/bare/rag-hyperdb/handlers/ingest.js +2 -1
  161. package/dist/server/bare/rag-hyperdb/handlers/search.js +2 -1
  162. package/dist/server/bare/rag-hyperdb/rag-workspace-manager.js +7 -0
  163. package/dist/server/bare/registry/model-config-utils.js +3 -0
  164. package/dist/server/bare/registry/model-registry.d.ts +2 -2
  165. package/dist/server/bare/registry/model-registry.js +1 -1
  166. package/dist/server/bare/registry/registry-client.js +21 -0
  167. package/dist/server/bare/runtime-lifecycle.d.ts +28 -0
  168. package/dist/server/bare/runtime-lifecycle.js +132 -0
  169. package/dist/server/bare/types/addon-responses.d.ts +62 -0
  170. package/dist/server/bare/types/addon-responses.js +1 -0
  171. package/dist/server/env.js +3 -1
  172. package/dist/server/rpc/delegate-transport.js +1 -1
  173. package/dist/server/rpc/handle-request.js +50 -4
  174. package/dist/server/rpc/handler-registry.js +20 -1
  175. package/dist/server/rpc/handler-utils.d.ts +5 -3
  176. package/dist/server/rpc/handler-utils.js +52 -50
  177. package/dist/server/rpc/handlers/delete-cache.js +3 -0
  178. package/dist/server/rpc/handlers/diffusion-stream.d.ts +3 -0
  179. package/dist/server/rpc/handlers/diffusion-stream.js +4 -0
  180. package/dist/server/rpc/handlers/finetune.d.ts +3 -0
  181. package/dist/server/rpc/handlers/finetune.js +16 -0
  182. package/dist/server/rpc/handlers/index.d.ts +10 -0
  183. package/dist/server/rpc/handlers/index.js +10 -0
  184. package/dist/server/rpc/handlers/load-model/hyperdrive.js +24 -9
  185. package/dist/server/rpc/handlers/load-model/registry.d.ts +11 -0
  186. package/dist/server/rpc/handlers/load-model/registry.js +31 -4
  187. package/dist/server/rpc/handlers/plugin-dispatch.d.ts +1 -13
  188. package/dist/server/rpc/handlers/plugin-dispatch.js +21 -21
  189. package/dist/server/rpc/handlers/resume.d.ts +3 -0
  190. package/dist/server/rpc/handlers/resume.js +11 -0
  191. package/dist/server/rpc/handlers/suspend.d.ts +3 -0
  192. package/dist/server/rpc/handlers/suspend.js +11 -0
  193. package/dist/server/rpc/handlers/transcribe-stream.d.ts +1 -1
  194. package/dist/server/rpc/handlers/transcribe-stream.js +2 -2
  195. package/dist/server/rpc/handlers/transcribe.d.ts +3 -0
  196. package/dist/server/rpc/handlers/transcribe.js +4 -0
  197. package/dist/server/rpc/profiling/operation-metrics.js +107 -16
  198. package/dist/server/rpc/profiling/operation-wrappers.d.ts +1 -1
  199. package/dist/server/rpc/profiling/operation-wrappers.js +51 -19
  200. package/dist/server/rpc/progress-throttle.d.ts +8 -0
  201. package/dist/server/rpc/progress-throttle.js +47 -0
  202. package/dist/server/rpc/rpc-utils.d.ts +5 -0
  203. package/dist/server/rpc/rpc-utils.js +11 -0
  204. package/dist/server/utils/path-security.js +1 -1
  205. package/dist/server/utils/tool-integration.js +10 -0
  206. package/dist/server/utils/tool-parser.js +9 -5
  207. package/dist/server/worker-core.d.ts +2 -0
  208. package/dist/server/worker-core.js +26 -23
  209. package/dist/server/worker.js +2 -1
  210. package/dist/utils/encoding.d.ts +9 -0
  211. package/dist/utils/encoding.js +14 -0
  212. package/dist/utils/errors-server.d.ts +6 -0
  213. package/dist/utils/errors-server.js +11 -0
  214. package/package.json +36 -25
  215. package/dist/server/rpc/delegate-utils.d.ts +0 -2
  216. package/dist/server/rpc/delegate-utils.js +0 -5
@@ -17,6 +17,7 @@ export const ttsSupertonicRuntimeConfigSchema = z.object({
17
17
  language: ttsLanguageSchema,
18
18
  ttsSpeed: z.number().optional(),
19
19
  ttsNumInferenceSteps: z.number().optional(),
20
+ ttsSupertonicMultilingual: z.boolean().optional(),
20
21
  });
21
22
  export const ttsRuntimeConfigSchema = z.union([
22
23
  ttsChatterboxRuntimeConfigSchema,
@@ -31,11 +32,13 @@ export const ttsChatterboxConfigSchema = ttsChatterboxRuntimeConfigSchema.extend
31
32
  referenceAudioSrc: modelSrcInputSchema,
32
33
  });
33
34
  export const ttsSupertonicConfigSchema = ttsSupertonicRuntimeConfigSchema.extend({
34
- ttsTokenizerSrc: modelSrcInputSchema,
35
35
  ttsTextEncoderSrc: modelSrcInputSchema,
36
- ttsLatentDenoiserSrc: modelSrcInputSchema,
37
- ttsVoiceDecoderSrc: modelSrcInputSchema,
38
- ttsVoiceSrc: modelSrcInputSchema,
36
+ ttsDurationPredictorSrc: modelSrcInputSchema,
37
+ ttsVectorEstimatorSrc: modelSrcInputSchema,
38
+ ttsVocoderSrc: modelSrcInputSchema,
39
+ ttsUnicodeIndexerSrc: modelSrcInputSchema,
40
+ ttsTtsConfigSrc: modelSrcInputSchema,
41
+ ttsVoiceStyleSrc: modelSrcInputSchema,
39
42
  });
40
43
  export const ttsConfigSchema = z.union([
41
44
  ttsChatterboxConfigSchema,
@@ -50,8 +53,13 @@ export const ttsClientParamsSchema = z.object({
50
53
  export const ttsRequestSchema = ttsClientParamsSchema.extend({
51
54
  type: z.literal("textToSpeech"),
52
55
  });
56
+ export const ttsStatsSchema = z.object({
57
+ audioDuration: z.number().optional(),
58
+ totalSamples: z.number().optional(),
59
+ });
53
60
  export const ttsResponseSchema = z.object({
54
61
  type: z.literal("textToSpeech"),
55
62
  buffer: z.array(z.number()),
56
63
  done: z.boolean().default(false),
64
+ stats: ttsStatsSchema.optional(),
57
65
  });
@@ -83,6 +83,7 @@ export declare const whisperConfigSchema: z.ZodObject<{
83
83
  "onnx-tts": "onnx-tts";
84
84
  "parakeet-transcription": "parakeet-transcription";
85
85
  "onnx-ocr": "onnx-ocr";
86
+ "sdcpp-generation": "sdcpp-generation";
86
87
  llm: "llm";
87
88
  whisper: "whisper";
88
89
  embeddings: "embeddings";
@@ -90,6 +91,7 @@ export declare const whisperConfigSchema: z.ZodObject<{
90
91
  parakeet: "parakeet";
91
92
  tts: "tts";
92
93
  ocr: "ocr";
94
+ diffusion: "diffusion";
93
95
  }>, z.ZodLiteral<"vad">]>>;
94
96
  }, z.core.$strip>]>>;
95
97
  }, z.core.$strip>;
@@ -144,6 +146,7 @@ export declare const parakeetConfigSchema: z.ZodObject<{
144
146
  "onnx-tts": "onnx-tts";
145
147
  "parakeet-transcription": "parakeet-transcription";
146
148
  "onnx-ocr": "onnx-ocr";
149
+ "sdcpp-generation": "sdcpp-generation";
147
150
  llm: "llm";
148
151
  whisper: "whisper";
149
152
  embeddings: "embeddings";
@@ -151,34 +154,7 @@ export declare const parakeetConfigSchema: z.ZodObject<{
151
154
  parakeet: "parakeet";
152
155
  tts: "tts";
153
156
  ocr: "ocr";
154
- }>, z.ZodLiteral<"vad">]>>;
155
- }, z.core.$strip>]>>;
156
- parakeetEncoderDataSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
157
- src: z.ZodString;
158
- name: z.ZodOptional<z.ZodString>;
159
- modelId: z.ZodOptional<z.ZodString>;
160
- registryPath: z.ZodOptional<z.ZodString>;
161
- registrySource: z.ZodOptional<z.ZodString>;
162
- blobCoreKey: z.ZodOptional<z.ZodString>;
163
- blobIndex: z.ZodOptional<z.ZodNumber>;
164
- engine: z.ZodOptional<z.ZodString>;
165
- expectedSize: z.ZodOptional<z.ZodNumber>;
166
- sha256Checksum: z.ZodOptional<z.ZodString>;
167
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
168
- "llamacpp-completion": "llamacpp-completion";
169
- "whispercpp-transcription": "whispercpp-transcription";
170
- "llamacpp-embedding": "llamacpp-embedding";
171
- "nmtcpp-translation": "nmtcpp-translation";
172
- "onnx-tts": "onnx-tts";
173
- "parakeet-transcription": "parakeet-transcription";
174
- "onnx-ocr": "onnx-ocr";
175
- llm: "llm";
176
- whisper: "whisper";
177
- embeddings: "embeddings";
178
- nmt: "nmt";
179
- parakeet: "parakeet";
180
- tts: "tts";
181
- ocr: "ocr";
157
+ diffusion: "diffusion";
182
158
  }>, z.ZodLiteral<"vad">]>>;
183
159
  }, z.core.$strip>]>>;
184
160
  parakeetDecoderSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -200,6 +176,7 @@ export declare const parakeetConfigSchema: z.ZodObject<{
200
176
  "onnx-tts": "onnx-tts";
201
177
  "parakeet-transcription": "parakeet-transcription";
202
178
  "onnx-ocr": "onnx-ocr";
179
+ "sdcpp-generation": "sdcpp-generation";
203
180
  llm: "llm";
204
181
  whisper: "whisper";
205
182
  embeddings: "embeddings";
@@ -207,6 +184,7 @@ export declare const parakeetConfigSchema: z.ZodObject<{
207
184
  parakeet: "parakeet";
208
185
  tts: "tts";
209
186
  ocr: "ocr";
187
+ diffusion: "diffusion";
210
188
  }>, z.ZodLiteral<"vad">]>>;
211
189
  }, z.core.$strip>]>>;
212
190
  parakeetVocabSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -228,6 +206,7 @@ export declare const parakeetConfigSchema: z.ZodObject<{
228
206
  "onnx-tts": "onnx-tts";
229
207
  "parakeet-transcription": "parakeet-transcription";
230
208
  "onnx-ocr": "onnx-ocr";
209
+ "sdcpp-generation": "sdcpp-generation";
231
210
  llm: "llm";
232
211
  whisper: "whisper";
233
212
  embeddings: "embeddings";
@@ -235,6 +214,7 @@ export declare const parakeetConfigSchema: z.ZodObject<{
235
214
  parakeet: "parakeet";
236
215
  tts: "tts";
237
216
  ocr: "ocr";
217
+ diffusion: "diffusion";
238
218
  }>, z.ZodLiteral<"vad">]>>;
239
219
  }, z.core.$strip>]>>;
240
220
  parakeetPreprocessorSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -256,6 +236,7 @@ export declare const parakeetConfigSchema: z.ZodObject<{
256
236
  "onnx-tts": "onnx-tts";
257
237
  "parakeet-transcription": "parakeet-transcription";
258
238
  "onnx-ocr": "onnx-ocr";
239
+ "sdcpp-generation": "sdcpp-generation";
259
240
  llm: "llm";
260
241
  whisper: "whisper";
261
242
  embeddings: "embeddings";
@@ -263,6 +244,7 @@ export declare const parakeetConfigSchema: z.ZodObject<{
263
244
  parakeet: "parakeet";
264
245
  tts: "tts";
265
246
  ocr: "ocr";
247
+ diffusion: "diffusion";
266
248
  }>, z.ZodLiteral<"vad">]>>;
267
249
  }, z.core.$strip>]>>;
268
250
  parakeetCtcModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -284,6 +266,7 @@ export declare const parakeetConfigSchema: z.ZodObject<{
284
266
  "onnx-tts": "onnx-tts";
285
267
  "parakeet-transcription": "parakeet-transcription";
286
268
  "onnx-ocr": "onnx-ocr";
269
+ "sdcpp-generation": "sdcpp-generation";
287
270
  llm: "llm";
288
271
  whisper: "whisper";
289
272
  embeddings: "embeddings";
@@ -291,34 +274,7 @@ export declare const parakeetConfigSchema: z.ZodObject<{
291
274
  parakeet: "parakeet";
292
275
  tts: "tts";
293
276
  ocr: "ocr";
294
- }>, z.ZodLiteral<"vad">]>>;
295
- }, z.core.$strip>]>>;
296
- parakeetCtcModelDataSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
297
- src: z.ZodString;
298
- name: z.ZodOptional<z.ZodString>;
299
- modelId: z.ZodOptional<z.ZodString>;
300
- registryPath: z.ZodOptional<z.ZodString>;
301
- registrySource: z.ZodOptional<z.ZodString>;
302
- blobCoreKey: z.ZodOptional<z.ZodString>;
303
- blobIndex: z.ZodOptional<z.ZodNumber>;
304
- engine: z.ZodOptional<z.ZodString>;
305
- expectedSize: z.ZodOptional<z.ZodNumber>;
306
- sha256Checksum: z.ZodOptional<z.ZodString>;
307
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
308
- "llamacpp-completion": "llamacpp-completion";
309
- "whispercpp-transcription": "whispercpp-transcription";
310
- "llamacpp-embedding": "llamacpp-embedding";
311
- "nmtcpp-translation": "nmtcpp-translation";
312
- "onnx-tts": "onnx-tts";
313
- "parakeet-transcription": "parakeet-transcription";
314
- "onnx-ocr": "onnx-ocr";
315
- llm: "llm";
316
- whisper: "whisper";
317
- embeddings: "embeddings";
318
- nmt: "nmt";
319
- parakeet: "parakeet";
320
- tts: "tts";
321
- ocr: "ocr";
277
+ diffusion: "diffusion";
322
278
  }>, z.ZodLiteral<"vad">]>>;
323
279
  }, z.core.$strip>]>>;
324
280
  parakeetTokenizerSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -340,6 +296,7 @@ export declare const parakeetConfigSchema: z.ZodObject<{
340
296
  "onnx-tts": "onnx-tts";
341
297
  "parakeet-transcription": "parakeet-transcription";
342
298
  "onnx-ocr": "onnx-ocr";
299
+ "sdcpp-generation": "sdcpp-generation";
343
300
  llm: "llm";
344
301
  whisper: "whisper";
345
302
  embeddings: "embeddings";
@@ -347,6 +304,7 @@ export declare const parakeetConfigSchema: z.ZodObject<{
347
304
  parakeet: "parakeet";
348
305
  tts: "tts";
349
306
  ocr: "ocr";
307
+ diffusion: "diffusion";
350
308
  }>, z.ZodLiteral<"vad">]>>;
351
309
  }, z.core.$strip>]>>;
352
310
  parakeetSortformerSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -368,6 +326,7 @@ export declare const parakeetConfigSchema: z.ZodObject<{
368
326
  "onnx-tts": "onnx-tts";
369
327
  "parakeet-transcription": "parakeet-transcription";
370
328
  "onnx-ocr": "onnx-ocr";
329
+ "sdcpp-generation": "sdcpp-generation";
371
330
  llm: "llm";
372
331
  whisper: "whisper";
373
332
  embeddings: "embeddings";
@@ -375,6 +334,7 @@ export declare const parakeetConfigSchema: z.ZodObject<{
375
334
  parakeet: "parakeet";
376
335
  tts: "tts";
377
336
  ocr: "ocr";
337
+ diffusion: "diffusion";
378
338
  }>, z.ZodLiteral<"vad">]>>;
379
339
  }, z.core.$strip>]>>;
380
340
  }, z.core.$strip>;
@@ -82,13 +82,11 @@ export const parakeetRuntimeConfigSchema = z.object({
82
82
  export const parakeetConfigSchema = parakeetRuntimeConfigSchema.extend({
83
83
  // TDT sources
84
84
  parakeetEncoderSrc: modelSrcInputSchema.optional(),
85
- parakeetEncoderDataSrc: modelSrcInputSchema.optional(),
86
85
  parakeetDecoderSrc: modelSrcInputSchema.optional(),
87
86
  parakeetVocabSrc: modelSrcInputSchema.optional(),
88
87
  parakeetPreprocessorSrc: modelSrcInputSchema.optional(),
89
88
  // CTC sources
90
89
  parakeetCtcModelSrc: modelSrcInputSchema.optional(),
91
- parakeetCtcModelDataSrc: modelSrcInputSchema.optional(),
92
90
  parakeetTokenizerSrc: modelSrcInputSchema.optional(),
93
91
  // Sortformer source
94
92
  parakeetSortformerSrc: modelSrcInputSchema.optional(),
@@ -8,6 +8,7 @@ export declare const audioInputSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
8
8
  }, z.core.$strip>], "type">;
9
9
  export declare const transcribeParamsSchema: z.ZodObject<{
10
10
  modelId: z.ZodString;
11
+ prompt: z.ZodOptional<z.ZodString>;
11
12
  audioChunk: z.ZodDiscriminatedUnion<[z.ZodObject<{
12
13
  type: z.ZodLiteral<"base64">;
13
14
  value: z.ZodString;
@@ -15,10 +16,22 @@ export declare const transcribeParamsSchema: z.ZodObject<{
15
16
  type: z.ZodLiteral<"filePath">;
16
17
  value: z.ZodString;
17
18
  }, z.core.$strip>], "type">;
18
- prompt: z.ZodOptional<z.ZodString>;
19
19
  }, z.core.$strip>;
20
- export declare const transcribeStreamRequestSchema: z.ZodObject<{
20
+ export declare const transcribeStatsSchema: z.ZodObject<{
21
+ audioDuration: z.ZodOptional<z.ZodNumber>;
22
+ realTimeFactor: z.ZodOptional<z.ZodNumber>;
23
+ tokensPerSecond: z.ZodOptional<z.ZodNumber>;
24
+ totalTokens: z.ZodOptional<z.ZodNumber>;
25
+ totalSegments: z.ZodOptional<z.ZodNumber>;
26
+ whisperEncodeTime: z.ZodOptional<z.ZodNumber>;
27
+ whisperDecodeTime: z.ZodOptional<z.ZodNumber>;
28
+ encoderTime: z.ZodOptional<z.ZodNumber>;
29
+ decoderTime: z.ZodOptional<z.ZodNumber>;
30
+ melSpecTime: z.ZodOptional<z.ZodNumber>;
31
+ }, z.core.$strip>;
32
+ export declare const transcribeRequestSchema: z.ZodObject<{
21
33
  modelId: z.ZodString;
34
+ prompt: z.ZodOptional<z.ZodString>;
22
35
  audioChunk: z.ZodDiscriminatedUnion<[z.ZodObject<{
23
36
  type: z.ZodLiteral<"base64">;
24
37
  value: z.ZodString;
@@ -26,14 +39,25 @@ export declare const transcribeStreamRequestSchema: z.ZodObject<{
26
39
  type: z.ZodLiteral<"filePath">;
27
40
  value: z.ZodString;
28
41
  }, z.core.$strip>], "type">;
29
- prompt: z.ZodOptional<z.ZodString>;
30
- type: z.ZodLiteral<"transcribeStream">;
42
+ type: z.ZodLiteral<"transcribe">;
31
43
  }, z.core.$strip>;
32
- export declare const transcribeStreamResponseSchema: z.ZodObject<{
33
- type: z.ZodLiteral<"transcribeStream">;
44
+ export declare const transcribeResponseSchema: z.ZodObject<{
34
45
  text: z.ZodOptional<z.ZodString>;
35
46
  done: z.ZodOptional<z.ZodBoolean>;
47
+ stats: z.ZodOptional<z.ZodObject<{
48
+ audioDuration: z.ZodOptional<z.ZodNumber>;
49
+ realTimeFactor: z.ZodOptional<z.ZodNumber>;
50
+ tokensPerSecond: z.ZodOptional<z.ZodNumber>;
51
+ totalTokens: z.ZodOptional<z.ZodNumber>;
52
+ totalSegments: z.ZodOptional<z.ZodNumber>;
53
+ whisperEncodeTime: z.ZodOptional<z.ZodNumber>;
54
+ whisperDecodeTime: z.ZodOptional<z.ZodNumber>;
55
+ encoderTime: z.ZodOptional<z.ZodNumber>;
56
+ decoderTime: z.ZodOptional<z.ZodNumber>;
57
+ melSpecTime: z.ZodOptional<z.ZodNumber>;
58
+ }, z.core.$strip>>;
36
59
  error: z.ZodOptional<z.ZodString>;
60
+ type: z.ZodLiteral<"transcribe">;
37
61
  }, z.core.$strip>;
38
62
  export type AudioInput = z.infer<typeof audioInputSchema>;
39
63
  export type TranscribeParams = z.infer<typeof transcribeParamsSchema>;
@@ -42,6 +66,42 @@ export type TranscribeClientParams = {
42
66
  audioChunk: string | Buffer;
43
67
  prompt?: string;
44
68
  };
69
+ export type TranscribeRequest = z.infer<typeof transcribeRequestSchema>;
70
+ export type TranscribeResponse = z.infer<typeof transcribeResponseSchema>;
71
+ export declare const transcribeStreamRequestSchema: z.ZodObject<{
72
+ modelId: z.ZodString;
73
+ prompt: z.ZodOptional<z.ZodString>;
74
+ type: z.ZodLiteral<"transcribeStream">;
75
+ }, z.core.$strip>;
76
+ export declare const transcribeStreamResponseSchema: z.ZodObject<{
77
+ text: z.ZodOptional<z.ZodString>;
78
+ done: z.ZodOptional<z.ZodBoolean>;
79
+ stats: z.ZodOptional<z.ZodObject<{
80
+ audioDuration: z.ZodOptional<z.ZodNumber>;
81
+ realTimeFactor: z.ZodOptional<z.ZodNumber>;
82
+ tokensPerSecond: z.ZodOptional<z.ZodNumber>;
83
+ totalTokens: z.ZodOptional<z.ZodNumber>;
84
+ totalSegments: z.ZodOptional<z.ZodNumber>;
85
+ whisperEncodeTime: z.ZodOptional<z.ZodNumber>;
86
+ whisperDecodeTime: z.ZodOptional<z.ZodNumber>;
87
+ encoderTime: z.ZodOptional<z.ZodNumber>;
88
+ decoderTime: z.ZodOptional<z.ZodNumber>;
89
+ melSpecTime: z.ZodOptional<z.ZodNumber>;
90
+ }, z.core.$strip>>;
91
+ error: z.ZodOptional<z.ZodString>;
92
+ type: z.ZodLiteral<"transcribeStream">;
93
+ }, z.core.$strip>;
45
94
  export type TranscribeStreamRequest = z.infer<typeof transcribeStreamRequestSchema>;
46
95
  export type TranscribeStreamResponse = z.infer<typeof transcribeStreamResponseSchema>;
96
+ export type TranscribeStreamClientParams = {
97
+ modelId: string;
98
+ prompt?: string;
99
+ };
100
+ export interface TranscribeStreamSession {
101
+ write(audioChunk: Buffer): void;
102
+ end(): void;
103
+ destroy(): void;
104
+ [Symbol.asyncIterator](): AsyncIterator<string>;
105
+ }
106
+ export type TranscribeStats = z.infer<typeof transcribeStatsSchema>;
47
107
  //# sourceMappingURL=transcription.d.ts.map
@@ -9,18 +9,40 @@ export const audioInputSchema = z.discriminatedUnion("type", [
9
9
  value: z.string(),
10
10
  }),
11
11
  ]);
12
- export const transcribeParamsSchema = z.object({
12
+ const transcribeBaseSchema = z.object({
13
13
  modelId: z.string(),
14
- audioChunk: audioInputSchema,
15
14
  prompt: z.string().optional(),
16
15
  });
17
- // Streaming Transcribe Schema (for real-time audio streaming)
18
- export const transcribeStreamRequestSchema = transcribeParamsSchema.extend({
19
- type: z.literal("transcribeStream"),
16
+ export const transcribeParamsSchema = transcribeBaseSchema.extend({
17
+ audioChunk: audioInputSchema,
20
18
  });
21
- export const transcribeStreamResponseSchema = z.object({
22
- type: z.literal("transcribeStream"),
19
+ export const transcribeStatsSchema = z.object({
20
+ audioDuration: z.number().optional(),
21
+ realTimeFactor: z.number().optional(),
22
+ tokensPerSecond: z.number().optional(),
23
+ totalTokens: z.number().optional(),
24
+ totalSegments: z.number().optional(),
25
+ whisperEncodeTime: z.number().optional(),
26
+ whisperDecodeTime: z.number().optional(),
27
+ encoderTime: z.number().optional(),
28
+ decoderTime: z.number().optional(),
29
+ melSpecTime: z.number().optional(),
30
+ });
31
+ export const transcribeRequestSchema = transcribeParamsSchema.extend({
32
+ type: z.literal("transcribe"),
33
+ });
34
+ const transcriptionResultBase = z.object({
23
35
  text: z.string().optional(),
24
36
  done: z.boolean().optional(),
37
+ stats: transcribeStatsSchema.optional(),
25
38
  error: z.string().optional(),
26
39
  });
40
+ export const transcribeResponseSchema = transcriptionResultBase.extend({
41
+ type: z.literal("transcribe"),
42
+ });
43
+ export const transcribeStreamRequestSchema = transcribeBaseSchema.extend({
44
+ type: z.literal("transcribeStream"),
45
+ });
46
+ export const transcribeStreamResponseSchema = transcriptionResultBase.extend({
47
+ type: z.literal("transcribeStream"),
48
+ });
@@ -22,8 +22,13 @@ declare const translateParamsSchema: z.ZodUnion<readonly [z.ZodObject<{
22
22
  context: z.ZodOptional<z.ZodString>;
23
23
  }, z.core.$strip>]>;
24
24
  export declare const translationStatsSchema: z.ZodObject<{
25
- processedTokens: z.ZodNumber;
26
- processingTime: z.ZodNumber;
25
+ totalTime: z.ZodOptional<z.ZodNumber>;
26
+ totalTokens: z.ZodOptional<z.ZodNumber>;
27
+ tokensPerSecond: z.ZodOptional<z.ZodNumber>;
28
+ timeToFirstToken: z.ZodOptional<z.ZodNumber>;
29
+ decodeTime: z.ZodOptional<z.ZodNumber>;
30
+ encodeTime: z.ZodOptional<z.ZodNumber>;
31
+ cacheTokens: z.ZodOptional<z.ZodNumber>;
27
32
  }, z.core.$strip>;
28
33
  export declare const translateRequestSchema: z.ZodUnion<readonly [z.ZodObject<{
29
34
  modelId: z.ZodString;
@@ -98,8 +103,13 @@ export declare const translateResponseSchema: z.ZodObject<{
98
103
  token: z.ZodString;
99
104
  done: z.ZodOptional<z.ZodBoolean>;
100
105
  stats: z.ZodOptional<z.ZodObject<{
101
- processedTokens: z.ZodNumber;
102
- processingTime: z.ZodNumber;
106
+ totalTime: z.ZodOptional<z.ZodNumber>;
107
+ totalTokens: z.ZodOptional<z.ZodNumber>;
108
+ tokensPerSecond: z.ZodOptional<z.ZodNumber>;
109
+ timeToFirstToken: z.ZodOptional<z.ZodNumber>;
110
+ decodeTime: z.ZodOptional<z.ZodNumber>;
111
+ encodeTime: z.ZodOptional<z.ZodNumber>;
112
+ cacheTokens: z.ZodOptional<z.ZodNumber>;
103
113
  }, z.core.$strip>>;
104
114
  error: z.ZodOptional<z.ZodString>;
105
115
  }, z.core.$strip>;
@@ -26,8 +26,16 @@ const translateParamsSchema = z.union([
26
26
  translateParamsLlmSchema,
27
27
  ]);
28
28
  export const translationStatsSchema = z.object({
29
- processedTokens: z.number(),
30
- processingTime: z.number(),
29
+ // Common stats
30
+ totalTime: z.number().optional(),
31
+ totalTokens: z.number().optional(),
32
+ tokensPerSecond: z.number().optional(),
33
+ timeToFirstToken: z.number().optional(),
34
+ // NMT-specific
35
+ decodeTime: z.number().optional(),
36
+ encodeTime: z.number().optional(),
37
+ // LLM-specific
38
+ cacheTokens: z.number().optional(),
31
39
  });
32
40
  export const translateRequestSchema = z.union([
33
41
  translateParamsNmtSchema.extend({ type: z.literal("translate") }),