@qvac/sdk 0.13.1 → 0.13.3
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.
|
@@ -1,12 +1,28 @@
|
|
|
1
1
|
import { z } from "zod";
|
|
2
|
-
export declare const
|
|
2
|
+
export declare const TTS_CHATTERBOX_LANGUAGES: readonly ["en", "es", "fr", "de", "it", "pt", "nl", "pl", "tr", "sv", "da", "fi", "no", "el", "ms", "sw", "ar", "ko"];
|
|
3
|
+
export declare const TTS_SUPERTONIC_LANGUAGES: readonly ["en", "es", "fr", "pt", "ko"];
|
|
4
|
+
export declare const TTS_LANGUAGES: readonly ["en", "es", "fr", "de", "it", "pt", "nl", "pl", "tr", "sv", "da", "fi", "no", "el", "ms", "sw", "ar", "ko"];
|
|
3
5
|
export declare const ttsChatterboxRuntimeConfigSchema: z.ZodObject<{
|
|
4
6
|
ttsEngine: z.ZodLiteral<"chatterbox">;
|
|
5
7
|
language: z.ZodEnum<{
|
|
6
8
|
en: "en";
|
|
9
|
+
ar: "ar";
|
|
7
10
|
de: "de";
|
|
8
11
|
es: "es";
|
|
12
|
+
fi: "fi";
|
|
13
|
+
fr: "fr";
|
|
9
14
|
it: "it";
|
|
15
|
+
ko: "ko";
|
|
16
|
+
nl: "nl";
|
|
17
|
+
pl: "pl";
|
|
18
|
+
pt: "pt";
|
|
19
|
+
da: "da";
|
|
20
|
+
el: "el";
|
|
21
|
+
ms: "ms";
|
|
22
|
+
no: "no";
|
|
23
|
+
sv: "sv";
|
|
24
|
+
tr: "tr";
|
|
25
|
+
sw: "sw";
|
|
10
26
|
}>;
|
|
11
27
|
voice: z.ZodOptional<z.ZodString>;
|
|
12
28
|
useGPU: z.ZodOptional<z.ZodBoolean>;
|
|
@@ -15,9 +31,10 @@ export declare const ttsSupertonicRuntimeConfigSchema: z.ZodObject<{
|
|
|
15
31
|
ttsEngine: z.ZodLiteral<"supertonic">;
|
|
16
32
|
language: z.ZodEnum<{
|
|
17
33
|
en: "en";
|
|
18
|
-
de: "de";
|
|
19
34
|
es: "es";
|
|
20
|
-
|
|
35
|
+
fr: "fr";
|
|
36
|
+
ko: "ko";
|
|
37
|
+
pt: "pt";
|
|
21
38
|
}>;
|
|
22
39
|
voice: z.ZodOptional<z.ZodString>;
|
|
23
40
|
ttsSpeed: z.ZodOptional<z.ZodNumber>;
|
|
@@ -28,9 +45,23 @@ export declare const ttsRuntimeConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
|
|
|
28
45
|
ttsEngine: z.ZodLiteral<"chatterbox">;
|
|
29
46
|
language: z.ZodEnum<{
|
|
30
47
|
en: "en";
|
|
48
|
+
ar: "ar";
|
|
31
49
|
de: "de";
|
|
32
50
|
es: "es";
|
|
51
|
+
fi: "fi";
|
|
52
|
+
fr: "fr";
|
|
33
53
|
it: "it";
|
|
54
|
+
ko: "ko";
|
|
55
|
+
nl: "nl";
|
|
56
|
+
pl: "pl";
|
|
57
|
+
pt: "pt";
|
|
58
|
+
da: "da";
|
|
59
|
+
el: "el";
|
|
60
|
+
ms: "ms";
|
|
61
|
+
no: "no";
|
|
62
|
+
sv: "sv";
|
|
63
|
+
tr: "tr";
|
|
64
|
+
sw: "sw";
|
|
34
65
|
}>;
|
|
35
66
|
voice: z.ZodOptional<z.ZodString>;
|
|
36
67
|
useGPU: z.ZodOptional<z.ZodBoolean>;
|
|
@@ -38,9 +69,10 @@ export declare const ttsRuntimeConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
|
|
|
38
69
|
ttsEngine: z.ZodLiteral<"supertonic">;
|
|
39
70
|
language: z.ZodEnum<{
|
|
40
71
|
en: "en";
|
|
41
|
-
de: "de";
|
|
42
72
|
es: "es";
|
|
43
|
-
|
|
73
|
+
fr: "fr";
|
|
74
|
+
ko: "ko";
|
|
75
|
+
pt: "pt";
|
|
44
76
|
}>;
|
|
45
77
|
voice: z.ZodOptional<z.ZodString>;
|
|
46
78
|
ttsSpeed: z.ZodOptional<z.ZodNumber>;
|
|
@@ -51,9 +83,23 @@ export declare const ttsChatterboxLoadConfigSchema: z.ZodObject<{
|
|
|
51
83
|
ttsEngine: z.ZodLiteral<"chatterbox">;
|
|
52
84
|
language: z.ZodEnum<{
|
|
53
85
|
en: "en";
|
|
86
|
+
ar: "ar";
|
|
54
87
|
de: "de";
|
|
55
88
|
es: "es";
|
|
89
|
+
fi: "fi";
|
|
90
|
+
fr: "fr";
|
|
56
91
|
it: "it";
|
|
92
|
+
ko: "ko";
|
|
93
|
+
nl: "nl";
|
|
94
|
+
pl: "pl";
|
|
95
|
+
pt: "pt";
|
|
96
|
+
da: "da";
|
|
97
|
+
el: "el";
|
|
98
|
+
ms: "ms";
|
|
99
|
+
no: "no";
|
|
100
|
+
sv: "sv";
|
|
101
|
+
tr: "tr";
|
|
102
|
+
sw: "sw";
|
|
57
103
|
}>;
|
|
58
104
|
voice: z.ZodOptional<z.ZodString>;
|
|
59
105
|
useGPU: z.ZodOptional<z.ZodBoolean>;
|
|
@@ -136,9 +182,10 @@ export declare const ttsSupertonicLoadConfigSchema: z.ZodObject<{
|
|
|
136
182
|
ttsEngine: z.ZodLiteral<"supertonic">;
|
|
137
183
|
language: z.ZodEnum<{
|
|
138
184
|
en: "en";
|
|
139
|
-
de: "de";
|
|
140
185
|
es: "es";
|
|
141
|
-
|
|
186
|
+
fr: "fr";
|
|
187
|
+
ko: "ko";
|
|
188
|
+
pt: "pt";
|
|
142
189
|
}>;
|
|
143
190
|
voice: z.ZodOptional<z.ZodString>;
|
|
144
191
|
ttsSpeed: z.ZodOptional<z.ZodNumber>;
|
|
@@ -149,9 +196,23 @@ export declare const ttsLoadConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
|
|
|
149
196
|
ttsEngine: z.ZodLiteral<"chatterbox">;
|
|
150
197
|
language: z.ZodEnum<{
|
|
151
198
|
en: "en";
|
|
199
|
+
ar: "ar";
|
|
152
200
|
de: "de";
|
|
153
201
|
es: "es";
|
|
202
|
+
fi: "fi";
|
|
203
|
+
fr: "fr";
|
|
154
204
|
it: "it";
|
|
205
|
+
ko: "ko";
|
|
206
|
+
nl: "nl";
|
|
207
|
+
pl: "pl";
|
|
208
|
+
pt: "pt";
|
|
209
|
+
da: "da";
|
|
210
|
+
el: "el";
|
|
211
|
+
ms: "ms";
|
|
212
|
+
no: "no";
|
|
213
|
+
sv: "sv";
|
|
214
|
+
tr: "tr";
|
|
215
|
+
sw: "sw";
|
|
155
216
|
}>;
|
|
156
217
|
voice: z.ZodOptional<z.ZodString>;
|
|
157
218
|
useGPU: z.ZodOptional<z.ZodBoolean>;
|
|
@@ -233,9 +294,10 @@ export declare const ttsLoadConfigSchema: z.ZodUnion<readonly [z.ZodObject<{
|
|
|
233
294
|
ttsEngine: z.ZodLiteral<"supertonic">;
|
|
234
295
|
language: z.ZodEnum<{
|
|
235
296
|
en: "en";
|
|
236
|
-
de: "de";
|
|
237
297
|
es: "es";
|
|
238
|
-
|
|
298
|
+
fr: "fr";
|
|
299
|
+
ko: "ko";
|
|
300
|
+
pt: "pt";
|
|
239
301
|
}>;
|
|
240
302
|
voice: z.ZodOptional<z.ZodString>;
|
|
241
303
|
ttsSpeed: z.ZodOptional<z.ZodNumber>;
|
|
@@ -319,6 +381,8 @@ export declare const textToSpeechStreamResponseSchema: z.ZodObject<{
|
|
|
319
381
|
sentenceChunk: z.ZodOptional<z.ZodString>;
|
|
320
382
|
}, z.core.$strip>;
|
|
321
383
|
export type TtsLanguage = (typeof TTS_LANGUAGES)[number];
|
|
384
|
+
export type TtsChatterboxLanguage = (typeof TTS_CHATTERBOX_LANGUAGES)[number];
|
|
385
|
+
export type TtsSupertonicLanguage = (typeof TTS_SUPERTONIC_LANGUAGES)[number];
|
|
322
386
|
export type TtsChatterboxLoadConfig = z.infer<typeof ttsChatterboxLoadConfigSchema>;
|
|
323
387
|
export type TtsSupertonicLoadConfig = z.infer<typeof ttsSupertonicLoadConfigSchema>;
|
|
324
388
|
export type TtsLoadConfig = z.infer<typeof ttsLoadConfigSchema>;
|
|
@@ -1,22 +1,49 @@
|
|
|
1
1
|
import { z } from "zod";
|
|
2
2
|
import { modelSrcInputSchema } from "./model-src-utils.js";
|
|
3
|
-
//
|
|
4
|
-
|
|
3
|
+
// Chatterbox multilingual supported languages (18). The engines support
|
|
4
|
+
// different language sets, so the language enum is validated per engine.
|
|
5
|
+
export const TTS_CHATTERBOX_LANGUAGES = [
|
|
5
6
|
"en", // English
|
|
6
7
|
"es", // Spanish
|
|
8
|
+
"fr", // French
|
|
7
9
|
"de", // German
|
|
8
10
|
"it", // Italian
|
|
11
|
+
"pt", // Portuguese
|
|
12
|
+
"nl", // Dutch
|
|
13
|
+
"pl", // Polish
|
|
14
|
+
"tr", // Turkish
|
|
15
|
+
"sv", // Swedish
|
|
16
|
+
"da", // Danish
|
|
17
|
+
"fi", // Finnish
|
|
18
|
+
"no", // Norwegian
|
|
19
|
+
"el", // Greek
|
|
20
|
+
"ms", // Malay
|
|
21
|
+
"sw", // Swahili
|
|
22
|
+
"ar", // Arabic
|
|
23
|
+
"ko", // Korean
|
|
9
24
|
];
|
|
10
|
-
|
|
25
|
+
// Supertonic supported languages (subset of the Chatterbox set).
|
|
26
|
+
export const TTS_SUPERTONIC_LANGUAGES = [
|
|
27
|
+
"en", // English
|
|
28
|
+
"es", // Spanish
|
|
29
|
+
"fr", // French
|
|
30
|
+
"pt", // Portuguese
|
|
31
|
+
"ko", // Korean
|
|
32
|
+
];
|
|
33
|
+
// Union of all TTS-supported languages across engines. Kept for backwards
|
|
34
|
+
// compatibility; prefer the engine-specific lists when validating a config.
|
|
35
|
+
export const TTS_LANGUAGES = [...TTS_CHATTERBOX_LANGUAGES];
|
|
36
|
+
const ttsChatterboxLanguageSchema = z.enum(TTS_CHATTERBOX_LANGUAGES);
|
|
37
|
+
const ttsSupertonicLanguageSchema = z.enum(TTS_SUPERTONIC_LANGUAGES);
|
|
11
38
|
export const ttsChatterboxRuntimeConfigSchema = z.object({
|
|
12
39
|
ttsEngine: z.literal("chatterbox"),
|
|
13
|
-
language:
|
|
40
|
+
language: ttsChatterboxLanguageSchema,
|
|
14
41
|
voice: z.string().optional(),
|
|
15
42
|
useGPU: z.boolean().optional(),
|
|
16
43
|
});
|
|
17
44
|
export const ttsSupertonicRuntimeConfigSchema = z.object({
|
|
18
45
|
ttsEngine: z.literal("supertonic"),
|
|
19
|
-
language:
|
|
46
|
+
language: ttsSupertonicLanguageSchema,
|
|
20
47
|
voice: z.string().optional(),
|
|
21
48
|
ttsSpeed: z.number().optional(),
|
|
22
49
|
ttsNumInferenceSteps: z.number().optional(),
|
|
@@ -10,14 +10,14 @@ export declare const ttsPlugin: {
|
|
|
10
10
|
}, import("zod/v4/core").$strict>]>;
|
|
11
11
|
resolveConfig(cfg: Record<string, unknown>, ctx: ResolveContext): Promise<ResolveResult<{
|
|
12
12
|
ttsEngine: "supertonic";
|
|
13
|
-
language: "en" | "
|
|
13
|
+
language: "en" | "es" | "fr" | "ko" | "pt";
|
|
14
14
|
voice?: string | undefined;
|
|
15
15
|
ttsSpeed?: number | undefined;
|
|
16
16
|
ttsNumInferenceSteps?: number | undefined;
|
|
17
17
|
useGPU?: boolean | undefined;
|
|
18
18
|
} | {
|
|
19
19
|
ttsEngine: "chatterbox";
|
|
20
|
-
language: "en" | "de" | "es" | "it";
|
|
20
|
+
language: "en" | "ar" | "de" | "es" | "fi" | "fr" | "it" | "ko" | "nl" | "pl" | "pt" | "da" | "el" | "ms" | "no" | "sv" | "tr" | "sw";
|
|
21
21
|
voice?: string | undefined;
|
|
22
22
|
useGPU?: boolean | undefined;
|
|
23
23
|
}, string>>;
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@qvac/sdk",
|
|
3
|
-
"version": "0.13.
|
|
3
|
+
"version": "0.13.3",
|
|
4
4
|
"license": "Apache-2.0",
|
|
5
5
|
"repository": {
|
|
6
6
|
"type": "git",
|
|
@@ -208,7 +208,7 @@
|
|
|
208
208
|
"dependencies": {
|
|
209
209
|
"@qvac/bci-whispercpp": "^0.2.0",
|
|
210
210
|
"@qvac/classification-ggml": "^0.3.1",
|
|
211
|
-
"@qvac/decoder-audio": "^0.
|
|
211
|
+
"@qvac/decoder-audio": "^0.5.0",
|
|
212
212
|
"@qvac/diffusion-cpp": "^0.11.2",
|
|
213
213
|
"@qvac/embed-llamacpp": "^0.19.1",
|
|
214
214
|
"@qvac/error": "^0.1.1",
|
|
@@ -216,7 +216,7 @@
|
|
|
216
216
|
"@qvac/llm-llamacpp": "^0.24.0",
|
|
217
217
|
"@qvac/logging": "^0.1.0",
|
|
218
218
|
"@qvac/ocr-onnx": "^0.6.0",
|
|
219
|
-
"@qvac/rag": "^0.6.
|
|
219
|
+
"@qvac/rag": "^0.6.4",
|
|
220
220
|
"@qvac/registry-client": "^0.6.0",
|
|
221
221
|
"@qvac/transcription-parakeet": "^0.7.1",
|
|
222
222
|
"@qvac/transcription-whispercpp": "^0.9.0",
|