@qvac/sdk 0.11.0 → 0.12.1

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 (211) hide show
  1. package/NOTICE +1423 -43
  2. package/README.md +2 -0
  3. package/dist/client/api/classify.d.ts +23 -0
  4. package/dist/client/api/classify.js +43 -0
  5. package/dist/client/api/get-model-info.d.ts +1 -1
  6. package/dist/client/api/index.d.ts +4 -0
  7. package/dist/client/api/index.js +4 -0
  8. package/dist/client/api/load-model.js +24 -1
  9. package/dist/client/api/transcribe.d.ts +3 -0
  10. package/dist/client/api/transcribe.js +13 -2
  11. package/dist/client/api/video.d.ts +64 -0
  12. package/dist/client/api/video.js +147 -0
  13. package/dist/client/api/vla-helpers.d.ts +45 -0
  14. package/dist/client/api/vla-helpers.js +130 -0
  15. package/dist/client/api/vla.d.ts +65 -0
  16. package/dist/client/api/vla.js +109 -0
  17. package/dist/client/config-loader/resolve-config.node.d.ts +16 -0
  18. package/dist/client/config-loader/resolve-config.node.js +77 -23
  19. package/dist/client/plugins-factory.d.ts +8 -0
  20. package/dist/client/plugins-factory.js +10 -0
  21. package/dist/client/rpc/bare-client.d.ts +1 -0
  22. package/dist/client/rpc/bare-client.js +66 -66
  23. package/dist/client/rpc/ensure-worker-ready.d.ts +9 -0
  24. package/dist/client/rpc/ensure-worker-ready.js +17 -0
  25. package/dist/client/rpc/expo-rpc-client.d.ts +1 -0
  26. package/dist/client/rpc/expo-rpc-client.js +4 -0
  27. package/dist/client/rpc/node-rpc-client.d.ts +1 -0
  28. package/dist/client/rpc/node-rpc-client.js +97 -5
  29. package/dist/client/rpc/rpc-client.js +115 -8
  30. package/dist/client/rpc/rpc-error.js +14 -1
  31. package/dist/commands/bundle/bare-pack.d.ts +13 -0
  32. package/dist/commands/bundle/bare-pack.js +54 -0
  33. package/dist/commands/bundle/constants.d.ts +7 -0
  34. package/dist/commands/bundle/constants.js +23 -0
  35. package/dist/commands/bundle/entry-gen.d.ts +2 -0
  36. package/dist/commands/bundle/entry-gen.js +48 -0
  37. package/dist/commands/bundle/index.d.ts +20 -0
  38. package/dist/commands/bundle/index.js +141 -0
  39. package/dist/commands/bundle/manifest.d.ts +21 -0
  40. package/dist/commands/bundle/manifest.js +193 -0
  41. package/dist/commands/bundle/plugins.d.ts +9 -0
  42. package/dist/commands/bundle/plugins.js +51 -0
  43. package/dist/commands/index.d.ts +5 -0
  44. package/dist/commands/index.js +2 -0
  45. package/dist/commands/verify/abi.d.ts +52 -0
  46. package/dist/commands/verify/abi.js +125 -0
  47. package/dist/commands/verify/addon-source.d.ts +43 -0
  48. package/dist/commands/verify/addon-source.js +75 -0
  49. package/dist/commands/verify/bundle-source.d.ts +12 -0
  50. package/dist/commands/verify/bundle-source.js +53 -0
  51. package/dist/commands/verify/index.d.ts +59 -0
  52. package/dist/commands/verify/index.js +371 -0
  53. package/dist/commands/verify/node-modules-source.d.ts +11 -0
  54. package/dist/commands/verify/node-modules-source.js +76 -0
  55. package/dist/commands/verify/prebuilds.d.ts +15 -0
  56. package/dist/commands/verify/prebuilds.js +39 -0
  57. package/dist/examples/classification/classify-image.d.ts +2 -0
  58. package/dist/examples/classification/classify-image.js +23 -0
  59. package/dist/examples/diffusion-txt2vid.d.ts +2 -0
  60. package/dist/examples/diffusion-txt2vid.js +81 -0
  61. package/dist/examples/plugins.js +3 -7
  62. package/dist/examples/tools/llamacpp-tools-gemma4.js +2 -5
  63. package/dist/examples/tools/llamacpp-tools-qwen35.js +2 -3
  64. package/dist/examples/transcription/parakeet-ctc-filesystem.js +18 -12
  65. package/dist/examples/transcription/parakeet-microphone-record.js +7 -12
  66. package/dist/examples/transcription/parakeet-microphone-stream.d.ts +2 -0
  67. package/dist/examples/transcription/parakeet-microphone-stream.js +79 -0
  68. package/dist/examples/transcription/parakeet-sortformer-streaming.d.ts +2 -0
  69. package/dist/examples/transcription/parakeet-sortformer-streaming.js +121 -0
  70. package/dist/examples/transcription/parakeet-sortformer.js +58 -55
  71. package/dist/examples/transcription/parakeet-tdt-filesystem.js +18 -15
  72. package/dist/examples/transcription/whispercpp-microphone-conversation.js +6 -1
  73. package/dist/examples/tts/chatterbox.js +7 -16
  74. package/dist/examples/tts/llm-to-tts-streaming.js +3 -10
  75. package/dist/examples/tts/supertonic-multilingual.js +4 -12
  76. package/dist/examples/tts/supertonic.js +5 -12
  77. package/dist/examples/vla-smolvla.d.ts +2 -0
  78. package/dist/examples/vla-smolvla.js +75 -0
  79. package/dist/examples/voice-assistant/voice-assistant.js +3 -10
  80. package/dist/expo/plugins/find-in-ancestor-node-modules.d.ts +14 -0
  81. package/dist/expo/plugins/find-in-ancestor-node-modules.js +24 -0
  82. package/dist/expo/plugins/get-project-root.d.ts +22 -0
  83. package/dist/expo/plugins/get-project-root.js +23 -0
  84. package/dist/expo/plugins/index.d.ts +4 -0
  85. package/dist/expo/plugins/index.js +4 -0
  86. package/dist/expo/plugins/resolve-sdk-package-dir.d.ts +9 -2
  87. package/dist/expo/plugins/resolve-sdk-package-dir.js +38 -11
  88. package/dist/expo/plugins/withDeviceInfo.js +2 -1
  89. package/dist/expo/plugins/withMobileBundle.d.ts +2 -17
  90. package/dist/expo/plugins/withMobileBundle.js +31 -85
  91. package/dist/index.d.ts +5 -3
  92. package/dist/index.js +9 -3
  93. package/dist/models/registry/models.d.ts +3711 -368
  94. package/dist/models/registry/models.js +7066 -3562
  95. package/dist/models/update-models/naming.js +22 -8
  96. package/dist/pear/pre.js +6 -2
  97. package/dist/schemas/cancel.d.ts +5 -0
  98. package/dist/schemas/cancel.js +1 -0
  99. package/dist/schemas/classification.d.ts +53 -0
  100. package/dist/schemas/classification.js +33 -0
  101. package/dist/schemas/common.d.ts +1001 -1739
  102. package/dist/schemas/common.js +6 -1
  103. package/dist/schemas/completion-event.d.ts +3 -0
  104. package/dist/schemas/completion-event.js +1 -0
  105. package/dist/schemas/completion-stream.d.ts +1 -0
  106. package/dist/schemas/download-asset.d.ts +13 -3
  107. package/dist/schemas/engine-addon-map.d.ts +3 -0
  108. package/dist/schemas/engine-addon-map.js +18 -5
  109. package/dist/schemas/get-model-info.d.ts +4 -0
  110. package/dist/schemas/get-model-info.js +2 -0
  111. package/dist/schemas/index.d.ts +3 -0
  112. package/dist/schemas/index.js +3 -0
  113. package/dist/schemas/llamacpp-config.d.ts +12 -2
  114. package/dist/schemas/load-model.d.ts +6152 -10764
  115. package/dist/schemas/load-model.js +75 -7
  116. package/dist/schemas/model-src-utils.d.ts +22 -2
  117. package/dist/schemas/model-types.d.ts +46 -6
  118. package/dist/schemas/model-types.js +28 -5
  119. package/dist/schemas/ocr.d.ts +5 -0
  120. package/dist/schemas/plugin.d.ts +20 -5
  121. package/dist/schemas/plugin.js +21 -4
  122. package/dist/schemas/registry.d.ts +44 -0
  123. package/dist/schemas/registry.js +5 -0
  124. package/dist/schemas/reload-config.d.ts +22 -2
  125. package/dist/schemas/sdcpp-config.d.ts +316 -1
  126. package/dist/schemas/sdcpp-config.js +232 -33
  127. package/dist/schemas/sdk-config.d.ts +654 -1324
  128. package/dist/schemas/sdk-config.js +16 -1
  129. package/dist/schemas/sdk-errors-client.d.ts +7 -0
  130. package/dist/schemas/sdk-errors-client.js +36 -1
  131. package/dist/schemas/sdk-errors-server.d.ts +3 -1
  132. package/dist/schemas/sdk-errors-server.js +20 -5
  133. package/dist/schemas/text-to-speech.d.ts +79 -696
  134. package/dist/schemas/text-to-speech.js +46 -19
  135. package/dist/schemas/transcription-config.d.ts +37 -226
  136. package/dist/schemas/transcription-config.js +76 -15
  137. package/dist/schemas/transcription.d.ts +63 -11
  138. package/dist/schemas/transcription.js +56 -2
  139. package/dist/schemas/translation-config.d.ts +60 -10
  140. package/dist/schemas/vla.d.ts +87 -0
  141. package/dist/schemas/vla.js +108 -0
  142. package/dist/scripts/run-unit-tests.js +8 -2
  143. package/dist/server/bare/ops/transcribe.d.ts +2 -1
  144. package/dist/server/bare/ops/transcribe.js +59 -34
  145. package/dist/server/bare/plugins/ggml-classification/ops/classify.d.ts +6 -0
  146. package/dist/server/bare/plugins/ggml-classification/ops/classify.js +30 -0
  147. package/dist/server/bare/plugins/ggml-classification/plugin.d.ts +32 -0
  148. package/dist/server/bare/plugins/ggml-classification/plugin.js +55 -0
  149. package/dist/server/bare/plugins/ggml-vla/ops/vla-hparams.d.ts +3 -0
  150. package/dist/server/bare/plugins/ggml-vla/ops/vla-hparams.js +12 -0
  151. package/dist/server/bare/plugins/ggml-vla/ops/vla-run.d.ts +3 -0
  152. package/dist/server/bare/plugins/ggml-vla/ops/vla-run.js +75 -0
  153. package/dist/server/bare/plugins/ggml-vla/plugin.d.ts +54 -0
  154. package/dist/server/bare/plugins/ggml-vla/plugin.js +57 -0
  155. package/dist/server/bare/plugins/index.d.ts +3 -1
  156. package/dist/server/bare/plugins/index.js +3 -1
  157. package/dist/server/bare/plugins/llamacpp-completion/ops/completion-stream.js +3 -0
  158. package/dist/server/bare/plugins/llamacpp-completion/ops/context-overflow.d.ts +29 -0
  159. package/dist/server/bare/plugins/llamacpp-completion/ops/context-overflow.js +65 -0
  160. package/dist/server/bare/plugins/llamacpp-completion/plugin.d.ts +6 -0
  161. package/dist/server/bare/plugins/llamacpp-completion/plugin.js +26 -0
  162. package/dist/server/bare/plugins/nmtcpp-translation/plugin.d.ts +25 -0
  163. package/dist/server/bare/plugins/onnx-ocr/plugin.d.ts +5 -0
  164. package/dist/server/bare/plugins/parakeet-transcription/plugin.d.ts +67 -246
  165. package/dist/server/bare/plugins/parakeet-transcription/plugin.js +66 -97
  166. package/dist/server/bare/plugins/sdcpp-generation/ops/video.d.ts +3 -0
  167. package/dist/server/bare/plugins/sdcpp-generation/ops/video.js +199 -0
  168. package/dist/server/bare/plugins/sdcpp-generation/plugin.d.ts +200 -2
  169. package/dist/server/bare/plugins/sdcpp-generation/plugin.js +61 -8
  170. package/dist/server/bare/plugins/{onnx-tts → tts-ggml}/ops/text-to-speech-stream.js +2 -15
  171. package/dist/server/bare/plugins/{onnx-tts → tts-ggml}/ops/text-to-speech.js +2 -1
  172. package/dist/server/bare/plugins/tts-ggml/plugin.d.ts +75 -0
  173. package/dist/server/bare/plugins/tts-ggml/plugin.js +189 -0
  174. package/dist/server/bare/plugins/whispercpp-transcription/plugin.d.ts +29 -4
  175. package/dist/server/bare/plugins/whispercpp-transcription/plugin.js +12 -1
  176. package/dist/server/bare/registry/model-config-utils.js +7 -0
  177. package/dist/server/bare/runtime/request-context.d.ts +1 -1
  178. package/dist/server/bare/runtime-lifecycle.js +9 -9
  179. package/dist/server/bare/types/addon-responses.d.ts +1 -0
  180. package/dist/server/env.js +4 -1
  181. package/dist/server/plugins/index.d.ts +1 -1
  182. package/dist/server/plugins/index.js +1 -1
  183. package/dist/server/plugins/registry.d.ts +1 -0
  184. package/dist/server/plugins/registry.js +5 -0
  185. package/dist/server/rpc/handler-registry.js +4 -0
  186. package/dist/server/rpc/handler-selection.d.ts +9 -0
  187. package/dist/server/rpc/handler-selection.js +14 -0
  188. package/dist/server/rpc/handler-utils.js +5 -13
  189. package/dist/server/rpc/handlers/classify.d.ts +3 -0
  190. package/dist/server/rpc/handlers/classify.js +4 -0
  191. package/dist/server/rpc/handlers/index.d.ts +2 -0
  192. package/dist/server/rpc/handlers/index.js +2 -0
  193. package/dist/server/rpc/handlers/load-model/handler.js +4 -1
  194. package/dist/server/rpc/handlers/load-model/resolve.js +9 -0
  195. package/dist/server/rpc/handlers/video-stream.d.ts +3 -0
  196. package/dist/server/rpc/handlers/video-stream.js +4 -0
  197. package/dist/server/rpc/profiling/operation-metrics.js +23 -0
  198. package/dist/server/utils/config-transform.js +1 -1
  199. package/dist/server/utils/tools/parsers/qwen35.js +7 -3
  200. package/dist/server/worker.js +14 -11
  201. package/dist/utils/errors-client.d.ts +23 -0
  202. package/dist/utils/errors-client.js +40 -0
  203. package/dist/utils/errors-server.d.ts +30 -2
  204. package/dist/utils/errors-server.js +51 -3
  205. package/package.json +66 -80
  206. package/dist/server/bare/plugins/onnx-tts/plugin.d.ts +0 -490
  207. package/dist/server/bare/plugins/onnx-tts/plugin.js +0 -272
  208. package/dist/server/bare/plugins/onnx-tts/wav-helper.d.ts +0 -14
  209. package/dist/server/bare/plugins/onnx-tts/wav-helper.js +0 -107
  210. /package/dist/server/bare/plugins/{onnx-tts → tts-ggml}/ops/text-to-speech-stream.d.ts +0 -0
  211. /package/dist/server/bare/plugins/{onnx-tts → tts-ggml}/ops/text-to-speech.d.ts +0 -0
package/dist/schemas/text-to-speech.d.ts CHANGED
@@ -8,6 +8,8 @@ export declare const ttsChatterboxRuntimeConfigSchema: z.ZodObject<{
8
8
  es: "es";
9
9
  it: "it";
10
10
  }>;
11
+ voice: z.ZodOptional<z.ZodString>;
12
+ useGPU: z.ZodOptional<z.ZodBoolean>;
11
13
  }, z.core.$strip>;
12
14
  export declare const ttsSupertonicRuntimeConfigSchema: z.ZodObject<{
13
15
  ttsEngine: z.ZodLiteral<"supertonic">;
@@ -17,9 +19,10 @@ export declare const ttsSupertonicRuntimeConfigSchema: z.ZodObject<{
17
19
  es: "es";
18
20
  it: "it";
19
21
  }>;
22
+ voice: z.ZodOptional<z.ZodString>;
20
23
  ttsSpeed: z.ZodOptional<z.ZodNumber>;
21
24
  ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
22
- ttsSupertonicMultilingual: z.ZodOptional<z.ZodBoolean>;
25
+ useGPU: z.ZodOptional<z.ZodBoolean>;
23
26
  }, z.core.$strip>;
24
27
  export declare const ttsRuntimeConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
25
28
  ttsEngine: z.ZodLiteral<"chatterbox">;
@@ -29,6 +32,8 @@ export declare const ttsRuntimeConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
29
32
  es: "es";
30
33
  it: "it";
31
34
  }>;
35
+ voice: z.ZodOptional<z.ZodString>;
36
+ useGPU: z.ZodOptional<z.ZodBoolean>;
32
37
  }, z.core.$strip>, z.ZodObject<{
33
38
  ttsEngine: z.ZodLiteral<"supertonic">;
34
39
  language: z.ZodEnum<{
@@ -37,11 +42,12 @@ export declare const ttsRuntimeConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
37
42
  es: "es";
38
43
  it: "it";
39
44
  }>;
45
+ voice: z.ZodOptional<z.ZodString>;
40
46
  ttsSpeed: z.ZodOptional<z.ZodNumber>;
41
47
  ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
42
- ttsSupertonicMultilingual: z.ZodOptional<z.ZodBoolean>;
48
+ useGPU: z.ZodOptional<z.ZodBoolean>;
43
49
  }, z.core.$strip>]>;
44
- export declare const ttsChatterboxConfigSchema: z.ZodObject<{
50
+ export declare const ttsChatterboxLoadConfigSchema: z.ZodObject<{
45
51
  ttsEngine: z.ZodLiteral<"chatterbox">;
46
52
  language: z.ZodEnum<{
47
53
  en: "en";
@@ -49,7 +55,9 @@ export declare const ttsChatterboxConfigSchema: z.ZodObject<{
49
55
  es: "es";
50
56
  it: "it";
51
57
  }>;
52
- ttsTokenizerSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
58
+ voice: z.ZodOptional<z.ZodString>;
59
+ useGPU: z.ZodOptional<z.ZodBoolean>;
60
+ s3genModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
53
61
  src: z.ZodString;
54
62
  name: z.ZodOptional<z.ZodString>;
55
63
  modelId: z.ZodOptional<z.ZodString>;
@@ -65,21 +73,26 @@ export declare const ttsChatterboxConfigSchema: z.ZodObject<{
65
73
  diffusion: "diffusion";
66
74
  tts: "tts";
67
75
  ocr: "ocr";
76
+ vla: "vla";
68
77
  "llamacpp-completion": "llamacpp-completion";
69
78
  "whispercpp-transcription": "whispercpp-transcription";
70
79
  "llamacpp-embedding": "llamacpp-embedding";
71
80
  "nmtcpp-translation": "nmtcpp-translation";
72
81
  "onnx-tts": "onnx-tts";
82
+ "tts-ggml": "tts-ggml";
73
83
  "parakeet-transcription": "parakeet-transcription";
74
84
  "onnx-ocr": "onnx-ocr";
75
85
  "sdcpp-generation": "sdcpp-generation";
86
+ "ggml-vla": "ggml-vla";
87
+ "ggml-classification": "ggml-classification";
76
88
  llm: "llm";
77
89
  whisper: "whisper";
78
90
  nmt: "nmt";
79
91
  parakeet: "parakeet";
92
+ classification: "classification";
80
93
  }>, z.ZodLiteral<"vad">]>>;
81
- }, z.core.$strip>]>;
82
- ttsSpeechEncoderSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
94
+ }, z.core.$strip>]>>;
95
+ referenceAudioSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
83
96
  src: z.ZodString;
84
97
  name: z.ZodOptional<z.ZodString>;
85
98
  modelId: z.ZodOptional<z.ZodString>;
@@ -95,492 +108,50 @@ export declare const ttsChatterboxConfigSchema: z.ZodObject<{
95
108
  diffusion: "diffusion";
96
109
  tts: "tts";
97
110
  ocr: "ocr";
111
+ vla: "vla";
98
112
  "llamacpp-completion": "llamacpp-completion";
99
113
  "whispercpp-transcription": "whispercpp-transcription";
100
114
  "llamacpp-embedding": "llamacpp-embedding";
101
115
  "nmtcpp-translation": "nmtcpp-translation";
102
116
  "onnx-tts": "onnx-tts";
117
+ "tts-ggml": "tts-ggml";
103
118
  "parakeet-transcription": "parakeet-transcription";
104
119
  "onnx-ocr": "onnx-ocr";
105
120
  "sdcpp-generation": "sdcpp-generation";
121
+ "ggml-vla": "ggml-vla";
122
+ "ggml-classification": "ggml-classification";
106
123
  llm: "llm";
107
124
  whisper: "whisper";
108
125
  nmt: "nmt";
109
126
  parakeet: "parakeet";
127
+ classification: "classification";
110
128
  }>, z.ZodLiteral<"vad">]>>;
111
- }, z.core.$strip>]>;
112
- ttsEmbedTokensSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
113
- src: z.ZodString;
114
- name: z.ZodOptional<z.ZodString>;
115
- modelId: z.ZodOptional<z.ZodString>;
116
- registryPath: z.ZodOptional<z.ZodString>;
117
- registrySource: z.ZodOptional<z.ZodString>;
118
- blobCoreKey: z.ZodOptional<z.ZodString>;
119
- blobIndex: z.ZodOptional<z.ZodNumber>;
120
- engine: z.ZodOptional<z.ZodString>;
121
- expectedSize: z.ZodOptional<z.ZodNumber>;
122
- sha256Checksum: z.ZodOptional<z.ZodString>;
123
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
124
- embeddings: "embeddings";
125
- diffusion: "diffusion";
126
- tts: "tts";
127
- ocr: "ocr";
128
- "llamacpp-completion": "llamacpp-completion";
129
- "whispercpp-transcription": "whispercpp-transcription";
130
- "llamacpp-embedding": "llamacpp-embedding";
131
- "nmtcpp-translation": "nmtcpp-translation";
132
- "onnx-tts": "onnx-tts";
133
- "parakeet-transcription": "parakeet-transcription";
134
- "onnx-ocr": "onnx-ocr";
135
- "sdcpp-generation": "sdcpp-generation";
136
- llm: "llm";
137
- whisper: "whisper";
138
- nmt: "nmt";
139
- parakeet: "parakeet";
140
- }>, z.ZodLiteral<"vad">]>>;
141
- }, z.core.$strip>]>;
142
- ttsConditionalDecoderSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
143
- src: z.ZodString;
144
- name: z.ZodOptional<z.ZodString>;
145
- modelId: z.ZodOptional<z.ZodString>;
146
- registryPath: z.ZodOptional<z.ZodString>;
147
- registrySource: z.ZodOptional<z.ZodString>;
148
- blobCoreKey: z.ZodOptional<z.ZodString>;
149
- blobIndex: z.ZodOptional<z.ZodNumber>;
150
- engine: z.ZodOptional<z.ZodString>;
151
- expectedSize: z.ZodOptional<z.ZodNumber>;
152
- sha256Checksum: z.ZodOptional<z.ZodString>;
153
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
154
- embeddings: "embeddings";
155
- diffusion: "diffusion";
156
- tts: "tts";
157
- ocr: "ocr";
158
- "llamacpp-completion": "llamacpp-completion";
159
- "whispercpp-transcription": "whispercpp-transcription";
160
- "llamacpp-embedding": "llamacpp-embedding";
161
- "nmtcpp-translation": "nmtcpp-translation";
162
- "onnx-tts": "onnx-tts";
163
- "parakeet-transcription": "parakeet-transcription";
164
- "onnx-ocr": "onnx-ocr";
165
- "sdcpp-generation": "sdcpp-generation";
166
- llm: "llm";
167
- whisper: "whisper";
168
- nmt: "nmt";
169
- parakeet: "parakeet";
170
- }>, z.ZodLiteral<"vad">]>>;
171
- }, z.core.$strip>]>;
172
- ttsLanguageModelSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
173
- src: z.ZodString;
174
- name: z.ZodOptional<z.ZodString>;
175
- modelId: z.ZodOptional<z.ZodString>;
176
- registryPath: z.ZodOptional<z.ZodString>;
177
- registrySource: z.ZodOptional<z.ZodString>;
178
- blobCoreKey: z.ZodOptional<z.ZodString>;
179
- blobIndex: z.ZodOptional<z.ZodNumber>;
180
- engine: z.ZodOptional<z.ZodString>;
181
- expectedSize: z.ZodOptional<z.ZodNumber>;
182
- sha256Checksum: z.ZodOptional<z.ZodString>;
183
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
184
- embeddings: "embeddings";
185
- diffusion: "diffusion";
186
- tts: "tts";
187
- ocr: "ocr";
188
- "llamacpp-completion": "llamacpp-completion";
189
- "whispercpp-transcription": "whispercpp-transcription";
190
- "llamacpp-embedding": "llamacpp-embedding";
191
- "nmtcpp-translation": "nmtcpp-translation";
192
- "onnx-tts": "onnx-tts";
193
- "parakeet-transcription": "parakeet-transcription";
194
- "onnx-ocr": "onnx-ocr";
195
- "sdcpp-generation": "sdcpp-generation";
196
- llm: "llm";
197
- whisper: "whisper";
198
- nmt: "nmt";
199
- parakeet: "parakeet";
200
- }>, z.ZodLiteral<"vad">]>>;
201
- }, z.core.$strip>]>;
202
- referenceAudioSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
203
- src: z.ZodString;
204
- name: z.ZodOptional<z.ZodString>;
205
- modelId: z.ZodOptional<z.ZodString>;
206
- registryPath: z.ZodOptional<z.ZodString>;
207
- registrySource: z.ZodOptional<z.ZodString>;
208
- blobCoreKey: z.ZodOptional<z.ZodString>;
209
- blobIndex: z.ZodOptional<z.ZodNumber>;
210
- engine: z.ZodOptional<z.ZodString>;
211
- expectedSize: z.ZodOptional<z.ZodNumber>;
212
- sha256Checksum: z.ZodOptional<z.ZodString>;
213
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
214
- embeddings: "embeddings";
215
- diffusion: "diffusion";
216
- tts: "tts";
217
- ocr: "ocr";
218
- "llamacpp-completion": "llamacpp-completion";
219
- "whispercpp-transcription": "whispercpp-transcription";
220
- "llamacpp-embedding": "llamacpp-embedding";
221
- "nmtcpp-translation": "nmtcpp-translation";
222
- "onnx-tts": "onnx-tts";
223
- "parakeet-transcription": "parakeet-transcription";
224
- "onnx-ocr": "onnx-ocr";
225
- "sdcpp-generation": "sdcpp-generation";
226
- llm: "llm";
227
- whisper: "whisper";
228
- nmt: "nmt";
229
- parakeet: "parakeet";
230
- }>, z.ZodLiteral<"vad">]>>;
231
- }, z.core.$strip>]>;
232
- }, z.core.$strip>;
233
- export declare const ttsSupertonicConfigSchema: z.ZodObject<{
234
- ttsEngine: z.ZodLiteral<"supertonic">;
235
- language: z.ZodEnum<{
236
- en: "en";
237
- de: "de";
238
- es: "es";
239
- it: "it";
240
- }>;
241
- ttsSpeed: z.ZodOptional<z.ZodNumber>;
242
- ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
243
- ttsSupertonicMultilingual: z.ZodOptional<z.ZodBoolean>;
244
- ttsTextEncoderSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
245
- src: z.ZodString;
246
- name: z.ZodOptional<z.ZodString>;
247
- modelId: z.ZodOptional<z.ZodString>;
248
- registryPath: z.ZodOptional<z.ZodString>;
249
- registrySource: z.ZodOptional<z.ZodString>;
250
- blobCoreKey: z.ZodOptional<z.ZodString>;
251
- blobIndex: z.ZodOptional<z.ZodNumber>;
252
- engine: z.ZodOptional<z.ZodString>;
253
- expectedSize: z.ZodOptional<z.ZodNumber>;
254
- sha256Checksum: z.ZodOptional<z.ZodString>;
255
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
256
- embeddings: "embeddings";
257
- diffusion: "diffusion";
258
- tts: "tts";
259
- ocr: "ocr";
260
- "llamacpp-completion": "llamacpp-completion";
261
- "whispercpp-transcription": "whispercpp-transcription";
262
- "llamacpp-embedding": "llamacpp-embedding";
263
- "nmtcpp-translation": "nmtcpp-translation";
264
- "onnx-tts": "onnx-tts";
265
- "parakeet-transcription": "parakeet-transcription";
266
- "onnx-ocr": "onnx-ocr";
267
- "sdcpp-generation": "sdcpp-generation";
268
- llm: "llm";
269
- whisper: "whisper";
270
- nmt: "nmt";
271
- parakeet: "parakeet";
272
- }>, z.ZodLiteral<"vad">]>>;
273
- }, z.core.$strip>]>;
274
- ttsDurationPredictorSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
275
- src: z.ZodString;
276
- name: z.ZodOptional<z.ZodString>;
277
- modelId: z.ZodOptional<z.ZodString>;
278
- registryPath: z.ZodOptional<z.ZodString>;
279
- registrySource: z.ZodOptional<z.ZodString>;
280
- blobCoreKey: z.ZodOptional<z.ZodString>;
281
- blobIndex: z.ZodOptional<z.ZodNumber>;
282
- engine: z.ZodOptional<z.ZodString>;
283
- expectedSize: z.ZodOptional<z.ZodNumber>;
284
- sha256Checksum: z.ZodOptional<z.ZodString>;
285
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
286
- embeddings: "embeddings";
287
- diffusion: "diffusion";
288
- tts: "tts";
289
- ocr: "ocr";
290
- "llamacpp-completion": "llamacpp-completion";
291
- "whispercpp-transcription": "whispercpp-transcription";
292
- "llamacpp-embedding": "llamacpp-embedding";
293
- "nmtcpp-translation": "nmtcpp-translation";
294
- "onnx-tts": "onnx-tts";
295
- "parakeet-transcription": "parakeet-transcription";
296
- "onnx-ocr": "onnx-ocr";
297
- "sdcpp-generation": "sdcpp-generation";
298
- llm: "llm";
299
- whisper: "whisper";
300
- nmt: "nmt";
301
- parakeet: "parakeet";
302
- }>, z.ZodLiteral<"vad">]>>;
303
- }, z.core.$strip>]>;
304
- ttsVectorEstimatorSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
305
- src: z.ZodString;
306
- name: z.ZodOptional<z.ZodString>;
307
- modelId: z.ZodOptional<z.ZodString>;
308
- registryPath: z.ZodOptional<z.ZodString>;
309
- registrySource: z.ZodOptional<z.ZodString>;
310
- blobCoreKey: z.ZodOptional<z.ZodString>;
311
- blobIndex: z.ZodOptional<z.ZodNumber>;
312
- engine: z.ZodOptional<z.ZodString>;
313
- expectedSize: z.ZodOptional<z.ZodNumber>;
314
- sha256Checksum: z.ZodOptional<z.ZodString>;
315
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
316
- embeddings: "embeddings";
317
- diffusion: "diffusion";
318
- tts: "tts";
319
- ocr: "ocr";
320
- "llamacpp-completion": "llamacpp-completion";
321
- "whispercpp-transcription": "whispercpp-transcription";
322
- "llamacpp-embedding": "llamacpp-embedding";
323
- "nmtcpp-translation": "nmtcpp-translation";
324
- "onnx-tts": "onnx-tts";
325
- "parakeet-transcription": "parakeet-transcription";
326
- "onnx-ocr": "onnx-ocr";
327
- "sdcpp-generation": "sdcpp-generation";
328
- llm: "llm";
329
- whisper: "whisper";
330
- nmt: "nmt";
331
- parakeet: "parakeet";
332
- }>, z.ZodLiteral<"vad">]>>;
333
- }, z.core.$strip>]>;
334
- ttsVocoderSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
335
- src: z.ZodString;
336
- name: z.ZodOptional<z.ZodString>;
337
- modelId: z.ZodOptional<z.ZodString>;
338
- registryPath: z.ZodOptional<z.ZodString>;
339
- registrySource: z.ZodOptional<z.ZodString>;
340
- blobCoreKey: z.ZodOptional<z.ZodString>;
341
- blobIndex: z.ZodOptional<z.ZodNumber>;
342
- engine: z.ZodOptional<z.ZodString>;
343
- expectedSize: z.ZodOptional<z.ZodNumber>;
344
- sha256Checksum: z.ZodOptional<z.ZodString>;
345
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
346
- embeddings: "embeddings";
347
- diffusion: "diffusion";
348
- tts: "tts";
349
- ocr: "ocr";
350
- "llamacpp-completion": "llamacpp-completion";
351
- "whispercpp-transcription": "whispercpp-transcription";
352
- "llamacpp-embedding": "llamacpp-embedding";
353
- "nmtcpp-translation": "nmtcpp-translation";
354
- "onnx-tts": "onnx-tts";
355
- "parakeet-transcription": "parakeet-transcription";
356
- "onnx-ocr": "onnx-ocr";
357
- "sdcpp-generation": "sdcpp-generation";
358
- llm: "llm";
359
- whisper: "whisper";
360
- nmt: "nmt";
361
- parakeet: "parakeet";
362
- }>, z.ZodLiteral<"vad">]>>;
363
- }, z.core.$strip>]>;
364
- ttsUnicodeIndexerSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
365
- src: z.ZodString;
366
- name: z.ZodOptional<z.ZodString>;
367
- modelId: z.ZodOptional<z.ZodString>;
368
- registryPath: z.ZodOptional<z.ZodString>;
369
- registrySource: z.ZodOptional<z.ZodString>;
370
- blobCoreKey: z.ZodOptional<z.ZodString>;
371
- blobIndex: z.ZodOptional<z.ZodNumber>;
372
- engine: z.ZodOptional<z.ZodString>;
373
- expectedSize: z.ZodOptional<z.ZodNumber>;
374
- sha256Checksum: z.ZodOptional<z.ZodString>;
375
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
376
- embeddings: "embeddings";
377
- diffusion: "diffusion";
378
- tts: "tts";
379
- ocr: "ocr";
380
- "llamacpp-completion": "llamacpp-completion";
381
- "whispercpp-transcription": "whispercpp-transcription";
382
- "llamacpp-embedding": "llamacpp-embedding";
383
- "nmtcpp-translation": "nmtcpp-translation";
384
- "onnx-tts": "onnx-tts";
385
- "parakeet-transcription": "parakeet-transcription";
386
- "onnx-ocr": "onnx-ocr";
387
- "sdcpp-generation": "sdcpp-generation";
388
- llm: "llm";
389
- whisper: "whisper";
390
- nmt: "nmt";
391
- parakeet: "parakeet";
392
- }>, z.ZodLiteral<"vad">]>>;
393
- }, z.core.$strip>]>;
394
- ttsTtsConfigSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
395
- src: z.ZodString;
396
- name: z.ZodOptional<z.ZodString>;
397
- modelId: z.ZodOptional<z.ZodString>;
398
- registryPath: z.ZodOptional<z.ZodString>;
399
- registrySource: z.ZodOptional<z.ZodString>;
400
- blobCoreKey: z.ZodOptional<z.ZodString>;
401
- blobIndex: z.ZodOptional<z.ZodNumber>;
402
- engine: z.ZodOptional<z.ZodString>;
403
- expectedSize: z.ZodOptional<z.ZodNumber>;
404
- sha256Checksum: z.ZodOptional<z.ZodString>;
405
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
406
- embeddings: "embeddings";
407
- diffusion: "diffusion";
408
- tts: "tts";
409
- ocr: "ocr";
410
- "llamacpp-completion": "llamacpp-completion";
411
- "whispercpp-transcription": "whispercpp-transcription";
412
- "llamacpp-embedding": "llamacpp-embedding";
413
- "nmtcpp-translation": "nmtcpp-translation";
414
- "onnx-tts": "onnx-tts";
415
- "parakeet-transcription": "parakeet-transcription";
416
- "onnx-ocr": "onnx-ocr";
417
- "sdcpp-generation": "sdcpp-generation";
418
- llm: "llm";
419
- whisper: "whisper";
420
- nmt: "nmt";
421
- parakeet: "parakeet";
422
- }>, z.ZodLiteral<"vad">]>>;
423
- }, z.core.$strip>]>;
424
- ttsVoiceStyleSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
425
- src: z.ZodString;
426
- name: z.ZodOptional<z.ZodString>;
427
- modelId: z.ZodOptional<z.ZodString>;
428
- registryPath: z.ZodOptional<z.ZodString>;
429
- registrySource: z.ZodOptional<z.ZodString>;
430
- blobCoreKey: z.ZodOptional<z.ZodString>;
431
- blobIndex: z.ZodOptional<z.ZodNumber>;
432
- engine: z.ZodOptional<z.ZodString>;
433
- expectedSize: z.ZodOptional<z.ZodNumber>;
434
- sha256Checksum: z.ZodOptional<z.ZodString>;
435
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
436
- embeddings: "embeddings";
437
- diffusion: "diffusion";
438
- tts: "tts";
439
- ocr: "ocr";
440
- "llamacpp-completion": "llamacpp-completion";
441
- "whispercpp-transcription": "whispercpp-transcription";
442
- "llamacpp-embedding": "llamacpp-embedding";
443
- "nmtcpp-translation": "nmtcpp-translation";
444
- "onnx-tts": "onnx-tts";
445
- "parakeet-transcription": "parakeet-transcription";
446
- "onnx-ocr": "onnx-ocr";
447
- "sdcpp-generation": "sdcpp-generation";
448
- llm: "llm";
449
- whisper: "whisper";
450
- nmt: "nmt";
451
- parakeet: "parakeet";
452
- }>, z.ZodLiteral<"vad">]>>;
453
- }, z.core.$strip>]>;
454
- }, z.core.$strip>;
455
- export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
456
- ttsEngine: z.ZodLiteral<"chatterbox">;
457
- language: z.ZodEnum<{
458
- en: "en";
459
- de: "de";
460
- es: "es";
461
- it: "it";
462
- }>;
463
- ttsTokenizerSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
464
- src: z.ZodString;
465
- name: z.ZodOptional<z.ZodString>;
466
- modelId: z.ZodOptional<z.ZodString>;
467
- registryPath: z.ZodOptional<z.ZodString>;
468
- registrySource: z.ZodOptional<z.ZodString>;
469
- blobCoreKey: z.ZodOptional<z.ZodString>;
470
- blobIndex: z.ZodOptional<z.ZodNumber>;
471
- engine: z.ZodOptional<z.ZodString>;
472
- expectedSize: z.ZodOptional<z.ZodNumber>;
473
- sha256Checksum: z.ZodOptional<z.ZodString>;
474
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
475
- embeddings: "embeddings";
476
- diffusion: "diffusion";
477
- tts: "tts";
478
- ocr: "ocr";
479
- "llamacpp-completion": "llamacpp-completion";
480
- "whispercpp-transcription": "whispercpp-transcription";
481
- "llamacpp-embedding": "llamacpp-embedding";
482
- "nmtcpp-translation": "nmtcpp-translation";
483
- "onnx-tts": "onnx-tts";
484
- "parakeet-transcription": "parakeet-transcription";
485
- "onnx-ocr": "onnx-ocr";
486
- "sdcpp-generation": "sdcpp-generation";
487
- llm: "llm";
488
- whisper: "whisper";
489
- nmt: "nmt";
490
- parakeet: "parakeet";
491
- }>, z.ZodLiteral<"vad">]>>;
492
- }, z.core.$strip>]>;
493
- ttsSpeechEncoderSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
494
- src: z.ZodString;
495
- name: z.ZodOptional<z.ZodString>;
496
- modelId: z.ZodOptional<z.ZodString>;
497
- registryPath: z.ZodOptional<z.ZodString>;
498
- registrySource: z.ZodOptional<z.ZodString>;
499
- blobCoreKey: z.ZodOptional<z.ZodString>;
500
- blobIndex: z.ZodOptional<z.ZodNumber>;
501
- engine: z.ZodOptional<z.ZodString>;
502
- expectedSize: z.ZodOptional<z.ZodNumber>;
503
- sha256Checksum: z.ZodOptional<z.ZodString>;
504
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
505
- embeddings: "embeddings";
506
- diffusion: "diffusion";
507
- tts: "tts";
508
- ocr: "ocr";
509
- "llamacpp-completion": "llamacpp-completion";
510
- "whispercpp-transcription": "whispercpp-transcription";
511
- "llamacpp-embedding": "llamacpp-embedding";
512
- "nmtcpp-translation": "nmtcpp-translation";
513
- "onnx-tts": "onnx-tts";
514
- "parakeet-transcription": "parakeet-transcription";
515
- "onnx-ocr": "onnx-ocr";
516
- "sdcpp-generation": "sdcpp-generation";
517
- llm: "llm";
518
- whisper: "whisper";
519
- nmt: "nmt";
520
- parakeet: "parakeet";
521
- }>, z.ZodLiteral<"vad">]>>;
522
- }, z.core.$strip>]>;
523
- ttsEmbedTokensSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
524
- src: z.ZodString;
525
- name: z.ZodOptional<z.ZodString>;
526
- modelId: z.ZodOptional<z.ZodString>;
527
- registryPath: z.ZodOptional<z.ZodString>;
528
- registrySource: z.ZodOptional<z.ZodString>;
529
- blobCoreKey: z.ZodOptional<z.ZodString>;
530
- blobIndex: z.ZodOptional<z.ZodNumber>;
531
- engine: z.ZodOptional<z.ZodString>;
532
- expectedSize: z.ZodOptional<z.ZodNumber>;
533
- sha256Checksum: z.ZodOptional<z.ZodString>;
534
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
535
- embeddings: "embeddings";
536
- diffusion: "diffusion";
537
- tts: "tts";
538
- ocr: "ocr";
539
- "llamacpp-completion": "llamacpp-completion";
540
- "whispercpp-transcription": "whispercpp-transcription";
541
- "llamacpp-embedding": "llamacpp-embedding";
542
- "nmtcpp-translation": "nmtcpp-translation";
543
- "onnx-tts": "onnx-tts";
544
- "parakeet-transcription": "parakeet-transcription";
545
- "onnx-ocr": "onnx-ocr";
546
- "sdcpp-generation": "sdcpp-generation";
547
- llm: "llm";
548
- whisper: "whisper";
549
- nmt: "nmt";
550
- parakeet: "parakeet";
551
- }>, z.ZodLiteral<"vad">]>>;
552
- }, z.core.$strip>]>;
553
- ttsConditionalDecoderSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
554
- src: z.ZodString;
555
- name: z.ZodOptional<z.ZodString>;
556
- modelId: z.ZodOptional<z.ZodString>;
557
- registryPath: z.ZodOptional<z.ZodString>;
558
- registrySource: z.ZodOptional<z.ZodString>;
559
- blobCoreKey: z.ZodOptional<z.ZodString>;
560
- blobIndex: z.ZodOptional<z.ZodNumber>;
561
- engine: z.ZodOptional<z.ZodString>;
562
- expectedSize: z.ZodOptional<z.ZodNumber>;
563
- sha256Checksum: z.ZodOptional<z.ZodString>;
564
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
565
- embeddings: "embeddings";
566
- diffusion: "diffusion";
567
- tts: "tts";
568
- ocr: "ocr";
569
- "llamacpp-completion": "llamacpp-completion";
570
- "whispercpp-transcription": "whispercpp-transcription";
571
- "llamacpp-embedding": "llamacpp-embedding";
572
- "nmtcpp-translation": "nmtcpp-translation";
573
- "onnx-tts": "onnx-tts";
574
- "parakeet-transcription": "parakeet-transcription";
575
- "onnx-ocr": "onnx-ocr";
576
- "sdcpp-generation": "sdcpp-generation";
577
- llm: "llm";
578
- whisper: "whisper";
579
- nmt: "nmt";
580
- parakeet: "parakeet";
581
- }>, z.ZodLiteral<"vad">]>>;
582
- }, z.core.$strip>]>;
583
- ttsLanguageModelSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
129
+ }, z.core.$strip>]>>;
130
+ }, z.core.$strip>;
131
+ export declare const ttsSupertonicLoadConfigSchema: z.ZodObject<{
132
+ ttsEngine: z.ZodLiteral<"supertonic">;
133
+ language: z.ZodEnum<{
134
+ en: "en";
135
+ de: "de";
136
+ es: "es";
137
+ it: "it";
138
+ }>;
139
+ voice: z.ZodOptional<z.ZodString>;
140
+ ttsSpeed: z.ZodOptional<z.ZodNumber>;
141
+ ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
142
+ useGPU: z.ZodOptional<z.ZodBoolean>;
143
+ }, z.core.$strip>;
144
+ export declare const ttsLoadConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
145
+ ttsEngine: z.ZodLiteral<"chatterbox">;
146
+ language: z.ZodEnum<{
147
+ en: "en";
148
+ de: "de";
149
+ es: "es";
150
+ it: "it";
151
+ }>;
152
+ voice: z.ZodOptional<z.ZodString>;
153
+ useGPU: z.ZodOptional<z.ZodBoolean>;
154
+ s3genModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
584
155
  src: z.ZodString;
585
156
  name: z.ZodOptional<z.ZodString>;
586
157
  modelId: z.ZodOptional<z.ZodString>;
@@ -596,21 +167,26 @@ export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
596
167
  diffusion: "diffusion";
597
168
  tts: "tts";
598
169
  ocr: "ocr";
170
+ vla: "vla";
599
171
  "llamacpp-completion": "llamacpp-completion";
600
172
  "whispercpp-transcription": "whispercpp-transcription";
601
173
  "llamacpp-embedding": "llamacpp-embedding";
602
174
  "nmtcpp-translation": "nmtcpp-translation";
603
175
  "onnx-tts": "onnx-tts";
176
+ "tts-ggml": "tts-ggml";
604
177
  "parakeet-transcription": "parakeet-transcription";
605
178
  "onnx-ocr": "onnx-ocr";
606
179
  "sdcpp-generation": "sdcpp-generation";
180
+ "ggml-vla": "ggml-vla";
181
+ "ggml-classification": "ggml-classification";
607
182
  llm: "llm";
608
183
  whisper: "whisper";
609
184
  nmt: "nmt";
610
185
  parakeet: "parakeet";
186
+ classification: "classification";
611
187
  }>, z.ZodLiteral<"vad">]>>;
612
- }, z.core.$strip>]>;
613
- referenceAudioSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
188
+ }, z.core.$strip>]>>;
189
+ referenceAudioSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
614
190
  src: z.ZodString;
615
191
  name: z.ZodOptional<z.ZodString>;
616
192
  modelId: z.ZodOptional<z.ZodString>;
@@ -626,20 +202,25 @@ export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
626
202
  diffusion: "diffusion";
627
203
  tts: "tts";
628
204
  ocr: "ocr";
205
+ vla: "vla";
629
206
  "llamacpp-completion": "llamacpp-completion";
630
207
  "whispercpp-transcription": "whispercpp-transcription";
631
208
  "llamacpp-embedding": "llamacpp-embedding";
632
209
  "nmtcpp-translation": "nmtcpp-translation";
633
210
  "onnx-tts": "onnx-tts";
211
+ "tts-ggml": "tts-ggml";
634
212
  "parakeet-transcription": "parakeet-transcription";
635
213
  "onnx-ocr": "onnx-ocr";
636
214
  "sdcpp-generation": "sdcpp-generation";
215
+ "ggml-vla": "ggml-vla";
216
+ "ggml-classification": "ggml-classification";
637
217
  llm: "llm";
638
218
  whisper: "whisper";
639
219
  nmt: "nmt";
640
220
  parakeet: "parakeet";
221
+ classification: "classification";
641
222
  }>, z.ZodLiteral<"vad">]>>;
642
- }, z.core.$strip>]>;
223
+ }, z.core.$strip>]>>;
643
224
  }, z.core.$strip>, z.ZodObject<{
644
225
  ttsEngine: z.ZodLiteral<"supertonic">;
645
226
  language: z.ZodEnum<{
@@ -648,220 +229,17 @@ export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
648
229
  es: "es";
649
230
  it: "it";
650
231
  }>;
232
+ voice: z.ZodOptional<z.ZodString>;
651
233
  ttsSpeed: z.ZodOptional<z.ZodNumber>;
652
234
  ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
653
- ttsSupertonicMultilingual: z.ZodOptional<z.ZodBoolean>;
654
- ttsTextEncoderSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
655
- src: z.ZodString;
656
- name: z.ZodOptional<z.ZodString>;
657
- modelId: z.ZodOptional<z.ZodString>;
658
- registryPath: z.ZodOptional<z.ZodString>;
659
- registrySource: z.ZodOptional<z.ZodString>;
660
- blobCoreKey: z.ZodOptional<z.ZodString>;
661
- blobIndex: z.ZodOptional<z.ZodNumber>;
662
- engine: z.ZodOptional<z.ZodString>;
663
- expectedSize: z.ZodOptional<z.ZodNumber>;
664
- sha256Checksum: z.ZodOptional<z.ZodString>;
665
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
666
- embeddings: "embeddings";
667
- diffusion: "diffusion";
668
- tts: "tts";
669
- ocr: "ocr";
670
- "llamacpp-completion": "llamacpp-completion";
671
- "whispercpp-transcription": "whispercpp-transcription";
672
- "llamacpp-embedding": "llamacpp-embedding";
673
- "nmtcpp-translation": "nmtcpp-translation";
674
- "onnx-tts": "onnx-tts";
675
- "parakeet-transcription": "parakeet-transcription";
676
- "onnx-ocr": "onnx-ocr";
677
- "sdcpp-generation": "sdcpp-generation";
678
- llm: "llm";
679
- whisper: "whisper";
680
- nmt: "nmt";
681
- parakeet: "parakeet";
682
- }>, z.ZodLiteral<"vad">]>>;
683
- }, z.core.$strip>]>;
684
- ttsDurationPredictorSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
685
- src: z.ZodString;
686
- name: z.ZodOptional<z.ZodString>;
687
- modelId: z.ZodOptional<z.ZodString>;
688
- registryPath: z.ZodOptional<z.ZodString>;
689
- registrySource: z.ZodOptional<z.ZodString>;
690
- blobCoreKey: z.ZodOptional<z.ZodString>;
691
- blobIndex: z.ZodOptional<z.ZodNumber>;
692
- engine: z.ZodOptional<z.ZodString>;
693
- expectedSize: z.ZodOptional<z.ZodNumber>;
694
- sha256Checksum: z.ZodOptional<z.ZodString>;
695
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
696
- embeddings: "embeddings";
697
- diffusion: "diffusion";
698
- tts: "tts";
699
- ocr: "ocr";
700
- "llamacpp-completion": "llamacpp-completion";
701
- "whispercpp-transcription": "whispercpp-transcription";
702
- "llamacpp-embedding": "llamacpp-embedding";
703
- "nmtcpp-translation": "nmtcpp-translation";
704
- "onnx-tts": "onnx-tts";
705
- "parakeet-transcription": "parakeet-transcription";
706
- "onnx-ocr": "onnx-ocr";
707
- "sdcpp-generation": "sdcpp-generation";
708
- llm: "llm";
709
- whisper: "whisper";
710
- nmt: "nmt";
711
- parakeet: "parakeet";
712
- }>, z.ZodLiteral<"vad">]>>;
713
- }, z.core.$strip>]>;
714
- ttsVectorEstimatorSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
715
- src: z.ZodString;
716
- name: z.ZodOptional<z.ZodString>;
717
- modelId: z.ZodOptional<z.ZodString>;
718
- registryPath: z.ZodOptional<z.ZodString>;
719
- registrySource: z.ZodOptional<z.ZodString>;
720
- blobCoreKey: z.ZodOptional<z.ZodString>;
721
- blobIndex: z.ZodOptional<z.ZodNumber>;
722
- engine: z.ZodOptional<z.ZodString>;
723
- expectedSize: z.ZodOptional<z.ZodNumber>;
724
- sha256Checksum: z.ZodOptional<z.ZodString>;
725
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
726
- embeddings: "embeddings";
727
- diffusion: "diffusion";
728
- tts: "tts";
729
- ocr: "ocr";
730
- "llamacpp-completion": "llamacpp-completion";
731
- "whispercpp-transcription": "whispercpp-transcription";
732
- "llamacpp-embedding": "llamacpp-embedding";
733
- "nmtcpp-translation": "nmtcpp-translation";
734
- "onnx-tts": "onnx-tts";
735
- "parakeet-transcription": "parakeet-transcription";
736
- "onnx-ocr": "onnx-ocr";
737
- "sdcpp-generation": "sdcpp-generation";
738
- llm: "llm";
739
- whisper: "whisper";
740
- nmt: "nmt";
741
- parakeet: "parakeet";
742
- }>, z.ZodLiteral<"vad">]>>;
743
- }, z.core.$strip>]>;
744
- ttsVocoderSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
745
- src: z.ZodString;
746
- name: z.ZodOptional<z.ZodString>;
747
- modelId: z.ZodOptional<z.ZodString>;
748
- registryPath: z.ZodOptional<z.ZodString>;
749
- registrySource: z.ZodOptional<z.ZodString>;
750
- blobCoreKey: z.ZodOptional<z.ZodString>;
751
- blobIndex: z.ZodOptional<z.ZodNumber>;
752
- engine: z.ZodOptional<z.ZodString>;
753
- expectedSize: z.ZodOptional<z.ZodNumber>;
754
- sha256Checksum: z.ZodOptional<z.ZodString>;
755
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
756
- embeddings: "embeddings";
757
- diffusion: "diffusion";
758
- tts: "tts";
759
- ocr: "ocr";
760
- "llamacpp-completion": "llamacpp-completion";
761
- "whispercpp-transcription": "whispercpp-transcription";
762
- "llamacpp-embedding": "llamacpp-embedding";
763
- "nmtcpp-translation": "nmtcpp-translation";
764
- "onnx-tts": "onnx-tts";
765
- "parakeet-transcription": "parakeet-transcription";
766
- "onnx-ocr": "onnx-ocr";
767
- "sdcpp-generation": "sdcpp-generation";
768
- llm: "llm";
769
- whisper: "whisper";
770
- nmt: "nmt";
771
- parakeet: "parakeet";
772
- }>, z.ZodLiteral<"vad">]>>;
773
- }, z.core.$strip>]>;
774
- ttsUnicodeIndexerSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
775
- src: z.ZodString;
776
- name: z.ZodOptional<z.ZodString>;
777
- modelId: z.ZodOptional<z.ZodString>;
778
- registryPath: z.ZodOptional<z.ZodString>;
779
- registrySource: z.ZodOptional<z.ZodString>;
780
- blobCoreKey: z.ZodOptional<z.ZodString>;
781
- blobIndex: z.ZodOptional<z.ZodNumber>;
782
- engine: z.ZodOptional<z.ZodString>;
783
- expectedSize: z.ZodOptional<z.ZodNumber>;
784
- sha256Checksum: z.ZodOptional<z.ZodString>;
785
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
786
- embeddings: "embeddings";
787
- diffusion: "diffusion";
788
- tts: "tts";
789
- ocr: "ocr";
790
- "llamacpp-completion": "llamacpp-completion";
791
- "whispercpp-transcription": "whispercpp-transcription";
792
- "llamacpp-embedding": "llamacpp-embedding";
793
- "nmtcpp-translation": "nmtcpp-translation";
794
- "onnx-tts": "onnx-tts";
795
- "parakeet-transcription": "parakeet-transcription";
796
- "onnx-ocr": "onnx-ocr";
797
- "sdcpp-generation": "sdcpp-generation";
798
- llm: "llm";
799
- whisper: "whisper";
800
- nmt: "nmt";
801
- parakeet: "parakeet";
802
- }>, z.ZodLiteral<"vad">]>>;
803
- }, z.core.$strip>]>;
804
- ttsTtsConfigSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
805
- src: z.ZodString;
806
- name: z.ZodOptional<z.ZodString>;
807
- modelId: z.ZodOptional<z.ZodString>;
808
- registryPath: z.ZodOptional<z.ZodString>;
809
- registrySource: z.ZodOptional<z.ZodString>;
810
- blobCoreKey: z.ZodOptional<z.ZodString>;
811
- blobIndex: z.ZodOptional<z.ZodNumber>;
812
- engine: z.ZodOptional<z.ZodString>;
813
- expectedSize: z.ZodOptional<z.ZodNumber>;
814
- sha256Checksum: z.ZodOptional<z.ZodString>;
815
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
816
- embeddings: "embeddings";
817
- diffusion: "diffusion";
818
- tts: "tts";
819
- ocr: "ocr";
820
- "llamacpp-completion": "llamacpp-completion";
821
- "whispercpp-transcription": "whispercpp-transcription";
822
- "llamacpp-embedding": "llamacpp-embedding";
823
- "nmtcpp-translation": "nmtcpp-translation";
824
- "onnx-tts": "onnx-tts";
825
- "parakeet-transcription": "parakeet-transcription";
826
- "onnx-ocr": "onnx-ocr";
827
- "sdcpp-generation": "sdcpp-generation";
828
- llm: "llm";
829
- whisper: "whisper";
830
- nmt: "nmt";
831
- parakeet: "parakeet";
832
- }>, z.ZodLiteral<"vad">]>>;
833
- }, z.core.$strip>]>;
834
- ttsVoiceStyleSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
835
- src: z.ZodString;
836
- name: z.ZodOptional<z.ZodString>;
837
- modelId: z.ZodOptional<z.ZodString>;
838
- registryPath: z.ZodOptional<z.ZodString>;
839
- registrySource: z.ZodOptional<z.ZodString>;
840
- blobCoreKey: z.ZodOptional<z.ZodString>;
841
- blobIndex: z.ZodOptional<z.ZodNumber>;
842
- engine: z.ZodOptional<z.ZodString>;
843
- expectedSize: z.ZodOptional<z.ZodNumber>;
844
- sha256Checksum: z.ZodOptional<z.ZodString>;
845
- addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
846
- embeddings: "embeddings";
847
- diffusion: "diffusion";
848
- tts: "tts";
849
- ocr: "ocr";
850
- "llamacpp-completion": "llamacpp-completion";
851
- "whispercpp-transcription": "whispercpp-transcription";
852
- "llamacpp-embedding": "llamacpp-embedding";
853
- "nmtcpp-translation": "nmtcpp-translation";
854
- "onnx-tts": "onnx-tts";
855
- "parakeet-transcription": "parakeet-transcription";
856
- "onnx-ocr": "onnx-ocr";
857
- "sdcpp-generation": "sdcpp-generation";
858
- llm: "llm";
859
- whisper: "whisper";
860
- nmt: "nmt";
861
- parakeet: "parakeet";
862
- }>, z.ZodLiteral<"vad">]>>;
863
- }, z.core.$strip>]>;
235
+ useGPU: z.ZodOptional<z.ZodBoolean>;
864
236
  }, z.core.$strip>]>;
237
+ export declare const LEGACY_TTS_ONNX_MODEL_CONFIG_FIELDS: readonly ["ttsSupertonicMultilingual", "ttsTokenizerSrc", "ttsSpeechEncoderSrc", "ttsEmbedTokensSrc", "ttsConditionalDecoderSrc", "ttsLanguageModelSrc", "ttsTextEncoderSrc", "ttsDurationPredictorSrc", "ttsVectorEstimatorSrc", "ttsVocoderSrc", "ttsUnicodeIndexerSrc", "ttsTtsConfigSrc", "ttsVoiceStyleSrc"];
238
+ export declare const ttsConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
239
+ [x: string]: z.ZodOptional<z.ZodUnknown>;
240
+ }, z.core.$strict>, z.ZodObject<{
241
+ [x: string]: z.ZodOptional<z.ZodUnknown>;
242
+ }, z.core.$strict>]>;
865
243
  export declare const ttsClientParamsSchema: z.ZodObject<{
866
244
  modelId: z.ZodString;
867
245
  inputType: z.ZodDefault<z.ZodString>;
@@ -933,12 +311,17 @@ export declare const textToSpeechStreamResponseSchema: z.ZodObject<{
933
311
  sentenceChunk: z.ZodOptional<z.ZodString>;
934
312
  }, z.core.$strip>;
935
313
  export type TtsLanguage = (typeof TTS_LANGUAGES)[number];
936
- export type TtsChatterboxConfig = z.infer<typeof ttsChatterboxConfigSchema>;
937
- export type TtsSupertonicConfig = z.infer<typeof ttsSupertonicConfigSchema>;
938
- export type TtsConfig = z.infer<typeof ttsConfigSchema>;
314
+ export type TtsChatterboxLoadConfig = z.infer<typeof ttsChatterboxLoadConfigSchema>;
315
+ export type TtsSupertonicLoadConfig = z.infer<typeof ttsSupertonicLoadConfigSchema>;
316
+ export type TtsLoadConfig = z.infer<typeof ttsLoadConfigSchema>;
317
+ /** @deprecated Use {@link TtsChatterboxLoadConfig} */
318
+ export type TtsChatterboxConfig = TtsChatterboxLoadConfig;
319
+ /** @deprecated Use {@link TtsSupertonicLoadConfig} */
320
+ export type TtsSupertonicConfig = TtsSupertonicLoadConfig;
939
321
  export type TtsChatterboxRuntimeConfig = z.infer<typeof ttsChatterboxRuntimeConfigSchema>;
940
322
  export type TtsSupertonicRuntimeConfig = z.infer<typeof ttsSupertonicRuntimeConfigSchema>;
941
323
  export type TtsRuntimeConfig = z.infer<typeof ttsRuntimeConfigSchema>;
324
+ export type TtsConfig = z.infer<typeof ttsConfigSchema>;
942
325
  export type TtsClientParamsInput = z.input<typeof ttsClientParamsSchema>;
943
326
  export type TtsClientParams = z.output<typeof ttsClientParamsSchema>;
944
327
  export type TtsRequest = z.infer<typeof ttsRequestSchema>;