@qvac/sdk 0.7.0 → 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 +14 -17
  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
@@ -2,7 +2,7 @@ import { z } from "zod";
2
2
  import { logLevelSchema } from "./logging-stream.js";
3
3
  import { ModelType } from "./model-types.js";
4
4
  import { llmConfigBaseSchema, embedConfigBaseSchema } from "./llamacpp-config.js";
5
- import { whisperConfigSchema } from "./whispercpp-config.js";
5
+ import { whisperConfigSchema, parakeetConfigSchema, } from "./transcription-config.js";
6
6
  import { ocrConfigSchema } from "./ocr.js";
7
7
  import { runtimeContextSchema } from "./runtime-context.js";
8
8
  // Alias keys for user convenience (maps to canonical types)
@@ -11,6 +11,7 @@ const AliasKeys = {
11
11
  whisper: "whisper",
12
12
  embeddings: "embeddings",
13
13
  nmt: "nmt",
14
+ parakeet: "parakeet",
14
15
  tts: "tts",
15
16
  ocr: "ocr",
16
17
  };
@@ -41,6 +42,9 @@ export const deviceConfigDefaultsSchema = z
41
42
  [ModelType.whispercppTranscription]: whisperConfigSchema
42
43
  .partial()
43
44
  .optional(),
45
+ [ModelType.parakeetTranscription]: parakeetConfigSchema
46
+ .partial()
47
+ .optional(),
44
48
  [ModelType.nmtcppTranslation]: z.record(z.string(), z.unknown()).optional(),
45
49
  [ModelType.onnxTts]: z.record(z.string(), z.unknown()).optional(),
46
50
  [ModelType.onnxOcr]: ocrConfigSchema.partial().optional(),
@@ -48,6 +52,7 @@ export const deviceConfigDefaultsSchema = z
48
52
  [AliasKeys.llm]: llmConfigBaseSchema.optional(),
49
53
  [AliasKeys.embeddings]: embedConfigBaseSchema.optional(),
50
54
  [AliasKeys.whisper]: whisperConfigSchema.partial().optional(),
55
+ [AliasKeys.parakeet]: parakeetConfigSchema.partial().optional(),
51
56
  [AliasKeys.nmt]: z.record(z.string(), z.unknown()).optional(),
52
57
  [AliasKeys.tts]: z.record(z.string(), z.unknown()).optional(),
53
58
  [AliasKeys.ocr]: ocrConfigSchema.partial().optional(),
@@ -24,6 +24,7 @@ export declare const SDK_CLIENT_ERROR_CODES: {
24
24
  readonly CONFIG_VALIDATION_FAILED: 50605;
25
25
  readonly PEAR_WORKER_ENTRY_REQUIRED: 50606;
26
26
  readonly MULTIPLE_SDK_INSTALLATIONS: 50607;
27
+ readonly PROFILER_INVALID_CAPACITY: 50800;
27
28
  };
28
29
  declare const clientErrorDefinitions: ErrorCodesMap;
29
30
  export { clientErrorDefinitions as SDK_CLIENT_ERROR_DEFINITIONS };
@@ -29,6 +29,8 @@ export const SDK_CLIENT_ERROR_CODES = {
29
29
  CONFIG_VALIDATION_FAILED: 50605,
30
30
  PEAR_WORKER_ENTRY_REQUIRED: 50606,
31
31
  MULTIPLE_SDK_INSTALLATIONS: 50607,
32
+ // Profiler Errors (50,800-50,899)
33
+ PROFILER_INVALID_CAPACITY: 50800,
32
34
  };
33
35
  const clientErrorDefinitions = {
34
36
  // Response Validation Errors (50,001-50,199)
@@ -131,6 +133,11 @@ const clientErrorDefinitions = {
131
133
  name: "PEAR_WORKER_ENTRY_REQUIRED",
132
134
  message: (workerEntry) => `No plugins registered. Pear apps must spawn ${workerEntry} as the worker entry. Run \`npx qvac bundle sdk\` to generate it, then spawn the generated file instead of your worker directly.`,
133
135
  },
136
+ // Profiler Errors (50,800-50,899)
137
+ [SDK_CLIENT_ERROR_CODES.PROFILER_INVALID_CAPACITY]: {
138
+ name: "PROFILER_INVALID_CAPACITY",
139
+ message: (minCapacity) => `Ring buffer capacity must be at least ${minCapacity}`,
140
+ },
134
141
  };
135
142
  addCodes(clientErrorDefinitions, { name: "qvac-sdk-client", version: "1.1.0" });
136
143
  export { clientErrorDefinitions as SDK_CLIENT_ERROR_DEFINITIONS };
@@ -13,6 +13,7 @@ export declare const SDK_SERVER_ERROR_CODES: {
13
13
  readonly VAD_MODEL_REQUIRED: 52205;
14
14
  readonly TTS_ARTIFACTS_REQUIRED: 52208;
15
15
  readonly TTS_REFERENCE_AUDIO_REQUIRED: 52209;
16
+ readonly PARAKEET_ARTIFACTS_REQUIRED: 52210;
16
17
  readonly MODEL_UNLOAD_FAILED: 52400;
17
18
  readonly EMBED_FAILED: 52401;
18
19
  readonly EMBED_NO_EMBEDDINGS: 52402;
@@ -78,6 +79,8 @@ export declare const SDK_SERVER_ERROR_CODES: {
78
79
  readonly PLUGIN_HANDLER_TYPE_MISMATCH: 53855;
79
80
  readonly PLUGIN_LOGGING_INVALID: 53856;
80
81
  readonly PLUGIN_DEFINITION_INVALID: 53857;
82
+ readonly PLUGIN_MODEL_TYPE_RESERVED: 53858;
83
+ readonly PLUGIN_LOAD_CONFIG_VALIDATION_FAILED: 53859;
81
84
  readonly PATH_TRAVERSAL: 53900;
82
85
  readonly QVAC_MODEL_REGISTRY_QUERY_FAILED: 53950;
83
86
  };
@@ -16,6 +16,7 @@ export const SDK_SERVER_ERROR_CODES = {
16
16
  VAD_MODEL_REQUIRED: 52205,
17
17
  TTS_ARTIFACTS_REQUIRED: 52208,
18
18
  TTS_REFERENCE_AUDIO_REQUIRED: 52209,
19
+ PARAKEET_ARTIFACTS_REQUIRED: 52210,
19
20
  // Model Operations (52,400-52,799)
20
21
  MODEL_UNLOAD_FAILED: 52400,
21
22
  EMBED_FAILED: 52401,
@@ -89,6 +90,8 @@ export const SDK_SERVER_ERROR_CODES = {
89
90
  PLUGIN_HANDLER_TYPE_MISMATCH: 53855,
90
91
  PLUGIN_LOGGING_INVALID: 53856,
91
92
  PLUGIN_DEFINITION_INVALID: 53857,
93
+ PLUGIN_MODEL_TYPE_RESERVED: 53858,
94
+ PLUGIN_LOAD_CONFIG_VALIDATION_FAILED: 53859,
92
95
  // Security (53,900-53,949)
93
96
  PATH_TRAVERSAL: 53900,
94
97
  // QVAC Model Registry Operations (53,950-54,000)
@@ -150,6 +153,10 @@ const serverErrorDefinitions = {
150
153
  name: "TTS_REFERENCE_AUDIO_REQUIRED",
151
154
  message: "TTS (Chatterbox) requires referenceAudioSrc (path or URL to a WAV file for voice cloning)",
152
155
  },
156
+ [SDK_SERVER_ERROR_CODES.PARAKEET_ARTIFACTS_REQUIRED]: {
157
+ name: "PARAKEET_ARTIFACTS_REQUIRED",
158
+ message: "Parakeet model sources are missing. TDT requires parakeetEncoderSrc, parakeetDecoderSrc, parakeetVocabSrc, parakeetPreprocessorSrc. CTC requires parakeetCtcModelSrc, parakeetTokenizerSrc. Sortformer requires parakeetSortformerSrc.",
159
+ },
153
160
  // Model Operations (52,400-52,799)
154
161
  [SDK_SERVER_ERROR_CODES.MODEL_UNLOAD_FAILED]: {
155
162
  name: "MODEL_UNLOAD_FAILED",
@@ -420,6 +427,14 @@ const serverErrorDefinitions = {
420
427
  name: "PLUGIN_DEFINITION_INVALID",
421
428
  message: (modelType, details) => `Plugin definition invalid for "${modelType}": ${details}`,
422
429
  },
430
+ [SDK_SERVER_ERROR_CODES.PLUGIN_MODEL_TYPE_RESERVED]: {
431
+ name: "PLUGIN_MODEL_TYPE_RESERVED",
432
+ message: (modelType) => `modelType "${modelType}" is reserved for built-in plugins`,
433
+ },
434
+ [SDK_SERVER_ERROR_CODES.PLUGIN_LOAD_CONFIG_VALIDATION_FAILED]: {
435
+ name: "PLUGIN_LOAD_CONFIG_VALIDATION_FAILED",
436
+ message: (modelType, details) => `modelConfig validation failed for "${modelType}": ${details}`,
437
+ },
423
438
  // Security (53,900-53,949)
424
439
  [SDK_SERVER_ERROR_CODES.PATH_TRAVERSAL]: {
425
440
  name: "PATH_TRAVERSAL",
@@ -1,5 +1,44 @@
1
1
  import { z } from "zod";
2
2
  export declare const TTS_LANGUAGES: readonly ["en", "es", "de", "it"];
3
+ export declare const ttsChatterboxRuntimeConfigSchema: z.ZodObject<{
4
+ ttsEngine: z.ZodLiteral<"chatterbox">;
5
+ language: z.ZodEnum<{
6
+ en: "en";
7
+ de: "de";
8
+ es: "es";
9
+ it: "it";
10
+ }>;
11
+ }, z.core.$strip>;
12
+ export declare const ttsSupertonicRuntimeConfigSchema: z.ZodObject<{
13
+ ttsEngine: z.ZodLiteral<"supertonic">;
14
+ language: z.ZodEnum<{
15
+ en: "en";
16
+ de: "de";
17
+ es: "es";
18
+ it: "it";
19
+ }>;
20
+ ttsSpeed: z.ZodOptional<z.ZodNumber>;
21
+ ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
22
+ }, z.core.$strip>;
23
+ export declare const ttsRuntimeConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
24
+ ttsEngine: z.ZodLiteral<"chatterbox">;
25
+ language: z.ZodEnum<{
26
+ en: "en";
27
+ de: "de";
28
+ es: "es";
29
+ it: "it";
30
+ }>;
31
+ }, z.core.$strip>, z.ZodObject<{
32
+ ttsEngine: z.ZodLiteral<"supertonic">;
33
+ language: z.ZodEnum<{
34
+ en: "en";
35
+ de: "de";
36
+ es: "es";
37
+ it: "it";
38
+ }>;
39
+ ttsSpeed: z.ZodOptional<z.ZodNumber>;
40
+ ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
41
+ }, z.core.$strip>]>;
3
42
  export declare const ttsChatterboxConfigSchema: z.ZodObject<{
4
43
  ttsEngine: z.ZodLiteral<"chatterbox">;
5
44
  language: z.ZodEnum<{
@@ -25,11 +64,13 @@ export declare const ttsChatterboxConfigSchema: z.ZodObject<{
25
64
  "llamacpp-embedding": "llamacpp-embedding";
26
65
  "nmtcpp-translation": "nmtcpp-translation";
27
66
  "onnx-tts": "onnx-tts";
67
+ "parakeet-transcription": "parakeet-transcription";
28
68
  "onnx-ocr": "onnx-ocr";
29
69
  llm: "llm";
30
70
  whisper: "whisper";
31
71
  embeddings: "embeddings";
32
72
  nmt: "nmt";
73
+ parakeet: "parakeet";
33
74
  tts: "tts";
34
75
  ocr: "ocr";
35
76
  }>, z.ZodLiteral<"vad">]>>;
@@ -51,11 +92,13 @@ export declare const ttsChatterboxConfigSchema: z.ZodObject<{
51
92
  "llamacpp-embedding": "llamacpp-embedding";
52
93
  "nmtcpp-translation": "nmtcpp-translation";
53
94
  "onnx-tts": "onnx-tts";
95
+ "parakeet-transcription": "parakeet-transcription";
54
96
  "onnx-ocr": "onnx-ocr";
55
97
  llm: "llm";
56
98
  whisper: "whisper";
57
99
  embeddings: "embeddings";
58
100
  nmt: "nmt";
101
+ parakeet: "parakeet";
59
102
  tts: "tts";
60
103
  ocr: "ocr";
61
104
  }>, z.ZodLiteral<"vad">]>>;
@@ -77,11 +120,13 @@ export declare const ttsChatterboxConfigSchema: z.ZodObject<{
77
120
  "llamacpp-embedding": "llamacpp-embedding";
78
121
  "nmtcpp-translation": "nmtcpp-translation";
79
122
  "onnx-tts": "onnx-tts";
123
+ "parakeet-transcription": "parakeet-transcription";
80
124
  "onnx-ocr": "onnx-ocr";
81
125
  llm: "llm";
82
126
  whisper: "whisper";
83
127
  embeddings: "embeddings";
84
128
  nmt: "nmt";
129
+ parakeet: "parakeet";
85
130
  tts: "tts";
86
131
  ocr: "ocr";
87
132
  }>, z.ZodLiteral<"vad">]>>;
@@ -103,11 +148,13 @@ export declare const ttsChatterboxConfigSchema: z.ZodObject<{
103
148
  "llamacpp-embedding": "llamacpp-embedding";
104
149
  "nmtcpp-translation": "nmtcpp-translation";
105
150
  "onnx-tts": "onnx-tts";
151
+ "parakeet-transcription": "parakeet-transcription";
106
152
  "onnx-ocr": "onnx-ocr";
107
153
  llm: "llm";
108
154
  whisper: "whisper";
109
155
  embeddings: "embeddings";
110
156
  nmt: "nmt";
157
+ parakeet: "parakeet";
111
158
  tts: "tts";
112
159
  ocr: "ocr";
113
160
  }>, z.ZodLiteral<"vad">]>>;
@@ -129,11 +176,13 @@ export declare const ttsChatterboxConfigSchema: z.ZodObject<{
129
176
  "llamacpp-embedding": "llamacpp-embedding";
130
177
  "nmtcpp-translation": "nmtcpp-translation";
131
178
  "onnx-tts": "onnx-tts";
179
+ "parakeet-transcription": "parakeet-transcription";
132
180
  "onnx-ocr": "onnx-ocr";
133
181
  llm: "llm";
134
182
  whisper: "whisper";
135
183
  embeddings: "embeddings";
136
184
  nmt: "nmt";
185
+ parakeet: "parakeet";
137
186
  tts: "tts";
138
187
  ocr: "ocr";
139
188
  }>, z.ZodLiteral<"vad">]>>;
@@ -155,11 +204,13 @@ export declare const ttsChatterboxConfigSchema: z.ZodObject<{
155
204
  "llamacpp-embedding": "llamacpp-embedding";
156
205
  "nmtcpp-translation": "nmtcpp-translation";
157
206
  "onnx-tts": "onnx-tts";
207
+ "parakeet-transcription": "parakeet-transcription";
158
208
  "onnx-ocr": "onnx-ocr";
159
209
  llm: "llm";
160
210
  whisper: "whisper";
161
211
  embeddings: "embeddings";
162
212
  nmt: "nmt";
213
+ parakeet: "parakeet";
163
214
  tts: "tts";
164
215
  ocr: "ocr";
165
216
  }>, z.ZodLiteral<"vad">]>>;
@@ -173,6 +224,8 @@ export declare const ttsSupertonicConfigSchema: z.ZodObject<{
173
224
  es: "es";
174
225
  it: "it";
175
226
  }>;
227
+ ttsSpeed: z.ZodOptional<z.ZodNumber>;
228
+ ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
176
229
  ttsTokenizerSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
177
230
  src: z.ZodString;
178
231
  name: z.ZodOptional<z.ZodString>;
@@ -190,11 +243,13 @@ export declare const ttsSupertonicConfigSchema: z.ZodObject<{
190
243
  "llamacpp-embedding": "llamacpp-embedding";
191
244
  "nmtcpp-translation": "nmtcpp-translation";
192
245
  "onnx-tts": "onnx-tts";
246
+ "parakeet-transcription": "parakeet-transcription";
193
247
  "onnx-ocr": "onnx-ocr";
194
248
  llm: "llm";
195
249
  whisper: "whisper";
196
250
  embeddings: "embeddings";
197
251
  nmt: "nmt";
252
+ parakeet: "parakeet";
198
253
  tts: "tts";
199
254
  ocr: "ocr";
200
255
  }>, z.ZodLiteral<"vad">]>>;
@@ -216,11 +271,13 @@ export declare const ttsSupertonicConfigSchema: z.ZodObject<{
216
271
  "llamacpp-embedding": "llamacpp-embedding";
217
272
  "nmtcpp-translation": "nmtcpp-translation";
218
273
  "onnx-tts": "onnx-tts";
274
+ "parakeet-transcription": "parakeet-transcription";
219
275
  "onnx-ocr": "onnx-ocr";
220
276
  llm: "llm";
221
277
  whisper: "whisper";
222
278
  embeddings: "embeddings";
223
279
  nmt: "nmt";
280
+ parakeet: "parakeet";
224
281
  tts: "tts";
225
282
  ocr: "ocr";
226
283
  }>, z.ZodLiteral<"vad">]>>;
@@ -242,11 +299,13 @@ export declare const ttsSupertonicConfigSchema: z.ZodObject<{
242
299
  "llamacpp-embedding": "llamacpp-embedding";
243
300
  "nmtcpp-translation": "nmtcpp-translation";
244
301
  "onnx-tts": "onnx-tts";
302
+ "parakeet-transcription": "parakeet-transcription";
245
303
  "onnx-ocr": "onnx-ocr";
246
304
  llm: "llm";
247
305
  whisper: "whisper";
248
306
  embeddings: "embeddings";
249
307
  nmt: "nmt";
308
+ parakeet: "parakeet";
250
309
  tts: "tts";
251
310
  ocr: "ocr";
252
311
  }>, z.ZodLiteral<"vad">]>>;
@@ -268,11 +327,13 @@ export declare const ttsSupertonicConfigSchema: z.ZodObject<{
268
327
  "llamacpp-embedding": "llamacpp-embedding";
269
328
  "nmtcpp-translation": "nmtcpp-translation";
270
329
  "onnx-tts": "onnx-tts";
330
+ "parakeet-transcription": "parakeet-transcription";
271
331
  "onnx-ocr": "onnx-ocr";
272
332
  llm: "llm";
273
333
  whisper: "whisper";
274
334
  embeddings: "embeddings";
275
335
  nmt: "nmt";
336
+ parakeet: "parakeet";
276
337
  tts: "tts";
277
338
  ocr: "ocr";
278
339
  }>, z.ZodLiteral<"vad">]>>;
@@ -294,17 +355,17 @@ export declare const ttsSupertonicConfigSchema: z.ZodObject<{
294
355
  "llamacpp-embedding": "llamacpp-embedding";
295
356
  "nmtcpp-translation": "nmtcpp-translation";
296
357
  "onnx-tts": "onnx-tts";
358
+ "parakeet-transcription": "parakeet-transcription";
297
359
  "onnx-ocr": "onnx-ocr";
298
360
  llm: "llm";
299
361
  whisper: "whisper";
300
362
  embeddings: "embeddings";
301
363
  nmt: "nmt";
364
+ parakeet: "parakeet";
302
365
  tts: "tts";
303
366
  ocr: "ocr";
304
367
  }>, z.ZodLiteral<"vad">]>>;
305
368
  }, z.core.$strip>]>;
306
- ttsSpeed: z.ZodOptional<z.ZodNumber>;
307
- ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
308
369
  }, z.core.$strip>;
309
370
  export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
310
371
  ttsEngine: z.ZodLiteral<"chatterbox">;
@@ -331,11 +392,13 @@ export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
331
392
  "llamacpp-embedding": "llamacpp-embedding";
332
393
  "nmtcpp-translation": "nmtcpp-translation";
333
394
  "onnx-tts": "onnx-tts";
395
+ "parakeet-transcription": "parakeet-transcription";
334
396
  "onnx-ocr": "onnx-ocr";
335
397
  llm: "llm";
336
398
  whisper: "whisper";
337
399
  embeddings: "embeddings";
338
400
  nmt: "nmt";
401
+ parakeet: "parakeet";
339
402
  tts: "tts";
340
403
  ocr: "ocr";
341
404
  }>, z.ZodLiteral<"vad">]>>;
@@ -357,11 +420,13 @@ export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
357
420
  "llamacpp-embedding": "llamacpp-embedding";
358
421
  "nmtcpp-translation": "nmtcpp-translation";
359
422
  "onnx-tts": "onnx-tts";
423
+ "parakeet-transcription": "parakeet-transcription";
360
424
  "onnx-ocr": "onnx-ocr";
361
425
  llm: "llm";
362
426
  whisper: "whisper";
363
427
  embeddings: "embeddings";
364
428
  nmt: "nmt";
429
+ parakeet: "parakeet";
365
430
  tts: "tts";
366
431
  ocr: "ocr";
367
432
  }>, z.ZodLiteral<"vad">]>>;
@@ -383,11 +448,13 @@ export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
383
448
  "llamacpp-embedding": "llamacpp-embedding";
384
449
  "nmtcpp-translation": "nmtcpp-translation";
385
450
  "onnx-tts": "onnx-tts";
451
+ "parakeet-transcription": "parakeet-transcription";
386
452
  "onnx-ocr": "onnx-ocr";
387
453
  llm: "llm";
388
454
  whisper: "whisper";
389
455
  embeddings: "embeddings";
390
456
  nmt: "nmt";
457
+ parakeet: "parakeet";
391
458
  tts: "tts";
392
459
  ocr: "ocr";
393
460
  }>, z.ZodLiteral<"vad">]>>;
@@ -409,11 +476,13 @@ export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
409
476
  "llamacpp-embedding": "llamacpp-embedding";
410
477
  "nmtcpp-translation": "nmtcpp-translation";
411
478
  "onnx-tts": "onnx-tts";
479
+ "parakeet-transcription": "parakeet-transcription";
412
480
  "onnx-ocr": "onnx-ocr";
413
481
  llm: "llm";
414
482
  whisper: "whisper";
415
483
  embeddings: "embeddings";
416
484
  nmt: "nmt";
485
+ parakeet: "parakeet";
417
486
  tts: "tts";
418
487
  ocr: "ocr";
419
488
  }>, z.ZodLiteral<"vad">]>>;
@@ -435,11 +504,13 @@ export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
435
504
  "llamacpp-embedding": "llamacpp-embedding";
436
505
  "nmtcpp-translation": "nmtcpp-translation";
437
506
  "onnx-tts": "onnx-tts";
507
+ "parakeet-transcription": "parakeet-transcription";
438
508
  "onnx-ocr": "onnx-ocr";
439
509
  llm: "llm";
440
510
  whisper: "whisper";
441
511
  embeddings: "embeddings";
442
512
  nmt: "nmt";
513
+ parakeet: "parakeet";
443
514
  tts: "tts";
444
515
  ocr: "ocr";
445
516
  }>, z.ZodLiteral<"vad">]>>;
@@ -461,11 +532,13 @@ export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
461
532
  "llamacpp-embedding": "llamacpp-embedding";
462
533
  "nmtcpp-translation": "nmtcpp-translation";
463
534
  "onnx-tts": "onnx-tts";
535
+ "parakeet-transcription": "parakeet-transcription";
464
536
  "onnx-ocr": "onnx-ocr";
465
537
  llm: "llm";
466
538
  whisper: "whisper";
467
539
  embeddings: "embeddings";
468
540
  nmt: "nmt";
541
+ parakeet: "parakeet";
469
542
  tts: "tts";
470
543
  ocr: "ocr";
471
544
  }>, z.ZodLiteral<"vad">]>>;
@@ -478,6 +551,8 @@ export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
478
551
  es: "es";
479
552
  it: "it";
480
553
  }>;
554
+ ttsSpeed: z.ZodOptional<z.ZodNumber>;
555
+ ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
481
556
  ttsTokenizerSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
482
557
  src: z.ZodString;
483
558
  name: z.ZodOptional<z.ZodString>;
@@ -495,11 +570,13 @@ export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
495
570
  "llamacpp-embedding": "llamacpp-embedding";
496
571
  "nmtcpp-translation": "nmtcpp-translation";
497
572
  "onnx-tts": "onnx-tts";
573
+ "parakeet-transcription": "parakeet-transcription";
498
574
  "onnx-ocr": "onnx-ocr";
499
575
  llm: "llm";
500
576
  whisper: "whisper";
501
577
  embeddings: "embeddings";
502
578
  nmt: "nmt";
579
+ parakeet: "parakeet";
503
580
  tts: "tts";
504
581
  ocr: "ocr";
505
582
  }>, z.ZodLiteral<"vad">]>>;
@@ -521,11 +598,13 @@ export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
521
598
  "llamacpp-embedding": "llamacpp-embedding";
522
599
  "nmtcpp-translation": "nmtcpp-translation";
523
600
  "onnx-tts": "onnx-tts";
601
+ "parakeet-transcription": "parakeet-transcription";
524
602
  "onnx-ocr": "onnx-ocr";
525
603
  llm: "llm";
526
604
  whisper: "whisper";
527
605
  embeddings: "embeddings";
528
606
  nmt: "nmt";
607
+ parakeet: "parakeet";
529
608
  tts: "tts";
530
609
  ocr: "ocr";
531
610
  }>, z.ZodLiteral<"vad">]>>;
@@ -547,11 +626,13 @@ export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
547
626
  "llamacpp-embedding": "llamacpp-embedding";
548
627
  "nmtcpp-translation": "nmtcpp-translation";
549
628
  "onnx-tts": "onnx-tts";
629
+ "parakeet-transcription": "parakeet-transcription";
550
630
  "onnx-ocr": "onnx-ocr";
551
631
  llm: "llm";
552
632
  whisper: "whisper";
553
633
  embeddings: "embeddings";
554
634
  nmt: "nmt";
635
+ parakeet: "parakeet";
555
636
  tts: "tts";
556
637
  ocr: "ocr";
557
638
  }>, z.ZodLiteral<"vad">]>>;
@@ -573,11 +654,13 @@ export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
573
654
  "llamacpp-embedding": "llamacpp-embedding";
574
655
  "nmtcpp-translation": "nmtcpp-translation";
575
656
  "onnx-tts": "onnx-tts";
657
+ "parakeet-transcription": "parakeet-transcription";
576
658
  "onnx-ocr": "onnx-ocr";
577
659
  llm: "llm";
578
660
  whisper: "whisper";
579
661
  embeddings: "embeddings";
580
662
  nmt: "nmt";
663
+ parakeet: "parakeet";
581
664
  tts: "tts";
582
665
  ocr: "ocr";
583
666
  }>, z.ZodLiteral<"vad">]>>;
@@ -599,78 +682,17 @@ export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
599
682
  "llamacpp-embedding": "llamacpp-embedding";
600
683
  "nmtcpp-translation": "nmtcpp-translation";
601
684
  "onnx-tts": "onnx-tts";
685
+ "parakeet-transcription": "parakeet-transcription";
602
686
  "onnx-ocr": "onnx-ocr";
603
687
  llm: "llm";
604
688
  whisper: "whisper";
605
689
  embeddings: "embeddings";
606
690
  nmt: "nmt";
691
+ parakeet: "parakeet";
607
692
  tts: "tts";
608
693
  ocr: "ocr";
609
694
  }>, z.ZodLiteral<"vad">]>>;
610
695
  }, z.core.$strip>]>;
611
- ttsSpeed: z.ZodOptional<z.ZodNumber>;
612
- ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
613
- }, z.core.$strip>]>;
614
- export declare const ttsChatterboxRequestConfigSchema: z.ZodObject<{
615
- ttsEngine: z.ZodLiteral<"chatterbox">;
616
- language: z.ZodEnum<{
617
- en: "en";
618
- de: "de";
619
- es: "es";
620
- it: "it";
621
- }>;
622
- ttsTokenizerSrc: z.ZodString;
623
- ttsSpeechEncoderSrc: z.ZodString;
624
- ttsEmbedTokensSrc: z.ZodString;
625
- ttsConditionalDecoderSrc: z.ZodString;
626
- ttsLanguageModelSrc: z.ZodString;
627
- referenceAudioSrc: z.ZodString;
628
- }, z.core.$strip>;
629
- export declare const ttsSupertonicRequestConfigSchema: z.ZodObject<{
630
- ttsEngine: z.ZodLiteral<"supertonic">;
631
- language: z.ZodEnum<{
632
- en: "en";
633
- de: "de";
634
- es: "es";
635
- it: "it";
636
- }>;
637
- ttsTokenizerSrc: z.ZodString;
638
- ttsTextEncoderSrc: z.ZodString;
639
- ttsLatentDenoiserSrc: z.ZodString;
640
- ttsVoiceDecoderSrc: z.ZodString;
641
- ttsVoiceSrc: z.ZodString;
642
- ttsSpeed: z.ZodOptional<z.ZodNumber>;
643
- ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
644
- }, z.core.$strip>;
645
- export declare const ttsRequestConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
646
- ttsEngine: z.ZodLiteral<"chatterbox">;
647
- language: z.ZodEnum<{
648
- en: "en";
649
- de: "de";
650
- es: "es";
651
- it: "it";
652
- }>;
653
- ttsTokenizerSrc: z.ZodString;
654
- ttsSpeechEncoderSrc: z.ZodString;
655
- ttsEmbedTokensSrc: z.ZodString;
656
- ttsConditionalDecoderSrc: z.ZodString;
657
- ttsLanguageModelSrc: z.ZodString;
658
- referenceAudioSrc: z.ZodString;
659
- }, z.core.$strip>, z.ZodObject<{
660
- ttsEngine: z.ZodLiteral<"supertonic">;
661
- language: z.ZodEnum<{
662
- en: "en";
663
- de: "de";
664
- es: "es";
665
- it: "it";
666
- }>;
667
- ttsTokenizerSrc: z.ZodString;
668
- ttsTextEncoderSrc: z.ZodString;
669
- ttsLatentDenoiserSrc: z.ZodString;
670
- ttsVoiceDecoderSrc: z.ZodString;
671
- ttsVoiceSrc: z.ZodString;
672
- ttsSpeed: z.ZodOptional<z.ZodNumber>;
673
- ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
674
696
  }, z.core.$strip>]>;
675
697
  export declare const ttsClientParamsSchema: z.ZodObject<{
676
698
  modelId: z.ZodString;
@@ -694,9 +716,9 @@ export type TtsLanguage = (typeof TTS_LANGUAGES)[number];
694
716
  export type TtsChatterboxConfig = z.infer<typeof ttsChatterboxConfigSchema>;
695
717
  export type TtsSupertonicConfig = z.infer<typeof ttsSupertonicConfigSchema>;
696
718
  export type TtsConfig = z.infer<typeof ttsConfigSchema>;
697
- export type TtsChatterboxRequestConfig = z.infer<typeof ttsChatterboxRequestConfigSchema>;
698
- export type TtsSupertonicRequestConfig = z.infer<typeof ttsSupertonicRequestConfigSchema>;
699
- export type TtsRequestConfig = z.infer<typeof ttsRequestConfigSchema>;
719
+ export type TtsChatterboxRuntimeConfig = z.infer<typeof ttsChatterboxRuntimeConfigSchema>;
720
+ export type TtsSupertonicRuntimeConfig = z.infer<typeof ttsSupertonicRuntimeConfigSchema>;
721
+ export type TtsRuntimeConfig = z.infer<typeof ttsRuntimeConfigSchema>;
700
722
  export type TtsClientParams = z.infer<typeof ttsClientParamsSchema>;
701
723
  export type TtsRequest = z.infer<typeof ttsRequestSchema>;
702
724
  export type TtsResponse = z.infer<typeof ttsResponseSchema>;
@@ -7,9 +7,22 @@ export const TTS_LANGUAGES = [
7
7
  "de", // German
8
8
  "it", // Italian
9
9
  ];
10
- export const ttsChatterboxConfigSchema = z.object({
10
+ const ttsLanguageSchema = z.enum(TTS_LANGUAGES);
11
+ export const ttsChatterboxRuntimeConfigSchema = z.object({
11
12
  ttsEngine: z.literal("chatterbox"),
12
- language: z.enum(TTS_LANGUAGES),
13
+ language: ttsLanguageSchema,
14
+ });
15
+ export const ttsSupertonicRuntimeConfigSchema = z.object({
16
+ ttsEngine: z.literal("supertonic"),
17
+ language: ttsLanguageSchema,
18
+ ttsSpeed: z.number().optional(),
19
+ ttsNumInferenceSteps: z.number().optional(),
20
+ });
21
+ export const ttsRuntimeConfigSchema = z.union([
22
+ ttsChatterboxRuntimeConfigSchema,
23
+ ttsSupertonicRuntimeConfigSchema,
24
+ ]);
25
+ export const ttsChatterboxConfigSchema = ttsChatterboxRuntimeConfigSchema.extend({
13
26
  ttsTokenizerSrc: modelSrcInputSchema,
14
27
  ttsSpeechEncoderSrc: modelSrcInputSchema,
15
28
  ttsEmbedTokensSrc: modelSrcInputSchema,
@@ -17,47 +30,17 @@ export const ttsChatterboxConfigSchema = z.object({
17
30
  ttsLanguageModelSrc: modelSrcInputSchema,
18
31
  referenceAudioSrc: modelSrcInputSchema,
19
32
  });
20
- export const ttsSupertonicConfigSchema = z.object({
21
- ttsEngine: z.literal("supertonic"),
22
- language: z.enum(TTS_LANGUAGES),
33
+ export const ttsSupertonicConfigSchema = ttsSupertonicRuntimeConfigSchema.extend({
23
34
  ttsTokenizerSrc: modelSrcInputSchema,
24
35
  ttsTextEncoderSrc: modelSrcInputSchema,
25
36
  ttsLatentDenoiserSrc: modelSrcInputSchema,
26
37
  ttsVoiceDecoderSrc: modelSrcInputSchema,
27
38
  ttsVoiceSrc: modelSrcInputSchema,
28
- ttsSpeed: z.number().optional(),
29
- ttsNumInferenceSteps: z.number().optional(),
30
39
  });
31
40
  export const ttsConfigSchema = z.union([
32
41
  ttsChatterboxConfigSchema,
33
42
  ttsSupertonicConfigSchema,
34
43
  ]);
35
- // Request-level schemas with string sources (after ModelSrcInput is resolved to string)
36
- export const ttsChatterboxRequestConfigSchema = z.object({
37
- ttsEngine: z.literal("chatterbox"),
38
- language: z.enum(TTS_LANGUAGES),
39
- ttsTokenizerSrc: z.string(),
40
- ttsSpeechEncoderSrc: z.string(),
41
- ttsEmbedTokensSrc: z.string(),
42
- ttsConditionalDecoderSrc: z.string(),
43
- ttsLanguageModelSrc: z.string(),
44
- referenceAudioSrc: z.string(),
45
- });
46
- export const ttsSupertonicRequestConfigSchema = z.object({
47
- ttsEngine: z.literal("supertonic"),
48
- language: z.enum(TTS_LANGUAGES),
49
- ttsTokenizerSrc: z.string(),
50
- ttsTextEncoderSrc: z.string(),
51
- ttsLatentDenoiserSrc: z.string(),
52
- ttsVoiceDecoderSrc: z.string(),
53
- ttsVoiceSrc: z.string(),
54
- ttsSpeed: z.number().optional(),
55
- ttsNumInferenceSteps: z.number().optional(),
56
- });
57
- export const ttsRequestConfigSchema = z.union([
58
- ttsChatterboxRequestConfigSchema,
59
- ttsSupertonicRequestConfigSchema,
60
- ]);
61
44
  export const ttsClientParamsSchema = z.object({
62
45
  modelId: z.string(),
63
46
  inputType: z.string().default("text"),