@wildix/wilma-voice-client 0.0.26 → 0.0.28
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-cjs/WilmaVoice.js +4 -4
- package/dist-cjs/commands/{PreviewSpeechCommand.js → ListVoicesCommand.js} +6 -6
- package/dist-cjs/commands/{ListSpeechVoicesCommand.js → TextToSpeechCommand.js} +6 -6
- package/dist-cjs/commands/index.js +2 -2
- package/dist-cjs/models/enums.js +1 -9
- package/dist-cjs/schemas/schemas_0.js +42 -66
- package/dist-es/WilmaVoice.js +4 -4
- package/dist-es/commands/{PreviewSpeechCommand.js → ListVoicesCommand.js} +5 -5
- package/dist-es/commands/{ListSpeechVoicesCommand.js → TextToSpeechCommand.js} +5 -5
- package/dist-es/commands/index.js +2 -2
- package/dist-es/models/enums.js +0 -8
- package/dist-es/schemas/schemas_0.js +40 -64
- package/dist-types/WilmaVoice.d.ts +12 -12
- package/dist-types/WilmaVoiceClient.d.ts +4 -4
- package/dist-types/commands/DescribeVoiceProtocolCommand.d.ts +1 -1
- package/dist-types/commands/ListVoicesCommand.d.ts +87 -0
- package/dist-types/commands/TextToSpeechCommand.d.ts +103 -0
- package/dist-types/commands/index.d.ts +2 -2
- package/dist-types/models/enums.d.ts +0 -24
- package/dist-types/models/models_0.d.ts +81 -165
- package/dist-types/schemas/schemas_0.d.ts +7 -8
- package/package.json +1 -1
- package/dist-types/commands/ListSpeechVoicesCommand.d.ts +0 -101
- package/dist-types/commands/PreviewSpeechCommand.d.ts +0 -119
|
@@ -1,9 +1,10 @@
|
|
|
1
|
-
import type {
|
|
1
|
+
import type { DocumentType as __DocumentType } from "@smithy/types";
|
|
2
|
+
import type { VoiceAudioEncoding, VoiceCommandType, VoiceControlAction, VoiceEventType, VoiceInputMode, VoicePronunciationEncoding } from "./enums";
|
|
2
3
|
/**
|
|
3
4
|
* A pronunciation rule for one phrase. Belongs to a LANGUAGE, not to the agent: 'Wildix' is
|
|
4
|
-
* respelled differently for an Italian and a German voice
|
|
5
|
-
* would fight itself
|
|
6
|
-
*
|
|
5
|
+
* respelled differently for an Italian and a German voice, so one list shared ACROSS languages
|
|
6
|
+
* would fight itself. That is why the rules hang off the language and nowhere else — there is no
|
|
7
|
+
* company-wide dictionary above them to inherit from, or to disagree with.
|
|
7
8
|
* @public
|
|
8
9
|
*/
|
|
9
10
|
export interface VoicePronunciation {
|
|
@@ -23,57 +24,6 @@ export interface VoicePronunciation {
|
|
|
23
24
|
*/
|
|
24
25
|
encoding?: VoicePronunciationEncoding | undefined;
|
|
25
26
|
}
|
|
26
|
-
/**
|
|
27
|
-
* Synthesis options, used at two levels: as provider-level defaults on VoiceSpeechSettings and as
|
|
28
|
-
* per-language overrides on VoiceLanguage. Absent means inherit; there is no way to express
|
|
29
|
-
* 'explicitly the same as the parent', which is the intended simplification — the merged result is
|
|
30
|
-
* identical either way. Which members apply is a property of the provider AND the model (Chirp 3:
|
|
31
|
-
* HD documents only speakingRate; eleven_v3 carries no stability, similarity or speaker boost), so
|
|
32
|
-
* the console renders them from a catalog rather than unconditionally.
|
|
33
|
-
* @public
|
|
34
|
-
*/
|
|
35
|
-
export interface VoiceSynthesisOptions {
|
|
36
|
-
/**
|
|
37
|
-
* Provider model id, e.g. 'eleven_flash_v2_5'.
|
|
38
|
-
* @public
|
|
39
|
-
*/
|
|
40
|
-
model?: string | undefined;
|
|
41
|
-
/**
|
|
42
|
-
* Speaking rate. ElevenLabs 0.7-1.2 (default 1.0); Google 0.25-2.0 (default 1.0).
|
|
43
|
-
* @public
|
|
44
|
-
*/
|
|
45
|
-
speed?: number | undefined;
|
|
46
|
-
/**
|
|
47
|
-
* ElevenLabs stability, 0-1 (default 0.5). Our clamp, not a vendor-stated bound.
|
|
48
|
-
* @public
|
|
49
|
-
*/
|
|
50
|
-
stability?: number | undefined;
|
|
51
|
-
/**
|
|
52
|
-
* ElevenLabs similarity boost, 0-1 (default 0.75). Our clamp.
|
|
53
|
-
* @public
|
|
54
|
-
*/
|
|
55
|
-
similarity?: number | undefined;
|
|
56
|
-
/**
|
|
57
|
-
* ElevenLabs style exaggeration, 0-1 (default 0). Our clamp.
|
|
58
|
-
* @public
|
|
59
|
-
*/
|
|
60
|
-
style?: number | undefined;
|
|
61
|
-
/**
|
|
62
|
-
* ElevenLabs speaker boost (default true).
|
|
63
|
-
* @public
|
|
64
|
-
*/
|
|
65
|
-
speakerBoost?: boolean | undefined;
|
|
66
|
-
/**
|
|
67
|
-
* Google pitch in semitones, -20 to 20. Not offered for Chirp 3: HD voices.
|
|
68
|
-
* @public
|
|
69
|
-
*/
|
|
70
|
-
pitch?: number | undefined;
|
|
71
|
-
/**
|
|
72
|
-
* Google volume gain in dB, -96 to 16 (default 0); the vendor recommends staying under +10. Not offered for Chirp 3: HD voices.
|
|
73
|
-
* @public
|
|
74
|
-
*/
|
|
75
|
-
volumeGain?: number | undefined;
|
|
76
|
-
}
|
|
77
27
|
/**
|
|
78
28
|
* A chunk of raw caller audio, base64-encoded in the session's negotiated input format
|
|
79
29
|
* (inputAudio from the connect handshake). Used by web/app clients in AUDIO input mode; the
|
|
@@ -485,7 +435,7 @@ export interface SayCommandData {
|
|
|
485
435
|
*/
|
|
486
436
|
interruptible?: boolean | undefined;
|
|
487
437
|
/**
|
|
488
|
-
*
|
|
438
|
+
* Voice URI the PBX should synthesize with. Set only for a provider whose synthesis runs ON THE PBX (`amazon_polly://`); ElevenLabs and Google are synthesized in the cloud and streamed as PCM through `play` instead, and leave this absent.
|
|
489
439
|
* @public
|
|
490
440
|
*/
|
|
491
441
|
voice?: string | undefined;
|
|
@@ -544,23 +494,23 @@ export interface SessionEndedWebsocketEvent {
|
|
|
544
494
|
}
|
|
545
495
|
/**
|
|
546
496
|
* Recognition intent for the call, sent on `started` at protocolVersion >= 2. NOT recogniser
|
|
547
|
-
* parameters: the PBX translates it into its
|
|
497
|
+
* parameters: the PBX translates it into its provider's API. Against Amazon Transcribe with
|
|
548
498
|
* autoDetect true that is identify-multiple-languages plus language-options and
|
|
549
499
|
* preferred-language; with autoDetect false, a single-language stream on preferredLanguage.
|
|
550
500
|
*
|
|
551
|
-
* Both members the server can get wrong are
|
|
501
|
+
* Both members the server can get wrong are RESOLVED here rather than forwarded: `engine` is
|
|
552
502
|
* always concrete (an absent one would let the PBX apply a different default than the console
|
|
553
503
|
* built its language list for), and `autoDetect` is clamped to false whenever the language count
|
|
554
|
-
* is below the
|
|
555
|
-
* with one candidate makes Amazon reject the stream outright — a dead call).
|
|
504
|
+
* is below the recogniser's minimum or the recogniser cannot identify languages at all
|
|
505
|
+
* (forwarding true with one candidate makes Amazon reject the stream outright — a dead call).
|
|
556
506
|
* @public
|
|
557
507
|
*/
|
|
558
508
|
export interface VoiceSpeechPayload {
|
|
559
509
|
/**
|
|
560
|
-
* Concrete
|
|
510
|
+
* Concrete recogniser URI, resolved from the agent or from the platform default; never absent. See TranscriptionUri.
|
|
561
511
|
* @public
|
|
562
512
|
*/
|
|
563
|
-
|
|
513
|
+
engine: string;
|
|
564
514
|
/**
|
|
565
515
|
* Every configured language, as candidates for identification.
|
|
566
516
|
* @public
|
|
@@ -572,7 +522,7 @@ export interface VoiceSpeechPayload {
|
|
|
572
522
|
*/
|
|
573
523
|
preferredLanguage: string;
|
|
574
524
|
/**
|
|
575
|
-
* Whether to identify the spoken language per utterance. Already clamped by the server.
|
|
525
|
+
* Whether to identify the spoken language per utterance. Already clamped by the server — see the shape's documentation.
|
|
576
526
|
* @public
|
|
577
527
|
*/
|
|
578
528
|
autoDetect: boolean;
|
|
@@ -637,6 +587,10 @@ export interface SessionStartedEvent {
|
|
|
637
587
|
/**
|
|
638
588
|
* Music-on-hold class the PBX should play under the call as ambient background.
|
|
639
589
|
*
|
|
590
|
+
* A BARE NAME, not the configuration object `VoiceBackgroundSoundSettings`: this is the RESOLVED
|
|
591
|
+
* outcome. The server has already applied `enabled` and the class default, so the PBX gets a name
|
|
592
|
+
* to play or nothing at all and needs no resolution rule of its own.
|
|
593
|
+
*
|
|
640
594
|
* TOP LEVEL, NOT INSIDE `speech`, because it is not recognition intent and must not inherit
|
|
641
595
|
* `speech`'s conditions: a call with no configured languages carries no `speech` payload and can
|
|
642
596
|
* still have background audio.
|
|
@@ -846,12 +800,12 @@ export interface DescribeVoiceProtocolOutput {
|
|
|
846
800
|
/**
|
|
847
801
|
* @public
|
|
848
802
|
*/
|
|
849
|
-
export interface
|
|
803
|
+
export interface ListVoicesInput {
|
|
850
804
|
/**
|
|
851
|
-
*
|
|
805
|
+
* Provider whose catalog to read: 'elevenlabs', 'google', 'amazon_polly'. This is the scheme of the URIs answered on `voices`.
|
|
852
806
|
* @public
|
|
853
807
|
*/
|
|
854
|
-
provider:
|
|
808
|
+
provider: string;
|
|
855
809
|
/**
|
|
856
810
|
* Server-side filter for language-bound providers: only voices whose language SET CONTAINS this code. Ignored for ElevenLabs, whose voices are not language bound.
|
|
857
811
|
* @public
|
|
@@ -859,124 +813,51 @@ export interface ListSpeechVoicesInput {
|
|
|
859
813
|
language?: string | undefined;
|
|
860
814
|
search?: string | undefined;
|
|
861
815
|
/**
|
|
862
|
-
*
|
|
816
|
+
* A reference to the company's own provider credential, so the picker lists that account's
|
|
817
|
+
* voices including cloned ones. Absent means the platform key.
|
|
818
|
+
*
|
|
819
|
+
* `secrets:\{secretId\}` ONLY, resolved under the CALLER's company — an id that does not resolve
|
|
820
|
+
* there is rejected. There is no form of this member that carries a key, which is what keeps a
|
|
821
|
+
* provider credential out of a URL on this stack: every one of them lives in `wilma-secrets`,
|
|
822
|
+
* and a reference in a query string is not a secret.
|
|
823
|
+
*
|
|
824
|
+
* A stored voice URI says the same thing the same way: its `apiKey` parameter is a reference,
|
|
825
|
+
* and `validateSpeech` refuses a literal key in one.
|
|
863
826
|
* @public
|
|
864
827
|
*/
|
|
865
828
|
credential?: string | undefined;
|
|
866
829
|
}
|
|
867
830
|
/**
|
|
868
|
-
* One voice
|
|
869
|
-
*
|
|
870
|
-
|
|
871
|
-
|
|
872
|
-
|
|
873
|
-
|
|
874
|
-
* @public
|
|
875
|
-
*/
|
|
876
|
-
id: string;
|
|
877
|
-
name: string;
|
|
878
|
-
/**
|
|
879
|
-
* ElevenLabs verified_languages, Google languageCodes, or Polly LanguageCode plus AdditionalLanguageCodes. For an unbound provider this is metadata, not a constraint.
|
|
880
|
-
* @public
|
|
881
|
-
*/
|
|
882
|
-
languages?: string[] | undefined;
|
|
883
|
-
gender?: string | undefined;
|
|
884
|
-
accent?: string | undefined;
|
|
885
|
-
/**
|
|
886
|
-
* ElevenLabs category, Google tier, or the Polly engines the voice supports.
|
|
887
|
-
* @public
|
|
888
|
-
*/
|
|
889
|
-
category?: string | undefined;
|
|
890
|
-
/**
|
|
891
|
-
* Playable sample published by the provider. ElevenLabs only; the others publish none.
|
|
892
|
-
* @public
|
|
893
|
-
*/
|
|
894
|
-
previewUrl?: string | undefined;
|
|
895
|
-
/**
|
|
896
|
-
* ElevenLabs high_quality_base_model_ids — which models support this voice.
|
|
897
|
-
* @public
|
|
898
|
-
*/
|
|
899
|
-
models?: string[] | undefined;
|
|
900
|
-
}
|
|
901
|
-
/**
|
|
902
|
-
* @public
|
|
903
|
-
*/
|
|
904
|
-
export interface ListSpeechVoicesOutput {
|
|
905
|
-
voices: SpeechVoice[];
|
|
906
|
-
/**
|
|
907
|
-
* The provider's FULL language set, independent of the `language` filter — this is what a form intersects with the transcription vendor's set to populate its language select. Empty when voices are not language bound, which means 'imposes no restriction'.
|
|
908
|
-
* @public
|
|
909
|
-
*/
|
|
910
|
-
languages?: string[] | undefined;
|
|
911
|
-
/**
|
|
912
|
-
* When this catalog was last read from the provider. `IsoTimestamp` rather than a bare `Timestamp`: restJson1 serialises the latter as epoch-seconds, which would make this the only Wilma operation answering with a numeric timestamp.
|
|
913
|
-
* @public
|
|
914
|
-
*/
|
|
915
|
-
refreshedAt: string;
|
|
916
|
-
/**
|
|
917
|
-
* True when the provider failed and an expired cache entry was served.
|
|
918
|
-
* @public
|
|
919
|
-
*/
|
|
920
|
-
stale?: boolean | undefined;
|
|
921
|
-
}
|
|
922
|
-
/**
|
|
831
|
+
* One voice a provider offers: the URI to store it by, a name to show, and the provider's own
|
|
832
|
+
* metadata.
|
|
833
|
+
*
|
|
834
|
+
* `uri` IS THE RESULT. It is exactly what goes into `VoiceLanguage.voice`, so a form stores what
|
|
835
|
+
* it was given instead of assembling a string out of a provider name and an id — the one place a
|
|
836
|
+
* client could get the convention wrong.
|
|
923
837
|
* @public
|
|
924
838
|
*/
|
|
925
|
-
export interface
|
|
926
|
-
/**
|
|
927
|
-
* Text-to-speech provider used to synthesize the agent's replies.
|
|
928
|
-
* @public
|
|
929
|
-
*/
|
|
930
|
-
provider: VoiceSpeechProvider;
|
|
931
|
-
voice: string;
|
|
932
|
-
/**
|
|
933
|
-
* BCP-47 language tag, optionally with a region subtag: 'en', 'en-US', 'pt-BR'. Deliberately a
|
|
934
|
-
* constrained string rather than an enum: the offerable set is a property of the selected
|
|
935
|
-
* transcription vendor and changes with it, so an enumeration here would cap the set at whatever
|
|
936
|
-
* was known when the model was released.
|
|
937
|
-
* @public
|
|
938
|
-
*/
|
|
939
|
-
language: string;
|
|
839
|
+
export interface Voice {
|
|
940
840
|
/**
|
|
941
|
-
*
|
|
841
|
+
* Voice URI, ready to store as `VoiceLanguage.voice` or send to TextToSpeech. Carries no parameters — the caller appends the synthesis parameters and, where applicable, the `apiKey=secrets:\{secretId\}` reference it wants.
|
|
942
842
|
* @public
|
|
943
843
|
*/
|
|
944
|
-
|
|
844
|
+
uri: string;
|
|
945
845
|
/**
|
|
946
|
-
*
|
|
846
|
+
* Display name as the provider publishes it, e.g. 'Rachel'. A member rather than part of `options` because it is the one thing every provider has and every picker shows.
|
|
947
847
|
* @public
|
|
948
848
|
*/
|
|
949
|
-
|
|
950
|
-
/**
|
|
951
|
-
* The selected language's overrides of the above.
|
|
952
|
-
* @public
|
|
953
|
-
*/
|
|
954
|
-
languageSynthesis?: VoiceSynthesisOptions | undefined;
|
|
955
|
-
/**
|
|
956
|
-
* The agent's OWN rules for this language, unmerged. The company dictionary for `language` is applied server-side, for the reason given on `synthesis`.
|
|
957
|
-
* @public
|
|
958
|
-
*/
|
|
959
|
-
pronunciations?: VoicePronunciation[] | undefined;
|
|
960
|
-
/**
|
|
961
|
-
* Company-dictionary phrases this language declines. Applied SERVER-SIDE with the rules above, by the same helper the call path uses.
|
|
962
|
-
* @public
|
|
963
|
-
*/
|
|
964
|
-
suppressions?: string[] | undefined;
|
|
849
|
+
name: string;
|
|
965
850
|
/**
|
|
966
|
-
*
|
|
851
|
+
* The provider's own metadata for this voice, for rendering and filtering only. See VoiceOptions.
|
|
967
852
|
* @public
|
|
968
853
|
*/
|
|
969
|
-
|
|
854
|
+
options?: __DocumentType | undefined;
|
|
970
855
|
}
|
|
971
856
|
/**
|
|
972
857
|
* @public
|
|
973
858
|
*/
|
|
974
|
-
export interface
|
|
975
|
-
|
|
976
|
-
* audio/mpeg. The call path streams raw PCM to the PBX; the preview returns MP3 so a browser <audio> element plays it with no client-side decoding.
|
|
977
|
-
* @public
|
|
978
|
-
*/
|
|
979
|
-
audio: Uint8Array;
|
|
859
|
+
export interface ListVoicesOutput {
|
|
860
|
+
voices: Voice[];
|
|
980
861
|
}
|
|
981
862
|
/**
|
|
982
863
|
* Terminate the session and hang up the call.
|
|
@@ -1039,7 +920,7 @@ export interface SendSayInput {
|
|
|
1039
920
|
*/
|
|
1040
921
|
interruptible?: boolean | undefined;
|
|
1041
922
|
/**
|
|
1042
|
-
*
|
|
923
|
+
* Voice URI the PBX should synthesize with. Set only for a provider whose synthesis runs ON THE PBX (`amazon_polly://`); ElevenLabs and Google are synthesized in the cloud and streamed as PCM through `play` instead, and leave this absent.
|
|
1043
924
|
* @public
|
|
1044
925
|
*/
|
|
1045
926
|
voice?: string | undefined;
|
|
@@ -1092,3 +973,38 @@ export interface SendTransferInput {
|
|
|
1092
973
|
*/
|
|
1093
974
|
export interface SendTransferOutput {
|
|
1094
975
|
}
|
|
976
|
+
/**
|
|
977
|
+
* @public
|
|
978
|
+
*/
|
|
979
|
+
export interface TextToSpeechInput {
|
|
980
|
+
/**
|
|
981
|
+
* Voice to speak with, as the URI a call would use — provider, voice id, synthesis parameters and, for a company's own account, `apiKey=secrets:\{secretId\}`. Passing the call's own value is what keeps an audition from diverging from the call it is auditioning.
|
|
982
|
+
* @public
|
|
983
|
+
*/
|
|
984
|
+
voice: string;
|
|
985
|
+
/**
|
|
986
|
+
* Language the phrase is in. Reaches the provider as the synthesis language, and is the language `pronunciations` was written for.
|
|
987
|
+
* @public
|
|
988
|
+
*/
|
|
989
|
+
language: string;
|
|
990
|
+
/**
|
|
991
|
+
* Text to speak, BEFORE pronunciation rules are applied.
|
|
992
|
+
* @public
|
|
993
|
+
*/
|
|
994
|
+
text: string;
|
|
995
|
+
/**
|
|
996
|
+
* Rules to apply to `text`, applied SERVER-SIDE by the same helper the call path uses — never pre-applied by the caller, or the form reimplements a rule the server owns and can silently diverge from the call. Pass the language's own list to hear exactly what a call would say.
|
|
997
|
+
* @public
|
|
998
|
+
*/
|
|
999
|
+
pronunciations?: VoicePronunciation[] | undefined;
|
|
1000
|
+
}
|
|
1001
|
+
/**
|
|
1002
|
+
* @public
|
|
1003
|
+
*/
|
|
1004
|
+
export interface TextToSpeechOutput {
|
|
1005
|
+
/**
|
|
1006
|
+
* audio/mpeg. The call path streams raw PCM to the PBX; this returns MP3 so a browser <audio> element plays it with no client-side decoding.
|
|
1007
|
+
* @public
|
|
1008
|
+
*/
|
|
1009
|
+
audio: Uint8Array;
|
|
1010
|
+
}
|
|
@@ -11,7 +11,6 @@ export declare var ValidationException$: StaticErrorSchema;
|
|
|
11
11
|
*/
|
|
12
12
|
export declare const errorTypeRegistries: TypeRegistry[];
|
|
13
13
|
export declare var VoicePronunciation$: StaticStructureSchema;
|
|
14
|
-
export declare var VoiceSynthesisOptions$: StaticStructureSchema;
|
|
15
14
|
export declare var AudioEvent$: StaticStructureSchema;
|
|
16
15
|
export declare var AudioWebsocketEvent$: StaticStructureSchema;
|
|
17
16
|
export declare var CompleteEvent$: StaticStructureSchema;
|
|
@@ -25,14 +24,12 @@ export declare var HangupCommandData$: StaticStructureSchema;
|
|
|
25
24
|
export declare var HangupWebsocketCommand$: StaticStructureSchema;
|
|
26
25
|
export declare var InterruptionEvent$: StaticStructureSchema;
|
|
27
26
|
export declare var InterruptionWebsocketEvent$: StaticStructureSchema;
|
|
28
|
-
export declare var
|
|
29
|
-
export declare var
|
|
27
|
+
export declare var ListVoicesInput$: StaticStructureSchema;
|
|
28
|
+
export declare var ListVoicesOutput$: StaticStructureSchema;
|
|
30
29
|
export declare var PlaybackEvent$: StaticStructureSchema;
|
|
31
30
|
export declare var PlaybackWebsocketEvent$: StaticStructureSchema;
|
|
32
31
|
export declare var PlayCommandData$: StaticStructureSchema;
|
|
33
32
|
export declare var PlayWebsocketCommand$: StaticStructureSchema;
|
|
34
|
-
export declare var PreviewSpeechInput$: StaticStructureSchema;
|
|
35
|
-
export declare var PreviewSpeechOutput$: StaticStructureSchema;
|
|
36
33
|
export declare var ReplyEvent$: StaticStructureSchema;
|
|
37
34
|
export declare var ReplyWebsocketEvent$: StaticStructureSchema;
|
|
38
35
|
export declare var SayCommandAudioOptions$: StaticStructureSchema;
|
|
@@ -52,22 +49,24 @@ export declare var SessionEndedEvent$: StaticStructureSchema;
|
|
|
52
49
|
export declare var SessionEndedWebsocketEvent$: StaticStructureSchema;
|
|
53
50
|
export declare var SessionStartedEvent$: StaticStructureSchema;
|
|
54
51
|
export declare var SessionStartedWebsocketEvent$: StaticStructureSchema;
|
|
55
|
-
export declare var SpeechVoice$: StaticStructureSchema;
|
|
56
52
|
export declare var StopCommandData$: StaticStructureSchema;
|
|
57
53
|
export declare var StopWebsocketCommand$: StaticStructureSchema;
|
|
54
|
+
export declare var TextToSpeechInput$: StaticStructureSchema;
|
|
55
|
+
export declare var TextToSpeechOutput$: StaticStructureSchema;
|
|
58
56
|
export declare var TranscriptionEvent$: StaticStructureSchema;
|
|
59
57
|
export declare var TranscriptionWebsocketEvent$: StaticStructureSchema;
|
|
60
58
|
export declare var TransferCommandData$: StaticStructureSchema;
|
|
61
59
|
export declare var TransferWebsocketCommand$: StaticStructureSchema;
|
|
60
|
+
export declare var Voice$: StaticStructureSchema;
|
|
62
61
|
export declare var VoiceAudioFormat$: StaticStructureSchema;
|
|
63
62
|
export declare var VoiceConnectParams$: StaticStructureSchema;
|
|
64
63
|
export declare var VoiceErrorEvent$: StaticStructureSchema;
|
|
65
64
|
export declare var VoiceSpeechPayload$: StaticStructureSchema;
|
|
66
65
|
export declare var DescribeVoiceProtocol$: StaticOperationSchema;
|
|
67
|
-
export declare var
|
|
68
|
-
export declare var PreviewSpeech$: StaticOperationSchema;
|
|
66
|
+
export declare var ListVoices$: StaticOperationSchema;
|
|
69
67
|
export declare var SendHangup$: StaticOperationSchema;
|
|
70
68
|
export declare var SendPlay$: StaticOperationSchema;
|
|
71
69
|
export declare var SendSay$: StaticOperationSchema;
|
|
72
70
|
export declare var SendStop$: StaticOperationSchema;
|
|
73
71
|
export declare var SendTransfer$: StaticOperationSchema;
|
|
72
|
+
export declare var TextToSpeech$: StaticOperationSchema;
|
package/package.json
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@wildix/wilma-voice-client",
|
|
3
3
|
"description": "@wildix/wilma-voice-client client",
|
|
4
|
-
"version": "0.0.
|
|
4
|
+
"version": "0.0.28",
|
|
5
5
|
"scripts": {
|
|
6
6
|
"build": "concurrently 'yarn:build:cjs' 'yarn:build:es' 'yarn:build:types'",
|
|
7
7
|
"build:cjs": "tsc -p tsconfig.cjs.json",
|
|
@@ -1,101 +0,0 @@
|
|
|
1
|
-
import { Command as $Command } from "@smithy/core/client";
|
|
2
|
-
import type { MetadataBearer as __MetadataBearer } from "@smithy/types";
|
|
3
|
-
import type { ListSpeechVoicesInput, ListSpeechVoicesOutput } from "../models/models_0";
|
|
4
|
-
import type { ServiceInputTypes, ServiceOutputTypes, WilmaVoiceClientResolvedConfig } from "../WilmaVoiceClient";
|
|
5
|
-
/**
|
|
6
|
-
* @public
|
|
7
|
-
*/
|
|
8
|
-
export type { __MetadataBearer };
|
|
9
|
-
export { $Command };
|
|
10
|
-
/**
|
|
11
|
-
* @public
|
|
12
|
-
*
|
|
13
|
-
* The input for {@link ListSpeechVoicesCommand}.
|
|
14
|
-
*/
|
|
15
|
-
export interface ListSpeechVoicesCommandInput extends ListSpeechVoicesInput {
|
|
16
|
-
}
|
|
17
|
-
/**
|
|
18
|
-
* @public
|
|
19
|
-
*
|
|
20
|
-
* The output of {@link ListSpeechVoicesCommand}.
|
|
21
|
-
*/
|
|
22
|
-
export interface ListSpeechVoicesCommandOutput extends ListSpeechVoicesOutput, __MetadataBearer {
|
|
23
|
-
}
|
|
24
|
-
declare const ListSpeechVoicesCommand_base: {
|
|
25
|
-
new (input: ListSpeechVoicesCommandInput): import("@smithy/core/client").CommandImpl<ListSpeechVoicesCommandInput, ListSpeechVoicesCommandOutput, WilmaVoiceClientResolvedConfig, ServiceInputTypes, ServiceOutputTypes>;
|
|
26
|
-
new (input: ListSpeechVoicesCommandInput): import("@smithy/core/client").CommandImpl<ListSpeechVoicesCommandInput, ListSpeechVoicesCommandOutput, WilmaVoiceClientResolvedConfig, ServiceInputTypes, ServiceOutputTypes>;
|
|
27
|
-
getEndpointParameterInstructions(): import("@smithy/types").EndpointParameterInstructions;
|
|
28
|
-
};
|
|
29
|
-
/**
|
|
30
|
-
* Lists the voices a synthesis provider offers, from the provider's own API, cached per company, provider, credential and language.
|
|
31
|
-
* @example
|
|
32
|
-
* Use a bare-bones client and the command you need to make an API call.
|
|
33
|
-
* ```javascript
|
|
34
|
-
* import { WilmaVoiceClient, ListSpeechVoicesCommand } from "@wildix/wilma-voice-client"; // ES Modules import
|
|
35
|
-
* // const { WilmaVoiceClient, ListSpeechVoicesCommand } = require("@wildix/wilma-voice-client"); // CommonJS import
|
|
36
|
-
* // import type { WilmaVoiceClientConfig } from "@wildix/wilma-voice-client";
|
|
37
|
-
* const config = {}; // type is WilmaVoiceClientConfig
|
|
38
|
-
* const client = new WilmaVoiceClient(config);
|
|
39
|
-
* const input = { // ListSpeechVoicesInput
|
|
40
|
-
* provider: "elevenlabs" || "google" || "amazon_polly", // required
|
|
41
|
-
* language: "STRING_VALUE",
|
|
42
|
-
* search: "STRING_VALUE",
|
|
43
|
-
* credential: "STRING_VALUE",
|
|
44
|
-
* };
|
|
45
|
-
* const command = new ListSpeechVoicesCommand(input);
|
|
46
|
-
* const response = await client.send(command);
|
|
47
|
-
* // { // ListSpeechVoicesOutput
|
|
48
|
-
* // voices: [ // SpeechVoiceList // required
|
|
49
|
-
* // { // SpeechVoice
|
|
50
|
-
* // id: "STRING_VALUE", // required
|
|
51
|
-
* // name: "STRING_VALUE", // required
|
|
52
|
-
* // languages: [ // ProviderLanguageTagList
|
|
53
|
-
* // "STRING_VALUE",
|
|
54
|
-
* // ],
|
|
55
|
-
* // gender: "STRING_VALUE",
|
|
56
|
-
* // accent: "STRING_VALUE",
|
|
57
|
-
* // category: "STRING_VALUE",
|
|
58
|
-
* // previewUrl: "STRING_VALUE",
|
|
59
|
-
* // models: [ // SpeechVoiceModelList
|
|
60
|
-
* // "STRING_VALUE",
|
|
61
|
-
* // ],
|
|
62
|
-
* // },
|
|
63
|
-
* // ],
|
|
64
|
-
* // languages: [
|
|
65
|
-
* // "STRING_VALUE",
|
|
66
|
-
* // ],
|
|
67
|
-
* // refreshedAt: "STRING_VALUE", // required
|
|
68
|
-
* // stale: true || false,
|
|
69
|
-
* // };
|
|
70
|
-
*
|
|
71
|
-
* ```
|
|
72
|
-
*
|
|
73
|
-
* @param ListSpeechVoicesCommandInput - {@link ListSpeechVoicesCommandInput}
|
|
74
|
-
* @returns {@link ListSpeechVoicesCommandOutput}
|
|
75
|
-
* @see {@link ListSpeechVoicesCommandInput} for command's `input` shape.
|
|
76
|
-
* @see {@link ListSpeechVoicesCommandOutput} for command's `response` shape.
|
|
77
|
-
* @see {@link WilmaVoiceClientResolvedConfig | config} for WilmaVoiceClient's `config` shape.
|
|
78
|
-
*
|
|
79
|
-
* @throws {@link ValidationException} (client fault)
|
|
80
|
-
*
|
|
81
|
-
* @throws {@link ForbiddenException} (client fault)
|
|
82
|
-
*
|
|
83
|
-
* @throws {@link WilmaVoiceServiceException}
|
|
84
|
-
* <p>Base exception class for all service exceptions from WilmaVoice service.</p>
|
|
85
|
-
*
|
|
86
|
-
*
|
|
87
|
-
* @public
|
|
88
|
-
*/
|
|
89
|
-
export declare class ListSpeechVoicesCommand extends ListSpeechVoicesCommand_base {
|
|
90
|
-
/** @internal type navigation helper, not in runtime. */
|
|
91
|
-
protected static __types: {
|
|
92
|
-
api: {
|
|
93
|
-
input: ListSpeechVoicesInput;
|
|
94
|
-
output: ListSpeechVoicesOutput;
|
|
95
|
-
};
|
|
96
|
-
sdk: {
|
|
97
|
-
input: ListSpeechVoicesCommandInput;
|
|
98
|
-
output: ListSpeechVoicesCommandOutput;
|
|
99
|
-
};
|
|
100
|
-
};
|
|
101
|
-
}
|
|
@@ -1,119 +0,0 @@
|
|
|
1
|
-
import { Command as $Command } from "@smithy/core/client";
|
|
2
|
-
import { Uint8ArrayBlobAdapter } from "@smithy/core/serde";
|
|
3
|
-
import type { MetadataBearer as __MetadataBearer } from "@smithy/types";
|
|
4
|
-
import type { PreviewSpeechInput, PreviewSpeechOutput } from "../models/models_0";
|
|
5
|
-
import type { ServiceInputTypes, ServiceOutputTypes, WilmaVoiceClientResolvedConfig } from "../WilmaVoiceClient";
|
|
6
|
-
/**
|
|
7
|
-
* @public
|
|
8
|
-
*/
|
|
9
|
-
export type { __MetadataBearer };
|
|
10
|
-
export { $Command };
|
|
11
|
-
/**
|
|
12
|
-
* @public
|
|
13
|
-
*
|
|
14
|
-
* The input for {@link PreviewSpeechCommand}.
|
|
15
|
-
*/
|
|
16
|
-
export interface PreviewSpeechCommandInput extends PreviewSpeechInput {
|
|
17
|
-
}
|
|
18
|
-
/**
|
|
19
|
-
* @public
|
|
20
|
-
*/
|
|
21
|
-
export type PreviewSpeechCommandOutputType = Omit<PreviewSpeechOutput, "audio"> & {
|
|
22
|
-
audio: Uint8ArrayBlobAdapter;
|
|
23
|
-
};
|
|
24
|
-
/**
|
|
25
|
-
* @public
|
|
26
|
-
*
|
|
27
|
-
* The output of {@link PreviewSpeechCommand}.
|
|
28
|
-
*/
|
|
29
|
-
export interface PreviewSpeechCommandOutput extends PreviewSpeechCommandOutputType, __MetadataBearer {
|
|
30
|
-
}
|
|
31
|
-
declare const PreviewSpeechCommand_base: {
|
|
32
|
-
new (input: PreviewSpeechCommandInput): import("@smithy/core/client").CommandImpl<PreviewSpeechCommandInput, PreviewSpeechCommandOutput, WilmaVoiceClientResolvedConfig, ServiceInputTypes, ServiceOutputTypes>;
|
|
33
|
-
new (input: PreviewSpeechCommandInput): import("@smithy/core/client").CommandImpl<PreviewSpeechCommandInput, PreviewSpeechCommandOutput, WilmaVoiceClientResolvedConfig, ServiceInputTypes, ServiceOutputTypes>;
|
|
34
|
-
getEndpointParameterInstructions(): import("@smithy/types").EndpointParameterInstructions;
|
|
35
|
-
};
|
|
36
|
-
/**
|
|
37
|
-
* Synthesizes one phrase with the supplied settings and returns MP3. Everything about the AGENT travels in the request because the settings being previewed are normally unsaved form state; the operation reads no stored agent. It does read the company's pronunciation dictionary for `language` — that is stored configuration rather than draft state, and it must be exactly what a call would use.
|
|
38
|
-
* @example
|
|
39
|
-
* Use a bare-bones client and the command you need to make an API call.
|
|
40
|
-
* ```javascript
|
|
41
|
-
* import { WilmaVoiceClient, PreviewSpeechCommand } from "@wildix/wilma-voice-client"; // ES Modules import
|
|
42
|
-
* // const { WilmaVoiceClient, PreviewSpeechCommand } = require("@wildix/wilma-voice-client"); // CommonJS import
|
|
43
|
-
* // import type { WilmaVoiceClientConfig } from "@wildix/wilma-voice-client";
|
|
44
|
-
* const config = {}; // type is WilmaVoiceClientConfig
|
|
45
|
-
* const client = new WilmaVoiceClient(config);
|
|
46
|
-
* const input = { // PreviewSpeechInput
|
|
47
|
-
* provider: "elevenlabs" || "google" || "amazon_polly", // required
|
|
48
|
-
* voice: "STRING_VALUE", // required
|
|
49
|
-
* language: "STRING_VALUE", // required
|
|
50
|
-
* text: "STRING_VALUE", // required
|
|
51
|
-
* synthesis: { // VoiceSynthesisOptions
|
|
52
|
-
* model: "STRING_VALUE",
|
|
53
|
-
* speed: Number("float"),
|
|
54
|
-
* stability: Number("float"),
|
|
55
|
-
* similarity: Number("float"),
|
|
56
|
-
* style: Number("float"),
|
|
57
|
-
* speakerBoost: true || false,
|
|
58
|
-
* pitch: Number("float"),
|
|
59
|
-
* volumeGain: Number("float"),
|
|
60
|
-
* },
|
|
61
|
-
* languageSynthesis: {
|
|
62
|
-
* model: "STRING_VALUE",
|
|
63
|
-
* speed: Number("float"),
|
|
64
|
-
* stability: Number("float"),
|
|
65
|
-
* similarity: Number("float"),
|
|
66
|
-
* style: Number("float"),
|
|
67
|
-
* speakerBoost: true || false,
|
|
68
|
-
* pitch: Number("float"),
|
|
69
|
-
* volumeGain: Number("float"),
|
|
70
|
-
* },
|
|
71
|
-
* pronunciations: [ // VoicePronunciationList
|
|
72
|
-
* { // VoicePronunciation
|
|
73
|
-
* phrase: "STRING_VALUE", // required
|
|
74
|
-
* as: "STRING_VALUE", // required
|
|
75
|
-
* encoding: "alias" || "ipa",
|
|
76
|
-
* },
|
|
77
|
-
* ],
|
|
78
|
-
* suppressions: [ // VoicePronunciationSuppressionList
|
|
79
|
-
* "STRING_VALUE",
|
|
80
|
-
* ],
|
|
81
|
-
* credential: "STRING_VALUE",
|
|
82
|
-
* };
|
|
83
|
-
* const command = new PreviewSpeechCommand(input);
|
|
84
|
-
* const response = await client.send(command);
|
|
85
|
-
* // { // PreviewSpeechOutput
|
|
86
|
-
* // audio: new Uint8Array(), // required
|
|
87
|
-
* // };
|
|
88
|
-
*
|
|
89
|
-
* ```
|
|
90
|
-
*
|
|
91
|
-
* @param PreviewSpeechCommandInput - {@link PreviewSpeechCommandInput}
|
|
92
|
-
* @returns {@link PreviewSpeechCommandOutput}
|
|
93
|
-
* @see {@link PreviewSpeechCommandInput} for command's `input` shape.
|
|
94
|
-
* @see {@link PreviewSpeechCommandOutput} for command's `response` shape.
|
|
95
|
-
* @see {@link WilmaVoiceClientResolvedConfig | config} for WilmaVoiceClient's `config` shape.
|
|
96
|
-
*
|
|
97
|
-
* @throws {@link ValidationException} (client fault)
|
|
98
|
-
*
|
|
99
|
-
* @throws {@link ForbiddenException} (client fault)
|
|
100
|
-
*
|
|
101
|
-
* @throws {@link WilmaVoiceServiceException}
|
|
102
|
-
* <p>Base exception class for all service exceptions from WilmaVoice service.</p>
|
|
103
|
-
*
|
|
104
|
-
*
|
|
105
|
-
* @public
|
|
106
|
-
*/
|
|
107
|
-
export declare class PreviewSpeechCommand extends PreviewSpeechCommand_base {
|
|
108
|
-
/** @internal type navigation helper, not in runtime. */
|
|
109
|
-
protected static __types: {
|
|
110
|
-
api: {
|
|
111
|
-
input: PreviewSpeechInput;
|
|
112
|
-
output: PreviewSpeechOutput;
|
|
113
|
-
};
|
|
114
|
-
sdk: {
|
|
115
|
-
input: PreviewSpeechCommandInput;
|
|
116
|
-
output: PreviewSpeechCommandOutput;
|
|
117
|
-
};
|
|
118
|
-
};
|
|
119
|
-
}
|