@qvac/sdk 0.7.1 → 0.8.0

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 (218) hide show
  1. package/NOTICE +57 -331
  2. package/README.md +2 -2
  3. package/bare-imports.json +9 -0
  4. package/dist/client/api/completion-stream.d.ts +2 -1
  5. package/dist/client/api/completion-stream.js +2 -1
  6. package/dist/client/api/download-asset.d.ts +3 -2
  7. package/dist/client/api/download-asset.js +4 -3
  8. package/dist/client/api/embed.d.ts +5 -2
  9. package/dist/client/api/embed.js +2 -2
  10. package/dist/client/api/get-model-info.d.ts +1 -1
  11. package/dist/client/api/invoke-plugin.d.ts +3 -2
  12. package/dist/client/api/invoke-plugin.js +4 -4
  13. package/dist/client/api/load-model.d.ts +12 -10
  14. package/dist/client/api/load-model.js +3 -3
  15. package/dist/client/api/rag.d.ts +10 -10
  16. package/dist/client/api/rag.js +21 -21
  17. package/dist/client/api/text-to-speech.d.ts +2 -2
  18. package/dist/client/api/text-to-speech.js +3 -3
  19. package/dist/client/api/transcribe.d.ts +5 -3
  20. package/dist/client/api/transcribe.js +6 -4
  21. package/dist/client/api/translate.d.ts +2 -2
  22. package/dist/client/api/translate.js +3 -3
  23. package/dist/client/rpc/node-rpc-client.d.ts +2 -7
  24. package/dist/client/rpc/node-rpc-client.js +2 -39
  25. package/dist/client/rpc/profiling/index.d.ts +2 -0
  26. package/dist/client/rpc/profiling/index.js +1 -0
  27. package/dist/client/rpc/profiling/profiler.d.ts +27 -0
  28. package/dist/client/rpc/profiling/profiler.js +139 -0
  29. package/dist/client/rpc/rpc-client.d.ts +2 -2
  30. package/dist/client/rpc/rpc-client.js +174 -12
  31. package/dist/examples/delegated-inference/consumer-profiled.d.ts +2 -0
  32. package/dist/examples/delegated-inference/consumer-profiled.js +67 -0
  33. package/dist/examples/llamacpp-multimodal.js +1 -1
  34. package/dist/examples/parallel-download.js +45 -13
  35. package/dist/examples/profiling/basic.d.ts +2 -0
  36. package/dist/examples/profiling/basic.js +71 -0
  37. package/dist/examples/profiling/per-call.d.ts +2 -0
  38. package/dist/examples/profiling/per-call.js +30 -0
  39. package/dist/examples/transcription/parakeet-ctc-filesystem.d.ts +2 -0
  40. package/dist/examples/transcription/parakeet-ctc-filesystem.js +40 -0
  41. package/dist/examples/transcription/parakeet-microphone-record.d.ts +2 -0
  42. package/dist/examples/transcription/parakeet-microphone-record.js +95 -0
  43. package/dist/examples/transcription/parakeet-sortformer.d.ts +2 -0
  44. package/dist/examples/transcription/parakeet-sortformer.js +115 -0
  45. package/dist/examples/transcription/parakeet-tdt-filesystem.d.ts +2 -0
  46. package/dist/examples/transcription/parakeet-tdt-filesystem.js +44 -0
  47. package/dist/examples/translation/translation-bergamot-pivot.d.ts +2 -0
  48. package/dist/examples/translation/translation-bergamot-pivot.js +73 -0
  49. package/dist/examples/translation/translation-llm-afriquegemma.d.ts +2 -0
  50. package/dist/examples/translation/translation-llm-afriquegemma.js +39 -0
  51. package/dist/index.d.ts +3 -1
  52. package/dist/index.js +2 -0
  53. package/dist/logging/index.d.ts +1 -1
  54. package/dist/logging/index.js +1 -1
  55. package/dist/logging/namespaces.d.ts +3 -9
  56. package/dist/logging/namespaces.js +2 -8
  57. package/dist/logging/registry.js +15 -1
  58. package/dist/models/registry/models.d.ts +1606 -245
  59. package/dist/models/registry/models.js +2127 -721
  60. package/dist/models/update-models/codegen.js +2 -2
  61. package/dist/models/update-models/history.d.ts +11 -0
  62. package/dist/models/update-models/history.js +24 -5
  63. package/dist/models/update-models/index.js +17 -16
  64. package/dist/models/update-models/naming.js +56 -0
  65. package/dist/profiling/aggregator.d.ts +15 -0
  66. package/dist/profiling/aggregator.js +100 -0
  67. package/dist/profiling/clock.d.ts +12 -0
  68. package/dist/profiling/clock.js +51 -0
  69. package/dist/profiling/controller.d.ts +33 -0
  70. package/dist/profiling/controller.js +123 -0
  71. package/dist/profiling/envelope.d.ts +10 -0
  72. package/dist/profiling/envelope.js +36 -0
  73. package/dist/profiling/events.d.ts +20 -0
  74. package/dist/profiling/events.js +37 -0
  75. package/dist/profiling/exporters.d.ts +10 -0
  76. package/dist/profiling/exporters.js +236 -0
  77. package/dist/profiling/index.d.ts +36 -0
  78. package/dist/profiling/index.js +31 -0
  79. package/dist/profiling/ring-buffer.d.ts +19 -0
  80. package/dist/profiling/ring-buffer.js +83 -0
  81. package/dist/profiling/types.d.ts +47 -0
  82. package/dist/profiling/types.js +1 -0
  83. package/dist/schemas/common.d.ts +1888 -133
  84. package/dist/schemas/common.js +2 -0
  85. package/dist/schemas/completion-stream.d.ts +31 -0
  86. package/dist/schemas/completion-stream.js +13 -0
  87. package/dist/schemas/download-asset.d.ts +14 -3
  88. package/dist/schemas/download-asset.js +3 -1
  89. package/dist/schemas/engine-addon-map.js +9 -6
  90. package/dist/schemas/get-model-info.d.ts +2 -0
  91. package/dist/schemas/get-model-info.js +1 -0
  92. package/dist/schemas/index.d.ts +2 -1
  93. package/dist/schemas/index.js +2 -1
  94. package/dist/schemas/llamacpp-config.d.ts +140 -14
  95. package/dist/schemas/llamacpp-config.js +11 -7
  96. package/dist/schemas/load-model.d.ts +9649 -2515
  97. package/dist/schemas/load-model.js +142 -182
  98. package/dist/schemas/model-src-utils.d.ts +10 -2
  99. package/dist/schemas/model-types.d.ts +19 -1
  100. package/dist/schemas/model-types.js +11 -0
  101. package/dist/schemas/ocr.d.ts +37 -0
  102. package/dist/schemas/ocr.js +5 -0
  103. package/dist/schemas/plugin.d.ts +51 -5
  104. package/dist/schemas/plugin.js +26 -0
  105. package/dist/schemas/profiling.d.ts +105 -0
  106. package/dist/schemas/profiling.js +66 -0
  107. package/dist/schemas/registry.d.ts +18 -0
  108. package/dist/schemas/registry.js +2 -0
  109. package/dist/schemas/reload-config.d.ts +126 -6
  110. package/dist/schemas/reload-config.js +1 -1
  111. package/dist/schemas/sdk-config.d.ts +2280 -24
  112. package/dist/schemas/sdk-config.js +6 -1
  113. package/dist/schemas/sdk-errors-client.d.ts +1 -0
  114. package/dist/schemas/sdk-errors-client.js +7 -0
  115. package/dist/schemas/sdk-errors-server.d.ts +3 -0
  116. package/dist/schemas/sdk-errors-server.js +15 -0
  117. package/dist/schemas/text-to-speech.d.ts +90 -68
  118. package/dist/schemas/text-to-speech.js +16 -33
  119. package/dist/schemas/transcription-config.d.ts +383 -0
  120. package/dist/schemas/{whispercpp-config.js → transcription-config.js} +29 -1
  121. package/dist/schemas/translation-config.d.ts +463 -8
  122. package/dist/schemas/translation-config.js +35 -2
  123. package/dist/server/bare/delegate-rpc-client.js +11 -1
  124. package/dist/server/bare/ops/load-model.d.ts +7 -1
  125. package/dist/server/bare/ops/load-model.js +41 -42
  126. package/dist/server/bare/ops/transcribe.d.ts +3 -0
  127. package/dist/server/bare/ops/transcribe.js +75 -0
  128. package/dist/server/bare/ops/translate.d.ts +2 -0
  129. package/dist/server/bare/ops/translate.js +12 -4
  130. package/dist/server/bare/plugins/index.d.ts +1 -0
  131. package/dist/server/bare/plugins/index.js +1 -0
  132. package/dist/server/bare/plugins/llamacpp-completion/ops/completion-stream.d.ts +2 -1
  133. package/dist/server/bare/plugins/llamacpp-completion/ops/completion-stream.js +8 -6
  134. package/dist/server/bare/plugins/llamacpp-completion/plugin.d.ts +108 -3
  135. package/dist/server/bare/plugins/llamacpp-completion/plugin.js +19 -6
  136. package/dist/server/bare/plugins/llamacpp-embedding/plugin.d.ts +32 -2
  137. package/dist/server/bare/plugins/llamacpp-embedding/plugin.js +31 -17
  138. package/dist/server/bare/plugins/nmtcpp-translation/plugin.d.ts +325 -3
  139. package/dist/server/bare/plugins/nmtcpp-translation/plugin.js +117 -10
  140. package/dist/server/bare/plugins/onnx-ocr/plugin.d.ts +50 -2
  141. package/dist/server/bare/plugins/onnx-ocr/plugin.js +46 -5
  142. package/dist/server/bare/plugins/onnx-tts/plugin.d.ts +351 -4
  143. package/dist/server/bare/plugins/onnx-tts/plugin.js +81 -61
  144. package/dist/server/bare/plugins/parakeet-transcription/plugin.d.ts +317 -0
  145. package/dist/server/bare/plugins/parakeet-transcription/plugin.js +154 -0
  146. package/dist/server/bare/plugins/whispercpp-transcription/plugin.d.ts +209 -3
  147. package/dist/server/bare/plugins/whispercpp-transcription/plugin.js +21 -9
  148. package/dist/server/bare/rag-hyperdb/rag-workspace-manager.js +2 -2
  149. package/dist/server/bare/registry/model-config-utils.js +3 -1
  150. package/dist/server/bare/utils/audio-input.d.ts +10 -0
  151. package/dist/server/bare/utils/audio-input.js +33 -0
  152. package/dist/server/error-handlers.d.ts +3 -2
  153. package/dist/server/error-handlers.js +11 -8
  154. package/dist/server/plugins/registry.js +5 -1
  155. package/dist/server/rpc/delegate-transport.d.ts +11 -3
  156. package/dist/server/rpc/delegate-transport.js +158 -3
  157. package/dist/server/rpc/handle-request.js +43 -4
  158. package/dist/server/rpc/handler-registry.js +7 -1
  159. package/dist/server/rpc/handler-utils.d.ts +5 -4
  160. package/dist/server/rpc/handler-utils.js +72 -31
  161. package/dist/server/rpc/handlers/completion-stream-delegated.d.ts +3 -1
  162. package/dist/server/rpc/handlers/completion-stream-delegated.js +11 -2
  163. package/dist/server/rpc/handlers/download-asset.js +55 -3
  164. package/dist/server/rpc/handlers/load-model/download-stats.d.ts +6 -0
  165. package/dist/server/rpc/handlers/load-model/download-stats.js +87 -0
  166. package/dist/server/rpc/handlers/load-model/handler.js +108 -123
  167. package/dist/server/rpc/handlers/load-model/http.d.ts +2 -1
  168. package/dist/server/rpc/handlers/load-model/http.js +21 -5
  169. package/dist/server/rpc/handlers/load-model/hyperdrive.d.ts +2 -1
  170. package/dist/server/rpc/handlers/load-model/hyperdrive.js +19 -14
  171. package/dist/server/rpc/handlers/load-model/registry.d.ts +2 -1
  172. package/dist/server/rpc/handlers/load-model/registry.js +123 -130
  173. package/dist/server/rpc/handlers/load-model/resolve.d.ts +2 -0
  174. package/dist/server/rpc/handlers/load-model/resolve.js +55 -22
  175. package/dist/server/rpc/handlers/load-model/types.d.ts +29 -0
  176. package/dist/server/rpc/handlers/load-model/types.js +1 -0
  177. package/dist/server/rpc/handlers/load-model-delegated.d.ts +5 -1
  178. package/dist/server/rpc/handlers/load-model-delegated.js +35 -7
  179. package/dist/server/rpc/handlers/plugin-dispatch.js +15 -10
  180. package/dist/server/rpc/handlers/plugin-invoke.js +56 -51
  181. package/dist/server/rpc/handlers/rag.js +24 -0
  182. package/dist/server/rpc/handlers/unload-model-delegated.d.ts +3 -0
  183. package/dist/server/rpc/handlers/unload-model-delegated.js +32 -0
  184. package/dist/server/rpc/profiling/context.d.ts +18 -0
  185. package/dist/server/rpc/profiling/context.js +38 -0
  186. package/dist/server/rpc/profiling/delegation-profiler.d.ts +32 -0
  187. package/dist/server/rpc/profiling/delegation-profiler.js +154 -0
  188. package/dist/server/rpc/profiling/index.d.ts +5 -0
  189. package/dist/server/rpc/profiling/index.js +4 -0
  190. package/dist/server/rpc/profiling/operation-metrics.d.ts +25 -0
  191. package/dist/server/rpc/profiling/operation-metrics.js +199 -0
  192. package/dist/server/rpc/profiling/operation-wrappers.d.ts +18 -0
  193. package/dist/server/rpc/profiling/operation-wrappers.js +124 -0
  194. package/dist/server/rpc/profiling/profiler.d.ts +13 -0
  195. package/dist/server/rpc/profiling/profiler.js +97 -0
  196. package/dist/server/utils/cache.d.ts +2 -1
  197. package/dist/server/utils/cache.js +5 -1
  198. package/dist/server/utils/checksum.d.ts +2 -0
  199. package/dist/server/utils/checksum.js +7 -0
  200. package/dist/server/utils/formatting.d.ts +6 -0
  201. package/dist/server/utils/formatting.js +35 -0
  202. package/dist/server/worker.js +2 -1
  203. package/dist/utils/errors-client.d.ts +3 -0
  204. package/dist/utils/errors-client.js +6 -0
  205. package/dist/utils/errors-server.d.ts +9 -0
  206. package/dist/utils/errors-server.js +15 -0
  207. package/dist/utils/global-singleton.d.ts +2 -0
  208. package/dist/utils/global-singleton.js +10 -0
  209. package/package.json +13 -16
  210. package/dist/schemas/whispercpp-config.d.ts +0 -70
  211. package/dist/server/bare/plugins/whispercpp-transcription/ops/transcribe-stream.d.ts +0 -3
  212. package/dist/server/bare/plugins/whispercpp-transcription/ops/transcribe-stream.js +0 -81
  213. /package/dist/examples/{whispercpp-filesystem.d.ts → transcription/whispercpp-filesystem.d.ts} +0 -0
  214. /package/dist/examples/{whispercpp-filesystem.js → transcription/whispercpp-filesystem.js} +0 -0
  215. /package/dist/examples/{whispercpp-microphone-record.d.ts → transcription/whispercpp-microphone-record.d.ts} +0 -0
  216. /package/dist/examples/{whispercpp-microphone-record.js → transcription/whispercpp-microphone-record.js} +0 -0
  217. /package/dist/examples/{whispercpp-prompt.d.ts → transcription/whispercpp-prompt.d.ts} +0 -0
  218. /package/dist/examples/{whispercpp-prompt.js → transcription/whispercpp-prompt.js} +0 -0
@@ -0,0 +1,383 @@
1
+ import { z } from "zod";
2
+ export declare const audioFormatSchema: z.ZodEnum<{
3
+ f32le: "f32le";
4
+ s16le: "s16le";
5
+ }>;
6
+ export type AudioFormat = z.infer<typeof audioFormatSchema>;
7
+ export declare const whisperConfigSchema: z.ZodObject<{
8
+ strategy: z.ZodOptional<z.ZodEnum<{
9
+ greedy: "greedy";
10
+ beam_search: "beam_search";
11
+ }>>;
12
+ n_threads: z.ZodOptional<z.ZodNumber>;
13
+ n_max_text_ctx: z.ZodOptional<z.ZodNumber>;
14
+ offset_ms: z.ZodOptional<z.ZodNumber>;
15
+ duration_ms: z.ZodOptional<z.ZodNumber>;
16
+ audio_ctx: z.ZodOptional<z.ZodNumber>;
17
+ translate: z.ZodOptional<z.ZodBoolean>;
18
+ no_context: z.ZodOptional<z.ZodBoolean>;
19
+ no_timestamps: z.ZodOptional<z.ZodBoolean>;
20
+ single_segment: z.ZodOptional<z.ZodBoolean>;
21
+ print_special: z.ZodOptional<z.ZodBoolean>;
22
+ print_progress: z.ZodOptional<z.ZodBoolean>;
23
+ print_realtime: z.ZodOptional<z.ZodBoolean>;
24
+ print_timestamps: z.ZodOptional<z.ZodBoolean>;
25
+ token_timestamps: z.ZodOptional<z.ZodBoolean>;
26
+ thold_pt: z.ZodOptional<z.ZodNumber>;
27
+ thold_ptsum: z.ZodOptional<z.ZodNumber>;
28
+ max_len: z.ZodOptional<z.ZodNumber>;
29
+ split_on_word: z.ZodOptional<z.ZodBoolean>;
30
+ max_tokens: z.ZodOptional<z.ZodNumber>;
31
+ debug_mode: z.ZodOptional<z.ZodBoolean>;
32
+ tdrz_enable: z.ZodOptional<z.ZodBoolean>;
33
+ suppress_regex: z.ZodOptional<z.ZodString>;
34
+ initial_prompt: z.ZodOptional<z.ZodString>;
35
+ language: z.ZodOptional<z.ZodString>;
36
+ detect_language: z.ZodOptional<z.ZodBoolean>;
37
+ suppress_blank: z.ZodOptional<z.ZodBoolean>;
38
+ suppress_nst: z.ZodOptional<z.ZodBoolean>;
39
+ temperature: z.ZodOptional<z.ZodNumber>;
40
+ length_penalty: z.ZodOptional<z.ZodNumber>;
41
+ temperature_inc: z.ZodOptional<z.ZodNumber>;
42
+ entropy_thold: z.ZodOptional<z.ZodNumber>;
43
+ logprob_thold: z.ZodOptional<z.ZodNumber>;
44
+ greedy_best_of: z.ZodOptional<z.ZodNumber>;
45
+ beam_search_beam_size: z.ZodOptional<z.ZodNumber>;
46
+ vad_params: z.ZodOptional<z.ZodObject<{
47
+ threshold: z.ZodOptional<z.ZodNumber>;
48
+ min_speech_duration_ms: z.ZodOptional<z.ZodNumber>;
49
+ min_silence_duration_ms: z.ZodOptional<z.ZodNumber>;
50
+ max_speech_duration_s: z.ZodOptional<z.ZodNumber>;
51
+ speech_pad_ms: z.ZodOptional<z.ZodNumber>;
52
+ samples_overlap: z.ZodOptional<z.ZodNumber>;
53
+ }, z.core.$strip>>;
54
+ audio_format: z.ZodOptional<z.ZodEnum<{
55
+ f32le: "f32le";
56
+ s16le: "s16le";
57
+ }>>;
58
+ contextParams: z.ZodOptional<z.ZodObject<{
59
+ model: z.ZodOptional<z.ZodString>;
60
+ use_gpu: z.ZodOptional<z.ZodBoolean>;
61
+ flash_attn: z.ZodOptional<z.ZodBoolean>;
62
+ gpu_device: z.ZodOptional<z.ZodNumber>;
63
+ }, z.core.$strip>>;
64
+ miscConfig: z.ZodOptional<z.ZodObject<{
65
+ caption_enabled: z.ZodOptional<z.ZodBoolean>;
66
+ }, z.core.$strip>>;
67
+ vadModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
68
+ src: z.ZodString;
69
+ name: z.ZodOptional<z.ZodString>;
70
+ modelId: z.ZodOptional<z.ZodString>;
71
+ registryPath: z.ZodOptional<z.ZodString>;
72
+ registrySource: z.ZodOptional<z.ZodString>;
73
+ blobCoreKey: z.ZodOptional<z.ZodString>;
74
+ blobIndex: z.ZodOptional<z.ZodNumber>;
75
+ engine: z.ZodOptional<z.ZodString>;
76
+ expectedSize: z.ZodOptional<z.ZodNumber>;
77
+ sha256Checksum: z.ZodOptional<z.ZodString>;
78
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
79
+ "llamacpp-completion": "llamacpp-completion";
80
+ "whispercpp-transcription": "whispercpp-transcription";
81
+ "llamacpp-embedding": "llamacpp-embedding";
82
+ "nmtcpp-translation": "nmtcpp-translation";
83
+ "onnx-tts": "onnx-tts";
84
+ "parakeet-transcription": "parakeet-transcription";
85
+ "onnx-ocr": "onnx-ocr";
86
+ llm: "llm";
87
+ whisper: "whisper";
88
+ embeddings: "embeddings";
89
+ nmt: "nmt";
90
+ parakeet: "parakeet";
91
+ tts: "tts";
92
+ ocr: "ocr";
93
+ }>, z.ZodLiteral<"vad">]>>;
94
+ }, z.core.$strip>]>>;
95
+ }, z.core.$strip>;
96
+ export type WhisperConfig = z.infer<typeof whisperConfigSchema>;
97
+ export declare const parakeetModelTypeEnumSchema: z.ZodEnum<{
98
+ tdt: "tdt";
99
+ ctc: "ctc";
100
+ sortformer: "sortformer";
101
+ }>;
102
+ export type ParakeetModelVariant = z.infer<typeof parakeetModelTypeEnumSchema>;
103
+ export declare const parakeetRuntimeConfigSchema: z.ZodObject<{
104
+ modelType: z.ZodDefault<z.ZodEnum<{
105
+ tdt: "tdt";
106
+ ctc: "ctc";
107
+ sortformer: "sortformer";
108
+ }>>;
109
+ maxThreads: z.ZodOptional<z.ZodNumber>;
110
+ useGPU: z.ZodOptional<z.ZodBoolean>;
111
+ sampleRate: z.ZodOptional<z.ZodNumber>;
112
+ channels: z.ZodOptional<z.ZodNumber>;
113
+ captionEnabled: z.ZodOptional<z.ZodBoolean>;
114
+ timestampsEnabled: z.ZodOptional<z.ZodBoolean>;
115
+ }, z.core.$strip>;
116
+ export declare const parakeetConfigSchema: z.ZodObject<{
117
+ modelType: z.ZodDefault<z.ZodEnum<{
118
+ tdt: "tdt";
119
+ ctc: "ctc";
120
+ sortformer: "sortformer";
121
+ }>>;
122
+ maxThreads: z.ZodOptional<z.ZodNumber>;
123
+ useGPU: z.ZodOptional<z.ZodBoolean>;
124
+ sampleRate: z.ZodOptional<z.ZodNumber>;
125
+ channels: z.ZodOptional<z.ZodNumber>;
126
+ captionEnabled: z.ZodOptional<z.ZodBoolean>;
127
+ timestampsEnabled: z.ZodOptional<z.ZodBoolean>;
128
+ parakeetEncoderSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
129
+ src: z.ZodString;
130
+ name: z.ZodOptional<z.ZodString>;
131
+ modelId: z.ZodOptional<z.ZodString>;
132
+ registryPath: z.ZodOptional<z.ZodString>;
133
+ registrySource: z.ZodOptional<z.ZodString>;
134
+ blobCoreKey: z.ZodOptional<z.ZodString>;
135
+ blobIndex: z.ZodOptional<z.ZodNumber>;
136
+ engine: z.ZodOptional<z.ZodString>;
137
+ expectedSize: z.ZodOptional<z.ZodNumber>;
138
+ sha256Checksum: z.ZodOptional<z.ZodString>;
139
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
140
+ "llamacpp-completion": "llamacpp-completion";
141
+ "whispercpp-transcription": "whispercpp-transcription";
142
+ "llamacpp-embedding": "llamacpp-embedding";
143
+ "nmtcpp-translation": "nmtcpp-translation";
144
+ "onnx-tts": "onnx-tts";
145
+ "parakeet-transcription": "parakeet-transcription";
146
+ "onnx-ocr": "onnx-ocr";
147
+ llm: "llm";
148
+ whisper: "whisper";
149
+ embeddings: "embeddings";
150
+ nmt: "nmt";
151
+ parakeet: "parakeet";
152
+ tts: "tts";
153
+ ocr: "ocr";
154
+ }>, z.ZodLiteral<"vad">]>>;
155
+ }, z.core.$strip>]>>;
156
+ parakeetEncoderDataSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
157
+ src: z.ZodString;
158
+ name: z.ZodOptional<z.ZodString>;
159
+ modelId: z.ZodOptional<z.ZodString>;
160
+ registryPath: z.ZodOptional<z.ZodString>;
161
+ registrySource: z.ZodOptional<z.ZodString>;
162
+ blobCoreKey: z.ZodOptional<z.ZodString>;
163
+ blobIndex: z.ZodOptional<z.ZodNumber>;
164
+ engine: z.ZodOptional<z.ZodString>;
165
+ expectedSize: z.ZodOptional<z.ZodNumber>;
166
+ sha256Checksum: z.ZodOptional<z.ZodString>;
167
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
168
+ "llamacpp-completion": "llamacpp-completion";
169
+ "whispercpp-transcription": "whispercpp-transcription";
170
+ "llamacpp-embedding": "llamacpp-embedding";
171
+ "nmtcpp-translation": "nmtcpp-translation";
172
+ "onnx-tts": "onnx-tts";
173
+ "parakeet-transcription": "parakeet-transcription";
174
+ "onnx-ocr": "onnx-ocr";
175
+ llm: "llm";
176
+ whisper: "whisper";
177
+ embeddings: "embeddings";
178
+ nmt: "nmt";
179
+ parakeet: "parakeet";
180
+ tts: "tts";
181
+ ocr: "ocr";
182
+ }>, z.ZodLiteral<"vad">]>>;
183
+ }, z.core.$strip>]>>;
184
+ parakeetDecoderSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
185
+ src: z.ZodString;
186
+ name: z.ZodOptional<z.ZodString>;
187
+ modelId: z.ZodOptional<z.ZodString>;
188
+ registryPath: z.ZodOptional<z.ZodString>;
189
+ registrySource: z.ZodOptional<z.ZodString>;
190
+ blobCoreKey: z.ZodOptional<z.ZodString>;
191
+ blobIndex: z.ZodOptional<z.ZodNumber>;
192
+ engine: z.ZodOptional<z.ZodString>;
193
+ expectedSize: z.ZodOptional<z.ZodNumber>;
194
+ sha256Checksum: z.ZodOptional<z.ZodString>;
195
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
196
+ "llamacpp-completion": "llamacpp-completion";
197
+ "whispercpp-transcription": "whispercpp-transcription";
198
+ "llamacpp-embedding": "llamacpp-embedding";
199
+ "nmtcpp-translation": "nmtcpp-translation";
200
+ "onnx-tts": "onnx-tts";
201
+ "parakeet-transcription": "parakeet-transcription";
202
+ "onnx-ocr": "onnx-ocr";
203
+ llm: "llm";
204
+ whisper: "whisper";
205
+ embeddings: "embeddings";
206
+ nmt: "nmt";
207
+ parakeet: "parakeet";
208
+ tts: "tts";
209
+ ocr: "ocr";
210
+ }>, z.ZodLiteral<"vad">]>>;
211
+ }, z.core.$strip>]>>;
212
+ parakeetVocabSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
213
+ src: z.ZodString;
214
+ name: z.ZodOptional<z.ZodString>;
215
+ modelId: z.ZodOptional<z.ZodString>;
216
+ registryPath: z.ZodOptional<z.ZodString>;
217
+ registrySource: z.ZodOptional<z.ZodString>;
218
+ blobCoreKey: z.ZodOptional<z.ZodString>;
219
+ blobIndex: z.ZodOptional<z.ZodNumber>;
220
+ engine: z.ZodOptional<z.ZodString>;
221
+ expectedSize: z.ZodOptional<z.ZodNumber>;
222
+ sha256Checksum: z.ZodOptional<z.ZodString>;
223
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
224
+ "llamacpp-completion": "llamacpp-completion";
225
+ "whispercpp-transcription": "whispercpp-transcription";
226
+ "llamacpp-embedding": "llamacpp-embedding";
227
+ "nmtcpp-translation": "nmtcpp-translation";
228
+ "onnx-tts": "onnx-tts";
229
+ "parakeet-transcription": "parakeet-transcription";
230
+ "onnx-ocr": "onnx-ocr";
231
+ llm: "llm";
232
+ whisper: "whisper";
233
+ embeddings: "embeddings";
234
+ nmt: "nmt";
235
+ parakeet: "parakeet";
236
+ tts: "tts";
237
+ ocr: "ocr";
238
+ }>, z.ZodLiteral<"vad">]>>;
239
+ }, z.core.$strip>]>>;
240
+ parakeetPreprocessorSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
241
+ src: z.ZodString;
242
+ name: z.ZodOptional<z.ZodString>;
243
+ modelId: z.ZodOptional<z.ZodString>;
244
+ registryPath: z.ZodOptional<z.ZodString>;
245
+ registrySource: z.ZodOptional<z.ZodString>;
246
+ blobCoreKey: z.ZodOptional<z.ZodString>;
247
+ blobIndex: z.ZodOptional<z.ZodNumber>;
248
+ engine: z.ZodOptional<z.ZodString>;
249
+ expectedSize: z.ZodOptional<z.ZodNumber>;
250
+ sha256Checksum: z.ZodOptional<z.ZodString>;
251
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
252
+ "llamacpp-completion": "llamacpp-completion";
253
+ "whispercpp-transcription": "whispercpp-transcription";
254
+ "llamacpp-embedding": "llamacpp-embedding";
255
+ "nmtcpp-translation": "nmtcpp-translation";
256
+ "onnx-tts": "onnx-tts";
257
+ "parakeet-transcription": "parakeet-transcription";
258
+ "onnx-ocr": "onnx-ocr";
259
+ llm: "llm";
260
+ whisper: "whisper";
261
+ embeddings: "embeddings";
262
+ nmt: "nmt";
263
+ parakeet: "parakeet";
264
+ tts: "tts";
265
+ ocr: "ocr";
266
+ }>, z.ZodLiteral<"vad">]>>;
267
+ }, z.core.$strip>]>>;
268
+ parakeetCtcModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
269
+ src: z.ZodString;
270
+ name: z.ZodOptional<z.ZodString>;
271
+ modelId: z.ZodOptional<z.ZodString>;
272
+ registryPath: z.ZodOptional<z.ZodString>;
273
+ registrySource: z.ZodOptional<z.ZodString>;
274
+ blobCoreKey: z.ZodOptional<z.ZodString>;
275
+ blobIndex: z.ZodOptional<z.ZodNumber>;
276
+ engine: z.ZodOptional<z.ZodString>;
277
+ expectedSize: z.ZodOptional<z.ZodNumber>;
278
+ sha256Checksum: z.ZodOptional<z.ZodString>;
279
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
280
+ "llamacpp-completion": "llamacpp-completion";
281
+ "whispercpp-transcription": "whispercpp-transcription";
282
+ "llamacpp-embedding": "llamacpp-embedding";
283
+ "nmtcpp-translation": "nmtcpp-translation";
284
+ "onnx-tts": "onnx-tts";
285
+ "parakeet-transcription": "parakeet-transcription";
286
+ "onnx-ocr": "onnx-ocr";
287
+ llm: "llm";
288
+ whisper: "whisper";
289
+ embeddings: "embeddings";
290
+ nmt: "nmt";
291
+ parakeet: "parakeet";
292
+ tts: "tts";
293
+ ocr: "ocr";
294
+ }>, z.ZodLiteral<"vad">]>>;
295
+ }, z.core.$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";
322
+ }>, z.ZodLiteral<"vad">]>>;
323
+ }, z.core.$strip>]>>;
324
+ parakeetTokenizerSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
325
+ src: z.ZodString;
326
+ name: z.ZodOptional<z.ZodString>;
327
+ modelId: z.ZodOptional<z.ZodString>;
328
+ registryPath: z.ZodOptional<z.ZodString>;
329
+ registrySource: z.ZodOptional<z.ZodString>;
330
+ blobCoreKey: z.ZodOptional<z.ZodString>;
331
+ blobIndex: z.ZodOptional<z.ZodNumber>;
332
+ engine: z.ZodOptional<z.ZodString>;
333
+ expectedSize: z.ZodOptional<z.ZodNumber>;
334
+ sha256Checksum: z.ZodOptional<z.ZodString>;
335
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
336
+ "llamacpp-completion": "llamacpp-completion";
337
+ "whispercpp-transcription": "whispercpp-transcription";
338
+ "llamacpp-embedding": "llamacpp-embedding";
339
+ "nmtcpp-translation": "nmtcpp-translation";
340
+ "onnx-tts": "onnx-tts";
341
+ "parakeet-transcription": "parakeet-transcription";
342
+ "onnx-ocr": "onnx-ocr";
343
+ llm: "llm";
344
+ whisper: "whisper";
345
+ embeddings: "embeddings";
346
+ nmt: "nmt";
347
+ parakeet: "parakeet";
348
+ tts: "tts";
349
+ ocr: "ocr";
350
+ }>, z.ZodLiteral<"vad">]>>;
351
+ }, z.core.$strip>]>>;
352
+ parakeetSortformerSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
353
+ src: z.ZodString;
354
+ name: z.ZodOptional<z.ZodString>;
355
+ modelId: z.ZodOptional<z.ZodString>;
356
+ registryPath: z.ZodOptional<z.ZodString>;
357
+ registrySource: z.ZodOptional<z.ZodString>;
358
+ blobCoreKey: z.ZodOptional<z.ZodString>;
359
+ blobIndex: z.ZodOptional<z.ZodNumber>;
360
+ engine: z.ZodOptional<z.ZodString>;
361
+ expectedSize: z.ZodOptional<z.ZodNumber>;
362
+ sha256Checksum: z.ZodOptional<z.ZodString>;
363
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
364
+ "llamacpp-completion": "llamacpp-completion";
365
+ "whispercpp-transcription": "whispercpp-transcription";
366
+ "llamacpp-embedding": "llamacpp-embedding";
367
+ "nmtcpp-translation": "nmtcpp-translation";
368
+ "onnx-tts": "onnx-tts";
369
+ "parakeet-transcription": "parakeet-transcription";
370
+ "onnx-ocr": "onnx-ocr";
371
+ llm: "llm";
372
+ whisper: "whisper";
373
+ embeddings: "embeddings";
374
+ nmt: "nmt";
375
+ parakeet: "parakeet";
376
+ tts: "tts";
377
+ ocr: "ocr";
378
+ }>, z.ZodLiteral<"vad">]>>;
379
+ }, z.core.$strip>]>>;
380
+ }, z.core.$strip>;
381
+ export type ParakeetRuntimeConfig = z.infer<typeof parakeetRuntimeConfigSchema>;
382
+ export type ParakeetConfig = z.infer<typeof parakeetConfigSchema>;
383
+ //# sourceMappingURL=transcription-config.d.ts.map
@@ -1,5 +1,8 @@
1
1
  import { z } from "zod";
2
+ import { modelSrcInputSchema } from "./model-src-utils.js";
3
+ // === Shared ===
2
4
  export const audioFormatSchema = z.enum(["f32le", "s16le"]);
5
+ // === Whisper (whisper.cpp) engine config ===
3
6
  const vadParamsSchema = z
4
7
  .object({
5
8
  threshold: z.number().optional(),
@@ -59,9 +62,34 @@ export const whisperConfigSchema = z.object({
59
62
  logprob_thold: z.number().optional(),
60
63
  greedy_best_of: z.number().int().optional(),
61
64
  beam_search_beam_size: z.number().int().optional(),
62
- vad_model_path: z.string().optional(),
63
65
  vad_params: vadParamsSchema,
64
66
  audio_format: audioFormatSchema.optional(),
65
67
  contextParams: contextParamsSchema,
66
68
  miscConfig: miscConfigSchema,
69
+ vadModelSrc: modelSrcInputSchema.optional(),
70
+ });
71
+ // === Parakeet (NVIDIA NeMo ONNX) engine config ===
72
+ export const parakeetModelTypeEnumSchema = z.enum(["tdt", "ctc", "sortformer"]);
73
+ export const parakeetRuntimeConfigSchema = z.object({
74
+ modelType: parakeetModelTypeEnumSchema.default("tdt"),
75
+ maxThreads: z.number().int().optional(),
76
+ useGPU: z.boolean().optional(),
77
+ sampleRate: z.number().int().optional(),
78
+ channels: z.number().int().optional(),
79
+ captionEnabled: z.boolean().optional(),
80
+ timestampsEnabled: z.boolean().optional(),
81
+ });
82
+ export const parakeetConfigSchema = parakeetRuntimeConfigSchema.extend({
83
+ // TDT sources
84
+ parakeetEncoderSrc: modelSrcInputSchema.optional(),
85
+ parakeetEncoderDataSrc: modelSrcInputSchema.optional(),
86
+ parakeetDecoderSrc: modelSrcInputSchema.optional(),
87
+ parakeetVocabSrc: modelSrcInputSchema.optional(),
88
+ parakeetPreprocessorSrc: modelSrcInputSchema.optional(),
89
+ // CTC sources
90
+ parakeetCtcModelSrc: modelSrcInputSchema.optional(),
91
+ parakeetCtcModelDataSrc: modelSrcInputSchema.optional(),
92
+ parakeetTokenizerSrc: modelSrcInputSchema.optional(),
93
+ // Sortformer source
94
+ parakeetSortformerSrc: modelSrcInputSchema.optional(),
67
95
  });