@speech-sdk/core 0.7.0 → 0.8.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +227 -108
- package/dist/__tests__/e2e/_save-audio.d.ts +0 -42
- package/dist/__tests__/e2e/_save-audio.d.ts.map +1 -1
- package/dist/__tests__/e2e/_save-audio.js +0 -59
- package/dist/__tests__/e2e/_save-audio.js.map +1 -1
- package/dist/audio-decode.d.ts +7 -0
- package/dist/audio-decode.d.ts.map +1 -0
- package/dist/audio-decode.js +109 -0
- package/dist/audio-decode.js.map +1 -0
- package/dist/audio-duration.d.ts +0 -5
- package/dist/audio-duration.d.ts.map +1 -1
- package/dist/audio-duration.js +5 -21
- package/dist/audio-duration.js.map +1 -1
- package/dist/audio-output.d.ts +39 -0
- package/dist/audio-output.d.ts.map +1 -0
- package/dist/audio-output.js +111 -0
- package/dist/audio-output.js.map +1 -0
- package/dist/audio-utils.d.ts +2 -10
- package/dist/audio-utils.d.ts.map +1 -1
- package/dist/audio-utils.js +57 -15
- package/dist/audio-utils.js.map +1 -1
- package/dist/captions.d.ts +0 -108
- package/dist/captions.d.ts.map +1 -1
- package/dist/captions.js +8 -98
- package/dist/captions.js.map +1 -1
- package/dist/conversation/attribute-timestamps.d.ts +26 -0
- package/dist/conversation/attribute-timestamps.d.ts.map +1 -0
- package/dist/conversation/attribute-timestamps.js +276 -0
- package/dist/conversation/attribute-timestamps.js.map +1 -0
- package/dist/conversation/dispatch.d.ts +5 -5
- package/dist/conversation/dispatch.d.ts.map +1 -1
- package/dist/conversation/dispatch.js +18 -8
- package/dist/conversation/dispatch.js.map +1 -1
- package/dist/conversation/errors.d.ts +3 -0
- package/dist/conversation/errors.d.ts.map +1 -1
- package/dist/conversation/errors.js +6 -0
- package/dist/conversation/errors.js.map +1 -1
- package/dist/conversation/pcm-concat.d.ts +0 -24
- package/dist/conversation/pcm-concat.d.ts.map +1 -1
- package/dist/conversation/pcm-concat.js +8 -183
- package/dist/conversation/pcm-concat.js.map +1 -1
- package/dist/conversation/proportional-fill.d.ts +10 -0
- package/dist/conversation/proportional-fill.d.ts.map +1 -0
- package/dist/conversation/proportional-fill.js +64 -0
- package/dist/conversation/proportional-fill.js.map +1 -0
- package/dist/conversation/silence-detection.d.ts +14 -0
- package/dist/conversation/silence-detection.d.ts.map +1 -0
- package/dist/conversation/silence-detection.js +52 -0
- package/dist/conversation/silence-detection.js.map +1 -0
- package/dist/conversation/stitch.d.ts +9 -6
- package/dist/conversation/stitch.d.ts.map +1 -1
- package/dist/conversation/stitch.js +72 -51
- package/dist/conversation/stitch.js.map +1 -1
- package/dist/conversation/types.d.ts +7 -37
- package/dist/conversation/types.d.ts.map +1 -1
- package/dist/conversation/validate.d.ts +1 -16
- package/dist/conversation/validate.d.ts.map +1 -1
- package/dist/conversation/validate.js +29 -29
- package/dist/conversation/validate.js.map +1 -1
- package/dist/default-stt-fallback.d.ts +3 -0
- package/dist/default-stt-fallback.d.ts.map +1 -0
- package/dist/default-stt-fallback.js +11 -0
- package/dist/default-stt-fallback.js.map +1 -0
- package/dist/derive-timestamps.d.ts +1 -5
- package/dist/derive-timestamps.d.ts.map +1 -1
- package/dist/derive-timestamps.js +1 -15
- package/dist/derive-timestamps.js.map +1 -1
- package/dist/encoders/mp3.d.ts +6 -0
- package/dist/encoders/mp3.d.ts.map +1 -0
- package/dist/encoders/mp3.js +54 -0
- package/dist/encoders/mp3.js.map +1 -0
- package/dist/errors.d.ts +20 -13
- package/dist/errors.d.ts.map +1 -1
- package/dist/errors.js +49 -15
- package/dist/errors.js.map +1 -1
- package/dist/generate-conversation.d.ts +5 -4
- package/dist/generate-conversation.d.ts.map +1 -1
- package/dist/generate-conversation.js +250 -93
- package/dist/generate-conversation.js.map +1 -1
- package/dist/generate-speech.d.ts +7 -28
- package/dist/generate-speech.d.ts.map +1 -1
- package/dist/generate-speech.js +185 -94
- package/dist/generate-speech.js.map +1 -1
- package/dist/index.d.ts +7 -11
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +6 -4
- package/dist/index.js.map +1 -1
- package/dist/logger.d.ts.map +1 -1
- package/dist/logger.js +2 -13
- package/dist/logger.js.map +1 -1
- package/dist/metadata.d.ts +0 -22
- package/dist/metadata.d.ts.map +1 -1
- package/dist/pronunciations/errors.d.ts +5 -0
- package/dist/pronunciations/errors.d.ts.map +1 -0
- package/dist/pronunciations/errors.js +8 -0
- package/dist/pronunciations/errors.js.map +1 -0
- package/dist/pronunciations/inverse-align.d.ts +4 -0
- package/dist/pronunciations/inverse-align.d.ts.map +1 -0
- package/dist/pronunciations/inverse-align.js +54 -0
- package/dist/pronunciations/inverse-align.js.map +1 -0
- package/dist/pronunciations/merge.d.ts +4 -0
- package/dist/pronunciations/merge.d.ts.map +1 -0
- package/dist/pronunciations/merge.js +13 -0
- package/dist/pronunciations/merge.js.map +1 -0
- package/dist/pronunciations/substitute.d.ts +6 -0
- package/dist/pronunciations/substitute.d.ts.map +1 -0
- package/dist/pronunciations/substitute.js +67 -0
- package/dist/pronunciations/substitute.js.map +1 -0
- package/dist/pronunciations/types.d.ts +18 -0
- package/dist/pronunciations/types.d.ts.map +1 -0
- package/dist/pronunciations/types.js +2 -0
- package/dist/pronunciations/types.js.map +1 -0
- package/dist/pronunciations/validate.d.ts +3 -0
- package/dist/pronunciations/validate.d.ts.map +1 -0
- package/dist/pronunciations/validate.js +26 -0
- package/dist/pronunciations/validate.js.map +1 -0
- package/dist/provider-utils.d.ts +4 -9
- package/dist/provider-utils.d.ts.map +1 -1
- package/dist/provider-utils.js +60 -51
- package/dist/provider-utils.js.map +1 -1
- package/dist/providers/cartesia/alignment.d.ts +0 -16
- package/dist/providers/cartesia/alignment.d.ts.map +1 -1
- package/dist/providers/cartesia/alignment.js +1 -6
- package/dist/providers/cartesia/alignment.js.map +1 -1
- package/dist/providers/cartesia/index.d.ts +29 -19
- package/dist/providers/cartesia/index.d.ts.map +1 -1
- package/dist/providers/cartesia/index.js +116 -80
- package/dist/providers/cartesia/index.js.map +1 -1
- package/dist/providers/deepgram/index.d.ts +23 -8
- package/dist/providers/deepgram/index.d.ts.map +1 -1
- package/dist/providers/deepgram/index.js +51 -18
- package/dist/providers/deepgram/index.js.map +1 -1
- package/dist/providers/elevenlabs/alignment.d.ts +7 -21
- package/dist/providers/elevenlabs/alignment.d.ts.map +1 -1
- package/dist/providers/elevenlabs/alignment.js +8 -9
- package/dist/providers/elevenlabs/alignment.js.map +1 -1
- package/dist/providers/elevenlabs/index.d.ts +14 -38
- package/dist/providers/elevenlabs/index.d.ts.map +1 -1
- package/dist/providers/elevenlabs/index.js +186 -169
- package/dist/providers/elevenlabs/index.js.map +1 -1
- package/dist/providers/fal/index.d.ts +11 -20
- package/dist/providers/fal/index.d.ts.map +1 -1
- package/dist/providers/fal/index.js +49 -37
- package/dist/providers/fal/index.js.map +1 -1
- package/dist/providers/fish-audio/index.d.ts +14 -8
- package/dist/providers/fish-audio/index.d.ts.map +1 -1
- package/dist/providers/fish-audio/index.js +47 -19
- package/dist/providers/fish-audio/index.js.map +1 -1
- package/dist/providers/gateway/index.d.ts +76 -0
- package/dist/providers/gateway/index.d.ts.map +1 -0
- package/dist/providers/gateway/index.js +251 -0
- package/dist/providers/gateway/index.js.map +1 -0
- package/dist/providers/google/index.d.ts +12 -20
- package/dist/providers/google/index.d.ts.map +1 -1
- package/dist/providers/google/index.js +180 -162
- package/dist/providers/google/index.js.map +1 -1
- package/dist/providers/hume/alignment.d.ts +30 -35
- package/dist/providers/hume/alignment.d.ts.map +1 -1
- package/dist/providers/hume/alignment.js +14 -8
- package/dist/providers/hume/alignment.js.map +1 -1
- package/dist/providers/hume/index.d.ts +16 -16
- package/dist/providers/hume/index.d.ts.map +1 -1
- package/dist/providers/hume/index.js +79 -65
- package/dist/providers/hume/index.js.map +1 -1
- package/dist/providers/inworld/alignment.d.ts +8 -22
- package/dist/providers/inworld/alignment.d.ts.map +1 -1
- package/dist/providers/inworld/alignment.js +9 -8
- package/dist/providers/inworld/alignment.js.map +1 -1
- package/dist/providers/inworld/index.d.ts +17 -20
- package/dist/providers/inworld/index.d.ts.map +1 -1
- package/dist/providers/inworld/index.js +79 -47
- package/dist/providers/inworld/index.js.map +1 -1
- package/dist/providers/mistral/index.d.ts +14 -8
- package/dist/providers/mistral/index.d.ts.map +1 -1
- package/dist/providers/mistral/index.js +63 -48
- package/dist/providers/mistral/index.js.map +1 -1
- package/dist/providers/murf/alignment.d.ts +10 -19
- package/dist/providers/murf/alignment.d.ts.map +1 -1
- package/dist/providers/murf/alignment.js +10 -5
- package/dist/providers/murf/alignment.js.map +1 -1
- package/dist/providers/murf/index.d.ts +15 -16
- package/dist/providers/murf/index.d.ts.map +1 -1
- package/dist/providers/murf/index.js +105 -58
- package/dist/providers/murf/index.js.map +1 -1
- package/dist/providers/openai/index.d.ts +43 -29
- package/dist/providers/openai/index.d.ts.map +1 -1
- package/dist/providers/openai/index.js +294 -106
- package/dist/providers/openai/index.js.map +1 -1
- package/dist/providers/resemble/alignment.d.ts +8 -29
- package/dist/providers/resemble/alignment.d.ts.map +1 -1
- package/dist/providers/resemble/alignment.js +9 -12
- package/dist/providers/resemble/alignment.js.map +1 -1
- package/dist/providers/resemble/index.d.ts +21 -11
- package/dist/providers/resemble/index.d.ts.map +1 -1
- package/dist/providers/resemble/index.js +89 -49
- package/dist/providers/resemble/index.js.map +1 -1
- package/dist/providers/smallest-ai/index.d.ts +47 -0
- package/dist/providers/smallest-ai/index.d.ts.map +1 -0
- package/dist/providers/smallest-ai/index.js +107 -0
- package/dist/providers/smallest-ai/index.js.map +1 -0
- package/dist/providers/xai/index.d.ts +25 -9
- package/dist/providers/xai/index.d.ts.map +1 -1
- package/dist/providers/xai/index.js +63 -40
- package/dist/providers/xai/index.js.map +1 -1
- package/dist/providers.d.ts +31 -0
- package/dist/providers.d.ts.map +1 -0
- package/dist/providers.js +16 -0
- package/dist/providers.js.map +1 -0
- package/dist/resolve-provider.d.ts.map +1 -1
- package/dist/resolve-provider.js +8 -51
- package/dist/resolve-provider.js.map +1 -1
- package/dist/retry-options.d.ts +6 -0
- package/dist/retry-options.d.ts.map +1 -0
- package/dist/retry-options.js +48 -0
- package/dist/retry-options.js.map +1 -0
- package/dist/speech-provider.d.ts +28 -53
- package/dist/speech-provider.d.ts.map +1 -1
- package/dist/speech-provider.js +5 -26
- package/dist/speech-provider.js.map +1 -1
- package/dist/speech-result.d.ts +8 -9
- package/dist/speech-result.d.ts.map +1 -1
- package/dist/speech-result.js.map +1 -1
- package/dist/speech-to-text-provider.d.ts +0 -12
- package/dist/speech-to-text-provider.d.ts.map +1 -1
- package/dist/stream-speech.d.ts +4 -2
- package/dist/stream-speech.d.ts.map +1 -1
- package/dist/stream-speech.js +36 -22
- package/dist/stream-speech.js.map +1 -1
- package/dist/timestamps.d.ts +3 -17
- package/dist/timestamps.d.ts.map +1 -1
- package/dist/turns.d.ts +9 -0
- package/dist/turns.d.ts.map +1 -0
- package/dist/turns.js +21 -0
- package/dist/turns.js.map +1 -0
- package/dist/types.d.ts +31 -0
- package/dist/types.d.ts.map +1 -1
- package/dist/volume-adjust.d.ts +0 -6
- package/dist/volume-adjust.d.ts.map +1 -1
- package/dist/volume-adjust.js +4 -16
- package/dist/volume-adjust.js.map +1 -1
- package/package.json +13 -66
- package/dist/stt-providers/openai/index.d.ts +0 -42
- package/dist/stt-providers/openai/index.d.ts.map +0 -1
- package/dist/stt-providers/openai/index.js +0 -184
- package/dist/stt-providers/openai/index.js.map +0 -1
|
@@ -1,28 +1,37 @@
|
|
|
1
|
+
import { z } from "zod";
|
|
1
2
|
import { ApiError, StreamingNotSupportedError } from "../../errors.js";
|
|
2
3
|
import { handleErrorResponse, resolveApiKey, SDK_USER_AGENT, } from "../../provider-utils.js";
|
|
4
|
+
const falJobResponseSchema = z.object({
|
|
5
|
+
audio: z.object({
|
|
6
|
+
url: z.string(),
|
|
7
|
+
content_type: z.string().optional(),
|
|
8
|
+
}),
|
|
9
|
+
});
|
|
10
|
+
export const FAL_PROVIDER_ID = "fal-ai";
|
|
11
|
+
export const FAL_MODELS = [
|
|
12
|
+
{
|
|
13
|
+
id: "f5-tts",
|
|
14
|
+
releaseDate: "2024-10-08",
|
|
15
|
+
languages: ["en", "zh", "fr", "it", "hi", "ja", "ru", "es", "fi"],
|
|
16
|
+
features: ["open-source", "inline-voice-cloning"],
|
|
17
|
+
},
|
|
18
|
+
{
|
|
19
|
+
id: "kokoro",
|
|
20
|
+
releaseDate: "2025-01-27",
|
|
21
|
+
languages: ["en", "fr", "ko", "ja", "zh"],
|
|
22
|
+
features: ["open-source"],
|
|
23
|
+
},
|
|
24
|
+
{
|
|
25
|
+
id: "orpheus-tts",
|
|
26
|
+
releaseDate: "2025-03-18",
|
|
27
|
+
languages: ["en", "es", "fr", "de", "it", "pt", "zh"],
|
|
28
|
+
features: ["open-source"],
|
|
29
|
+
},
|
|
30
|
+
];
|
|
3
31
|
export class FalSpeechProvider {
|
|
4
|
-
id =
|
|
32
|
+
id = FAL_PROVIDER_ID;
|
|
5
33
|
defaultModel = "";
|
|
6
|
-
models =
|
|
7
|
-
{
|
|
8
|
-
id: "f5-tts",
|
|
9
|
-
releaseDate: "2024-10-08",
|
|
10
|
-
languages: ["en", "zh", "fr", "it", "hi", "ja", "ru", "es", "fi"],
|
|
11
|
-
features: ["open-source", "inline-voice-cloning"],
|
|
12
|
-
},
|
|
13
|
-
{
|
|
14
|
-
id: "kokoro",
|
|
15
|
-
releaseDate: "2025-01-27",
|
|
16
|
-
languages: ["en", "fr", "ko", "ja", "zh"],
|
|
17
|
-
features: ["open-source"],
|
|
18
|
-
},
|
|
19
|
-
{
|
|
20
|
-
id: "orpheus-tts",
|
|
21
|
-
releaseDate: "2025-03-18",
|
|
22
|
-
languages: ["en", "es", "fr", "de", "it", "pt", "zh"],
|
|
23
|
-
features: ["open-source"],
|
|
24
|
-
},
|
|
25
|
-
];
|
|
34
|
+
models = FAL_MODELS;
|
|
26
35
|
apiKey;
|
|
27
36
|
baseURL;
|
|
28
37
|
fetchFn;
|
|
@@ -59,8 +68,8 @@ export class FalSpeechProvider {
|
|
|
59
68
|
body: JSON.stringify(body),
|
|
60
69
|
signal: options.abortSignal,
|
|
61
70
|
});
|
|
62
|
-
await handleErrorResponse(response
|
|
63
|
-
const json = (await response.json());
|
|
71
|
+
await handleErrorResponse(response);
|
|
72
|
+
const json = falJobResponseSchema.parse(await response.json());
|
|
64
73
|
return await this.fetchAudio(json, options);
|
|
65
74
|
}
|
|
66
75
|
async fetchAudio(json, options) {
|
|
@@ -70,14 +79,11 @@ export class FalSpeechProvider {
|
|
|
70
79
|
if (!audioResponse.ok) {
|
|
71
80
|
throw new ApiError(`API error: ${audioResponse.status}`, {
|
|
72
81
|
statusCode: audioResponse.status,
|
|
73
|
-
model: `fal-ai/${options.modelId}`,
|
|
74
82
|
responseBody: await audioResponse.text().catch(() => undefined),
|
|
75
83
|
});
|
|
76
84
|
}
|
|
77
85
|
const arrayBuffer = await audioResponse.arrayBuffer();
|
|
78
|
-
// fal
|
|
79
|
-
// Fall back to the CDN response header, then to audio/wav (the format
|
|
80
|
-
// every currently-listed model emits).
|
|
86
|
+
// Authoritative content_type lives in fal's JSON; CDN header and audio/wav are fallbacks.
|
|
81
87
|
const mediaType = json.audio.content_type ??
|
|
82
88
|
audioResponse.headers.get("content-type") ??
|
|
83
89
|
"audio/wav";
|
|
@@ -86,25 +92,31 @@ export class FalSpeechProvider {
|
|
|
86
92
|
stream(options) {
|
|
87
93
|
return Promise.reject(new StreamingNotSupportedError(`fal-ai/${options.modelId}`));
|
|
88
94
|
}
|
|
89
|
-
|
|
90
|
-
|
|
91
|
-
|
|
92
|
-
|
|
93
|
-
|
|
94
|
-
|
|
95
|
-
|
|
96
|
-
|
|
97
|
-
|
|
95
|
+
// fal exposes no format selector and uses path-style model IDs (e.g.
|
|
96
|
+
// "kokoro/american-english"); the registered models[] list only enumerates
|
|
97
|
+
// the well-known prefixes, so we accept any modelId here. All current fal
|
|
98
|
+
// TTS endpoints return WAV.
|
|
99
|
+
getStitchOptions(_modelId) {
|
|
100
|
+
return {
|
|
101
|
+
providerOptions: {},
|
|
102
|
+
mediaType: "audio/wav",
|
|
103
|
+
};
|
|
104
|
+
}
|
|
105
|
+
resolveOutputFormat(_modelId, output) {
|
|
106
|
+
if (output.format === "wav") {
|
|
107
|
+
return { providerOptions: {}, expectedMediaType: "audio/wav" };
|
|
98
108
|
}
|
|
99
|
-
return
|
|
109
|
+
return;
|
|
100
110
|
}
|
|
101
111
|
}
|
|
102
112
|
export function createFal(config = {}) {
|
|
103
113
|
const provider = new FalSpeechProvider(config);
|
|
114
|
+
const fallbackSTT = config.fallbackSTT;
|
|
104
115
|
return function fal(modelId) {
|
|
105
116
|
return {
|
|
106
117
|
provider,
|
|
107
118
|
modelId: modelId ?? provider.defaultModel,
|
|
119
|
+
...(fallbackSTT && { fallbackSTT }),
|
|
108
120
|
};
|
|
109
121
|
};
|
|
110
122
|
}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/providers/fal/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,QAAQ,EAAE,0BAA0B,EAAE,MAAM,iBAAiB,CAAC;AACvE,OAAO,EACL,mBAAmB,EACnB,aAAa,EACb,cAAc,GACf,MAAM,yBAAyB,CAAC;AASjC,MAAM,
|
|
1
|
+
{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/providers/fal/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAExB,OAAO,EAAE,QAAQ,EAAE,0BAA0B,EAAE,MAAM,iBAAiB,CAAC;AACvE,OAAO,EACL,mBAAmB,EACnB,aAAa,EACb,cAAc,GACf,MAAM,yBAAyB,CAAC;AASjC,MAAM,oBAAoB,GAAG,CAAC,CAAC,MAAM,CAAC;IACpC,KAAK,EAAE,CAAC,CAAC,MAAM,CAAC;QACd,GAAG,EAAE,CAAC,CAAC,MAAM,EAAE;QACf,YAAY,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE;KACpC,CAAC;CACH,CAAC,CAAC;AASH,MAAM,CAAC,MAAM,eAAe,GAAG,QAAiB,CAAC;AAEjD,MAAM,CAAC,MAAM,UAAU,GAAyB;IAC9C;QACE,EAAE,EAAE,QAAQ;QACZ,WAAW,EAAE,YAAY;QACzB,SAAS,EAAE,CAAC,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,CAAC;QACjE,QAAQ,EAAE,CAAC,aAAa,EAAE,sBAAsB,CAAC;KAClD;IACD;QACE,EAAE,EAAE,QAAQ;QACZ,WAAW,EAAE,YAAY;QACzB,SAAS,EAAE,CAAC,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,CAAC;QACzC,QAAQ,EAAE,CAAC,aAAa,CAAC;KAC1B;IACD;QACE,EAAE,EAAE,aAAa;QACjB,WAAW,EAAE,YAAY;QACzB,SAAS,EAAE,CAAC,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,CAAC;QACrD,QAAQ,EAAE,CAAC,aAAa,CAAC;KAC1B;CACO,CAAC;AAEX,MAAM,OAAO,iBAAiB;IAGnB,EAAE,GAAG,eAAe,CAAC;IACrB,YAAY,GAAG,EAAE,CAAC;IAElB,MAAM,GAAG,UAAU,CAAC;IAEZ,MAAM,CAAqB;IAC3B,OAAO,CAAS;IAChB,OAAO,CAA0B;IAElD,YAAY,MAA+B;QACzC,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC,OAAO,IAAI,iBAAiB,CAAC;QACnD,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC,KAAK,IAAI,UAAU,CAAC,KAAK,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC;IACnE,CAAC;IAED,KAAK,CAAC,QAAQ,CAAC,OAOd;QAKC,IAAI,CAAC,OAAO,CAAC,OAAO,EAAE,CAAC;YACrB,MAAM,IAAI,KAAK,CACb,yFAAyF,CAC1F,CAAC;QACJ,CAAC;QAED,MAAM,GAAG,GAAG,GAAG,IAAI,CAAC,OAAO,WAAW,OAAO,CAAC,OAAO,EAAE,CAAC;QAExD,MAAM,IAAI,GAA4B;YACpC,GAAG,OAAO,CAAC,eAAe;YAC1B,IAAI,EAAE,OAAO,CAAC,IAAI;SACnB,CAAC;QAEF,IAAI,OAAO,CAAC,KAAK,IAAI,IAAI,EAAE,CAAC;YAC1B,IAAI,OAAO,OAAO,CAAC,KAAK,KAAK,QAAQ,EAAE,CAAC;gBACtC,IAAI,CAAC,KAAK,GAAG,OAAO,CAAC,KAAK,CAAC;YAC7B,CAAC;iBAAM,IAAI,KAAK,IAAI,OAAO,CAAC,KAAK,EAAE,CAAC;gBAClC,IAAI,CAAC,SAAS,GAAG,OAAO,CAAC,KAAK,CAAC,GAAG,CAAC;YACrC,CAAC;QACH,CAAC;QAED,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,GAAG,EAAE;YACvC,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,cAAc,EAAE,kBAAkB;gBAClC,aAAa,EAAE,OAAO,aAAa,CAAC,IAAI,CAAC,MAAM,EAAE,aAAa,EAAE,KAAK,CAAC,EAAE;gBACxE,cAAc,EAAE,cAAc;gBAC9B,GAAG,OAAO,CAAC,OAAO;aACnB;YACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC;YAC1B,MAAM,EAAE,OAAO,CAAC,WAAW;SAC5B,CAAC,CAAC;QAEH,MAAM,mBAAmB,CAAC,QAAQ,CAAC,CAAC;QAEpC,MAAM,IAAI,GAAG,oBAAoB,CAAC,KAAK,CAAC,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC,CAAC;QAC/D,OAAO,MAAM,IAAI,CAAC,UAAU,CAAC,IAAI,EAAE,OAAO,CAAC,CAAC;IAC9C,CAAC;IAEO,KAAK,CAAC,UAAU,CACtB,IAA0C,EAC1C,OAAsC;QAEtC,MAAM,aAAa,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,IAAI,CAAC,KAAK,CAAC,GAAG,EAAE;YACvD,MAAM,EAAE,OAAO,CAAC,WAAW;SAC5B,CAAC,CAAC;QAEH,IAAI,CAAC,aAAa,CAAC,EAAE,EAAE,CAAC;YACtB,MAAM,IAAI,QAAQ,CAAC,cAAc,aAAa,CAAC,MAAM,EAAE,EAAE;gBACvD,UAAU,EAAE,aAAa,CAAC,MAAM;gBAChC,YAAY,EAAE,MAAM,aAAa,CAAC,IAAI,EAAE,CAAC,KAAK,CAAC,GAAG,EAAE,CAAC,SAAS,CAAC;aAChE,CAAC,CAAC;QACL,CAAC;QAED,MAAM,WAAW,GAAG,MAAM,aAAa,CAAC,WAAW,EAAE,CAAC;QACtD,0FAA0F;QAC1F,MAAM,SAAS,GACb,IAAI,CAAC,KAAK,CAAC,YAAY;YACvB,aAAa,CAAC,OAAO,CAAC,GAAG,CAAC,cAAc,CAAC;YACzC,WAAW,CAAC;QAEd,OAAO,EAAE,KAAK,EAAE,IAAI,UAAU,CAAC,WAAW,CAAC,EAAE,SAAS,EAAE,CAAC;IAC3D,CAAC;IAED,MAAM,CAAC,OAA4B;QACjC,OAAO,OAAO,CAAC,MAAM,CACnB,IAAI,0BAA0B,CAAC,UAAU,OAAO,CAAC,OAAO,EAAE,CAAC,CAC5D,CAAC;IACJ,CAAC;IAED,qEAAqE;IACrE,2EAA2E;IAC3E,0EAA0E;IAC1E,4BAA4B;IAC5B,gBAAgB,CAAC,QAAgB;QAC/B,OAAO;YACL,eAAe,EAAE,EAAE;YACnB,SAAS,EAAE,WAAW;SACvB,CAAC;IACJ,CAAC;IAED,mBAAmB,CACjB,QAAgB,EAChB,MAAmB;QAEnB,IAAI,MAAM,CAAC,MAAM,KAAK,KAAK,EAAE,CAAC;YAC5B,OAAO,EAAE,eAAe,EAAE,EAAE,EAAE,iBAAiB,EAAE,WAAW,EAAE,CAAC;QACjE,CAAC;QACD,OAAO;IACT,CAAC;CACF;AAED,MAAM,UAAU,SAAS,CAAC,SAAkC,EAAE;IAC5D,MAAM,QAAQ,GAAG,IAAI,iBAAiB,CAAC,MAAM,CAAC,CAAC;IAC/C,MAAM,WAAW,GAAG,MAAM,CAAC,WAAW,CAAC;IAEvC,OAAO,SAAS,GAAG,CACjB,OAAgB;QAEhB,OAAO;YACL,QAAQ;YACR,OAAO,EAAE,OAAO,IAAI,QAAQ,CAAC,YAAY;YACzC,GAAG,CAAC,WAAW,IAAI,EAAE,WAAW,EAAE,CAAC;SACpC,CAAC;IACJ,CAAC,CAAC;AACJ,CAAC"}
|
|
@@ -1,18 +1,18 @@
|
|
|
1
|
-
import
|
|
1
|
+
import type { AudioOutput } from "../../audio-output.js";
|
|
2
|
+
import { type ModelInfo, type ResolvedModel, type SpeechProvider } from "../../speech-provider.js";
|
|
3
|
+
import type { ResolvedSTTModel } from "../../speech-to-text-provider.js";
|
|
2
4
|
export interface FishAudioSpeechProviderConfig {
|
|
3
5
|
apiKey?: string;
|
|
4
6
|
baseURL?: string;
|
|
7
|
+
fallbackSTT?: ResolvedSTTModel;
|
|
5
8
|
fetch?: typeof globalThis.fetch;
|
|
6
9
|
}
|
|
10
|
+
export declare const FISH_AUDIO_PROVIDER_ID: "fish-audio";
|
|
11
|
+
export declare const FISH_AUDIO_MODELS: readonly ModelInfo[];
|
|
7
12
|
export declare class FishAudioSpeechProvider implements SpeechProvider<string, string> {
|
|
8
|
-
readonly id
|
|
13
|
+
readonly id: "fish-audio";
|
|
9
14
|
readonly defaultModel = "s2-pro";
|
|
10
|
-
readonly models: readonly [
|
|
11
|
-
readonly id: "s2-pro";
|
|
12
|
-
readonly releaseDate: "2026-03-09";
|
|
13
|
-
readonly languages: readonly ["ja", "en", "zh", "ko", "es", "pt", "ar", "ru", "fr", "de"];
|
|
14
|
-
readonly features: readonly ["streaming", "audio-tags", "open-source", "inline-voice-cloning"];
|
|
15
|
-
}];
|
|
15
|
+
readonly models: readonly ModelInfo[];
|
|
16
16
|
private readonly apiKey;
|
|
17
17
|
private readonly baseURL;
|
|
18
18
|
private readonly fetchFn;
|
|
@@ -51,6 +51,12 @@ export declare class FishAudioSpeechProvider implements SpeechProvider<string, s
|
|
|
51
51
|
};
|
|
52
52
|
mediaType: string;
|
|
53
53
|
} | undefined;
|
|
54
|
+
resolveOutputFormat(modelId: string, output: AudioOutput): {
|
|
55
|
+
providerOptions: {
|
|
56
|
+
format: string;
|
|
57
|
+
};
|
|
58
|
+
expectedMediaType: string;
|
|
59
|
+
} | undefined;
|
|
54
60
|
dialogueCapabilities(modelId: string): {
|
|
55
61
|
minVoices: number;
|
|
56
62
|
maxVoices: number;
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/providers/fish-audio/index.ts"],"names":[],"mappings":"
|
|
1
|
+
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/providers/fish-audio/index.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,uBAAuB,CAAC;AAOzD,OAAO,EAEL,KAAK,SAAS,EACd,KAAK,aAAa,EAClB,KAAK,cAAc,EACpB,MAAM,0BAA0B,CAAC;AAClC,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,kCAAkC,CAAC;AAEzE,MAAM,WAAW,6BAA6B;IAC5C,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,WAAW,CAAC,EAAE,gBAAgB,CAAC;IAC/B,KAAK,CAAC,EAAE,OAAO,UAAU,CAAC,KAAK,CAAC;CACjC;AAED,eAAO,MAAM,sBAAsB,EAAG,YAAqB,CAAC;AAE5D,eAAO,MAAM,iBAAiB,EAAE,SAAS,SAAS,EAYxC,CAAC;AAEX,qBAAa,uBAAwB,YAAW,cAAc,CAAC,MAAM,EAAE,MAAM,CAAC;IAC5E,QAAQ,CAAC,EAAE,eAA0B;IACrC,QAAQ,CAAC,YAAY,YAAY;IAEjC,QAAQ,CAAC,MAAM,uBAAqB;IAEpC,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAqB;IAC5C,OAAO,CAAC,QAAQ,CAAC,OAAO,CAAS;IACjC,OAAO,CAAC,QAAQ,CAAC,OAAO,CAA0B;gBAEtC,MAAM,EAAE,6BAA6B;IAMjD,gBAAgB,CACd,IAAI,EAAE,MAAM,EACZ,OAAO,EAAE,MAAM,GACd;QAAE,IAAI,EAAE,MAAM,CAAC;QAAC,QAAQ,EAAE,MAAM,EAAE,CAAA;KAAE;IASjC,QAAQ,CAAC,OAAO,EAAE;QACtB,OAAO,EAAE,MAAM,CAAC;QAChB,IAAI,EAAE,MAAM,CAAC;QACb,KAAK,CAAC,EAAE,MAAM,CAAC;QACf,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;QAC1C,WAAW,CAAC,EAAE,WAAW,CAAC;QAC1B,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;KAClC,GAAG,OAAO,CAAC;QACV,KAAK,EAAE,UAAU,CAAC;QAClB,SAAS,EAAE,MAAM,CAAC;QAClB,gBAAgB,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;KAC5C,CAAC;IAoCI,MAAM,CAAC,OAAO,EAAE;QACpB,OAAO,EAAE,MAAM,CAAC;QAChB,IAAI,EAAE,MAAM,CAAC;QACb,KAAK,CAAC,EAAE,MAAM,CAAC;QACf,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;QAC1C,WAAW,CAAC,EAAE,WAAW,CAAC;QAC1B,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;KAClC,GAAG,OAAO,CAAC;QACV,MAAM,EAAE,cAAc,CAAC,UAAU,CAAC,CAAC;QACnC,SAAS,EAAE,MAAM,CAAC;QAClB,gBAAgB,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;KAC5C,CAAC;IAoCF,gBAAgB,CAAC,OAAO,EAAE,MAAM;;;;;;IAUhC,mBAAmB,CAAC,OAAO,EAAE,MAAM,EAAE,MAAM,EAAE,WAAW;;;;;;IAyBxD,oBAAoB,CAAC,OAAO,EAAE,MAAM;;;;IAO9B,gBAAgB,CAAC,OAAO,EAAE;QAC9B,OAAO,EAAE,MAAM,CAAC;QAChB,KAAK,EAAE,SAAS;YAAE,KAAK,EAAE,MAAM,CAAC;YAAC,IAAI,EAAE,MAAM,CAAA;SAAE,EAAE,CAAC;QAClD,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;QAC1C,WAAW,CAAC,EAAE,WAAW,CAAC;QAC1B,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;KAClC,GAAG,OAAO,CAAC;QACV,KAAK,EAAE,UAAU,CAAC;QAClB,SAAS,EAAE,MAAM,CAAC;QAClB,gBAAgB,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;KAC5C,CAAC;CA8CH;AAED,wBAAgB,eAAe,CAAC,MAAM,GAAE,6BAAkC,IAI9C,UAAU,MAAM,KAAG,aAAa,CAAC,MAAM,CAAC,CAOnE"}
|
|
@@ -1,22 +1,24 @@
|
|
|
1
1
|
import { stripAudioTags } from "../../audio-tags.js";
|
|
2
2
|
import { handleErrorResponse, resolveApiKey, SDK_USER_AGENT, } from "../../provider-utils.js";
|
|
3
3
|
import { hasFeature, } from "../../speech-provider.js";
|
|
4
|
+
export const FISH_AUDIO_PROVIDER_ID = "fish-audio";
|
|
5
|
+
export const FISH_AUDIO_MODELS = [
|
|
6
|
+
{
|
|
7
|
+
id: "s2-pro",
|
|
8
|
+
releaseDate: "2026-03-09",
|
|
9
|
+
languages: ["ja", "en", "zh", "ko", "es", "pt", "ar", "ru", "fr", "de"],
|
|
10
|
+
features: [
|
|
11
|
+
"streaming",
|
|
12
|
+
"audio-tags",
|
|
13
|
+
"open-source",
|
|
14
|
+
"inline-voice-cloning",
|
|
15
|
+
],
|
|
16
|
+
},
|
|
17
|
+
];
|
|
4
18
|
export class FishAudioSpeechProvider {
|
|
5
|
-
id =
|
|
19
|
+
id = FISH_AUDIO_PROVIDER_ID;
|
|
6
20
|
defaultModel = "s2-pro";
|
|
7
|
-
models =
|
|
8
|
-
{
|
|
9
|
-
id: "s2-pro",
|
|
10
|
-
releaseDate: "2026-03-09",
|
|
11
|
-
languages: ["ja", "en", "zh", "ko", "es", "pt", "ar", "ru", "fr", "de"],
|
|
12
|
-
features: [
|
|
13
|
-
"streaming",
|
|
14
|
-
"audio-tags",
|
|
15
|
-
"open-source",
|
|
16
|
-
"inline-voice-cloning",
|
|
17
|
-
],
|
|
18
|
-
},
|
|
19
|
-
];
|
|
21
|
+
models = FISH_AUDIO_MODELS;
|
|
20
22
|
apiKey;
|
|
21
23
|
baseURL;
|
|
22
24
|
fetchFn;
|
|
@@ -52,7 +54,7 @@ export class FishAudioSpeechProvider {
|
|
|
52
54
|
body: JSON.stringify(body),
|
|
53
55
|
signal: options.abortSignal,
|
|
54
56
|
});
|
|
55
|
-
await handleErrorResponse(response
|
|
57
|
+
await handleErrorResponse(response);
|
|
56
58
|
const arrayBuffer = await response.arrayBuffer();
|
|
57
59
|
const mediaType = response.headers.get("content-type") ?? "audio/mpeg";
|
|
58
60
|
return {
|
|
@@ -81,7 +83,7 @@ export class FishAudioSpeechProvider {
|
|
|
81
83
|
body: JSON.stringify(body),
|
|
82
84
|
signal: options.abortSignal,
|
|
83
85
|
});
|
|
84
|
-
await handleErrorResponse(response
|
|
86
|
+
await handleErrorResponse(response);
|
|
85
87
|
if (!response.body) {
|
|
86
88
|
throw new Error(`fish-audio/${options.modelId}: response has no body`);
|
|
87
89
|
}
|
|
@@ -97,13 +99,37 @@ export class FishAudioSpeechProvider {
|
|
|
97
99
|
mediaType: "audio/wav",
|
|
98
100
|
};
|
|
99
101
|
}
|
|
100
|
-
return
|
|
102
|
+
return;
|
|
103
|
+
}
|
|
104
|
+
resolveOutputFormat(modelId, output) {
|
|
105
|
+
if (!this.models.some((m) => m.id === modelId)) {
|
|
106
|
+
return;
|
|
107
|
+
}
|
|
108
|
+
switch (output.format) {
|
|
109
|
+
case "wav":
|
|
110
|
+
return {
|
|
111
|
+
providerOptions: { format: "wav" },
|
|
112
|
+
expectedMediaType: "audio/wav",
|
|
113
|
+
};
|
|
114
|
+
case "mp3":
|
|
115
|
+
return {
|
|
116
|
+
providerOptions: { format: "mp3" },
|
|
117
|
+
expectedMediaType: "audio/mpeg",
|
|
118
|
+
};
|
|
119
|
+
case "pcm":
|
|
120
|
+
return {
|
|
121
|
+
providerOptions: { format: "wav" },
|
|
122
|
+
expectedMediaType: "audio/wav",
|
|
123
|
+
};
|
|
124
|
+
default:
|
|
125
|
+
return;
|
|
126
|
+
}
|
|
101
127
|
}
|
|
102
128
|
dialogueCapabilities(modelId) {
|
|
103
129
|
if (modelId === "s2-pro") {
|
|
104
130
|
return { minVoices: 1, maxVoices: 4 };
|
|
105
131
|
}
|
|
106
|
-
return
|
|
132
|
+
return;
|
|
107
133
|
}
|
|
108
134
|
async generateDialogue(options) {
|
|
109
135
|
if (options.modelId !== "s2-pro") {
|
|
@@ -138,7 +164,7 @@ export class FishAudioSpeechProvider {
|
|
|
138
164
|
body: JSON.stringify(body),
|
|
139
165
|
signal: options.abortSignal,
|
|
140
166
|
});
|
|
141
|
-
await handleErrorResponse(response
|
|
167
|
+
await handleErrorResponse(response);
|
|
142
168
|
return {
|
|
143
169
|
audio: new Uint8Array(await response.arrayBuffer()),
|
|
144
170
|
mediaType: response.headers.get("content-type") ?? "audio/mpeg",
|
|
@@ -147,10 +173,12 @@ export class FishAudioSpeechProvider {
|
|
|
147
173
|
}
|
|
148
174
|
export function createFishAudio(config = {}) {
|
|
149
175
|
const provider = new FishAudioSpeechProvider(config);
|
|
176
|
+
const fallbackSTT = config.fallbackSTT;
|
|
150
177
|
return function fishAudio(modelId) {
|
|
151
178
|
return {
|
|
152
179
|
provider,
|
|
153
180
|
modelId: modelId ?? provider.defaultModel,
|
|
181
|
+
...(fallbackSTT && { fallbackSTT }),
|
|
154
182
|
};
|
|
155
183
|
};
|
|
156
184
|
}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/providers/fish-audio/index.ts"],"names":[],"mappings":"
|
|
1
|
+
{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/providers/fish-audio/index.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,cAAc,EAAE,MAAM,qBAAqB,CAAC;AACrD,OAAO,EACL,mBAAmB,EACnB,aAAa,EACb,cAAc,GACf,MAAM,yBAAyB,CAAC;AACjC,OAAO,EACL,UAAU,GAIX,MAAM,0BAA0B,CAAC;AAUlC,MAAM,CAAC,MAAM,sBAAsB,GAAG,YAAqB,CAAC;AAE5D,MAAM,CAAC,MAAM,iBAAiB,GAAyB;IACrD;QACE,EAAE,EAAE,QAAQ;QACZ,WAAW,EAAE,YAAY;QACzB,SAAS,EAAE,CAAC,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,CAAC;QACvE,QAAQ,EAAE;YACR,WAAW;YACX,YAAY;YACZ,aAAa;YACb,sBAAsB;SACvB;KACF;CACO,CAAC;AAEX,MAAM,OAAO,uBAAuB;IACzB,EAAE,GAAG,sBAAsB,CAAC;IAC5B,YAAY,GAAG,QAAQ,CAAC;IAExB,MAAM,GAAG,iBAAiB,CAAC;IAEnB,MAAM,CAAqB;IAC3B,OAAO,CAAS;IAChB,OAAO,CAA0B;IAElD,YAAY,MAAqC;QAC/C,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC,OAAO,IAAI,wBAAwB,CAAC;QAC1D,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC,KAAK,IAAI,UAAU,CAAC,KAAK,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC;IACnE,CAAC;IAED,gBAAgB,CACd,IAAY,EACZ,OAAe;QAEf,IACE,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,OAAO,IAAI,UAAU,CAAC,CAAC,EAAE,YAAY,CAAC,CAAC,EACxE,CAAC;YACD,OAAO,EAAE,IAAI,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC;QAChC,CAAC;QACD,OAAO,cAAc,CAAC,IAAI,EAAE,cAAc,OAAO,EAAE,CAAC,CAAC;IACvD,CAAC;IAED,KAAK,CAAC,QAAQ,CAAC,OAOd;QAKC,MAAM,GAAG,GAAG,GAAG,IAAI,CAAC,OAAO,SAAS,CAAC;QAErC,MAAM,IAAI,GAA4B;YACpC,GAAG,OAAO,CAAC,eAAe;YAC1B,IAAI,EAAE,OAAO,CAAC,IAAI;SACnB,CAAC;QAEF,IAAI,OAAO,CAAC,KAAK,EAAE,CAAC;YAClB,IAAI,CAAC,YAAY,GAAG,OAAO,CAAC,KAAK,CAAC;QACpC,CAAC;QAED,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,GAAG,EAAE;YACvC,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,cAAc,EAAE,kBAAkB;gBAClC,aAAa,EAAE,UAAU,aAAa,CAAC,IAAI,CAAC,MAAM,EAAE,oBAAoB,EAAE,YAAY,CAAC,EAAE;gBACzF,KAAK,EAAE,OAAO,CAAC,OAAO;gBACtB,cAAc,EAAE,cAAc;gBAC9B,GAAG,OAAO,CAAC,OAAO;aACnB;YACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC;YAC1B,MAAM,EAAE,OAAO,CAAC,WAAW;SAC5B,CAAC,CAAC;QAEH,MAAM,mBAAmB,CAAC,QAAQ,CAAC,CAAC;QAEpC,MAAM,WAAW,GAAG,MAAM,QAAQ,CAAC,WAAW,EAAE,CAAC;QACjD,MAAM,SAAS,GAAG,QAAQ,CAAC,OAAO,CAAC,GAAG,CAAC,cAAc,CAAC,IAAI,YAAY,CAAC;QAEvE,OAAO;YACL,KAAK,EAAE,IAAI,UAAU,CAAC,WAAW,CAAC;YAClC,SAAS;SACV,CAAC;IACJ,CAAC;IAED,KAAK,CAAC,MAAM,CAAC,OAOZ;QAKC,MAAM,GAAG,GAAG,GAAG,IAAI,CAAC,OAAO,SAAS,CAAC;QAErC,MAAM,IAAI,GAA4B;YACpC,GAAG,OAAO,CAAC,eAAe;YAC1B,IAAI,EAAE,OAAO,CAAC,IAAI;SACnB,CAAC;QACF,IAAI,OAAO,CAAC,KAAK,EAAE,CAAC;YAClB,IAAI,CAAC,YAAY,GAAG,OAAO,CAAC,KAAK,CAAC;QACpC,CAAC;QAED,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,GAAG,EAAE;YACvC,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,cAAc,EAAE,kBAAkB;gBAClC,aAAa,EAAE,UAAU,aAAa,CAAC,IAAI,CAAC,MAAM,EAAE,oBAAoB,EAAE,YAAY,CAAC,EAAE;gBACzF,KAAK,EAAE,OAAO,CAAC,OAAO;gBACtB,cAAc,EAAE,cAAc;gBAC9B,GAAG,OAAO,CAAC,OAAO;aACnB;YACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC;YAC1B,MAAM,EAAE,OAAO,CAAC,WAAW;SAC5B,CAAC,CAAC;QAEH,MAAM,mBAAmB,CAAC,QAAQ,CAAC,CAAC;QAEpC,IAAI,CAAC,QAAQ,CAAC,IAAI,EAAE,CAAC;YACnB,MAAM,IAAI,KAAK,CAAC,cAAc,OAAO,CAAC,OAAO,wBAAwB,CAAC,CAAC;QACzE,CAAC;QAED,OAAO;YACL,MAAM,EAAE,QAAQ,CAAC,IAAI;YACrB,SAAS,EAAE,QAAQ,CAAC,OAAO,CAAC,GAAG,CAAC,cAAc,CAAC,IAAI,YAAY;SAChE,CAAC;IACJ,CAAC;IAED,gBAAgB,CAAC,OAAe;QAC9B,IAAI,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,OAAO,CAAC,EAAE,CAAC;YAC9C,OAAO;gBACL,eAAe,EAAE,EAAE,MAAM,EAAE,KAAK,EAAE;gBAClC,SAAS,EAAE,WAAW;aACvB,CAAC;QACJ,CAAC;QACD,OAAO;IACT,CAAC;IAED,mBAAmB,CAAC,OAAe,EAAE,MAAmB;QACtD,IAAI,CAAC,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,OAAO,CAAC,EAAE,CAAC;YAC/C,OAAO;QACT,CAAC;QACD,QAAQ,MAAM,CAAC,MAAM,EAAE,CAAC;YACtB,KAAK,KAAK;gBACR,OAAO;oBACL,eAAe,EAAE,EAAE,MAAM,EAAE,KAAK,EAAE;oBAClC,iBAAiB,EAAE,WAAW;iBAC/B,CAAC;YACJ,KAAK,KAAK;gBACR,OAAO;oBACL,eAAe,EAAE,EAAE,MAAM,EAAE,KAAK,EAAE;oBAClC,iBAAiB,EAAE,YAAY;iBAChC,CAAC;YACJ,KAAK,KAAK;gBACR,OAAO;oBACL,eAAe,EAAE,EAAE,MAAM,EAAE,KAAK,EAAE;oBAClC,iBAAiB,EAAE,WAAW;iBAC/B,CAAC;YACJ;gBACE,OAAO;QACX,CAAC;IACH,CAAC;IAED,oBAAoB,CAAC,OAAe;QAClC,IAAI,OAAO,KAAK,QAAQ,EAAE,CAAC;YACzB,OAAO,EAAE,SAAS,EAAE,CAAC,EAAE,SAAS,EAAE,CAAC,EAAE,CAAC;QACxC,CAAC;QACD,OAAO;IACT,CAAC;IAED,KAAK,CAAC,gBAAgB,CAAC,OAMtB;QAKC,IAAI,OAAO,CAAC,OAAO,KAAK,QAAQ,EAAE,CAAC;YACjC,MAAM,IAAI,KAAK,CACb,cAAc,OAAO,CAAC,OAAO,gDAAgD,CAC9E,CAAC;QACJ,CAAC;QAED,MAAM,YAAY,GAAG,IAAI,GAAG,EAAkB,CAAC;QAC/C,MAAM,MAAM,GAAa,EAAE,CAAC;QAC5B,KAAK,MAAM,CAAC,IAAI,OAAO,CAAC,KAAK,EAAE,CAAC;YAC9B,IAAI,GAAG,GAAG,YAAY,CAAC,GAAG,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC;YACpC,IAAI,GAAG,KAAK,SAAS,EAAE,CAAC;gBACtB,GAAG,GAAG,YAAY,CAAC,IAAI,CAAC;gBACxB,YAAY,CAAC,GAAG,CAAC,CAAC,CAAC,KAAK,EAAE,GAAG,CAAC,CAAC;YACjC,CAAC;YACD,MAAM,CAAC,IAAI,CAAC,aAAa,GAAG,KAAK,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC;QAC7C,CAAC;QACD,MAAM,IAAI,GAAG,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QAC/B,MAAM,YAAY,GAAG,KAAK,CAAC,IAAI,CAAC,YAAY,CAAC,IAAI,EAAE,CAAC,CAAC;QAErD,MAAM,IAAI,GAA4B;YACpC,GAAG,OAAO,CAAC,eAAe;YAC1B,IAAI;YACJ,YAAY,EAAE,YAAY;SAC3B,CAAC;QAEF,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,GAAG,IAAI,CAAC,OAAO,SAAS,EAAE;YAC5D,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,cAAc,EAAE,kBAAkB;gBAClC,aAAa,EAAE,UAAU,aAAa,CAAC,IAAI,CAAC,MAAM,EAAE,oBAAoB,EAAE,YAAY,CAAC,EAAE;gBACzF,KAAK,EAAE,OAAO,CAAC,OAAO;gBACtB,cAAc,EAAE,cAAc;gBAC9B,GAAG,OAAO,CAAC,OAAO;aACnB;YACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC;YAC1B,MAAM,EAAE,OAAO,CAAC,WAAW;SAC5B,CAAC,CAAC;QAEH,MAAM,mBAAmB,CAAC,QAAQ,CAAC,CAAC;QAEpC,OAAO;YACL,KAAK,EAAE,IAAI,UAAU,CAAC,MAAM,QAAQ,CAAC,WAAW,EAAE,CAAC;YACnD,SAAS,EAAE,QAAQ,CAAC,OAAO,CAAC,GAAG,CAAC,cAAc,CAAC,IAAI,YAAY;SAChE,CAAC;IACJ,CAAC;CACF;AAED,MAAM,UAAU,eAAe,CAAC,SAAwC,EAAE;IACxE,MAAM,QAAQ,GAAG,IAAI,uBAAuB,CAAC,MAAM,CAAC,CAAC;IACrD,MAAM,WAAW,GAAG,MAAM,CAAC,WAAW,CAAC;IAEvC,OAAO,SAAS,SAAS,CAAC,OAAgB;QACxC,OAAO;YACL,QAAQ;YACR,OAAO,EAAE,OAAO,IAAI,QAAQ,CAAC,YAAY;YACzC,GAAG,CAAC,WAAW,IAAI,EAAE,WAAW,EAAE,CAAC;SACpC,CAAC;IACJ,CAAC,CAAC;AACJ,CAAC"}
|
|
@@ -0,0 +1,76 @@
|
|
|
1
|
+
import type { AudioOutput } from "../../audio-output.js";
|
|
2
|
+
import type { PronunciationsInput } from "../../pronunciations/types.js";
|
|
3
|
+
import type { ModelInfo, ResolvedModel, SpeechProvider } from "../../speech-provider.js";
|
|
4
|
+
import type { ConversationWordTimestamp, WordTimestamp } from "../../timestamps.js";
|
|
5
|
+
export declare const SPEECH_GATEWAY_PROVIDER_ID: "speech-gateway";
|
|
6
|
+
export interface SpeechGatewayProviderConfig {
|
|
7
|
+
apiKey?: string;
|
|
8
|
+
baseURL?: string;
|
|
9
|
+
fetch?: typeof globalThis.fetch;
|
|
10
|
+
}
|
|
11
|
+
export declare class SpeechGatewayProvider implements SpeechProvider<string, string> {
|
|
12
|
+
readonly id: "speech-gateway";
|
|
13
|
+
readonly defaultModel = "";
|
|
14
|
+
readonly models: readonly ModelInfo[];
|
|
15
|
+
private readonly apiKey;
|
|
16
|
+
private readonly baseURL;
|
|
17
|
+
private readonly fetchFn;
|
|
18
|
+
constructor(config: SpeechGatewayProviderConfig);
|
|
19
|
+
private resolveKey;
|
|
20
|
+
generate(options: {
|
|
21
|
+
modelId: string;
|
|
22
|
+
text: string;
|
|
23
|
+
voice?: string;
|
|
24
|
+
providerOptions?: Record<string, unknown>;
|
|
25
|
+
abortSignal?: AbortSignal;
|
|
26
|
+
headers?: Record<string, string>;
|
|
27
|
+
includeTimestamps?: boolean;
|
|
28
|
+
volumeDbfs?: number;
|
|
29
|
+
output?: AudioOutput;
|
|
30
|
+
pronunciations?: PronunciationsInput;
|
|
31
|
+
}): Promise<{
|
|
32
|
+
audio: Uint8Array;
|
|
33
|
+
mediaType: string;
|
|
34
|
+
timestamps?: WordTimestamp[];
|
|
35
|
+
warnings?: string[];
|
|
36
|
+
}>;
|
|
37
|
+
stream(options: {
|
|
38
|
+
modelId: string;
|
|
39
|
+
text: string;
|
|
40
|
+
voice?: string;
|
|
41
|
+
providerOptions?: Record<string, unknown>;
|
|
42
|
+
abortSignal?: AbortSignal;
|
|
43
|
+
headers?: Record<string, string>;
|
|
44
|
+
pronunciations?: PronunciationsInput;
|
|
45
|
+
}): Promise<{
|
|
46
|
+
audioDurationMs?: number;
|
|
47
|
+
stream: ReadableStream<Uint8Array>;
|
|
48
|
+
mediaType: string;
|
|
49
|
+
providerMetadata?: Record<string, unknown>;
|
|
50
|
+
}>;
|
|
51
|
+
generateConversation(options: {
|
|
52
|
+
modelId?: string;
|
|
53
|
+
turns: readonly {
|
|
54
|
+
model?: string;
|
|
55
|
+
voice: string;
|
|
56
|
+
text: string;
|
|
57
|
+
providerOptions?: Record<string, unknown>;
|
|
58
|
+
}[];
|
|
59
|
+
gapMs?: number;
|
|
60
|
+
volumeDbfs?: number;
|
|
61
|
+
providerOptions?: Record<string, unknown>;
|
|
62
|
+
abortSignal?: AbortSignal;
|
|
63
|
+
headers?: Record<string, string>;
|
|
64
|
+
includeTimestamps?: boolean;
|
|
65
|
+
output?: AudioOutput;
|
|
66
|
+
pronunciations?: PronunciationsInput;
|
|
67
|
+
}): Promise<{
|
|
68
|
+
audio: Uint8Array;
|
|
69
|
+
mediaType: string;
|
|
70
|
+
timestamps?: ConversationWordTimestamp[];
|
|
71
|
+
warnings?: string[];
|
|
72
|
+
providerMetadata?: Record<string, unknown>;
|
|
73
|
+
}>;
|
|
74
|
+
}
|
|
75
|
+
export declare function createSpeechGateway(config?: SpeechGatewayProviderConfig): (modelId: string) => ResolvedModel<string>;
|
|
76
|
+
//# sourceMappingURL=index.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/providers/gateway/index.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,uBAAuB,CAAC;AAMzD,OAAO,KAAK,EAAE,mBAAmB,EAAE,MAAM,+BAA+B,CAAC;AAEzE,OAAO,KAAK,EACV,SAAS,EACT,aAAa,EACb,cAAc,EACf,MAAM,0BAA0B,CAAC;AAClC,OAAO,KAAK,EACV,yBAAyB,EACzB,aAAa,EACd,MAAM,qBAAqB,CAAC;AAE7B,eAAO,MAAM,0BAA0B,EAAG,gBAAyB,CAAC;AAEpE,MAAM,WAAW,2BAA2B;IAC1C,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,KAAK,CAAC,EAAE,OAAO,UAAU,CAAC,KAAK,CAAC;CACjC;AA+BD,qBAAa,qBAAsB,YAAW,cAAc,CAAC,MAAM,EAAE,MAAM,CAAC;IAC1E,QAAQ,CAAC,EAAE,mBAA8B;IACzC,QAAQ,CAAC,YAAY,MAAM;IAE3B,QAAQ,CAAC,MAAM,EAAE,SAAS,SAAS,EAAE,CAAM;IAE3C,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAqB;IAC5C,OAAO,CAAC,QAAQ,CAAC,OAAO,CAAS;IACjC,OAAO,CAAC,QAAQ,CAAC,OAAO,CAA0B;gBAEtC,MAAM,EAAE,2BAA2B;IAM/C,OAAO,CAAC,UAAU;IAkBZ,QAAQ,CAAC,OAAO,EAAE;QACtB,OAAO,EAAE,MAAM,CAAC;QAChB,IAAI,EAAE,MAAM,CAAC;QACb,KAAK,CAAC,EAAE,MAAM,CAAC;QACf,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;QAC1C,WAAW,CAAC,EAAE,WAAW,CAAC;QAC1B,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;QACjC,iBAAiB,CAAC,EAAE,OAAO,CAAC;QAC5B,UAAU,CAAC,EAAE,MAAM,CAAC;QACpB,MAAM,CAAC,EAAE,WAAW,CAAC;QACrB,cAAc,CAAC,EAAE,mBAAmB,CAAC;KACtC,GAAG,OAAO,CAAC;QACV,KAAK,EAAE,UAAU,CAAC;QAClB,SAAS,EAAE,MAAM,CAAC;QAClB,UAAU,CAAC,EAAE,aAAa,EAAE,CAAC;QAC7B,QAAQ,CAAC,EAAE,MAAM,EAAE,CAAC;KACrB,CAAC;IAkEI,MAAM,CAAC,OAAO,EAAE;QACpB,OAAO,EAAE,MAAM,CAAC;QAChB,IAAI,EAAE,MAAM,CAAC;QACb,KAAK,CAAC,EAAE,MAAM,CAAC;QACf,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;QAC1C,WAAW,CAAC,EAAE,WAAW,CAAC;QAC1B,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;QACjC,cAAc,CAAC,EAAE,mBAAmB,CAAC;KACtC,GAAG,OAAO,CAAC;QACV,eAAe,CAAC,EAAE,MAAM,CAAC;QACzB,MAAM,EAAE,cAAc,CAAC,UAAU,CAAC,CAAC;QACnC,SAAS,EAAE,MAAM,CAAC;QAClB,gBAAgB,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;KAC5C,CAAC;IAuDI,oBAAoB,CAAC,OAAO,EAAE;QAClC,OAAO,CAAC,EAAE,MAAM,CAAC;QACjB,KAAK,EAAE,SAAS;YACd,KAAK,CAAC,EAAE,MAAM,CAAC;YACf,KAAK,EAAE,MAAM,CAAC;YACd,IAAI,EAAE,MAAM,CAAC;YACb,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;SAC3C,EAAE,CAAC;QACJ,KAAK,CAAC,EAAE,MAAM,CAAC;QACf,UAAU,CAAC,EAAE,MAAM,CAAC;QACpB,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;QAC1C,WAAW,CAAC,EAAE,WAAW,CAAC;QAC1B,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;QACjC,iBAAiB,CAAC,EAAE,OAAO,CAAC;QAC5B,MAAM,CAAC,EAAE,WAAW,CAAC;QACrB,cAAc,CAAC,EAAE,mBAAmB,CAAC;KACtC,GAAG,OAAO,CAAC;QACV,KAAK,EAAE,UAAU,CAAC;QAClB,SAAS,EAAE,MAAM,CAAC;QAClB,UAAU,CAAC,EAAE,yBAAyB,EAAE,CAAC;QACzC,QAAQ,CAAC,EAAE,MAAM,EAAE,CAAC;QACpB,gBAAgB,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;KAC5C,CAAC;CA2FH;AAED,wBAAgB,mBAAmB,CAAC,MAAM,GAAE,2BAAgC,IAE5C,SAAS,MAAM,KAAG,aAAa,CAAC,MAAM,CAAC,CAQtE"}
|