@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
|
@@ -0,0 +1,251 @@
|
|
|
1
|
+
import { z } from "zod";
|
|
2
|
+
import { ApiError, GatewayInputError, MissingApiKeyError, } from "../../errors.js";
|
|
3
|
+
import { handleErrorResponse, SDK_USER_AGENT } from "../../provider-utils.js";
|
|
4
|
+
export const SPEECH_GATEWAY_PROVIDER_ID = "speech-gateway";
|
|
5
|
+
// audioDurationMs computed client-side via mediabunny for path consistency across gateway + direct providers.
|
|
6
|
+
const wordTimestampSchema = z.object({
|
|
7
|
+
text: z.string(),
|
|
8
|
+
start: z.number(),
|
|
9
|
+
end: z.number(),
|
|
10
|
+
});
|
|
11
|
+
const conversationWordTimestampSchema = wordTimestampSchema.extend({
|
|
12
|
+
turnIndex: z.number(),
|
|
13
|
+
});
|
|
14
|
+
const gatewayJsonResponseSchema = z.object({
|
|
15
|
+
audio: z.string(),
|
|
16
|
+
mediaType: z.string(),
|
|
17
|
+
timestamps: z.array(wordTimestampSchema).default([]),
|
|
18
|
+
warnings: z.array(z.string()).default([]),
|
|
19
|
+
});
|
|
20
|
+
const gatewayConversationJsonResponseSchema = z.object({
|
|
21
|
+
audio: z.string(),
|
|
22
|
+
mediaType: z.string(),
|
|
23
|
+
timestamps: z.array(conversationWordTimestampSchema).default([]),
|
|
24
|
+
warnings: z.array(z.string()).default([]),
|
|
25
|
+
providerMetadata: z.record(z.string(), z.unknown()).optional(),
|
|
26
|
+
});
|
|
27
|
+
const GATEWAY_401_MESSAGE = "Speech Gateway rejected your API key (401). Get a key at https://wavform.ai/ or verify your SPEECH_GATEWAY_API_KEY environment variable.";
|
|
28
|
+
export class SpeechGatewayProvider {
|
|
29
|
+
id = SPEECH_GATEWAY_PROVIDER_ID;
|
|
30
|
+
defaultModel = "";
|
|
31
|
+
// Gateway server is the source of truth for model capabilities; feature checks deferred to the wire.
|
|
32
|
+
models = [];
|
|
33
|
+
apiKey;
|
|
34
|
+
baseURL;
|
|
35
|
+
fetchFn;
|
|
36
|
+
constructor(config) {
|
|
37
|
+
this.apiKey = config.apiKey;
|
|
38
|
+
this.baseURL = config.baseURL ?? "https://api.speechgateway.com/v1";
|
|
39
|
+
this.fetchFn = config.fetch ?? globalThis.fetch.bind(globalThis);
|
|
40
|
+
}
|
|
41
|
+
resolveKey() {
|
|
42
|
+
const key = this.apiKey ??
|
|
43
|
+
(typeof process === "undefined"
|
|
44
|
+
? undefined
|
|
45
|
+
: process.env?.SPEECH_GATEWAY_API_KEY);
|
|
46
|
+
if (!key) {
|
|
47
|
+
const err = new MissingApiKeyError({
|
|
48
|
+
providerName: "Speech Gateway",
|
|
49
|
+
envVar: "SPEECH_GATEWAY_API_KEY",
|
|
50
|
+
});
|
|
51
|
+
err.message =
|
|
52
|
+
"To use the Speech Gateway, a Wavform AI api key is required. Sign up at https://wavform.ai/ to get a key, then pass it via the `apiKey` option or set the SPEECH_GATEWAY_API_KEY environment variable.";
|
|
53
|
+
throw err;
|
|
54
|
+
}
|
|
55
|
+
return key;
|
|
56
|
+
}
|
|
57
|
+
async generate(options) {
|
|
58
|
+
if (!options.voice) {
|
|
59
|
+
throw new GatewayInputError(`speech-gateway/${options.modelId}: "voice" is required when routing through the speech gateway in inline mode.`);
|
|
60
|
+
}
|
|
61
|
+
const body = {
|
|
62
|
+
mode: "inline",
|
|
63
|
+
model: options.modelId,
|
|
64
|
+
voice: options.voice,
|
|
65
|
+
text: options.text,
|
|
66
|
+
};
|
|
67
|
+
if (options.volumeDbfs != null) {
|
|
68
|
+
body.volumeDbfs = options.volumeDbfs;
|
|
69
|
+
}
|
|
70
|
+
if (options.providerOptions) {
|
|
71
|
+
body.providerOptions = options.providerOptions;
|
|
72
|
+
}
|
|
73
|
+
if (options.output) {
|
|
74
|
+
body.output = options.output;
|
|
75
|
+
}
|
|
76
|
+
if (options.pronunciations) {
|
|
77
|
+
body.pronunciations = options.pronunciations;
|
|
78
|
+
}
|
|
79
|
+
// Binary vs JSON-with-timestamps lives at separate URLs; no Accept-header content negotiation.
|
|
80
|
+
const url = options.includeTimestamps
|
|
81
|
+
? `${this.baseURL}/audio/speech/with-timestamps`
|
|
82
|
+
: `${this.baseURL}/audio/speech`;
|
|
83
|
+
const response = await this.fetchFn(url, {
|
|
84
|
+
method: "POST",
|
|
85
|
+
headers: {
|
|
86
|
+
...options.headers,
|
|
87
|
+
"Content-Type": "application/json",
|
|
88
|
+
Authorization: `Bearer ${this.resolveKey()}`,
|
|
89
|
+
"X-User-Agent": SDK_USER_AGENT,
|
|
90
|
+
},
|
|
91
|
+
body: JSON.stringify(body),
|
|
92
|
+
signal: options.abortSignal,
|
|
93
|
+
});
|
|
94
|
+
if (response.status === 401) {
|
|
95
|
+
throw new ApiError(GATEWAY_401_MESSAGE, { statusCode: 401 });
|
|
96
|
+
}
|
|
97
|
+
await handleErrorResponse(response);
|
|
98
|
+
if (options.includeTimestamps) {
|
|
99
|
+
const payload = gatewayJsonResponseSchema.parse(await response.json());
|
|
100
|
+
return {
|
|
101
|
+
audio: decodeBase64(payload.audio),
|
|
102
|
+
mediaType: payload.mediaType,
|
|
103
|
+
timestamps: payload.timestamps,
|
|
104
|
+
warnings: payload.warnings,
|
|
105
|
+
};
|
|
106
|
+
}
|
|
107
|
+
const arrayBuffer = await response.arrayBuffer();
|
|
108
|
+
return {
|
|
109
|
+
audio: new Uint8Array(arrayBuffer),
|
|
110
|
+
mediaType: mediaTypeFromHeaders(response.headers),
|
|
111
|
+
};
|
|
112
|
+
}
|
|
113
|
+
async stream(options) {
|
|
114
|
+
if (!options.voice) {
|
|
115
|
+
throw new GatewayInputError(`speech-gateway/${options.modelId}: "voice" is required when routing through the speech gateway in inline mode.`);
|
|
116
|
+
}
|
|
117
|
+
const body = {
|
|
118
|
+
mode: "inline",
|
|
119
|
+
model: options.modelId,
|
|
120
|
+
voice: options.voice,
|
|
121
|
+
text: options.text,
|
|
122
|
+
};
|
|
123
|
+
if (options.providerOptions) {
|
|
124
|
+
body.providerOptions = options.providerOptions;
|
|
125
|
+
}
|
|
126
|
+
if (options.pronunciations) {
|
|
127
|
+
body.pronunciations = options.pronunciations;
|
|
128
|
+
}
|
|
129
|
+
const url = `${this.baseURL}/audio/speech`;
|
|
130
|
+
const response = await this.fetchFn(url, {
|
|
131
|
+
method: "POST",
|
|
132
|
+
headers: {
|
|
133
|
+
...options.headers,
|
|
134
|
+
"Content-Type": "application/json",
|
|
135
|
+
Authorization: `Bearer ${this.resolveKey()}`,
|
|
136
|
+
"X-User-Agent": SDK_USER_AGENT,
|
|
137
|
+
},
|
|
138
|
+
body: JSON.stringify(body),
|
|
139
|
+
signal: options.abortSignal,
|
|
140
|
+
});
|
|
141
|
+
if (response.status === 401) {
|
|
142
|
+
throw new ApiError(GATEWAY_401_MESSAGE, { statusCode: 401 });
|
|
143
|
+
}
|
|
144
|
+
await handleErrorResponse(response);
|
|
145
|
+
if (!response.body) {
|
|
146
|
+
throw new Error(`speech-gateway/${options.modelId}: response has no body`);
|
|
147
|
+
}
|
|
148
|
+
return {
|
|
149
|
+
stream: response.body,
|
|
150
|
+
mediaType: mediaTypeFromHeaders(response.headers),
|
|
151
|
+
};
|
|
152
|
+
}
|
|
153
|
+
// Server handles stitching/normalization/alignment so callers never need their own STT key.
|
|
154
|
+
// Pick one of two wire shapes — mutually exclusive at the request level:
|
|
155
|
+
// 1. shared model — `modelId` set, every turn omits `model`
|
|
156
|
+
// 2. per-turn model — every turn declares its own `model`, `modelId` omitted (one conversation across providers)
|
|
157
|
+
async generateConversation(options) {
|
|
158
|
+
if (options.turns.length === 0) {
|
|
159
|
+
throw new GatewayInputError("speech-gateway/conversation: at least one turn is required.");
|
|
160
|
+
}
|
|
161
|
+
if (!options.turns.every((t) => t.voice)) {
|
|
162
|
+
throw new GatewayInputError('speech-gateway/conversation: every turn must specify a "voice" when routing through the speech gateway.');
|
|
163
|
+
}
|
|
164
|
+
const sharedModel = options.modelId;
|
|
165
|
+
const anyTurnModel = options.turns.some((t) => t.model != null);
|
|
166
|
+
if (sharedModel != null && anyTurnModel) {
|
|
167
|
+
throw new GatewayInputError("speech-gateway/conversation: pass either a shared `modelId` or per-turn `model` on every turn — not both.");
|
|
168
|
+
}
|
|
169
|
+
if (sharedModel == null && !options.turns.every((t) => t.model)) {
|
|
170
|
+
throw new GatewayInputError('speech-gateway/conversation: when no shared `modelId` is set, every turn must declare its own "model".');
|
|
171
|
+
}
|
|
172
|
+
// gapMs/volumeDbfs sent explicitly each call (don't rely on server defaults).
|
|
173
|
+
const body = {
|
|
174
|
+
mode: "conversation",
|
|
175
|
+
...(sharedModel != null && { model: sharedModel }),
|
|
176
|
+
turns: options.turns.map((t) => ({
|
|
177
|
+
...(t.model != null && { model: t.model }),
|
|
178
|
+
voice: t.voice,
|
|
179
|
+
text: t.text,
|
|
180
|
+
...(t.providerOptions && { providerOptions: t.providerOptions }),
|
|
181
|
+
})),
|
|
182
|
+
gapMs: options.gapMs ?? 300,
|
|
183
|
+
volumeDbfs: options.volumeDbfs ?? -20,
|
|
184
|
+
};
|
|
185
|
+
if (options.providerOptions) {
|
|
186
|
+
body.providerOptions = options.providerOptions;
|
|
187
|
+
}
|
|
188
|
+
if (options.output) {
|
|
189
|
+
body.output = options.output;
|
|
190
|
+
}
|
|
191
|
+
if (options.pronunciations) {
|
|
192
|
+
body.pronunciations = options.pronunciations;
|
|
193
|
+
}
|
|
194
|
+
const url = options.includeTimestamps
|
|
195
|
+
? `${this.baseURL}/audio/conversation/with-timestamps`
|
|
196
|
+
: `${this.baseURL}/audio/conversation`;
|
|
197
|
+
const response = await this.fetchFn(url, {
|
|
198
|
+
method: "POST",
|
|
199
|
+
headers: {
|
|
200
|
+
...options.headers,
|
|
201
|
+
"Content-Type": "application/json",
|
|
202
|
+
Authorization: `Bearer ${this.resolveKey()}`,
|
|
203
|
+
"X-User-Agent": SDK_USER_AGENT,
|
|
204
|
+
},
|
|
205
|
+
body: JSON.stringify(body),
|
|
206
|
+
signal: options.abortSignal,
|
|
207
|
+
});
|
|
208
|
+
if (response.status === 401) {
|
|
209
|
+
throw new ApiError(GATEWAY_401_MESSAGE, { statusCode: 401 });
|
|
210
|
+
}
|
|
211
|
+
await handleErrorResponse(response);
|
|
212
|
+
if (options.includeTimestamps) {
|
|
213
|
+
const payload = gatewayConversationJsonResponseSchema.parse(await response.json());
|
|
214
|
+
return {
|
|
215
|
+
audio: decodeBase64(payload.audio),
|
|
216
|
+
mediaType: payload.mediaType,
|
|
217
|
+
timestamps: payload.timestamps,
|
|
218
|
+
warnings: payload.warnings,
|
|
219
|
+
...(payload.providerMetadata !== undefined && {
|
|
220
|
+
providerMetadata: payload.providerMetadata,
|
|
221
|
+
}),
|
|
222
|
+
};
|
|
223
|
+
}
|
|
224
|
+
const arrayBuffer = await response.arrayBuffer();
|
|
225
|
+
return {
|
|
226
|
+
audio: new Uint8Array(arrayBuffer),
|
|
227
|
+
mediaType: mediaTypeFromHeaders(response.headers),
|
|
228
|
+
};
|
|
229
|
+
}
|
|
230
|
+
}
|
|
231
|
+
export function createSpeechGateway(config = {}) {
|
|
232
|
+
const provider = new SpeechGatewayProvider(config);
|
|
233
|
+
return function speechGateway(modelId) {
|
|
234
|
+
if (!modelId) {
|
|
235
|
+
throw new GatewayInputError('Speech Gateway requires a model ID (e.g., "openai/gpt-4o-mini-tts"). The gateway routes to upstream providers and has no default model.');
|
|
236
|
+
}
|
|
237
|
+
return { provider, modelId };
|
|
238
|
+
};
|
|
239
|
+
}
|
|
240
|
+
function mediaTypeFromHeaders(headers) {
|
|
241
|
+
return headers.get("content-type") ?? "audio/mpeg";
|
|
242
|
+
}
|
|
243
|
+
function decodeBase64(value) {
|
|
244
|
+
const binaryString = atob(value);
|
|
245
|
+
const bytes = new Uint8Array(binaryString.length);
|
|
246
|
+
for (let i = 0; i < binaryString.length; i++) {
|
|
247
|
+
bytes[i] = binaryString.charCodeAt(i);
|
|
248
|
+
}
|
|
249
|
+
return bytes;
|
|
250
|
+
}
|
|
251
|
+
//# sourceMappingURL=index.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/providers/gateway/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAExB,OAAO,EACL,QAAQ,EACR,iBAAiB,EACjB,kBAAkB,GACnB,MAAM,iBAAiB,CAAC;AAEzB,OAAO,EAAE,mBAAmB,EAAE,cAAc,EAAE,MAAM,yBAAyB,CAAC;AAW9E,MAAM,CAAC,MAAM,0BAA0B,GAAG,gBAAyB,CAAC;AAQpE,8GAA8G;AAC9G,MAAM,mBAAmB,GAAG,CAAC,CAAC,MAAM,CAAC;IACnC,IAAI,EAAE,CAAC,CAAC,MAAM,EAAE;IAChB,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE;IACjB,GAAG,EAAE,CAAC,CAAC,MAAM,EAAE;CAChB,CAAC,CAAC;AAEH,MAAM,+BAA+B,GAAG,mBAAmB,CAAC,MAAM,CAAC;IACjE,SAAS,EAAE,CAAC,CAAC,MAAM,EAAE;CACtB,CAAC,CAAC;AAEH,MAAM,yBAAyB,GAAG,CAAC,CAAC,MAAM,CAAC;IACzC,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE;IACjB,SAAS,EAAE,CAAC,CAAC,MAAM,EAAE;IACrB,UAAU,EAAE,CAAC,CAAC,KAAK,CAAC,mBAAmB,CAAC,CAAC,OAAO,CAAC,EAAE,CAAC;IACpD,QAAQ,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,MAAM,EAAE,CAAC,CAAC,OAAO,CAAC,EAAE,CAAC;CAC1C,CAAC,CAAC;AAEH,MAAM,qCAAqC,GAAG,CAAC,CAAC,MAAM,CAAC;IACrD,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE;IACjB,SAAS,EAAE,CAAC,CAAC,MAAM,EAAE;IACrB,UAAU,EAAE,CAAC,CAAC,KAAK,CAAC,+BAA+B,CAAC,CAAC,OAAO,CAAC,EAAE,CAAC;IAChE,QAAQ,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,MAAM,EAAE,CAAC,CAAC,OAAO,CAAC,EAAE,CAAC;IACzC,gBAAgB,EAAE,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,MAAM,EAAE,EAAE,CAAC,CAAC,OAAO,EAAE,CAAC,CAAC,QAAQ,EAAE;CAC/D,CAAC,CAAC;AAEH,MAAM,mBAAmB,GACvB,0IAA0I,CAAC;AAE7I,MAAM,OAAO,qBAAqB;IACvB,EAAE,GAAG,0BAA0B,CAAC;IAChC,YAAY,GAAG,EAAE,CAAC;IAC3B,qGAAqG;IAC5F,MAAM,GAAyB,EAAE,CAAC;IAE1B,MAAM,CAAqB;IAC3B,OAAO,CAAS;IAChB,OAAO,CAA0B;IAElD,YAAY,MAAmC;QAC7C,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC,OAAO,IAAI,kCAAkC,CAAC;QACpE,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC,KAAK,IAAI,UAAU,CAAC,KAAK,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC;IACnE,CAAC;IAEO,UAAU;QAChB,MAAM,GAAG,GACP,IAAI,CAAC,MAAM;YACX,CAAC,OAAO,OAAO,KAAK,WAAW;gBAC7B,CAAC,CAAC,SAAS;gBACX,CAAC,CAAC,OAAO,CAAC,GAAG,EAAE,sBAAsB,CAAC,CAAC;QAC3C,IAAI,CAAC,GAAG,EAAE,CAAC;YACT,MAAM,GAAG,GAAG,IAAI,kBAAkB,CAAC;gBACjC,YAAY,EAAE,gBAAgB;gBAC9B,MAAM,EAAE,wBAAwB;aACjC,CAAC,CAAC;YACH,GAAG,CAAC,OAAO;gBACT,wMAAwM,CAAC;YAC3M,MAAM,GAAG,CAAC;QACZ,CAAC;QACD,OAAO,GAAG,CAAC;IACb,CAAC;IAED,KAAK,CAAC,QAAQ,CAAC,OAWd;QAMC,IAAI,CAAC,OAAO,CAAC,KAAK,EAAE,CAAC;YACnB,MAAM,IAAI,iBAAiB,CACzB,kBAAkB,OAAO,CAAC,OAAO,+EAA+E,CACjH,CAAC;QACJ,CAAC;QAED,MAAM,IAAI,GAA4B;YACpC,IAAI,EAAE,QAAQ;YACd,KAAK,EAAE,OAAO,CAAC,OAAO;YACtB,KAAK,EAAE,OAAO,CAAC,KAAK;YACpB,IAAI,EAAE,OAAO,CAAC,IAAI;SACnB,CAAC;QACF,IAAI,OAAO,CAAC,UAAU,IAAI,IAAI,EAAE,CAAC;YAC/B,IAAI,CAAC,UAAU,GAAG,OAAO,CAAC,UAAU,CAAC;QACvC,CAAC;QACD,IAAI,OAAO,CAAC,eAAe,EAAE,CAAC;YAC5B,IAAI,CAAC,eAAe,GAAG,OAAO,CAAC,eAAe,CAAC;QACjD,CAAC;QACD,IAAI,OAAO,CAAC,MAAM,EAAE,CAAC;YACnB,IAAI,CAAC,MAAM,GAAG,OAAO,CAAC,MAAM,CAAC;QAC/B,CAAC;QACD,IAAI,OAAO,CAAC,cAAc,EAAE,CAAC;YAC3B,IAAI,CAAC,cAAc,GAAG,OAAO,CAAC,cAAc,CAAC;QAC/C,CAAC;QAED,+FAA+F;QAC/F,MAAM,GAAG,GAAG,OAAO,CAAC,iBAAiB;YACnC,CAAC,CAAC,GAAG,IAAI,CAAC,OAAO,+BAA+B;YAChD,CAAC,CAAC,GAAG,IAAI,CAAC,OAAO,eAAe,CAAC;QAEnC,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,GAAG,EAAE;YACvC,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,GAAG,OAAO,CAAC,OAAO;gBAClB,cAAc,EAAE,kBAAkB;gBAClC,aAAa,EAAE,UAAU,IAAI,CAAC,UAAU,EAAE,EAAE;gBAC5C,cAAc,EAAE,cAAc;aAC/B;YACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC;YAC1B,MAAM,EAAE,OAAO,CAAC,WAAW;SAC5B,CAAC,CAAC;QAEH,IAAI,QAAQ,CAAC,MAAM,KAAK,GAAG,EAAE,CAAC;YAC5B,MAAM,IAAI,QAAQ,CAAC,mBAAmB,EAAE,EAAE,UAAU,EAAE,GAAG,EAAE,CAAC,CAAC;QAC/D,CAAC;QACD,MAAM,mBAAmB,CAAC,QAAQ,CAAC,CAAC;QAEpC,IAAI,OAAO,CAAC,iBAAiB,EAAE,CAAC;YAC9B,MAAM,OAAO,GAAG,yBAAyB,CAAC,KAAK,CAAC,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC,CAAC;YACvE,OAAO;gBACL,KAAK,EAAE,YAAY,CAAC,OAAO,CAAC,KAAK,CAAC;gBAClC,SAAS,EAAE,OAAO,CAAC,SAAS;gBAC5B,UAAU,EAAE,OAAO,CAAC,UAAU;gBAC9B,QAAQ,EAAE,OAAO,CAAC,QAAQ;aAC3B,CAAC;QACJ,CAAC;QAED,MAAM,WAAW,GAAG,MAAM,QAAQ,CAAC,WAAW,EAAE,CAAC;QAEjD,OAAO;YACL,KAAK,EAAE,IAAI,UAAU,CAAC,WAAW,CAAC;YAClC,SAAS,EAAE,oBAAoB,CAAC,QAAQ,CAAC,OAAO,CAAC;SAClD,CAAC;IACJ,CAAC;IAED,KAAK,CAAC,MAAM,CAAC,OAQZ;QAMC,IAAI,CAAC,OAAO,CAAC,KAAK,EAAE,CAAC;YACnB,MAAM,IAAI,iBAAiB,CACzB,kBAAkB,OAAO,CAAC,OAAO,+EAA+E,CACjH,CAAC;QACJ,CAAC;QAED,MAAM,IAAI,GAA4B;YACpC,IAAI,EAAE,QAAQ;YACd,KAAK,EAAE,OAAO,CAAC,OAAO;YACtB,KAAK,EAAE,OAAO,CAAC,KAAK;YACpB,IAAI,EAAE,OAAO,CAAC,IAAI;SACnB,CAAC;QACF,IAAI,OAAO,CAAC,eAAe,EAAE,CAAC;YAC5B,IAAI,CAAC,eAAe,GAAG,OAAO,CAAC,eAAe,CAAC;QACjD,CAAC;QACD,IAAI,OAAO,CAAC,cAAc,EAAE,CAAC;YAC3B,IAAI,CAAC,cAAc,GAAG,OAAO,CAAC,cAAc,CAAC;QAC/C,CAAC;QAED,MAAM,GAAG,GAAG,GAAG,IAAI,CAAC,OAAO,eAAe,CAAC;QAE3C,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,GAAG,EAAE;YACvC,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,GAAG,OAAO,CAAC,OAAO;gBAClB,cAAc,EAAE,kBAAkB;gBAClC,aAAa,EAAE,UAAU,IAAI,CAAC,UAAU,EAAE,EAAE;gBAC5C,cAAc,EAAE,cAAc;aAC/B;YACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC;YAC1B,MAAM,EAAE,OAAO,CAAC,WAAW;SAC5B,CAAC,CAAC;QAEH,IAAI,QAAQ,CAAC,MAAM,KAAK,GAAG,EAAE,CAAC;YAC5B,MAAM,IAAI,QAAQ,CAAC,mBAAmB,EAAE,EAAE,UAAU,EAAE,GAAG,EAAE,CAAC,CAAC;QAC/D,CAAC;QACD,MAAM,mBAAmB,CAAC,QAAQ,CAAC,CAAC;QAEpC,IAAI,CAAC,QAAQ,CAAC,IAAI,EAAE,CAAC;YACnB,MAAM,IAAI,KAAK,CACb,kBAAkB,OAAO,CAAC,OAAO,wBAAwB,CAC1D,CAAC;QACJ,CAAC;QAED,OAAO;YACL,MAAM,EAAE,QAAQ,CAAC,IAAI;YACrB,SAAS,EAAE,oBAAoB,CAAC,QAAQ,CAAC,OAAO,CAAC;SAClD,CAAC;IACJ,CAAC;IAED,4FAA4F;IAC5F,yEAAyE;IACzE,iEAAiE;IACjE,oHAAoH;IACpH,KAAK,CAAC,oBAAoB,CAAC,OAgB1B;QAOC,IAAI,OAAO,CAAC,KAAK,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YAC/B,MAAM,IAAI,iBAAiB,CACzB,6DAA6D,CAC9D,CAAC;QACJ,CAAC;QACD,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,KAAK,CAAC,EAAE,CAAC;YACzC,MAAM,IAAI,iBAAiB,CACzB,yGAAyG,CAC1G,CAAC;QACJ,CAAC;QAED,MAAM,WAAW,GAAG,OAAO,CAAC,OAAO,CAAC;QACpC,MAAM,YAAY,GAAG,OAAO,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,KAAK,IAAI,IAAI,CAAC,CAAC;QAChE,IAAI,WAAW,IAAI,IAAI,IAAI,YAAY,EAAE,CAAC;YACxC,MAAM,IAAI,iBAAiB,CACzB,2GAA2G,CAC5G,CAAC;QACJ,CAAC;QACD,IAAI,WAAW,IAAI,IAAI,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,KAAK,CAAC,EAAE,CAAC;YAChE,MAAM,IAAI,iBAAiB,CACzB,wGAAwG,CACzG,CAAC;QACJ,CAAC;QAED,8EAA8E;QAC9E,MAAM,IAAI,GAA4B;YACpC,IAAI,EAAE,cAAc;YACpB,GAAG,CAAC,WAAW,IAAI,IAAI,IAAI,EAAE,KAAK,EAAE,WAAW,EAAE,CAAC;YAClD,KAAK,EAAE,OAAO,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;gBAC/B,GAAG,CAAC,CAAC,CAAC,KAAK,IAAI,IAAI,IAAI,EAAE,KAAK,EAAE,CAAC,CAAC,KAAK,EAAE,CAAC;gBAC1C,KAAK,EAAE,CAAC,CAAC,KAAK;gBACd,IAAI,EAAE,CAAC,CAAC,IAAI;gBACZ,GAAG,CAAC,CAAC,CAAC,eAAe,IAAI,EAAE,eAAe,EAAE,CAAC,CAAC,eAAe,EAAE,CAAC;aACjE,CAAC,CAAC;YACH,KAAK,EAAE,OAAO,CAAC,KAAK,IAAI,GAAG;YAC3B,UAAU,EAAE,OAAO,CAAC,UAAU,IAAI,CAAC,EAAE;SACtC,CAAC;QACF,IAAI,OAAO,CAAC,eAAe,EAAE,CAAC;YAC5B,IAAI,CAAC,eAAe,GAAG,OAAO,CAAC,eAAe,CAAC;QACjD,CAAC;QACD,IAAI,OAAO,CAAC,MAAM,EAAE,CAAC;YACnB,IAAI,CAAC,MAAM,GAAG,OAAO,CAAC,MAAM,CAAC;QAC/B,CAAC;QACD,IAAI,OAAO,CAAC,cAAc,EAAE,CAAC;YAC3B,IAAI,CAAC,cAAc,GAAG,OAAO,CAAC,cAAc,CAAC;QAC/C,CAAC;QAED,MAAM,GAAG,GAAG,OAAO,CAAC,iBAAiB;YACnC,CAAC,CAAC,GAAG,IAAI,CAAC,OAAO,qCAAqC;YACtD,CAAC,CAAC,GAAG,IAAI,CAAC,OAAO,qBAAqB,CAAC;QAEzC,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,GAAG,EAAE;YACvC,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,GAAG,OAAO,CAAC,OAAO;gBAClB,cAAc,EAAE,kBAAkB;gBAClC,aAAa,EAAE,UAAU,IAAI,CAAC,UAAU,EAAE,EAAE;gBAC5C,cAAc,EAAE,cAAc;aAC/B;YACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC;YAC1B,MAAM,EAAE,OAAO,CAAC,WAAW;SAC5B,CAAC,CAAC;QAEH,IAAI,QAAQ,CAAC,MAAM,KAAK,GAAG,EAAE,CAAC;YAC5B,MAAM,IAAI,QAAQ,CAAC,mBAAmB,EAAE,EAAE,UAAU,EAAE,GAAG,EAAE,CAAC,CAAC;QAC/D,CAAC;QACD,MAAM,mBAAmB,CAAC,QAAQ,CAAC,CAAC;QAEpC,IAAI,OAAO,CAAC,iBAAiB,EAAE,CAAC;YAC9B,MAAM,OAAO,GAAG,qCAAqC,CAAC,KAAK,CACzD,MAAM,QAAQ,CAAC,IAAI,EAAE,CACtB,CAAC;YACF,OAAO;gBACL,KAAK,EAAE,YAAY,CAAC,OAAO,CAAC,KAAK,CAAC;gBAClC,SAAS,EAAE,OAAO,CAAC,SAAS;gBAC5B,UAAU,EAAE,OAAO,CAAC,UAAU;gBAC9B,QAAQ,EAAE,OAAO,CAAC,QAAQ;gBAC1B,GAAG,CAAC,OAAO,CAAC,gBAAgB,KAAK,SAAS,IAAI;oBAC5C,gBAAgB,EAAE,OAAO,CAAC,gBAAgB;iBAC3C,CAAC;aACH,CAAC;QACJ,CAAC;QAED,MAAM,WAAW,GAAG,MAAM,QAAQ,CAAC,WAAW,EAAE,CAAC;QAEjD,OAAO;YACL,KAAK,EAAE,IAAI,UAAU,CAAC,WAAW,CAAC;YAClC,SAAS,EAAE,oBAAoB,CAAC,QAAQ,CAAC,OAAO,CAAC;SAClD,CAAC;IACJ,CAAC;CACF;AAED,MAAM,UAAU,mBAAmB,CAAC,SAAsC,EAAE;IAC1E,MAAM,QAAQ,GAAG,IAAI,qBAAqB,CAAC,MAAM,CAAC,CAAC;IACnD,OAAO,SAAS,aAAa,CAAC,OAAe;QAC3C,IAAI,CAAC,OAAO,EAAE,CAAC;YACb,MAAM,IAAI,iBAAiB,CACzB,yIAAyI,CAC1I,CAAC;QACJ,CAAC;QACD,OAAO,EAAE,QAAQ,EAAE,OAAO,EAAE,CAAC;IAC/B,CAAC,CAAC;AACJ,CAAC;AAED,SAAS,oBAAoB,CAAC,OAAgB;IAC5C,OAAO,OAAO,CAAC,GAAG,CAAC,cAAc,CAAC,IAAI,YAAY,CAAC;AACrD,CAAC;AAED,SAAS,YAAY,CAAC,KAAa;IACjC,MAAM,YAAY,GAAG,IAAI,CAAC,KAAK,CAAC,CAAC;IACjC,MAAM,KAAK,GAAG,IAAI,UAAU,CAAC,YAAY,CAAC,MAAM,CAAC,CAAC;IAClD,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,YAAY,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;QAC7C,KAAK,CAAC,CAAC,CAAC,GAAG,YAAY,CAAC,UAAU,CAAC,CAAC,CAAC,CAAC;IACxC,CAAC;IACD,OAAO,KAAK,CAAC;AACf,CAAC"}
|
|
@@ -1,30 +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 GoogleSpeechProviderConfig {
|
|
3
5
|
apiKey?: string;
|
|
4
6
|
baseURL?: string;
|
|
7
|
+
fallbackSTT?: ResolvedSTTModel;
|
|
5
8
|
fetch?: typeof globalThis.fetch;
|
|
6
9
|
}
|
|
10
|
+
export declare const GOOGLE_PROVIDER_ID: "google";
|
|
11
|
+
export declare const GOOGLE_MODELS: readonly ModelInfo[];
|
|
7
12
|
export declare class GoogleSpeechProvider implements SpeechProvider<string, string> {
|
|
8
|
-
readonly id
|
|
13
|
+
readonly id: "google";
|
|
9
14
|
readonly defaultModel = "gemini-2.5-flash-preview-tts";
|
|
10
|
-
|
|
11
|
-
private static readonly GEMINI_3_1_LANGUAGES;
|
|
12
|
-
readonly models: readonly [{
|
|
13
|
-
readonly id: "gemini-3.1-flash-tts-preview";
|
|
14
|
-
readonly releaseDate: "2026-04-15";
|
|
15
|
-
readonly languages: readonly ["af", "am", "ar", "az", "be", "bg", "bn", "ca", "ceb", "cmn", "cs", "da", "de", "el", "en", "es", "et", "eu", "fa", "fi", "fil", "fr", "gl", "gu", "he", "hi", "hr", "ht", "hu", "hy", "id", "is", "it", "ja", "jv", "ka", "kn", "ko", "kok", "la", "lb", "lo", "lt", "lv", "mai", "mg", "mk", "ml", "mn", "mr", "ms", "my", "nb", "ne", "nl", "nn", "or", "pa", "pl", "ps", "pt", "ro", "ru", "sd", "si", "sk", "sl", "sq", "sr", "sv", "sw", "ta", "te", "th", "tr", "uk", "ur", "vi"];
|
|
16
|
-
readonly features: readonly ["streaming", "audio-tags"];
|
|
17
|
-
}, {
|
|
18
|
-
readonly id: "gemini-2.5-flash-preview-tts";
|
|
19
|
-
readonly releaseDate: "2025-05-01";
|
|
20
|
-
readonly languages: readonly ["en", "fr", "de", "es", "pt", "zh", "ja", "ko", "hi", "it", "nl", "pl", "ru", "sv", "tr", "id", "ar", "cs", "da", "fi", "el", "hu", "ro", "uk"];
|
|
21
|
-
readonly features: readonly ["streaming"];
|
|
22
|
-
}, {
|
|
23
|
-
readonly id: "gemini-2.5-pro-preview-tts";
|
|
24
|
-
readonly releaseDate: "2025-05-01";
|
|
25
|
-
readonly languages: readonly ["en", "fr", "de", "es", "pt", "zh", "ja", "ko", "hi", "it", "nl", "pl", "ru", "sv", "tr", "id", "ar", "cs", "da", "fi", "el", "hu", "ro", "uk"];
|
|
26
|
-
readonly features: readonly ["streaming"];
|
|
27
|
-
}];
|
|
15
|
+
readonly models: readonly ModelInfo[];
|
|
28
16
|
private readonly apiKey;
|
|
29
17
|
private readonly baseURL;
|
|
30
18
|
private readonly fetchFn;
|
|
@@ -62,6 +50,10 @@ export declare class GoogleSpeechProvider implements SpeechProvider<string, stri
|
|
|
62
50
|
providerOptions: {};
|
|
63
51
|
mediaType: string;
|
|
64
52
|
} | undefined;
|
|
53
|
+
resolveOutputFormat(modelId: string, output: AudioOutput): {
|
|
54
|
+
providerOptions: {};
|
|
55
|
+
expectedMediaType: string;
|
|
56
|
+
} | undefined;
|
|
65
57
|
dialogueCapabilities(modelId: string): {
|
|
66
58
|
minVoices: number;
|
|
67
59
|
maxVoices: number;
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/providers/google/index.ts"],"names":[],"mappings":"
|
|
1
|
+
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/providers/google/index.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,uBAAuB,CAAC;AAazD,OAAO,EAEL,KAAK,SAAS,EACd,KAAK,aAAa,EAClB,KAAK,cAAc,EACpB,MAAM,0BAA0B,CAAC;AAClC,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,kCAAkC,CAAC;AA2BzE,MAAM,WAAW,0BAA0B;IACzC,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,kBAAkB,EAAG,QAAiB,CAAC;AA8GpD,eAAO,MAAM,aAAa,EAAE,SAAS,SAAS,EAmBpC,CAAC;AAEX,qBAAa,oBAAqB,YAAW,cAAc,CAAC,MAAM,EAAE,MAAM,CAAC;IACzE,QAAQ,CAAC,EAAE,WAAsB;IACjC,QAAQ,CAAC,YAAY,kCAAkC;IAEvD,QAAQ,CAAC,MAAM,uBAAiB;IAEhC,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAqB;IAC5C,OAAO,CAAC,QAAQ,CAAC,OAAO,CAAS;IACjC,OAAO,CAAC,QAAQ,CAAC,OAAO,CAA0B;gBAEtC,MAAM,EAAE,0BAA0B;IAQ9C,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,eAAe,CAAC,EAAE,MAAM,CAAC;QACzB,SAAS,EAAE,MAAM,CAAC;QAClB,gBAAgB,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;KAC5C,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;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;IAWF,gBAAgB,CAAC,OAAO,EAAE,MAAM;;;;IAWhC,mBAAmB,CAAC,OAAO,EAAE,MAAM,EAAE,MAAM,EAAE,WAAW;;;;IAmBxD,oBAAoB,CAAC,OAAO,EAAE,MAAM;;;;IAQ9B,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;CAyEH;AAED,wBAAgB,YAAY,CAAC,MAAM,GAAE,0BAA+B,IAI3C,UAAU,MAAM,KAAG,aAAa,CAAC,MAAM,CAAC,CAOhE"}
|