@volley/recognition-client-sdk 0.1.803 → 0.1.806
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/dist/browser.bundled.d.ts +71 -1
- package/dist/config-builder.d.ts +10 -1
- package/dist/config-builder.d.ts.map +1 -1
- package/dist/index.bundled.d.ts +133 -32
- package/dist/index.d.ts +2 -2
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +89 -9
- package/dist/index.js.map +4 -4
- package/dist/recog-client-sdk.browser.js +38 -9
- package/dist/recog-client-sdk.browser.js.map +4 -4
- package/dist/recognition-client.d.ts.map +1 -1
- package/dist/recognition-client.types.d.ts +9 -1
- package/dist/recognition-client.types.d.ts.map +1 -1
- package/dist/simplified-vgf-recognition-client.d.ts +53 -0
- package/dist/simplified-vgf-recognition-client.d.ts.map +1 -1
- package/dist/utils/url-builder.d.ts +3 -1
- package/dist/utils/url-builder.d.ts.map +1 -1
- package/package.json +4 -4
- package/src/config-builder.spec.ts +14 -0
- package/src/config-builder.ts +19 -1
- package/src/index.ts +6 -0
- package/src/recognition-client.ts +2 -0
- package/src/recognition-client.types.ts +12 -1
- package/src/simplified-vgf-recognition-client.spec.ts +74 -1
- package/src/simplified-vgf-recognition-client.ts +72 -0
- package/src/utils/url-builder.spec.ts +18 -0
- package/src/utils/url-builder.ts +9 -1
|
@@ -1,9 +1,40 @@
|
|
|
1
1
|
import { z } from 'zod';
|
|
2
2
|
|
|
3
|
+
/**
|
|
4
|
+
* Gemini Model Types
|
|
5
|
+
* Based on available models as of January 2025
|
|
6
|
+
*
|
|
7
|
+
* API Version Notes:
|
|
8
|
+
* - Gemini 2.5+ models: Use v1beta API (early access features)
|
|
9
|
+
* - Gemini 2.0 models: Use v1beta API (early access features)
|
|
10
|
+
* - Gemini 1.5 models: Use v1 API (stable, production-ready)
|
|
11
|
+
*
|
|
12
|
+
* @see https://ai.google.dev/gemini-api/docs/models
|
|
13
|
+
* @see https://ai.google.dev/gemini-api/docs/api-versions
|
|
14
|
+
*/
|
|
15
|
+
declare enum GeminiModel {
|
|
16
|
+
GEMINI_2_5_PRO = "gemini-2.5-pro",
|
|
17
|
+
GEMINI_2_5_FLASH = "gemini-2.5-flash",
|
|
18
|
+
GEMINI_2_5_FLASH_LITE = "gemini-2.5-flash-lite",
|
|
19
|
+
GEMINI_2_0_FLASH_LATEST = "gemini-2.0-flash-latest",
|
|
20
|
+
GEMINI_2_0_FLASH_EXP = "gemini-2.0-flash-exp"
|
|
21
|
+
}
|
|
22
|
+
|
|
23
|
+
/**
|
|
24
|
+
* OpenAI Batch API Model Types (HTTP API)
|
|
25
|
+
* @see https://platform.openai.com/docs/guides/speech-to-text
|
|
26
|
+
*
|
|
27
|
+
* Note: For openai-realtime provider models, see OpenAIRealtimeModel in provider.types.ts
|
|
28
|
+
*/
|
|
29
|
+
declare enum OpenAIModel {
|
|
30
|
+
WHISPER_1 = "whisper-1"
|
|
31
|
+
}
|
|
32
|
+
|
|
3
33
|
/**
|
|
4
34
|
* Provider types and enums for recognition services
|
|
5
35
|
* NOTE_TO_AI: DO NOT CHANGE THIS UNLESS EXPLICITLY ASKED. Always ask before making any changes.
|
|
6
36
|
*/
|
|
37
|
+
|
|
7
38
|
/**
|
|
8
39
|
* Supported speech recognition providers
|
|
9
40
|
*/
|
|
@@ -45,6 +76,17 @@ declare enum DeepgramModel {
|
|
|
45
76
|
NOVA_3 = "nova-3",
|
|
46
77
|
FLUX_GENERAL_EN = "flux-general-en"
|
|
47
78
|
}
|
|
79
|
+
/**
|
|
80
|
+
* AssemblyAI streaming speech models
|
|
81
|
+
* @see https://www.assemblyai.com/docs/streaming/universal-streaming
|
|
82
|
+
* @see https://www.assemblyai.com/docs/streaming/universal-3-pro
|
|
83
|
+
*/
|
|
84
|
+
declare enum AssemblyAIModel {
|
|
85
|
+
DEFAULT = "default",
|
|
86
|
+
UNIVERSAL_STREAMING_ENGLISH = "universal-streaming-english",
|
|
87
|
+
UNIVERSAL_STREAMING_MULTILINGUAL = "universal-streaming-multilingual",
|
|
88
|
+
U3_RT_PRO = "u3-rt-pro"
|
|
89
|
+
}
|
|
48
90
|
/**
|
|
49
91
|
* Google Cloud Speech models
|
|
50
92
|
* @see https://cloud.google.com/speech-to-text/docs/transcription-model
|
|
@@ -169,7 +211,7 @@ declare enum SelfServeVllmModel {
|
|
|
169
211
|
/**
|
|
170
212
|
* Type alias for any model from any provider
|
|
171
213
|
*/
|
|
172
|
-
type RecognitionModel = DeepgramModel | GoogleModel | FireworksModel | GladiaModel | ElevenLabsModel | OpenAIRealtimeModel | MistralVoxtralModel | CartesiaModel | DashScopeModel | InworldSttModel | SelfServeVllmModel | BedrockModel | AwsTranscribeModel | AmazonNovaSonicModel | string;
|
|
214
|
+
type RecognitionModel = AssemblyAIModel | DeepgramModel | GoogleModel | FireworksModel | GladiaModel | ElevenLabsModel | OpenAIModel | GeminiModel | OpenAIRealtimeModel | MistralVoxtralModel | CartesiaModel | DashScopeModel | InworldSttModel | SelfServeVllmModel | BedrockModel | AwsTranscribeModel | AmazonNovaSonicModel | string;
|
|
173
215
|
|
|
174
216
|
/**
|
|
175
217
|
* Recognition Result Types V1
|
|
@@ -797,6 +839,26 @@ declare const GameContextSchemaV1: z.ZodObject<{
|
|
|
797
839
|
}>;
|
|
798
840
|
type GameContextV1 = z.infer<typeof GameContextSchemaV1>;
|
|
799
841
|
|
|
842
|
+
/**
|
|
843
|
+
* Microphone Source Types
|
|
844
|
+
*
|
|
845
|
+
* Identifies which client-side audio capture path produced the audio.
|
|
846
|
+
* Paired with the upstream client capture enum — the recognition service
|
|
847
|
+
* always stores/uses this enum internally (never the raw wire string).
|
|
848
|
+
*
|
|
849
|
+
* Clients send the value as a string query param; normalize it to the enum
|
|
850
|
+
* at the SDK / server boundary via {@link coerceMicrophoneSourceType}.
|
|
851
|
+
*/
|
|
852
|
+
/**
|
|
853
|
+
* Audio capture source path reported by the client.
|
|
854
|
+
*/
|
|
855
|
+
declare enum MicrophoneSourceType {
|
|
856
|
+
/** Browser / legacy web capture path (default). */
|
|
857
|
+
WEB = "web",
|
|
858
|
+
/** Mobile native bridge path (MobileNativeSource) when supported. */
|
|
859
|
+
NATIVE = "native"
|
|
860
|
+
}
|
|
861
|
+
|
|
800
862
|
/**
|
|
801
863
|
* Unified ASR Request Configuration
|
|
802
864
|
*
|
|
@@ -1327,6 +1389,14 @@ interface IRecognitionClientConfig {
|
|
|
1327
1389
|
accountId?: string;
|
|
1328
1390
|
/** Question answer identifier for tracking Q&A sessions (optional and tracking purpose only) */
|
|
1329
1391
|
questionAnswerId?: string;
|
|
1392
|
+
/** Client identifier for downstream tracking/audio-labeling (optional and tracking purpose only) */
|
|
1393
|
+
clientId?: string;
|
|
1394
|
+
/**
|
|
1395
|
+
* Client audio capture path (optional). Accepts the {@link MicrophoneSourceType}
|
|
1396
|
+
* enum or its string value ('web' | 'native'). Defaults to 'web' on the server
|
|
1397
|
+
* when omitted. The service normalizes this to the enum at the boundary.
|
|
1398
|
+
*/
|
|
1399
|
+
microphoneSourceType?: MicrophoneSourceType | string;
|
|
1330
1400
|
/** Platform for audio recording device (optional, e.g., 'ios', 'android', 'web', 'unity') */
|
|
1331
1401
|
platform?: string;
|
|
1332
1402
|
/** Experiment cohort (optional). Defaults to 'control' if not provided. */
|
package/dist/config-builder.d.ts
CHANGED
|
@@ -4,7 +4,7 @@
|
|
|
4
4
|
* Simple builder pattern for RealTimeTwoWayWebSocketRecognitionClientConfig
|
|
5
5
|
*/
|
|
6
6
|
import type { RealTimeTwoWayWebSocketRecognitionClientConfig, RecognitionCallbackUrl } from './recognition-client.types.js';
|
|
7
|
-
import type { ASRRequestConfig, GameContextV1, TranscriptionResultV1, MetadataResultV1, SessionConfiguredV1, AudioMetricsResultV1, ErrorResultV1, Stage } from '@recog/shared-types';
|
|
7
|
+
import type { ASRRequestConfig, GameContextV1, TranscriptionResultV1, MetadataResultV1, SessionConfiguredV1, AudioMetricsResultV1, ErrorResultV1, Stage, MicrophoneSourceType } from '@recog/shared-types';
|
|
8
8
|
/**
|
|
9
9
|
* Builder for RealTimeTwoWayWebSocketRecognitionClientConfig
|
|
10
10
|
*
|
|
@@ -82,6 +82,15 @@ export declare class ConfigBuilder {
|
|
|
82
82
|
* Set question answer ID
|
|
83
83
|
*/
|
|
84
84
|
questionAnswerId(id: string): this;
|
|
85
|
+
/**
|
|
86
|
+
* Set client ID (for downstream tracking/audio-labeling)
|
|
87
|
+
*/
|
|
88
|
+
clientId(id: string): this;
|
|
89
|
+
/**
|
|
90
|
+
* Set the client audio capture path (MicrophoneSourceType enum or 'web' | 'native' string).
|
|
91
|
+
* Defaults to 'web' on the server when omitted.
|
|
92
|
+
*/
|
|
93
|
+
microphoneSourceType(source: MicrophoneSourceType | string): this;
|
|
85
94
|
/**
|
|
86
95
|
* Set platform
|
|
87
96
|
*/
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"config-builder.d.ts","sourceRoot":"","sources":["../src/config-builder.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AAEH,OAAO,KAAK,EACV,8CAA8C,EAC9C,sBAAsB,EACvB,MAAM,+BAA+B,CAAC;AACvC,OAAO,KAAK,EACV,gBAAgB,EAChB,aAAa,EACb,qBAAqB,EACrB,gBAAgB,EAChB,mBAAmB,EACnB,oBAAoB,EACpB,aAAa,EACb,KAAK,
|
|
1
|
+
{"version":3,"file":"config-builder.d.ts","sourceRoot":"","sources":["../src/config-builder.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AAEH,OAAO,KAAK,EACV,8CAA8C,EAC9C,sBAAsB,EACvB,MAAM,+BAA+B,CAAC;AACvC,OAAO,KAAK,EACV,gBAAgB,EAChB,aAAa,EACb,qBAAqB,EACrB,gBAAgB,EAChB,mBAAmB,EACnB,oBAAoB,EACpB,aAAa,EACb,KAAK,EACL,oBAAoB,EACrB,MAAM,qBAAqB,CAAC;AAE7B;;;;;;;;;;;;;;;;;;GAkBG;AACH,qBAAa,aAAa;IACxB,OAAO,CAAC,MAAM,CAA+D;IAE7E;;;OAGG;IACH,GAAG,CAAC,GAAG,EAAE,MAAM,GAAG,IAAI;IAKtB;;;;;;;;OAQG;IACH,KAAK,CAAC,KAAK,EAAE,KAAK,GAAG,MAAM,GAAG,IAAI;IAKlC;;OAEG;IACH,gBAAgB,CAAC,MAAM,EAAE,gBAAgB,GAAG,IAAI;IAKhD;;OAEG;IACH,WAAW,CAAC,OAAO,EAAE,aAAa,GAAG,IAAI;IAKzC;;;OAGG;IACH,MAAM,CAAC,EAAE,EAAE,MAAM,GAAG,IAAI;IAKxB;;OAEG;IACH,gBAAgB,CAAC,EAAE,EAAE,MAAM,GAAG,IAAI;IAKlC;;OAEG;IACH,YAAY,CAAC,IAAI,EAAE,sBAAsB,EAAE,GAAG,IAAI;IAKlD;;OAEG;IACH,MAAM,CAAC,EAAE,EAAE,MAAM,GAAG,IAAI;IAKxB;;OAEG;IACH,aAAa,CAAC,EAAE,EAAE,MAAM,GAAG,IAAI;IAK/B;;OAEG;IACH,QAAQ,CAAC,EAAE,EAAE,MAAM,GAAG,IAAI;IAK1B;;OAEG;IACH,SAAS,CAAC,EAAE,EAAE,MAAM,GAAG,IAAI;IAK3B;;OAEG;IACH,gBAAgB,CAAC,EAAE,EAAE,MAAM,GAAG,IAAI;IAKlC;;OAEG;IACH,QAAQ,CAAC,EAAE,EAAE,MAAM,GAAG,IAAI;IAK1B;;;OAGG;IACH,oBAAoB,CAAC,MAAM,EAAE,oBAAoB,GAAG,MAAM,GAAG,IAAI;IAKjE;;OAEG;IACH,QAAQ,CAAC,QAAQ,EAAE,MAAM,GAAG,IAAI;IAKhC;;OAEG;IACH,gBAAgB,CAAC,MAAM,EAAE,WAAW,GAAG,SAAS,GAAG,IAAI;IAKvD;;;OAGG;IACH,sBAAsB,CAAC,OAAO,EAAE,MAAM,GAAG,IAAI;IAK7C;;OAEG;IACH,YAAY,CAAC,QAAQ,EAAE,CAAC,MAAM,EAAE,qBAAqB,KAAK,IAAI,GAAG,IAAI;IAKrE;;OAEG;IACH,UAAU,CAAC,QAAQ,EAAE,CAAC,QAAQ,EAAE,gBAAgB,KAAK,IAAI,GAAG,IAAI;IAKhE;;OAEG;IACH,mBAAmB,CAAC,QAAQ,EAAE,CAAC,MAAM,EAAE,mBAAmB,KAAK,IAAI,GAAG,IAAI;IAK1E;;;OAGG;IACH,cAAc,CAAC,QAAQ,EAAE,CAAC,OAAO,EAAE,oBAAoB,KAAK,IAAI,GAAG,IAAI;IAKvE;;OAEG;IACH,OAAO,CAAC,QAAQ,EAAE,CAAC,KAAK,EAAE,aAAa,KAAK,IAAI,GAAG,IAAI;IAKvD;;OAEG;IACH,WAAW,CAAC,QAAQ,EAAE,MAAM,IAAI,GAAG,IAAI;IAKvC;;OAEG;IACH,cAAc,CAAC,QAAQ,EAAE,CAAC,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,KAAK,IAAI,GAAG,IAAI;IAKtE;;OAEG;IACH,aAAa,CAAC,KAAK,EAAE,MAAM,GAAG,IAAI;IAKlC;;OAEG;IACH,YAAY,CAAC,KAAK,EAAE,MAAM,GAAG,IAAI;IAKjC;;OAEG;IACH,oBAAoB,CAAC,OAAO,EAAE,MAAM,GAAG,IAAI;IAK3C;;OAEG;IACH,eAAe,CAAC,MAAM,EAAE,MAAM,GAAG,IAAI;IAKrC;;OAEG;IAEH,MAAM,CAAC,MAAM,EAAE,CAAC,KAAK,EAAE,OAAO,GAAG,MAAM,GAAG,MAAM,GAAG,OAAO,EAAE,OAAO,EAAE,MAAM,EAAE,IAAI,CAAC,EAAE,GAAG,KAAK,IAAI,GAAG,IAAI;IAKvG;;OAEG;IACH,KAAK,IAAI,8CAA8C;CAGxD"}
|
package/dist/index.bundled.d.ts
CHANGED
|
@@ -1,9 +1,40 @@
|
|
|
1
1
|
import { z } from 'zod';
|
|
2
2
|
|
|
3
|
+
/**
|
|
4
|
+
* Gemini Model Types
|
|
5
|
+
* Based on available models as of January 2025
|
|
6
|
+
*
|
|
7
|
+
* API Version Notes:
|
|
8
|
+
* - Gemini 2.5+ models: Use v1beta API (early access features)
|
|
9
|
+
* - Gemini 2.0 models: Use v1beta API (early access features)
|
|
10
|
+
* - Gemini 1.5 models: Use v1 API (stable, production-ready)
|
|
11
|
+
*
|
|
12
|
+
* @see https://ai.google.dev/gemini-api/docs/models
|
|
13
|
+
* @see https://ai.google.dev/gemini-api/docs/api-versions
|
|
14
|
+
*/
|
|
15
|
+
declare enum GeminiModel {
|
|
16
|
+
GEMINI_2_5_PRO = "gemini-2.5-pro",
|
|
17
|
+
GEMINI_2_5_FLASH = "gemini-2.5-flash",
|
|
18
|
+
GEMINI_2_5_FLASH_LITE = "gemini-2.5-flash-lite",
|
|
19
|
+
GEMINI_2_0_FLASH_LATEST = "gemini-2.0-flash-latest",
|
|
20
|
+
GEMINI_2_0_FLASH_EXP = "gemini-2.0-flash-exp"
|
|
21
|
+
}
|
|
22
|
+
|
|
23
|
+
/**
|
|
24
|
+
* OpenAI Batch API Model Types (HTTP API)
|
|
25
|
+
* @see https://platform.openai.com/docs/guides/speech-to-text
|
|
26
|
+
*
|
|
27
|
+
* Note: For openai-realtime provider models, see OpenAIRealtimeModel in provider.types.ts
|
|
28
|
+
*/
|
|
29
|
+
declare enum OpenAIModel {
|
|
30
|
+
WHISPER_1 = "whisper-1"
|
|
31
|
+
}
|
|
32
|
+
|
|
3
33
|
/**
|
|
4
34
|
* Provider types and enums for recognition services
|
|
5
35
|
* NOTE_TO_AI: DO NOT CHANGE THIS UNLESS EXPLICITLY ASKED. Always ask before making any changes.
|
|
6
36
|
*/
|
|
37
|
+
|
|
7
38
|
/**
|
|
8
39
|
* Supported speech recognition providers
|
|
9
40
|
*/
|
|
@@ -45,6 +76,17 @@ declare enum DeepgramModel {
|
|
|
45
76
|
NOVA_3 = "nova-3",
|
|
46
77
|
FLUX_GENERAL_EN = "flux-general-en"
|
|
47
78
|
}
|
|
79
|
+
/**
|
|
80
|
+
* AssemblyAI streaming speech models
|
|
81
|
+
* @see https://www.assemblyai.com/docs/streaming/universal-streaming
|
|
82
|
+
* @see https://www.assemblyai.com/docs/streaming/universal-3-pro
|
|
83
|
+
*/
|
|
84
|
+
declare enum AssemblyAIModel {
|
|
85
|
+
DEFAULT = "default",
|
|
86
|
+
UNIVERSAL_STREAMING_ENGLISH = "universal-streaming-english",
|
|
87
|
+
UNIVERSAL_STREAMING_MULTILINGUAL = "universal-streaming-multilingual",
|
|
88
|
+
U3_RT_PRO = "u3-rt-pro"
|
|
89
|
+
}
|
|
48
90
|
/**
|
|
49
91
|
* Google Cloud Speech models
|
|
50
92
|
* @see https://cloud.google.com/speech-to-text/docs/transcription-model
|
|
@@ -169,7 +211,7 @@ declare enum SelfServeVllmModel {
|
|
|
169
211
|
/**
|
|
170
212
|
* Type alias for any model from any provider
|
|
171
213
|
*/
|
|
172
|
-
type RecognitionModel = DeepgramModel | GoogleModel | FireworksModel | GladiaModel | ElevenLabsModel | OpenAIRealtimeModel | MistralVoxtralModel | CartesiaModel | DashScopeModel | InworldSttModel | SelfServeVllmModel | BedrockModel | AwsTranscribeModel | AmazonNovaSonicModel | string;
|
|
214
|
+
type RecognitionModel = AssemblyAIModel | DeepgramModel | GoogleModel | FireworksModel | GladiaModel | ElevenLabsModel | OpenAIModel | GeminiModel | OpenAIRealtimeModel | MistralVoxtralModel | CartesiaModel | DashScopeModel | InworldSttModel | SelfServeVllmModel | BedrockModel | AwsTranscribeModel | AmazonNovaSonicModel | string;
|
|
173
215
|
|
|
174
216
|
/**
|
|
175
217
|
* Recognition Result Types V1
|
|
@@ -1527,6 +1569,26 @@ declare const ASRRequestSchemaV1: z.ZodObject<{
|
|
|
1527
1569
|
}>;
|
|
1528
1570
|
type ASRRequestV1 = z.infer<typeof ASRRequestSchemaV1>;
|
|
1529
1571
|
|
|
1572
|
+
/**
|
|
1573
|
+
* Microphone Source Types
|
|
1574
|
+
*
|
|
1575
|
+
* Identifies which client-side audio capture path produced the audio.
|
|
1576
|
+
* Paired with the upstream client capture enum — the recognition service
|
|
1577
|
+
* always stores/uses this enum internally (never the raw wire string).
|
|
1578
|
+
*
|
|
1579
|
+
* Clients send the value as a string query param; normalize it to the enum
|
|
1580
|
+
* at the SDK / server boundary via {@link coerceMicrophoneSourceType}.
|
|
1581
|
+
*/
|
|
1582
|
+
/**
|
|
1583
|
+
* Audio capture source path reported by the client.
|
|
1584
|
+
*/
|
|
1585
|
+
declare enum MicrophoneSourceType {
|
|
1586
|
+
/** Browser / legacy web capture path (default). */
|
|
1587
|
+
WEB = "web",
|
|
1588
|
+
/** Mobile native bridge path (MobileNativeSource) when supported. */
|
|
1589
|
+
NATIVE = "native"
|
|
1590
|
+
}
|
|
1591
|
+
|
|
1530
1592
|
/**
|
|
1531
1593
|
* Unified ASR Request Configuration
|
|
1532
1594
|
*
|
|
@@ -1811,36 +1873,6 @@ type PartialASRRequestConfig = Partial<ASRRequestConfig>;
|
|
|
1811
1873
|
*/
|
|
1812
1874
|
declare function createDefaultASRConfig(overrides?: PartialASRRequestConfig): ASRRequestConfig;
|
|
1813
1875
|
|
|
1814
|
-
/**
|
|
1815
|
-
* Gemini Model Types
|
|
1816
|
-
* Based on available models as of January 2025
|
|
1817
|
-
*
|
|
1818
|
-
* API Version Notes:
|
|
1819
|
-
* - Gemini 2.5+ models: Use v1beta API (early access features)
|
|
1820
|
-
* - Gemini 2.0 models: Use v1beta API (early access features)
|
|
1821
|
-
* - Gemini 1.5 models: Use v1 API (stable, production-ready)
|
|
1822
|
-
*
|
|
1823
|
-
* @see https://ai.google.dev/gemini-api/docs/models
|
|
1824
|
-
* @see https://ai.google.dev/gemini-api/docs/api-versions
|
|
1825
|
-
*/
|
|
1826
|
-
declare enum GeminiModel {
|
|
1827
|
-
GEMINI_2_5_PRO = "gemini-2.5-pro",
|
|
1828
|
-
GEMINI_2_5_FLASH = "gemini-2.5-flash",
|
|
1829
|
-
GEMINI_2_5_FLASH_LITE = "gemini-2.5-flash-lite",
|
|
1830
|
-
GEMINI_2_0_FLASH_LATEST = "gemini-2.0-flash-latest",
|
|
1831
|
-
GEMINI_2_0_FLASH_EXP = "gemini-2.0-flash-exp"
|
|
1832
|
-
}
|
|
1833
|
-
|
|
1834
|
-
/**
|
|
1835
|
-
* OpenAI Batch API Model Types (HTTP API)
|
|
1836
|
-
* @see https://platform.openai.com/docs/guides/speech-to-text
|
|
1837
|
-
*
|
|
1838
|
-
* Note: For openai-realtime provider models, see OpenAIRealtimeModel in provider.types.ts
|
|
1839
|
-
*/
|
|
1840
|
-
declare enum OpenAIModel {
|
|
1841
|
-
WHISPER_1 = "whisper-1"
|
|
1842
|
-
}
|
|
1843
|
-
|
|
1844
1876
|
/**
|
|
1845
1877
|
* Standard stage/environment constants used across all services
|
|
1846
1878
|
*/
|
|
@@ -2096,6 +2128,14 @@ interface IRecognitionClientConfig {
|
|
|
2096
2128
|
accountId?: string;
|
|
2097
2129
|
/** Question answer identifier for tracking Q&A sessions (optional and tracking purpose only) */
|
|
2098
2130
|
questionAnswerId?: string;
|
|
2131
|
+
/** Client identifier for downstream tracking/audio-labeling (optional and tracking purpose only) */
|
|
2132
|
+
clientId?: string;
|
|
2133
|
+
/**
|
|
2134
|
+
* Client audio capture path (optional). Accepts the {@link MicrophoneSourceType}
|
|
2135
|
+
* enum or its string value ('web' | 'native'). Defaults to 'web' on the server
|
|
2136
|
+
* when omitted. The service normalizes this to the enum at the boundary.
|
|
2137
|
+
*/
|
|
2138
|
+
microphoneSourceType?: MicrophoneSourceType | string;
|
|
2099
2139
|
/** Platform for audio recording device (optional, e.g., 'ios', 'android', 'web', 'unity') */
|
|
2100
2140
|
platform?: string;
|
|
2101
2141
|
/** Experiment cohort (optional). Defaults to 'control' if not provided. */
|
|
@@ -2596,6 +2636,15 @@ declare class ConfigBuilder {
|
|
|
2596
2636
|
* Set question answer ID
|
|
2597
2637
|
*/
|
|
2598
2638
|
questionAnswerId(id: string): this;
|
|
2639
|
+
/**
|
|
2640
|
+
* Set client ID (for downstream tracking/audio-labeling)
|
|
2641
|
+
*/
|
|
2642
|
+
clientId(id: string): this;
|
|
2643
|
+
/**
|
|
2644
|
+
* Set the client audio capture path (MicrophoneSourceType enum or 'web' | 'native' string).
|
|
2645
|
+
* Defaults to 'web' on the server when omitted.
|
|
2646
|
+
*/
|
|
2647
|
+
microphoneSourceType(source: MicrophoneSourceType | string): this;
|
|
2599
2648
|
/**
|
|
2600
2649
|
* Set platform
|
|
2601
2650
|
*/
|
|
@@ -3024,6 +3073,40 @@ declare class SimplifiedVGFRecognitionClient implements ISimplifiedVGFRecognitio
|
|
|
3024
3073
|
getVGFState(): RecognitionState;
|
|
3025
3074
|
private notifyStateChange;
|
|
3026
3075
|
}
|
|
3076
|
+
/**
|
|
3077
|
+
* Fluent builder for {@link SimplifiedVGFClientConfig}.
|
|
3078
|
+
*
|
|
3079
|
+
* Extends {@link ConfigBuilder} with the two VGF-specific fields
|
|
3080
|
+
* (`onStateChange`, `initialState`) so the simplified client has the same
|
|
3081
|
+
* builder ergonomics as the base RealTimeTwoWayWebSocketRecognitionClient.
|
|
3082
|
+
*
|
|
3083
|
+
* Example:
|
|
3084
|
+
* ```typescript
|
|
3085
|
+
* import { STAGES, RecognitionProvider } from '@recog/shared-types';
|
|
3086
|
+
*
|
|
3087
|
+
* const config = new SimplifiedVGFConfigBuilder()
|
|
3088
|
+
* .stage(STAGES.STAGING)
|
|
3089
|
+
* .asrRequestConfig({ provider: RecognitionProvider.DEEPGRAM, language: 'en' })
|
|
3090
|
+
* .onStateChange((state) => setVGFState(state))
|
|
3091
|
+
* .build();
|
|
3092
|
+
* ```
|
|
3093
|
+
*/
|
|
3094
|
+
declare class SimplifiedVGFConfigBuilder extends ConfigBuilder {
|
|
3095
|
+
private vgfConfig;
|
|
3096
|
+
/**
|
|
3097
|
+
* Set the VGF state-change callback, fired whenever the VGF state updates.
|
|
3098
|
+
*/
|
|
3099
|
+
onStateChange(callback: (state: RecognitionState) => void): this;
|
|
3100
|
+
/**
|
|
3101
|
+
* Set the initial VGF state to restore from a previous session.
|
|
3102
|
+
* The audioUtteranceId is extracted from it when valid.
|
|
3103
|
+
*/
|
|
3104
|
+
initialState(state: RecognitionState): this;
|
|
3105
|
+
/**
|
|
3106
|
+
* Build the SimplifiedVGFClientConfig (base fields + VGF fields).
|
|
3107
|
+
*/
|
|
3108
|
+
build(): SimplifiedVGFClientConfig;
|
|
3109
|
+
}
|
|
3027
3110
|
/**
|
|
3028
3111
|
* Factory function for creating simplified client
|
|
3029
3112
|
* Usage examples:
|
|
@@ -3068,6 +3151,24 @@ declare class SimplifiedVGFRecognitionClient implements ISimplifiedVGFRecognitio
|
|
|
3068
3151
|
* // VGF state automatically updates based on transcription results
|
|
3069
3152
|
*/
|
|
3070
3153
|
declare function createSimplifiedVGFClient(config: SimplifiedVGFClientConfig): ISimplifiedVGFRecognitionClient;
|
|
3154
|
+
/**
|
|
3155
|
+
* Create a simplified VGF client using the builder pattern.
|
|
3156
|
+
*
|
|
3157
|
+
* Mirror of `createClientWithBuilder` for the base client.
|
|
3158
|
+
*
|
|
3159
|
+
* Example:
|
|
3160
|
+
* ```typescript
|
|
3161
|
+
* import { STAGES, RecognitionProvider } from '@recog/shared-types';
|
|
3162
|
+
*
|
|
3163
|
+
* const client = createSimplifiedVGFClientWithBuilder((builder) =>
|
|
3164
|
+
* builder
|
|
3165
|
+
* .stage(STAGES.STAGING)
|
|
3166
|
+
* .asrRequestConfig({ provider: RecognitionProvider.DEEPGRAM, language: 'en' })
|
|
3167
|
+
* .onStateChange((state) => setVGFState(state))
|
|
3168
|
+
* );
|
|
3169
|
+
* ```
|
|
3170
|
+
*/
|
|
3171
|
+
declare function createSimplifiedVGFClientWithBuilder(configure: (builder: SimplifiedVGFConfigBuilder) => SimplifiedVGFConfigBuilder): ISimplifiedVGFRecognitionClient;
|
|
3071
3172
|
|
|
3072
3173
|
/**
|
|
3073
3174
|
* VGF Recognition Mapper
|
|
@@ -3134,5 +3235,5 @@ declare function getRecognitionConductorHttpBase(stage?: Stage | string | null |
|
|
|
3134
3235
|
declare function getRecognitionConductorWsBase(stage?: Stage | string | null | undefined): string;
|
|
3135
3236
|
declare function getRecognitionConductorHost(stage?: Stage | string | null | undefined): string;
|
|
3136
3237
|
|
|
3137
|
-
export { AmazonNovaSonicModel, AudioEncoding, AwsTranscribeModel, BedrockModel, CartesiaModel, ClientControlActionV1, ClientState, ConfigBuilder, ConnectionError, ControlSignalTypeV1 as ControlSignal, ControlSignalTypeV1, DashScopeModel, DeepgramModel, ElevenLabsModel, ErrorTypeV1, FinalTranscriptStability, FireworksModel, GeminiModel, GladiaModel, GoogleModel, Language, MistralVoxtralModel, OpenAIModel, OpenAIRealtimeModel, RECOGNITION_CONDUCTOR_BASES, RECOGNITION_SERVICE_BASES, RealTimeTwoWayWebSocketRecognitionClient, RecognitionContextTypeV1, RecognitionError, RecognitionProvider, RecognitionResultTypeV1, RecognitionVGFStateSchema, RecordingStatus, STAGES, SampleRate, SelfServeVllmModel, SimplifiedVGFRecognitionClient, TimeoutError, TranscriptionStatus, ValidationError, createClient, createClientWithBuilder, createDefaultASRConfig, createInitialRecognitionState, createSimplifiedVGFClient, getRecognitionConductorBase, getRecognitionConductorHost, getRecognitionConductorHttpBase, getRecognitionConductorWsBase, getRecognitionServiceBase, getRecognitionServiceHost, getRecognitionServiceHttpBase, getRecognitionServiceWsBase, getUserFriendlyMessage, isExceptionImmediatelyAvailable, isNormalDisconnection, isTerminal, isValidRecordingStatusTransition, normalizeStage, resetRecognitionVGFState };
|
|
3238
|
+
export { AmazonNovaSonicModel, AssemblyAIModel, AudioEncoding, AwsTranscribeModel, BedrockModel, CartesiaModel, ClientControlActionV1, ClientState, ConfigBuilder, ConnectionError, ControlSignalTypeV1 as ControlSignal, ControlSignalTypeV1, DashScopeModel, DeepgramModel, ElevenLabsModel, ErrorTypeV1, FinalTranscriptStability, FireworksModel, GeminiModel, GladiaModel, GoogleModel, Language, MicrophoneSourceType, MistralVoxtralModel, OpenAIModel, OpenAIRealtimeModel, RECOGNITION_CONDUCTOR_BASES, RECOGNITION_SERVICE_BASES, RealTimeTwoWayWebSocketRecognitionClient, RecognitionContextTypeV1, RecognitionError, RecognitionProvider, RecognitionResultTypeV1, RecognitionVGFStateSchema, RecordingStatus, STAGES, SampleRate, SelfServeVllmModel, SimplifiedVGFConfigBuilder, SimplifiedVGFRecognitionClient, TimeoutError, TranscriptionStatus, ValidationError, createClient, createClientWithBuilder, createDefaultASRConfig, createInitialRecognitionState, createSimplifiedVGFClient, createSimplifiedVGFClientWithBuilder, getRecognitionConductorBase, getRecognitionConductorHost, getRecognitionConductorHttpBase, getRecognitionConductorWsBase, getRecognitionServiceBase, getRecognitionServiceHost, getRecognitionServiceHttpBase, getRecognitionServiceWsBase, getUserFriendlyMessage, isExceptionImmediatelyAvailable, isNormalDisconnection, isTerminal, isValidRecordingStatusTransition, normalizeStage, resetRecognitionVGFState };
|
|
3138
3239
|
export type { ASRRequestConfig, ASRRequestV1, AudioMetricsResultV1, AuthenticationException, ConnectionException, ErrorResultV1, FunctionCallResultV1, GameContextV1, IRecognitionClient, IRecognitionClientConfig, IRecognitionClientStats, ISimplifiedVGFRecognitionClient, MetadataResultV1, ProviderException, QuotaExceededException, RealTimeTwoWayWebSocketRecognitionClientConfig, RecognitionCallbackUrl, RecognitionException, RecognitionState, RecordingStatusType, SimplifiedVGFClientConfig, SlotMap, Stage, TimeoutException, TranscriptionResult, TranscriptionResultV1, TranscriptionStatusType, UnknownException, ValidationException };
|
package/dist/index.d.ts
CHANGED
|
@@ -6,11 +6,11 @@ export { RecognitionError, ConnectionError, TimeoutError, ValidationError } from
|
|
|
6
6
|
export { ErrorTypeV1 } from '@recog/shared-types';
|
|
7
7
|
export type { RecognitionException, ConnectionException, TimeoutException, ValidationException, AuthenticationException, ProviderException, QuotaExceededException, UnknownException } from '@recog/shared-types';
|
|
8
8
|
export { isExceptionImmediatelyAvailable, getUserFriendlyMessage } from '@recog/shared-types';
|
|
9
|
-
export { SimplifiedVGFRecognitionClient, createSimplifiedVGFClient, type ISimplifiedVGFRecognitionClient, type SimplifiedVGFClientConfig } from './simplified-vgf-recognition-client.js';
|
|
9
|
+
export { SimplifiedVGFRecognitionClient, SimplifiedVGFConfigBuilder, createSimplifiedVGFClient, createSimplifiedVGFClientWithBuilder, type ISimplifiedVGFRecognitionClient, type SimplifiedVGFClientConfig } from './simplified-vgf-recognition-client.js';
|
|
10
10
|
export { type RecognitionState, RecognitionVGFStateSchema, RecordingStatus, TranscriptionStatus, type RecordingStatusType, type TranscriptionStatusType, createInitialRecognitionState, isTerminal, isValidRecordingStatusTransition } from './vgf-recognition-state.js';
|
|
11
11
|
export { resetRecognitionVGFState } from './vgf-recognition-mapper.js';
|
|
12
12
|
export { AudioEncoding } from '@recog/websocket';
|
|
13
13
|
export { type GameContextV1, type SlotMap, RecognitionContextTypeV1, ControlSignalTypeV1, ControlSignalTypeV1 as ControlSignal, // Alias for backward compatibility
|
|
14
|
-
type TranscriptionResultV1, type FunctionCallResultV1, type MetadataResultV1, type AudioMetricsResultV1, type ErrorResultV1, RecognitionResultTypeV1, ClientControlActionV1, type ASRRequestConfig, type ASRRequestV1, FinalTranscriptStability, createDefaultASRConfig, RecognitionProvider, DeepgramModel, ElevenLabsModel, FireworksModel, GladiaModel, GoogleModel, GeminiModel, OpenAIModel, SelfServeVllmModel, OpenAIRealtimeModel, MistralVoxtralModel, CartesiaModel, DashScopeModel, BedrockModel, AwsTranscribeModel, AmazonNovaSonicModel, Language, SampleRate, STAGES, type Stage } from '@recog/shared-types';
|
|
14
|
+
type TranscriptionResultV1, type FunctionCallResultV1, type MetadataResultV1, type AudioMetricsResultV1, type ErrorResultV1, RecognitionResultTypeV1, ClientControlActionV1, type ASRRequestConfig, type ASRRequestV1, FinalTranscriptStability, createDefaultASRConfig, RecognitionProvider, AssemblyAIModel, DeepgramModel, ElevenLabsModel, FireworksModel, GladiaModel, GoogleModel, GeminiModel, OpenAIModel, SelfServeVllmModel, OpenAIRealtimeModel, MistralVoxtralModel, CartesiaModel, DashScopeModel, BedrockModel, AwsTranscribeModel, AmazonNovaSonicModel, Language, SampleRate, MicrophoneSourceType, STAGES, type Stage } from '@recog/shared-types';
|
|
15
15
|
export { getRecognitionServiceBase, getRecognitionServiceHttpBase, getRecognitionServiceWsBase, getRecognitionServiceHost, getRecognitionConductorBase, getRecognitionConductorHttpBase, getRecognitionConductorWsBase, getRecognitionConductorHost, normalizeStage, RECOGNITION_SERVICE_BASES, RECOGNITION_CONDUCTOR_BASES } from '@recog/shared-config';
|
|
16
16
|
//# sourceMappingURL=index.d.ts.map
|
package/dist/index.d.ts.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AACA,OAAO,EACL,wCAAwC,EACxC,KAAK,8CAA8C,EACnD,KAAK,mBAAmB,EACxB,qBAAqB,EACtB,MAAM,yBAAyB,CAAC;AAGjC,OAAO,EACL,KAAK,kBAAkB,EACvB,KAAK,wBAAwB,EAC7B,KAAK,uBAAuB,EAC5B,KAAK,sBAAsB,EAC3B,WAAW,EACZ,MAAM,+BAA+B,CAAC;AAGvC,OAAO,EAAE,aAAa,EAAE,MAAM,qBAAqB,CAAC;AAGpD,OAAO,EAAE,YAAY,EAAE,uBAAuB,EAAE,MAAM,cAAc,CAAC;AAGrE,OAAO,EACL,gBAAgB,EAChB,eAAe,EACf,YAAY,EACZ,eAAe,EAChB,MAAM,aAAa,CAAC;AAGrB,OAAO,EAAE,WAAW,EAAE,MAAM,qBAAqB,CAAC;AAGlD,YAAY,EACV,oBAAoB,EACpB,mBAAmB,EACnB,gBAAgB,EAChB,mBAAmB,EACnB,uBAAuB,EACvB,iBAAiB,EACjB,sBAAsB,EACtB,gBAAgB,EACjB,MAAM,qBAAqB,CAAC;AAG7B,OAAO,EACL,+BAA+B,EAC/B,sBAAsB,EACvB,MAAM,qBAAqB,CAAC;AAG7B,OAAO,EACL,8BAA8B,EAC9B,yBAAyB,EACzB,KAAK,+BAA+B,EACpC,KAAK,yBAAyB,EAC/B,MAAM,wCAAwC,CAAC;AAEhD,OAAO,EACL,KAAK,gBAAgB,EACrB,yBAAyB,EACzB,eAAe,EACf,mBAAmB,EACnB,KAAK,mBAAmB,EACxB,KAAK,uBAAuB,EAC5B,6BAA6B,EAC7B,UAAU,EACV,gCAAgC,EACjC,MAAM,4BAA4B,CAAC;AAEpC,OAAO,EAAE,wBAAwB,EAAE,MAAM,6BAA6B,CAAC;AAGvE,OAAO,EAAE,aAAa,EAAE,MAAM,kBAAkB,CAAC;AAGjD,OAAO,EAEL,KAAK,aAAa,EAClB,KAAK,OAAO,EACZ,wBAAwB,EACxB,mBAAmB,EACnB,mBAAmB,IAAI,aAAa,EAAG,mCAAmC;AAG1E,KAAK,qBAAqB,EAC1B,KAAK,oBAAoB,EACzB,KAAK,gBAAgB,EACrB,KAAK,oBAAoB,EACzB,KAAK,aAAa,EAClB,uBAAuB,EACvB,qBAAqB,EAGrB,KAAK,gBAAgB,EACrB,KAAK,YAAY,EACjB,wBAAwB,EACxB,sBAAsB,EACtB,mBAAmB,EACnB,aAAa,EACb,eAAe,EACf,cAAc,EACd,WAAW,EACX,WAAW,EACX,WAAW,EACX,WAAW,EACX,kBAAkB,EAClB,mBAAmB,EACnB,mBAAmB,EACnB,aAAa,EACb,cAAc,EACd,YAAY,EACZ,kBAAkB,EAClB,oBAAoB,EACpB,QAAQ,EACR,UAAU,EAGV,MAAM,EACN,KAAK,KAAK,EACX,MAAM,qBAAqB,CAAC;AAG7B,OAAO,EACL,yBAAyB,EACzB,6BAA6B,EAC7B,2BAA2B,EAC3B,yBAAyB,EACzB,2BAA2B,EAC3B,+BAA+B,EAC/B,6BAA6B,EAC7B,2BAA2B,EAC3B,cAAc,EACd,yBAAyB,EACzB,2BAA2B,EAC5B,MAAM,sBAAsB,CAAC"}
|
|
1
|
+
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AACA,OAAO,EACL,wCAAwC,EACxC,KAAK,8CAA8C,EACnD,KAAK,mBAAmB,EACxB,qBAAqB,EACtB,MAAM,yBAAyB,CAAC;AAGjC,OAAO,EACL,KAAK,kBAAkB,EACvB,KAAK,wBAAwB,EAC7B,KAAK,uBAAuB,EAC5B,KAAK,sBAAsB,EAC3B,WAAW,EACZ,MAAM,+BAA+B,CAAC;AAGvC,OAAO,EAAE,aAAa,EAAE,MAAM,qBAAqB,CAAC;AAGpD,OAAO,EAAE,YAAY,EAAE,uBAAuB,EAAE,MAAM,cAAc,CAAC;AAGrE,OAAO,EACL,gBAAgB,EAChB,eAAe,EACf,YAAY,EACZ,eAAe,EAChB,MAAM,aAAa,CAAC;AAGrB,OAAO,EAAE,WAAW,EAAE,MAAM,qBAAqB,CAAC;AAGlD,YAAY,EACV,oBAAoB,EACpB,mBAAmB,EACnB,gBAAgB,EAChB,mBAAmB,EACnB,uBAAuB,EACvB,iBAAiB,EACjB,sBAAsB,EACtB,gBAAgB,EACjB,MAAM,qBAAqB,CAAC;AAG7B,OAAO,EACL,+BAA+B,EAC/B,sBAAsB,EACvB,MAAM,qBAAqB,CAAC;AAG7B,OAAO,EACL,8BAA8B,EAC9B,0BAA0B,EAC1B,yBAAyB,EACzB,oCAAoC,EACpC,KAAK,+BAA+B,EACpC,KAAK,yBAAyB,EAC/B,MAAM,wCAAwC,CAAC;AAEhD,OAAO,EACL,KAAK,gBAAgB,EACrB,yBAAyB,EACzB,eAAe,EACf,mBAAmB,EACnB,KAAK,mBAAmB,EACxB,KAAK,uBAAuB,EAC5B,6BAA6B,EAC7B,UAAU,EACV,gCAAgC,EACjC,MAAM,4BAA4B,CAAC;AAEpC,OAAO,EAAE,wBAAwB,EAAE,MAAM,6BAA6B,CAAC;AAGvE,OAAO,EAAE,aAAa,EAAE,MAAM,kBAAkB,CAAC;AAGjD,OAAO,EAEL,KAAK,aAAa,EAClB,KAAK,OAAO,EACZ,wBAAwB,EACxB,mBAAmB,EACnB,mBAAmB,IAAI,aAAa,EAAG,mCAAmC;AAG1E,KAAK,qBAAqB,EAC1B,KAAK,oBAAoB,EACzB,KAAK,gBAAgB,EACrB,KAAK,oBAAoB,EACzB,KAAK,aAAa,EAClB,uBAAuB,EACvB,qBAAqB,EAGrB,KAAK,gBAAgB,EACrB,KAAK,YAAY,EACjB,wBAAwB,EACxB,sBAAsB,EACtB,mBAAmB,EACnB,eAAe,EACf,aAAa,EACb,eAAe,EACf,cAAc,EACd,WAAW,EACX,WAAW,EACX,WAAW,EACX,WAAW,EACX,kBAAkB,EAClB,mBAAmB,EACnB,mBAAmB,EACnB,aAAa,EACb,cAAc,EACd,YAAY,EACZ,kBAAkB,EAClB,oBAAoB,EACpB,QAAQ,EACR,UAAU,EAGV,oBAAoB,EAGpB,MAAM,EACN,KAAK,KAAK,EACX,MAAM,qBAAqB,CAAC;AAG7B,OAAO,EACL,yBAAyB,EACzB,6BAA6B,EAC7B,2BAA2B,EAC3B,yBAAyB,EACzB,2BAA2B,EAC3B,+BAA+B,EAC/B,6BAA6B,EAC7B,2BAA2B,EAC3B,cAAc,EACd,yBAAyB,EACzB,2BAA2B,EAC5B,MAAM,sBAAsB,CAAC"}
|
package/dist/index.js
CHANGED
|
@@ -3769,6 +3769,13 @@ var DeepgramModel;
|
|
|
3769
3769
|
DeepgramModel2["NOVA_3"] = "nova-3";
|
|
3770
3770
|
DeepgramModel2["FLUX_GENERAL_EN"] = "flux-general-en";
|
|
3771
3771
|
})(DeepgramModel || (DeepgramModel = {}));
|
|
3772
|
+
var AssemblyAIModel;
|
|
3773
|
+
(function(AssemblyAIModel2) {
|
|
3774
|
+
AssemblyAIModel2["DEFAULT"] = "default";
|
|
3775
|
+
AssemblyAIModel2["UNIVERSAL_STREAMING_ENGLISH"] = "universal-streaming-english";
|
|
3776
|
+
AssemblyAIModel2["UNIVERSAL_STREAMING_MULTILINGUAL"] = "universal-streaming-multilingual";
|
|
3777
|
+
AssemblyAIModel2["U3_RT_PRO"] = "u3-rt-pro";
|
|
3778
|
+
})(AssemblyAIModel || (AssemblyAIModel = {}));
|
|
3772
3779
|
var GoogleModel;
|
|
3773
3780
|
(function(GoogleModel2) {
|
|
3774
3781
|
GoogleModel2["CHIRP_3"] = "chirp_3";
|
|
@@ -4499,12 +4506,26 @@ var RecognitionContextSchemaV1 = z.discriminatedUnion("type", [
|
|
|
4499
4506
|
ASRRequestSchemaV1
|
|
4500
4507
|
]);
|
|
4501
4508
|
|
|
4509
|
+
// ../../libs/types/dist/microphone-source.types.js
|
|
4510
|
+
var MicrophoneSourceType;
|
|
4511
|
+
(function(MicrophoneSourceType2) {
|
|
4512
|
+
MicrophoneSourceType2["WEB"] = "web";
|
|
4513
|
+
MicrophoneSourceType2["NATIVE"] = "native";
|
|
4514
|
+
})(MicrophoneSourceType || (MicrophoneSourceType = {}));
|
|
4515
|
+
var DEFAULT_MICROPHONE_SOURCE_TYPE = MicrophoneSourceType.WEB;
|
|
4516
|
+
var NAME_TO_ENUM = {
|
|
4517
|
+
web: MicrophoneSourceType.WEB,
|
|
4518
|
+
native: MicrophoneSourceType.NATIVE
|
|
4519
|
+
};
|
|
4520
|
+
|
|
4502
4521
|
// ../../libs/types/dist/recognition-query.types.js
|
|
4503
4522
|
var RecognitionGameInfoSchema = z.object({
|
|
4504
4523
|
userId: z.string().optional(),
|
|
4505
4524
|
gameSessionId: z.string().optional(),
|
|
4506
4525
|
deviceId: z.string().optional(),
|
|
4507
4526
|
accountId: z.string().optional(),
|
|
4527
|
+
clientId: z.string().optional(),
|
|
4528
|
+
microphoneSourceType: z.nativeEnum(MicrophoneSourceType).optional(),
|
|
4508
4529
|
gameId: z.string().optional(),
|
|
4509
4530
|
gamePhase: z.string().optional(),
|
|
4510
4531
|
questionAskedId: z.string().optional(),
|
|
@@ -4545,7 +4566,7 @@ var AudioFormat;
|
|
|
4545
4566
|
[5, AudioFormat2.OPUS],
|
|
4546
4567
|
[6, AudioFormat2.PCM]
|
|
4547
4568
|
]);
|
|
4548
|
-
const
|
|
4569
|
+
const NAME_TO_ENUM2 = /* @__PURE__ */ new Map([
|
|
4549
4570
|
["WAV", AudioFormat2.WAV],
|
|
4550
4571
|
["MP3", AudioFormat2.MP3],
|
|
4551
4572
|
["FLAC", AudioFormat2.FLAC],
|
|
@@ -4566,7 +4587,7 @@ var AudioFormat;
|
|
|
4566
4587
|
}
|
|
4567
4588
|
AudioFormat2.fromId = fromId;
|
|
4568
4589
|
function fromName(nameStr) {
|
|
4569
|
-
return
|
|
4590
|
+
return NAME_TO_ENUM2.get(nameStr.toUpperCase());
|
|
4570
4591
|
}
|
|
4571
4592
|
AudioFormat2.fromName = fromName;
|
|
4572
4593
|
function toId(format) {
|
|
@@ -4582,7 +4603,7 @@ var AudioFormat;
|
|
|
4582
4603
|
}
|
|
4583
4604
|
AudioFormat2.isIdValid = isIdValid;
|
|
4584
4605
|
function isNameValid(nameStr) {
|
|
4585
|
-
return
|
|
4606
|
+
return NAME_TO_ENUM2.has(nameStr.toUpperCase());
|
|
4586
4607
|
}
|
|
4587
4608
|
AudioFormat2.isNameValid = isNameValid;
|
|
4588
4609
|
})(AudioFormat || (AudioFormat = {}));
|
|
@@ -4604,7 +4625,7 @@ var AudioEncoding;
|
|
|
4604
4625
|
[4, AudioEncoding2.MULAW],
|
|
4605
4626
|
[5, AudioEncoding2.ALAW]
|
|
4606
4627
|
]);
|
|
4607
|
-
const
|
|
4628
|
+
const NAME_TO_ENUM2 = /* @__PURE__ */ new Map([
|
|
4608
4629
|
["ENCODING_UNSPECIFIED", AudioEncoding2.ENCODING_UNSPECIFIED],
|
|
4609
4630
|
["LINEAR16", AudioEncoding2.LINEAR16],
|
|
4610
4631
|
["OGG_OPUS", AudioEncoding2.OGG_OPUS],
|
|
@@ -4625,7 +4646,7 @@ var AudioEncoding;
|
|
|
4625
4646
|
}
|
|
4626
4647
|
AudioEncoding2.fromId = fromId;
|
|
4627
4648
|
function fromName(nameStr) {
|
|
4628
|
-
return
|
|
4649
|
+
return NAME_TO_ENUM2.get(nameStr.toUpperCase());
|
|
4629
4650
|
}
|
|
4630
4651
|
AudioEncoding2.fromName = fromName;
|
|
4631
4652
|
function toId(encoding) {
|
|
@@ -4641,7 +4662,7 @@ var AudioEncoding;
|
|
|
4641
4662
|
}
|
|
4642
4663
|
AudioEncoding2.isIdValid = isIdValid;
|
|
4643
4664
|
function isNameValid(nameStr) {
|
|
4644
|
-
return
|
|
4665
|
+
return NAME_TO_ENUM2.has(nameStr.toUpperCase());
|
|
4645
4666
|
}
|
|
4646
4667
|
AudioEncoding2.isNameValid = isNameValid;
|
|
4647
4668
|
function coerce2(value, onStringInput) {
|
|
@@ -4681,7 +4702,7 @@ var SampleRate;
|
|
|
4681
4702
|
[44100, SampleRate2.RATE_44100],
|
|
4682
4703
|
[48e3, SampleRate2.RATE_48000]
|
|
4683
4704
|
]);
|
|
4684
|
-
const
|
|
4705
|
+
const NAME_TO_ENUM2 = /* @__PURE__ */ new Map([
|
|
4685
4706
|
["RATE_8000", SampleRate2.RATE_8000],
|
|
4686
4707
|
["RATE_16000", SampleRate2.RATE_16000],
|
|
4687
4708
|
["RATE_22050", SampleRate2.RATE_22050],
|
|
@@ -4704,7 +4725,7 @@ var SampleRate;
|
|
|
4704
4725
|
}
|
|
4705
4726
|
SampleRate2.fromHz = fromHz;
|
|
4706
4727
|
function fromName(nameStr) {
|
|
4707
|
-
return
|
|
4728
|
+
return NAME_TO_ENUM2.get(nameStr.toUpperCase());
|
|
4708
4729
|
}
|
|
4709
4730
|
SampleRate2.fromName = fromName;
|
|
4710
4731
|
function toHz(rate) {
|
|
@@ -4720,7 +4741,7 @@ var SampleRate;
|
|
|
4720
4741
|
}
|
|
4721
4742
|
SampleRate2.isHzValid = isHzValid;
|
|
4722
4743
|
function isNameValid(nameStr) {
|
|
4723
|
-
return
|
|
4744
|
+
return NAME_TO_ENUM2.has(nameStr.toUpperCase());
|
|
4724
4745
|
}
|
|
4725
4746
|
SampleRate2.isNameValid = isNameValid;
|
|
4726
4747
|
})(SampleRate || (SampleRate = {}));
|
|
@@ -5190,6 +5211,12 @@ function buildWebSocketUrl(config) {
|
|
|
5190
5211
|
if (config.questionAnswerId) {
|
|
5191
5212
|
url.searchParams.set("questionAnswerId", config.questionAnswerId);
|
|
5192
5213
|
}
|
|
5214
|
+
if (config.clientId) {
|
|
5215
|
+
url.searchParams.set("clientId", config.clientId);
|
|
5216
|
+
}
|
|
5217
|
+
if (config.microphoneSourceType) {
|
|
5218
|
+
url.searchParams.set("microphoneSourceType", config.microphoneSourceType);
|
|
5219
|
+
}
|
|
5193
5220
|
if (config.platform) {
|
|
5194
5221
|
url.searchParams.set("platform", config.platform);
|
|
5195
5222
|
}
|
|
@@ -5535,6 +5562,8 @@ var RealTimeTwoWayWebSocketRecognitionClient = class _RealTimeTwoWayWebSocketRec
|
|
|
5535
5562
|
...config.deviceId && { deviceId: config.deviceId },
|
|
5536
5563
|
...config.accountId && { accountId: config.accountId },
|
|
5537
5564
|
...config.questionAnswerId && { questionAnswerId: config.questionAnswerId },
|
|
5565
|
+
...config.clientId && { clientId: config.clientId },
|
|
5566
|
+
...config.microphoneSourceType && { microphoneSourceType: config.microphoneSourceType },
|
|
5538
5567
|
...config.platform && { platform: config.platform },
|
|
5539
5568
|
...config.gameContext && { gameContext: config.gameContext },
|
|
5540
5569
|
...config.gameId && { gameId: config.gameId },
|
|
@@ -6292,6 +6321,21 @@ var ConfigBuilder = class {
|
|
|
6292
6321
|
this.config.questionAnswerId = id;
|
|
6293
6322
|
return this;
|
|
6294
6323
|
}
|
|
6324
|
+
/**
|
|
6325
|
+
* Set client ID (for downstream tracking/audio-labeling)
|
|
6326
|
+
*/
|
|
6327
|
+
clientId(id) {
|
|
6328
|
+
this.config.clientId = id;
|
|
6329
|
+
return this;
|
|
6330
|
+
}
|
|
6331
|
+
/**
|
|
6332
|
+
* Set the client audio capture path (MicrophoneSourceType enum or 'web' | 'native' string).
|
|
6333
|
+
* Defaults to 'web' on the server when omitted.
|
|
6334
|
+
*/
|
|
6335
|
+
microphoneSourceType(source) {
|
|
6336
|
+
this.config.microphoneSourceType = source;
|
|
6337
|
+
return this;
|
|
6338
|
+
}
|
|
6295
6339
|
/**
|
|
6296
6340
|
* Set platform
|
|
6297
6341
|
*/
|
|
@@ -6923,11 +6967,44 @@ var SimplifiedVGFRecognitionClient = class {
|
|
|
6923
6967
|
this.stateChangeCallback({ ...this.state });
|
|
6924
6968
|
}
|
|
6925
6969
|
};
|
|
6970
|
+
var SimplifiedVGFConfigBuilder = class extends ConfigBuilder {
|
|
6971
|
+
constructor() {
|
|
6972
|
+
super(...arguments);
|
|
6973
|
+
this.vgfConfig = {};
|
|
6974
|
+
}
|
|
6975
|
+
/**
|
|
6976
|
+
* Set the VGF state-change callback, fired whenever the VGF state updates.
|
|
6977
|
+
*/
|
|
6978
|
+
onStateChange(callback) {
|
|
6979
|
+
this.vgfConfig.onStateChange = callback;
|
|
6980
|
+
return this;
|
|
6981
|
+
}
|
|
6982
|
+
/**
|
|
6983
|
+
* Set the initial VGF state to restore from a previous session.
|
|
6984
|
+
* The audioUtteranceId is extracted from it when valid.
|
|
6985
|
+
*/
|
|
6986
|
+
initialState(state) {
|
|
6987
|
+
this.vgfConfig.initialState = state;
|
|
6988
|
+
return this;
|
|
6989
|
+
}
|
|
6990
|
+
/**
|
|
6991
|
+
* Build the SimplifiedVGFClientConfig (base fields + VGF fields).
|
|
6992
|
+
*/
|
|
6993
|
+
build() {
|
|
6994
|
+
return { ...super.build(), ...this.vgfConfig };
|
|
6995
|
+
}
|
|
6996
|
+
};
|
|
6926
6997
|
function createSimplifiedVGFClient(config) {
|
|
6927
6998
|
return new SimplifiedVGFRecognitionClient(config);
|
|
6928
6999
|
}
|
|
7000
|
+
function createSimplifiedVGFClientWithBuilder(configure) {
|
|
7001
|
+
const builder = new SimplifiedVGFConfigBuilder();
|
|
7002
|
+
const config = configure(builder).build();
|
|
7003
|
+
return new SimplifiedVGFRecognitionClient(config);
|
|
7004
|
+
}
|
|
6929
7005
|
export {
|
|
6930
7006
|
AmazonNovaSonicModel,
|
|
7007
|
+
AssemblyAIModel,
|
|
6931
7008
|
AudioEncoding,
|
|
6932
7009
|
AwsTranscribeModel,
|
|
6933
7010
|
BedrockModel,
|
|
@@ -6948,6 +7025,7 @@ export {
|
|
|
6948
7025
|
GladiaModel,
|
|
6949
7026
|
GoogleModel,
|
|
6950
7027
|
Language,
|
|
7028
|
+
MicrophoneSourceType,
|
|
6951
7029
|
MistralVoxtralModel,
|
|
6952
7030
|
OpenAIModel,
|
|
6953
7031
|
OpenAIRealtimeModel,
|
|
@@ -6963,6 +7041,7 @@ export {
|
|
|
6963
7041
|
STAGES,
|
|
6964
7042
|
SampleRate,
|
|
6965
7043
|
SelfServeVllmModel,
|
|
7044
|
+
SimplifiedVGFConfigBuilder,
|
|
6966
7045
|
SimplifiedVGFRecognitionClient,
|
|
6967
7046
|
TimeoutError,
|
|
6968
7047
|
TranscriptionStatus,
|
|
@@ -6972,6 +7051,7 @@ export {
|
|
|
6972
7051
|
createDefaultASRConfig,
|
|
6973
7052
|
createInitialRecognitionState,
|
|
6974
7053
|
createSimplifiedVGFClient,
|
|
7054
|
+
createSimplifiedVGFClientWithBuilder,
|
|
6975
7055
|
getRecognitionConductorBase,
|
|
6976
7056
|
getRecognitionConductorHost,
|
|
6977
7057
|
getRecognitionConductorHttpBase,
|