@qvac/sdk 1.0.0 → 1.1.0

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.
@@ -47,16 +47,65 @@ export declare const loadModelOptionsSchema: z.ZodPipe<z.ZodUnion<readonly [z.Zo
47
47
  json: "json";
48
48
  }>>>;
49
49
  min_seconds: z.ZodOptional<z.ZodDefault<z.ZodNumber>>;
50
- max_seconds: z.ZodOptional<z.ZodDefault<z.ZodNumber>>;
51
- audio_format: z.ZodOptional<z.ZodDefault<z.ZodString>>;
52
50
  update_frequency: z.ZodOptional<z.ZodDefault<z.ZodString>>;
53
- vadConfig: z.ZodOptional<z.ZodOptional<z.ZodObject<{
54
- vad_threshold: z.ZodOptional<z.ZodNumber>;
55
- vad_min_speech_duration_ms: z.ZodOptional<z.ZodNumber>;
56
- vad_min_silence_duration_ms: z.ZodOptional<z.ZodNumber>;
57
- vad_max_speech_duration_s: z.ZodOptional<z.ZodNumber>;
58
- vad_speech_pad_ms: z.ZodOptional<z.ZodNumber>;
59
- vad_samples_overlap: z.ZodOptional<z.ZodNumber>;
51
+ strategy: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
52
+ greedy: "greedy";
53
+ beam_search: "beam_search";
54
+ }>>>;
55
+ n_threads: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
56
+ n_max_text_ctx: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
57
+ offset_ms: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
58
+ duration_ms: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
59
+ audio_ctx: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
60
+ translate: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
61
+ no_context: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
62
+ no_timestamps: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
63
+ single_segment: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
64
+ print_special: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
65
+ print_progress: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
66
+ print_realtime: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
67
+ print_timestamps: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
68
+ token_timestamps: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
69
+ thold_pt: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
70
+ thold_ptsum: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
71
+ max_len: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
72
+ split_on_word: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
73
+ max_tokens: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
74
+ debug_mode: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
75
+ tdrz_enable: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
76
+ suppress_regex: z.ZodOptional<z.ZodOptional<z.ZodString>>;
77
+ initial_prompt: z.ZodOptional<z.ZodOptional<z.ZodString>>;
78
+ language: z.ZodOptional<z.ZodOptional<z.ZodString>>;
79
+ detect_language: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
80
+ suppress_blank: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
81
+ suppress_nst: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
82
+ temperature: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
83
+ length_penalty: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
84
+ temperature_inc: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
85
+ entropy_thold: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
86
+ logprob_thold: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
87
+ greedy_best_of: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
88
+ beam_search_beam_size: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
89
+ vadModelPath: z.ZodOptional<z.ZodOptional<z.ZodString>>;
90
+ vad_params: z.ZodOptional<z.ZodOptional<z.ZodObject<{
91
+ threshold: z.ZodOptional<z.ZodNumber>;
92
+ min_speech_duration_ms: z.ZodOptional<z.ZodNumber>;
93
+ min_silence_duration_ms: z.ZodOptional<z.ZodNumber>;
94
+ max_speech_duration_s: z.ZodOptional<z.ZodNumber>;
95
+ speech_pad_ms: z.ZodOptional<z.ZodNumber>;
96
+ samples_overlap: z.ZodOptional<z.ZodNumber>;
97
+ }, z.core.$strip>>>;
98
+ max_seconds: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
99
+ chunk_length_s: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
100
+ audio_format: z.ZodOptional<z.ZodOptional<z.ZodString>>;
101
+ contextParams: z.ZodOptional<z.ZodOptional<z.ZodObject<{
102
+ model: z.ZodOptional<z.ZodString>;
103
+ use_gpu: z.ZodOptional<z.ZodBoolean>;
104
+ flash_attn: z.ZodOptional<z.ZodBoolean>;
105
+ gpu_device: z.ZodOptional<z.ZodNumber>;
106
+ }, z.core.$strip>>>;
107
+ miscConfig: z.ZodOptional<z.ZodOptional<z.ZodObject<{
108
+ caption_enabled: z.ZodOptional<z.ZodBoolean>;
60
109
  }, z.core.$strip>>>;
61
110
  }, z.core.$strict>>;
62
111
  seed: z.ZodOptional<z.ZodBoolean>;
@@ -231,16 +280,62 @@ export declare const loadModelOptionsSchema: z.ZodPipe<z.ZodUnion<readonly [z.Zo
231
280
  mode?: "caption" | "batch" | undefined;
232
281
  output_format?: "plaintext" | "json" | undefined;
233
282
  min_seconds?: number | undefined;
283
+ update_frequency?: string | undefined;
284
+ strategy?: "greedy" | "beam_search" | undefined;
285
+ n_threads?: number | undefined;
286
+ n_max_text_ctx?: number | undefined;
287
+ offset_ms?: number | undefined;
288
+ duration_ms?: number | undefined;
289
+ audio_ctx?: number | undefined;
290
+ translate?: boolean | undefined;
291
+ no_context?: boolean | undefined;
292
+ no_timestamps?: boolean | undefined;
293
+ single_segment?: boolean | undefined;
294
+ print_special?: boolean | undefined;
295
+ print_progress?: boolean | undefined;
296
+ print_realtime?: boolean | undefined;
297
+ print_timestamps?: boolean | undefined;
298
+ token_timestamps?: boolean | undefined;
299
+ thold_pt?: number | undefined;
300
+ thold_ptsum?: number | undefined;
301
+ max_len?: number | undefined;
302
+ split_on_word?: boolean | undefined;
303
+ max_tokens?: number | undefined;
304
+ debug_mode?: boolean | undefined;
305
+ tdrz_enable?: boolean | undefined;
306
+ suppress_regex?: string | undefined;
307
+ initial_prompt?: string | undefined;
308
+ language?: string | undefined;
309
+ detect_language?: boolean | undefined;
310
+ suppress_blank?: boolean | undefined;
311
+ suppress_nst?: boolean | undefined;
312
+ temperature?: number | undefined;
313
+ length_penalty?: number | undefined;
314
+ temperature_inc?: number | undefined;
315
+ entropy_thold?: number | undefined;
316
+ logprob_thold?: number | undefined;
317
+ greedy_best_of?: number | undefined;
318
+ beam_search_beam_size?: number | undefined;
319
+ vadModelPath?: string | undefined;
320
+ vad_params?: {
321
+ threshold?: number | undefined;
322
+ min_speech_duration_ms?: number | undefined;
323
+ min_silence_duration_ms?: number | undefined;
324
+ max_speech_duration_s?: number | undefined;
325
+ speech_pad_ms?: number | undefined;
326
+ samples_overlap?: number | undefined;
327
+ } | undefined;
234
328
  max_seconds?: number | undefined;
329
+ chunk_length_s?: number | undefined;
235
330
  audio_format?: string | undefined;
236
- update_frequency?: string | undefined;
237
- vadConfig?: {
238
- vad_threshold?: number | undefined;
239
- vad_min_speech_duration_ms?: number | undefined;
240
- vad_min_silence_duration_ms?: number | undefined;
241
- vad_max_speech_duration_s?: number | undefined;
242
- vad_speech_pad_ms?: number | undefined;
243
- vad_samples_overlap?: number | undefined;
331
+ contextParams?: {
332
+ model?: string | undefined;
333
+ use_gpu?: boolean | undefined;
334
+ flash_attn?: boolean | undefined;
335
+ gpu_device?: number | undefined;
336
+ } | undefined;
337
+ miscConfig?: {
338
+ caption_enabled?: boolean | undefined;
244
339
  } | undefined;
245
340
  } | undefined;
246
341
  vadModelSrc?: string | undefined;
@@ -332,16 +427,62 @@ export declare const loadModelOptionsSchema: z.ZodPipe<z.ZodUnion<readonly [z.Zo
332
427
  mode?: "caption" | "batch" | undefined;
333
428
  output_format?: "plaintext" | "json" | undefined;
334
429
  min_seconds?: number | undefined;
430
+ update_frequency?: string | undefined;
431
+ strategy?: "greedy" | "beam_search" | undefined;
432
+ n_threads?: number | undefined;
433
+ n_max_text_ctx?: number | undefined;
434
+ offset_ms?: number | undefined;
435
+ duration_ms?: number | undefined;
436
+ audio_ctx?: number | undefined;
437
+ translate?: boolean | undefined;
438
+ no_context?: boolean | undefined;
439
+ no_timestamps?: boolean | undefined;
440
+ single_segment?: boolean | undefined;
441
+ print_special?: boolean | undefined;
442
+ print_progress?: boolean | undefined;
443
+ print_realtime?: boolean | undefined;
444
+ print_timestamps?: boolean | undefined;
445
+ token_timestamps?: boolean | undefined;
446
+ thold_pt?: number | undefined;
447
+ thold_ptsum?: number | undefined;
448
+ max_len?: number | undefined;
449
+ split_on_word?: boolean | undefined;
450
+ max_tokens?: number | undefined;
451
+ debug_mode?: boolean | undefined;
452
+ tdrz_enable?: boolean | undefined;
453
+ suppress_regex?: string | undefined;
454
+ initial_prompt?: string | undefined;
455
+ language?: string | undefined;
456
+ detect_language?: boolean | undefined;
457
+ suppress_blank?: boolean | undefined;
458
+ suppress_nst?: boolean | undefined;
459
+ temperature?: number | undefined;
460
+ length_penalty?: number | undefined;
461
+ temperature_inc?: number | undefined;
462
+ entropy_thold?: number | undefined;
463
+ logprob_thold?: number | undefined;
464
+ greedy_best_of?: number | undefined;
465
+ beam_search_beam_size?: number | undefined;
466
+ vadModelPath?: string | undefined;
467
+ vad_params?: {
468
+ threshold?: number | undefined;
469
+ min_speech_duration_ms?: number | undefined;
470
+ min_silence_duration_ms?: number | undefined;
471
+ max_speech_duration_s?: number | undefined;
472
+ speech_pad_ms?: number | undefined;
473
+ samples_overlap?: number | undefined;
474
+ } | undefined;
335
475
  max_seconds?: number | undefined;
476
+ chunk_length_s?: number | undefined;
336
477
  audio_format?: string | undefined;
337
- update_frequency?: string | undefined;
338
- vadConfig?: {
339
- vad_threshold?: number | undefined;
340
- vad_min_speech_duration_ms?: number | undefined;
341
- vad_min_silence_duration_ms?: number | undefined;
342
- vad_max_speech_duration_s?: number | undefined;
343
- vad_speech_pad_ms?: number | undefined;
344
- vad_samples_overlap?: number | undefined;
478
+ contextParams?: {
479
+ model?: string | undefined;
480
+ use_gpu?: boolean | undefined;
481
+ flash_attn?: boolean | undefined;
482
+ gpu_device?: number | undefined;
483
+ } | undefined;
484
+ miscConfig?: {
485
+ caption_enabled?: boolean | undefined;
345
486
  } | undefined;
346
487
  } | undefined;
347
488
  seed?: boolean | undefined;
@@ -460,16 +601,65 @@ export declare const loadWhisperModelRequestSchema: z.ZodObject<{
460
601
  json: "json";
461
602
  }>>;
462
603
  min_seconds: z.ZodDefault<z.ZodNumber>;
463
- max_seconds: z.ZodDefault<z.ZodNumber>;
464
- audio_format: z.ZodDefault<z.ZodString>;
465
604
  update_frequency: z.ZodDefault<z.ZodString>;
466
- vadConfig: z.ZodOptional<z.ZodObject<{
467
- vad_threshold: z.ZodOptional<z.ZodNumber>;
468
- vad_min_speech_duration_ms: z.ZodOptional<z.ZodNumber>;
469
- vad_min_silence_duration_ms: z.ZodOptional<z.ZodNumber>;
470
- vad_max_speech_duration_s: z.ZodOptional<z.ZodNumber>;
471
- vad_speech_pad_ms: z.ZodOptional<z.ZodNumber>;
472
- vad_samples_overlap: z.ZodOptional<z.ZodNumber>;
605
+ strategy: z.ZodOptional<z.ZodEnum<{
606
+ greedy: "greedy";
607
+ beam_search: "beam_search";
608
+ }>>;
609
+ n_threads: z.ZodOptional<z.ZodNumber>;
610
+ n_max_text_ctx: z.ZodOptional<z.ZodNumber>;
611
+ offset_ms: z.ZodOptional<z.ZodNumber>;
612
+ duration_ms: z.ZodOptional<z.ZodNumber>;
613
+ audio_ctx: z.ZodOptional<z.ZodNumber>;
614
+ translate: z.ZodOptional<z.ZodBoolean>;
615
+ no_context: z.ZodOptional<z.ZodBoolean>;
616
+ no_timestamps: z.ZodOptional<z.ZodBoolean>;
617
+ single_segment: z.ZodOptional<z.ZodBoolean>;
618
+ print_special: z.ZodOptional<z.ZodBoolean>;
619
+ print_progress: z.ZodOptional<z.ZodBoolean>;
620
+ print_realtime: z.ZodOptional<z.ZodBoolean>;
621
+ print_timestamps: z.ZodOptional<z.ZodBoolean>;
622
+ token_timestamps: z.ZodOptional<z.ZodBoolean>;
623
+ thold_pt: z.ZodOptional<z.ZodNumber>;
624
+ thold_ptsum: z.ZodOptional<z.ZodNumber>;
625
+ max_len: z.ZodOptional<z.ZodNumber>;
626
+ split_on_word: z.ZodOptional<z.ZodBoolean>;
627
+ max_tokens: z.ZodOptional<z.ZodNumber>;
628
+ debug_mode: z.ZodOptional<z.ZodBoolean>;
629
+ tdrz_enable: z.ZodOptional<z.ZodBoolean>;
630
+ suppress_regex: z.ZodOptional<z.ZodString>;
631
+ initial_prompt: z.ZodOptional<z.ZodString>;
632
+ language: z.ZodOptional<z.ZodString>;
633
+ detect_language: z.ZodOptional<z.ZodBoolean>;
634
+ suppress_blank: z.ZodOptional<z.ZodBoolean>;
635
+ suppress_nst: z.ZodOptional<z.ZodBoolean>;
636
+ temperature: z.ZodOptional<z.ZodNumber>;
637
+ length_penalty: z.ZodOptional<z.ZodNumber>;
638
+ temperature_inc: z.ZodOptional<z.ZodNumber>;
639
+ entropy_thold: z.ZodOptional<z.ZodNumber>;
640
+ logprob_thold: z.ZodOptional<z.ZodNumber>;
641
+ greedy_best_of: z.ZodOptional<z.ZodNumber>;
642
+ beam_search_beam_size: z.ZodOptional<z.ZodNumber>;
643
+ vadModelPath: z.ZodOptional<z.ZodString>;
644
+ vad_params: z.ZodOptional<z.ZodObject<{
645
+ threshold: z.ZodOptional<z.ZodNumber>;
646
+ min_speech_duration_ms: z.ZodOptional<z.ZodNumber>;
647
+ min_silence_duration_ms: z.ZodOptional<z.ZodNumber>;
648
+ max_speech_duration_s: z.ZodOptional<z.ZodNumber>;
649
+ speech_pad_ms: z.ZodOptional<z.ZodNumber>;
650
+ samples_overlap: z.ZodOptional<z.ZodNumber>;
651
+ }, z.core.$strip>>;
652
+ max_seconds: z.ZodOptional<z.ZodNumber>;
653
+ chunk_length_s: z.ZodOptional<z.ZodNumber>;
654
+ audio_format: z.ZodOptional<z.ZodString>;
655
+ contextParams: z.ZodOptional<z.ZodObject<{
656
+ model: z.ZodOptional<z.ZodString>;
657
+ use_gpu: z.ZodOptional<z.ZodBoolean>;
658
+ flash_attn: z.ZodOptional<z.ZodBoolean>;
659
+ gpu_device: z.ZodOptional<z.ZodNumber>;
660
+ }, z.core.$strip>>;
661
+ miscConfig: z.ZodOptional<z.ZodObject<{
662
+ caption_enabled: z.ZodOptional<z.ZodBoolean>;
473
663
  }, z.core.$strip>>;
474
664
  }, z.core.$strip>;
475
665
  }, z.core.$strip>;
@@ -674,16 +864,65 @@ export declare const loadModelRequestSchema: z.ZodPipe<z.ZodDiscriminatedUnion<[
674
864
  json: "json";
675
865
  }>>;
676
866
  min_seconds: z.ZodDefault<z.ZodNumber>;
677
- max_seconds: z.ZodDefault<z.ZodNumber>;
678
- audio_format: z.ZodDefault<z.ZodString>;
679
867
  update_frequency: z.ZodDefault<z.ZodString>;
680
- vadConfig: z.ZodOptional<z.ZodObject<{
681
- vad_threshold: z.ZodOptional<z.ZodNumber>;
682
- vad_min_speech_duration_ms: z.ZodOptional<z.ZodNumber>;
683
- vad_min_silence_duration_ms: z.ZodOptional<z.ZodNumber>;
684
- vad_max_speech_duration_s: z.ZodOptional<z.ZodNumber>;
685
- vad_speech_pad_ms: z.ZodOptional<z.ZodNumber>;
686
- vad_samples_overlap: z.ZodOptional<z.ZodNumber>;
868
+ strategy: z.ZodOptional<z.ZodEnum<{
869
+ greedy: "greedy";
870
+ beam_search: "beam_search";
871
+ }>>;
872
+ n_threads: z.ZodOptional<z.ZodNumber>;
873
+ n_max_text_ctx: z.ZodOptional<z.ZodNumber>;
874
+ offset_ms: z.ZodOptional<z.ZodNumber>;
875
+ duration_ms: z.ZodOptional<z.ZodNumber>;
876
+ audio_ctx: z.ZodOptional<z.ZodNumber>;
877
+ translate: z.ZodOptional<z.ZodBoolean>;
878
+ no_context: z.ZodOptional<z.ZodBoolean>;
879
+ no_timestamps: z.ZodOptional<z.ZodBoolean>;
880
+ single_segment: z.ZodOptional<z.ZodBoolean>;
881
+ print_special: z.ZodOptional<z.ZodBoolean>;
882
+ print_progress: z.ZodOptional<z.ZodBoolean>;
883
+ print_realtime: z.ZodOptional<z.ZodBoolean>;
884
+ print_timestamps: z.ZodOptional<z.ZodBoolean>;
885
+ token_timestamps: z.ZodOptional<z.ZodBoolean>;
886
+ thold_pt: z.ZodOptional<z.ZodNumber>;
887
+ thold_ptsum: z.ZodOptional<z.ZodNumber>;
888
+ max_len: z.ZodOptional<z.ZodNumber>;
889
+ split_on_word: z.ZodOptional<z.ZodBoolean>;
890
+ max_tokens: z.ZodOptional<z.ZodNumber>;
891
+ debug_mode: z.ZodOptional<z.ZodBoolean>;
892
+ tdrz_enable: z.ZodOptional<z.ZodBoolean>;
893
+ suppress_regex: z.ZodOptional<z.ZodString>;
894
+ initial_prompt: z.ZodOptional<z.ZodString>;
895
+ language: z.ZodOptional<z.ZodString>;
896
+ detect_language: z.ZodOptional<z.ZodBoolean>;
897
+ suppress_blank: z.ZodOptional<z.ZodBoolean>;
898
+ suppress_nst: z.ZodOptional<z.ZodBoolean>;
899
+ temperature: z.ZodOptional<z.ZodNumber>;
900
+ length_penalty: z.ZodOptional<z.ZodNumber>;
901
+ temperature_inc: z.ZodOptional<z.ZodNumber>;
902
+ entropy_thold: z.ZodOptional<z.ZodNumber>;
903
+ logprob_thold: z.ZodOptional<z.ZodNumber>;
904
+ greedy_best_of: z.ZodOptional<z.ZodNumber>;
905
+ beam_search_beam_size: z.ZodOptional<z.ZodNumber>;
906
+ vadModelPath: z.ZodOptional<z.ZodString>;
907
+ vad_params: z.ZodOptional<z.ZodObject<{
908
+ threshold: z.ZodOptional<z.ZodNumber>;
909
+ min_speech_duration_ms: z.ZodOptional<z.ZodNumber>;
910
+ min_silence_duration_ms: z.ZodOptional<z.ZodNumber>;
911
+ max_speech_duration_s: z.ZodOptional<z.ZodNumber>;
912
+ speech_pad_ms: z.ZodOptional<z.ZodNumber>;
913
+ samples_overlap: z.ZodOptional<z.ZodNumber>;
914
+ }, z.core.$strip>>;
915
+ max_seconds: z.ZodOptional<z.ZodNumber>;
916
+ chunk_length_s: z.ZodOptional<z.ZodNumber>;
917
+ audio_format: z.ZodOptional<z.ZodString>;
918
+ contextParams: z.ZodOptional<z.ZodObject<{
919
+ model: z.ZodOptional<z.ZodString>;
920
+ use_gpu: z.ZodOptional<z.ZodBoolean>;
921
+ flash_attn: z.ZodOptional<z.ZodBoolean>;
922
+ gpu_device: z.ZodOptional<z.ZodNumber>;
923
+ }, z.core.$strip>>;
924
+ miscConfig: z.ZodOptional<z.ZodObject<{
925
+ caption_enabled: z.ZodOptional<z.ZodBoolean>;
687
926
  }, z.core.$strip>>;
688
927
  }, z.core.$strip>;
689
928
  }, z.core.$strip>, z.ZodObject<{
@@ -868,16 +1107,62 @@ export declare const loadModelRequestSchema: z.ZodPipe<z.ZodDiscriminatedUnion<[
868
1107
  mode: "caption" | "batch";
869
1108
  output_format: "plaintext" | "json";
870
1109
  min_seconds: number;
871
- max_seconds: number;
872
- audio_format: string;
873
1110
  update_frequency: string;
874
- vadConfig?: {
875
- vad_threshold?: number | undefined;
876
- vad_min_speech_duration_ms?: number | undefined;
877
- vad_min_silence_duration_ms?: number | undefined;
878
- vad_max_speech_duration_s?: number | undefined;
879
- vad_speech_pad_ms?: number | undefined;
880
- vad_samples_overlap?: number | undefined;
1111
+ strategy?: "greedy" | "beam_search" | undefined;
1112
+ n_threads?: number | undefined;
1113
+ n_max_text_ctx?: number | undefined;
1114
+ offset_ms?: number | undefined;
1115
+ duration_ms?: number | undefined;
1116
+ audio_ctx?: number | undefined;
1117
+ translate?: boolean | undefined;
1118
+ no_context?: boolean | undefined;
1119
+ no_timestamps?: boolean | undefined;
1120
+ single_segment?: boolean | undefined;
1121
+ print_special?: boolean | undefined;
1122
+ print_progress?: boolean | undefined;
1123
+ print_realtime?: boolean | undefined;
1124
+ print_timestamps?: boolean | undefined;
1125
+ token_timestamps?: boolean | undefined;
1126
+ thold_pt?: number | undefined;
1127
+ thold_ptsum?: number | undefined;
1128
+ max_len?: number | undefined;
1129
+ split_on_word?: boolean | undefined;
1130
+ max_tokens?: number | undefined;
1131
+ debug_mode?: boolean | undefined;
1132
+ tdrz_enable?: boolean | undefined;
1133
+ suppress_regex?: string | undefined;
1134
+ initial_prompt?: string | undefined;
1135
+ language?: string | undefined;
1136
+ detect_language?: boolean | undefined;
1137
+ suppress_blank?: boolean | undefined;
1138
+ suppress_nst?: boolean | undefined;
1139
+ temperature?: number | undefined;
1140
+ length_penalty?: number | undefined;
1141
+ temperature_inc?: number | undefined;
1142
+ entropy_thold?: number | undefined;
1143
+ logprob_thold?: number | undefined;
1144
+ greedy_best_of?: number | undefined;
1145
+ beam_search_beam_size?: number | undefined;
1146
+ vadModelPath?: string | undefined;
1147
+ vad_params?: {
1148
+ threshold?: number | undefined;
1149
+ min_speech_duration_ms?: number | undefined;
1150
+ min_silence_duration_ms?: number | undefined;
1151
+ max_speech_duration_s?: number | undefined;
1152
+ speech_pad_ms?: number | undefined;
1153
+ samples_overlap?: number | undefined;
1154
+ } | undefined;
1155
+ max_seconds?: number | undefined;
1156
+ chunk_length_s?: number | undefined;
1157
+ audio_format?: string | undefined;
1158
+ contextParams?: {
1159
+ model?: string | undefined;
1160
+ use_gpu?: boolean | undefined;
1161
+ flash_attn?: boolean | undefined;
1162
+ gpu_device?: number | undefined;
1163
+ } | undefined;
1164
+ miscConfig?: {
1165
+ caption_enabled?: boolean | undefined;
881
1166
  } | undefined;
882
1167
  };
883
1168
  projectionModelSrc?: string | undefined;
@@ -990,16 +1275,62 @@ export declare const loadModelRequestSchema: z.ZodPipe<z.ZodDiscriminatedUnion<[
990
1275
  mode: "caption" | "batch";
991
1276
  output_format: "plaintext" | "json";
992
1277
  min_seconds: number;
993
- max_seconds: number;
994
- audio_format: string;
995
1278
  update_frequency: string;
996
- vadConfig?: {
997
- vad_threshold?: number | undefined;
998
- vad_min_speech_duration_ms?: number | undefined;
999
- vad_min_silence_duration_ms?: number | undefined;
1000
- vad_max_speech_duration_s?: number | undefined;
1001
- vad_speech_pad_ms?: number | undefined;
1002
- vad_samples_overlap?: number | undefined;
1279
+ strategy?: "greedy" | "beam_search" | undefined;
1280
+ n_threads?: number | undefined;
1281
+ n_max_text_ctx?: number | undefined;
1282
+ offset_ms?: number | undefined;
1283
+ duration_ms?: number | undefined;
1284
+ audio_ctx?: number | undefined;
1285
+ translate?: boolean | undefined;
1286
+ no_context?: boolean | undefined;
1287
+ no_timestamps?: boolean | undefined;
1288
+ single_segment?: boolean | undefined;
1289
+ print_special?: boolean | undefined;
1290
+ print_progress?: boolean | undefined;
1291
+ print_realtime?: boolean | undefined;
1292
+ print_timestamps?: boolean | undefined;
1293
+ token_timestamps?: boolean | undefined;
1294
+ thold_pt?: number | undefined;
1295
+ thold_ptsum?: number | undefined;
1296
+ max_len?: number | undefined;
1297
+ split_on_word?: boolean | undefined;
1298
+ max_tokens?: number | undefined;
1299
+ debug_mode?: boolean | undefined;
1300
+ tdrz_enable?: boolean | undefined;
1301
+ suppress_regex?: string | undefined;
1302
+ initial_prompt?: string | undefined;
1303
+ language?: string | undefined;
1304
+ detect_language?: boolean | undefined;
1305
+ suppress_blank?: boolean | undefined;
1306
+ suppress_nst?: boolean | undefined;
1307
+ temperature?: number | undefined;
1308
+ length_penalty?: number | undefined;
1309
+ temperature_inc?: number | undefined;
1310
+ entropy_thold?: number | undefined;
1311
+ logprob_thold?: number | undefined;
1312
+ greedy_best_of?: number | undefined;
1313
+ beam_search_beam_size?: number | undefined;
1314
+ vadModelPath?: string | undefined;
1315
+ vad_params?: {
1316
+ threshold?: number | undefined;
1317
+ min_speech_duration_ms?: number | undefined;
1318
+ min_silence_duration_ms?: number | undefined;
1319
+ max_speech_duration_s?: number | undefined;
1320
+ speech_pad_ms?: number | undefined;
1321
+ samples_overlap?: number | undefined;
1322
+ } | undefined;
1323
+ max_seconds?: number | undefined;
1324
+ chunk_length_s?: number | undefined;
1325
+ audio_format?: string | undefined;
1326
+ contextParams?: {
1327
+ model?: string | undefined;
1328
+ use_gpu?: boolean | undefined;
1329
+ flash_attn?: boolean | undefined;
1330
+ gpu_device?: number | undefined;
1331
+ } | undefined;
1332
+ miscConfig?: {
1333
+ caption_enabled?: boolean | undefined;
1003
1334
  } | undefined;
1004
1335
  };
1005
1336
  projectionModelSrc?: string | undefined;
@@ -1139,16 +1470,65 @@ export declare const loadModelServerParamsSchema: z.ZodObject<{
1139
1470
  json: "json";
1140
1471
  }>>>;
1141
1472
  min_seconds: z.ZodOptional<z.ZodDefault<z.ZodNumber>>;
1142
- max_seconds: z.ZodOptional<z.ZodDefault<z.ZodNumber>>;
1143
- audio_format: z.ZodOptional<z.ZodDefault<z.ZodString>>;
1144
1473
  update_frequency: z.ZodOptional<z.ZodDefault<z.ZodString>>;
1145
- vadConfig: z.ZodOptional<z.ZodOptional<z.ZodObject<{
1146
- vad_threshold: z.ZodOptional<z.ZodNumber>;
1147
- vad_min_speech_duration_ms: z.ZodOptional<z.ZodNumber>;
1148
- vad_min_silence_duration_ms: z.ZodOptional<z.ZodNumber>;
1149
- vad_max_speech_duration_s: z.ZodOptional<z.ZodNumber>;
1150
- vad_speech_pad_ms: z.ZodOptional<z.ZodNumber>;
1151
- vad_samples_overlap: z.ZodOptional<z.ZodNumber>;
1474
+ strategy: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
1475
+ greedy: "greedy";
1476
+ beam_search: "beam_search";
1477
+ }>>>;
1478
+ n_threads: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
1479
+ n_max_text_ctx: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
1480
+ offset_ms: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
1481
+ duration_ms: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
1482
+ audio_ctx: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
1483
+ translate: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1484
+ no_context: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1485
+ no_timestamps: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1486
+ single_segment: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1487
+ print_special: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1488
+ print_progress: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1489
+ print_realtime: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1490
+ print_timestamps: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1491
+ token_timestamps: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1492
+ thold_pt: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
1493
+ thold_ptsum: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
1494
+ max_len: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
1495
+ split_on_word: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1496
+ max_tokens: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
1497
+ debug_mode: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1498
+ tdrz_enable: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1499
+ suppress_regex: z.ZodOptional<z.ZodOptional<z.ZodString>>;
1500
+ initial_prompt: z.ZodOptional<z.ZodOptional<z.ZodString>>;
1501
+ language: z.ZodOptional<z.ZodOptional<z.ZodString>>;
1502
+ detect_language: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1503
+ suppress_blank: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1504
+ suppress_nst: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1505
+ temperature: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
1506
+ length_penalty: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
1507
+ temperature_inc: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
1508
+ entropy_thold: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
1509
+ logprob_thold: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
1510
+ greedy_best_of: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
1511
+ beam_search_beam_size: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
1512
+ vadModelPath: z.ZodOptional<z.ZodOptional<z.ZodString>>;
1513
+ vad_params: z.ZodOptional<z.ZodOptional<z.ZodObject<{
1514
+ threshold: z.ZodOptional<z.ZodNumber>;
1515
+ min_speech_duration_ms: z.ZodOptional<z.ZodNumber>;
1516
+ min_silence_duration_ms: z.ZodOptional<z.ZodNumber>;
1517
+ max_speech_duration_s: z.ZodOptional<z.ZodNumber>;
1518
+ speech_pad_ms: z.ZodOptional<z.ZodNumber>;
1519
+ samples_overlap: z.ZodOptional<z.ZodNumber>;
1520
+ }, z.core.$strip>>>;
1521
+ max_seconds: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
1522
+ chunk_length_s: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
1523
+ audio_format: z.ZodOptional<z.ZodOptional<z.ZodString>>;
1524
+ contextParams: z.ZodOptional<z.ZodOptional<z.ZodObject<{
1525
+ model: z.ZodOptional<z.ZodString>;
1526
+ use_gpu: z.ZodOptional<z.ZodBoolean>;
1527
+ flash_attn: z.ZodOptional<z.ZodBoolean>;
1528
+ gpu_device: z.ZodOptional<z.ZodNumber>;
1529
+ }, z.core.$strip>>>;
1530
+ miscConfig: z.ZodOptional<z.ZodOptional<z.ZodObject<{
1531
+ caption_enabled: z.ZodOptional<z.ZodBoolean>;
1152
1532
  }, z.core.$strip>>>;
1153
1533
  }, z.core.$strict>>;
1154
1534
  seed: z.ZodOptional<z.ZodBoolean>;
@@ -1323,16 +1703,62 @@ export declare const loadModelServerParamsSchema: z.ZodObject<{
1323
1703
  mode?: "caption" | "batch" | undefined;
1324
1704
  output_format?: "plaintext" | "json" | undefined;
1325
1705
  min_seconds?: number | undefined;
1706
+ update_frequency?: string | undefined;
1707
+ strategy?: "greedy" | "beam_search" | undefined;
1708
+ n_threads?: number | undefined;
1709
+ n_max_text_ctx?: number | undefined;
1710
+ offset_ms?: number | undefined;
1711
+ duration_ms?: number | undefined;
1712
+ audio_ctx?: number | undefined;
1713
+ translate?: boolean | undefined;
1714
+ no_context?: boolean | undefined;
1715
+ no_timestamps?: boolean | undefined;
1716
+ single_segment?: boolean | undefined;
1717
+ print_special?: boolean | undefined;
1718
+ print_progress?: boolean | undefined;
1719
+ print_realtime?: boolean | undefined;
1720
+ print_timestamps?: boolean | undefined;
1721
+ token_timestamps?: boolean | undefined;
1722
+ thold_pt?: number | undefined;
1723
+ thold_ptsum?: number | undefined;
1724
+ max_len?: number | undefined;
1725
+ split_on_word?: boolean | undefined;
1726
+ max_tokens?: number | undefined;
1727
+ debug_mode?: boolean | undefined;
1728
+ tdrz_enable?: boolean | undefined;
1729
+ suppress_regex?: string | undefined;
1730
+ initial_prompt?: string | undefined;
1731
+ language?: string | undefined;
1732
+ detect_language?: boolean | undefined;
1733
+ suppress_blank?: boolean | undefined;
1734
+ suppress_nst?: boolean | undefined;
1735
+ temperature?: number | undefined;
1736
+ length_penalty?: number | undefined;
1737
+ temperature_inc?: number | undefined;
1738
+ entropy_thold?: number | undefined;
1739
+ logprob_thold?: number | undefined;
1740
+ greedy_best_of?: number | undefined;
1741
+ beam_search_beam_size?: number | undefined;
1742
+ vadModelPath?: string | undefined;
1743
+ vad_params?: {
1744
+ threshold?: number | undefined;
1745
+ min_speech_duration_ms?: number | undefined;
1746
+ min_silence_duration_ms?: number | undefined;
1747
+ max_speech_duration_s?: number | undefined;
1748
+ speech_pad_ms?: number | undefined;
1749
+ samples_overlap?: number | undefined;
1750
+ } | undefined;
1326
1751
  max_seconds?: number | undefined;
1752
+ chunk_length_s?: number | undefined;
1327
1753
  audio_format?: string | undefined;
1328
- update_frequency?: string | undefined;
1329
- vadConfig?: {
1330
- vad_threshold?: number | undefined;
1331
- vad_min_speech_duration_ms?: number | undefined;
1332
- vad_min_silence_duration_ms?: number | undefined;
1333
- vad_max_speech_duration_s?: number | undefined;
1334
- vad_speech_pad_ms?: number | undefined;
1335
- vad_samples_overlap?: number | undefined;
1754
+ contextParams?: {
1755
+ model?: string | undefined;
1756
+ use_gpu?: boolean | undefined;
1757
+ flash_attn?: boolean | undefined;
1758
+ gpu_device?: number | undefined;
1759
+ } | undefined;
1760
+ miscConfig?: {
1761
+ caption_enabled?: boolean | undefined;
1336
1762
  } | undefined;
1337
1763
  } | undefined;
1338
1764
  vadModelSrc?: string | undefined;
@@ -1424,16 +1850,62 @@ export declare const loadModelServerParamsSchema: z.ZodObject<{
1424
1850
  mode?: "caption" | "batch" | undefined;
1425
1851
  output_format?: "plaintext" | "json" | undefined;
1426
1852
  min_seconds?: number | undefined;
1853
+ update_frequency?: string | undefined;
1854
+ strategy?: "greedy" | "beam_search" | undefined;
1855
+ n_threads?: number | undefined;
1856
+ n_max_text_ctx?: number | undefined;
1857
+ offset_ms?: number | undefined;
1858
+ duration_ms?: number | undefined;
1859
+ audio_ctx?: number | undefined;
1860
+ translate?: boolean | undefined;
1861
+ no_context?: boolean | undefined;
1862
+ no_timestamps?: boolean | undefined;
1863
+ single_segment?: boolean | undefined;
1864
+ print_special?: boolean | undefined;
1865
+ print_progress?: boolean | undefined;
1866
+ print_realtime?: boolean | undefined;
1867
+ print_timestamps?: boolean | undefined;
1868
+ token_timestamps?: boolean | undefined;
1869
+ thold_pt?: number | undefined;
1870
+ thold_ptsum?: number | undefined;
1871
+ max_len?: number | undefined;
1872
+ split_on_word?: boolean | undefined;
1873
+ max_tokens?: number | undefined;
1874
+ debug_mode?: boolean | undefined;
1875
+ tdrz_enable?: boolean | undefined;
1876
+ suppress_regex?: string | undefined;
1877
+ initial_prompt?: string | undefined;
1878
+ language?: string | undefined;
1879
+ detect_language?: boolean | undefined;
1880
+ suppress_blank?: boolean | undefined;
1881
+ suppress_nst?: boolean | undefined;
1882
+ temperature?: number | undefined;
1883
+ length_penalty?: number | undefined;
1884
+ temperature_inc?: number | undefined;
1885
+ entropy_thold?: number | undefined;
1886
+ logprob_thold?: number | undefined;
1887
+ greedy_best_of?: number | undefined;
1888
+ beam_search_beam_size?: number | undefined;
1889
+ vadModelPath?: string | undefined;
1890
+ vad_params?: {
1891
+ threshold?: number | undefined;
1892
+ min_speech_duration_ms?: number | undefined;
1893
+ min_silence_duration_ms?: number | undefined;
1894
+ max_speech_duration_s?: number | undefined;
1895
+ speech_pad_ms?: number | undefined;
1896
+ samples_overlap?: number | undefined;
1897
+ } | undefined;
1427
1898
  max_seconds?: number | undefined;
1899
+ chunk_length_s?: number | undefined;
1428
1900
  audio_format?: string | undefined;
1429
- update_frequency?: string | undefined;
1430
- vadConfig?: {
1431
- vad_threshold?: number | undefined;
1432
- vad_min_speech_duration_ms?: number | undefined;
1433
- vad_min_silence_duration_ms?: number | undefined;
1434
- vad_max_speech_duration_s?: number | undefined;
1435
- vad_speech_pad_ms?: number | undefined;
1436
- vad_samples_overlap?: number | undefined;
1901
+ contextParams?: {
1902
+ model?: string | undefined;
1903
+ use_gpu?: boolean | undefined;
1904
+ flash_attn?: boolean | undefined;
1905
+ gpu_device?: number | undefined;
1906
+ } | undefined;
1907
+ miscConfig?: {
1908
+ caption_enabled?: boolean | undefined;
1437
1909
  } | undefined;
1438
1910
  } | undefined;
1439
1911
  seed?: boolean | undefined;