@wildix/wilma-voice-client 0.0.40 → 0.0.41

package/dist-types/models/models_0.d.ts

@@ -1,5 +1,5 @@
 import type { DocumentType as __DocumentType, StreamingBlobTypes } from "@smithy/types";
-import type { VoiceAudioEncoding, VoiceCallEndReason, VoiceCallToolPhase, VoiceCommandType, VoiceControlAction, VoiceEventType, VoiceInputMode, VoicePronunciationEncoding } from "./enums";
+import type { MediaAudioBeginFrameType, MediaAudioEncoding, MediaAudioEndFrameType, MediaDtmfFrameType, MediaDuckFrameType, MediaErrorFrameType, MediaFinishedFrameType, MediaFinishFrameType, MediaFinishReason, MediaHangupFrameType, MediaLanguageFrameType, MediaParty, MediaPlaybackEvent, MediaPlaybackFrameType, MediaPlaybackStopCause, MediaRestoreFrameType, MediaRunPurpose, MediaStartedFrameType, MediaStartFrameType, MediaStopFrameType, MediaTranscriptionFrameType, MediaTransferFrameType, MediaVadFrameType, VoiceAudioEncoding, VoiceCallEndReason, VoiceCallToolPhase, VoiceCommandType, VoiceControlAction, VoiceEventType, VoiceInputMode, VoicePronunciationEncoding } from "./enums";
 /**
  * A pronunciation rule for one phrase. Belongs to a LANGUAGE, not to the agent: 'Wildix' is
  *     respelled differently for an Italian and a German voice, so one list shared ACROSS languages
@@ -764,6 +764,586 @@ export interface InterruptionWebsocketEvent {
      */
     event: InterruptionEvent;
 }
+/**
+ * Opens an agent audio stream. `MediaCloudAudio` messages of the stream follow until `audio.end`. Streams do not interleave.
+ * @public
+ */
+export interface MediaAudioBeginFrame {
+    type: MediaAudioBeginFrameType;
+    /**
+     * Id of a reply, a stream, an utterance or a DTMF input, unique in the run.
+     * @public
+     */
+    replyId: string;
+    /**
+     * Id of a reply, a stream, an utterance or a DTMF input, unique in the run.
+     * @public
+     */
+    streamId: string;
+    /**
+     * Sample rate of the stream's `pcm_s16le` mono audio in Hz: 8000, 16000, 22050 or 24000.
+     * @public
+     */
+    sampleRate: number;
+}
+/**
+ * Closes an agent audio stream.
+ * @public
+ */
+export interface MediaAudioEndFrame {
+    type: MediaAudioEndFrameType;
+    /**
+     * Id of a reply, a stream, an utterance or a DTMF input, unique in the run.
+     * @public
+     */
+    streamId: string;
+}
+/**
+ * Lower the volume of a playing reply.
+ * @public
+ */
+export interface MediaDuckFrame {
+    type: MediaDuckFrameType;
+    /**
+     * Id of a reply, a stream, an utterance or a DTMF input, unique in the run.
+     * @public
+     */
+    replyId: string;
+    /**
+     * Target gain in dB.
+     * @public
+     */
+    volumeDb: number;
+    /**
+     * Ramp duration in ms.
+     * @public
+     */
+    rampMs: number;
+}
+/**
+ * An error on the run. When `retryable` is false, the socket closes after it.
+ * @public
+ */
+export interface MediaErrorFrame {
+    type: MediaErrorFrameType;
+    /**
+     * Stable error code, for example `START_TIMEOUT`, `PROTOCOL_ERROR` or `STT_FAILED`.
+     * @public
+     */
+    code: string;
+    retryable: boolean;
+    message: string;
+}
+/**
+ * Answer to `finish`: the cloud processed all audio. The socket closes with 1000 after it.
+ * @public
+ */
+export interface MediaFinishedFrame {
+    type: MediaFinishedFrameType;
+}
+/**
+ * Hang up the call after the queued agent audio plays. Terminal: the PBX answers with `finish`.
+ * @public
+ */
+export interface MediaHangupFrame {
+    type: MediaHangupFrameType;
+}
+/**
+ * The call continues in another language.
+ * @public
+ */
+export interface MediaLanguageFrame {
+    type: MediaLanguageFrameType;
+    /**
+     * Language that the call continues in.
+     * @public
+     */
+    language: string;
+}
+/**
+ * Return a reply to full volume.
+ * @public
+ */
+export interface MediaRestoreFrame {
+    type: MediaRestoreFrameType;
+    /**
+     * Id of a reply, a stream, an utterance or a DTMF input, unique in the run.
+     * @public
+     */
+    replyId: string;
+    /**
+     * Ramp duration in ms.
+     * @public
+     */
+    rampMs: number;
+}
+/**
+ * The agent that answers the run.
+ * @public
+ */
+export interface MediaStartedAgent {
+    name: string;
+}
+/**
+ * Caller voice detection of a run, with every value resolved.
+ * @public
+ */
+export interface MediaVoiceDetection {
+    /**
+     * Silence in ms that ends the caller speech.
+     * @public
+     */
+    silenceMs: number;
+    /**
+     * Level in dBFS below which caller audio is silence.
+     * @public
+     */
+    silenceLevelDb: number;
+    /**
+     * Mean frame level in dBFS from which caller audio is speech.
+     * @public
+     */
+    speechLevelDb: number;
+}
+/**
+ * The cloud accepted `start`. Sent once per connection.
+ * @public
+ */
+export interface MediaStartedFrame {
+    type: MediaStartedFrameType;
+    /**
+     * Id of the cloud stream of this run. Cloud logs carry it; use it to match PBX and cloud logs.
+     * @public
+     */
+    streamId: string;
+    /**
+     * Present when the run purpose is `agent`.
+     * @public
+     */
+    agent?: MediaStartedAgent | undefined;
+    /**
+     * Group of the agent. Absent when the agent has none.
+     * @public
+     */
+    groupId?: string | undefined;
+    /**
+     * Music-on-hold class that the PBX plays after each caller turn until agent audio starts. Absent: none.
+     * @public
+     */
+    backgroundSound?: string | undefined;
+    /**
+     * Caller voice detection that the PBX uses for the `vad` frames of this run. Absent: the PBX keeps its defaults.
+     * @public
+     */
+    voiceDetection?: MediaVoiceDetection | undefined;
+}
+/**
+ * Stop every agent audio stream of the run: the one that plays and the ones that wait.
+ * @public
+ */
+export interface MediaStopFrame {
+    type: MediaStopFrameType;
+}
+/**
+ * Words of one party. Partials and the final of one utterance share `id`. For the agent, the words that played.
+ * @public
+ */
+export interface MediaTranscriptionFrame {
+    type: MediaTranscriptionFrameType;
+    /**
+     * Party that spoke. The agent's own words are `callee`.
+     * @public
+     */
+    party: MediaParty;
+    /**
+     * Id of a reply, a stream, an utterance or a DTMF input, unique in the run.
+     * @public
+     */
+    id: string;
+    final: boolean;
+    text: string;
+    /**
+     * Run-clock time or duration in ms of PBX audio since the run started. It continues across reconnects.
+     * @public
+     */
+    startMs: number;
+    /**
+     * Run-clock time or duration in ms of PBX audio since the run started. It continues across reconnects.
+     * @public
+     */
+    endMs: number;
+    /**
+     * BCP-47 language tag, optionally with a region subtag: 'en', 'en-US', 'pt-BR'. Deliberately a
+     *     constrained string rather than an enum: the offerable set is a property of the selected
+     *     recogniser and changes with it, so an enumeration here would cap the set at whatever was known
+     *     when the model was released.
+     * @public
+     */
+    language: string;
+    /**
+     * Recogniser confidence from 0 to 1. Absent when the text is not from a recogniser, or the engine gives none.
+     * @public
+     */
+    confidence?: number | undefined;
+}
+/**
+ * Transfer the call after the queued agent audio plays. Terminal: the PBX answers with `finish`.
+ * @public
+ */
+export interface MediaTransferFrame {
+    type: MediaTransferFrameType;
+    /**
+     * Dialplan context of the target.
+     * @public
+     */
+    context: string;
+    /**
+     * Extension of the target.
+     * @public
+     */
+    extension: string;
+}
+/**
+ * One example of each message that the cloud sends.
+ * @public
+ */
+export interface MediaCloudFrames {
+    /**
+     * The cloud accepted `start`. Sent once per connection.
+     * @public
+     */
+    started?: MediaStartedFrame | undefined;
+    /**
+     * Words of one party. Partials and the final of one utterance share `id`. For the agent, the words that played.
+     * @public
+     */
+    transcription?: MediaTranscriptionFrame | undefined;
+    /**
+     * Opens an agent audio stream. `MediaCloudAudio` messages of the stream follow until `audio.end`. Streams do not interleave.
+     * @public
+     */
+    audioBegin?: MediaAudioBeginFrame | undefined;
+    /**
+     * One binary audio message. In the describe answer it is base64; on the socket it is raw bytes.
+     * @public
+     */
+    audio?: Uint8Array | undefined;
+    /**
+     * Closes an agent audio stream.
+     * @public
+     */
+    audioEnd?: MediaAudioEndFrame | undefined;
+    /**
+     * Stop every agent audio stream of the run: the one that plays and the ones that wait.
+     * @public
+     */
+    stop?: MediaStopFrame | undefined;
+    /**
+     * Lower the volume of a playing reply.
+     * @public
+     */
+    duck?: MediaDuckFrame | undefined;
+    /**
+     * Return a reply to full volume.
+     * @public
+     */
+    restore?: MediaRestoreFrame | undefined;
+    /**
+     * The call continues in another language.
+     * @public
+     */
+    language?: MediaLanguageFrame | undefined;
+    /**
+     * Transfer the call after the queued agent audio plays. Terminal: the PBX answers with `finish`.
+     * @public
+     */
+    transfer?: MediaTransferFrame | undefined;
+    /**
+     * Hang up the call after the queued agent audio plays. Terminal: the PBX answers with `finish`.
+     * @public
+     */
+    hangup?: MediaHangupFrame | undefined;
+    /**
+     * An error on the run. When `retryable` is false, the socket closes after it.
+     * @public
+     */
+    error?: MediaErrorFrame | undefined;
+    /**
+     * Answer to `finish`: the cloud processed all audio. The socket closes with 1000 after it.
+     * @public
+     */
+    finished?: MediaFinishedFrame | undefined;
+}
+/**
+ * DTMF input that the PBX collected. The cloud takes `digits` as the caller's reply.
+ * @public
+ */
+export interface MediaDtmfFrame {
+    type: MediaDtmfFrameType;
+    /**
+     * Id of a reply, a stream, an utterance or a DTMF input, unique in the run.
+     * @public
+     */
+    id: string;
+    digits: string;
+}
+/**
+ * The run ended on the PBX. The PBX sends no audio after it and waits up to 3 s for `finished`.
+ * @public
+ */
+export interface MediaFinishFrame {
+    type: MediaFinishFrameType;
+    /**
+     * Who ended the run on the PBX.
+     * @public
+     */
+    reason: MediaFinishReason;
+}
+/**
+ * An agent audio stream changed state on the PBX.
+ * @public
+ */
+export interface MediaPlaybackFrame {
+    type: MediaPlaybackFrameType;
+    /**
+     * Id of a reply, a stream, an utterance or a DTMF input, unique in the run.
+     * @public
+     */
+    replyId: string;
+    /**
+     * Id of a reply, a stream, an utterance or a DTMF input, unique in the run.
+     * @public
+     */
+    streamId: string;
+    /**
+     * Event of an agent audio stream.
+     * @public
+     */
+    event: MediaPlaybackEvent;
+    /**
+     * Run-clock time or duration in ms of PBX audio since the run started. It continues across reconnects.
+     * @public
+     */
+    atMs: number;
+    /**
+     * Audio of the stream that played, in ms. Present with `finished` and `stopped`.
+     * @public
+     */
+    playedMs?: number | undefined;
+    /**
+     * Present with `stopped`.
+     * @public
+     */
+    cause?: MediaPlaybackStopCause | undefined;
+}
+/**
+ * The agent of a media run. At most one of `draft` and `version` is set.
+ * @public
+ */
+export interface MediaStartAgent {
+    /**
+     * Agent id.
+     * @public
+     */
+    id: string;
+    /**
+     * True to run the unpublished draft. Absent or null: false.
+     * @public
+     */
+    draft?: boolean | undefined;
+    /**
+     * Published version to run. Absent or null: the version the agent routing selects.
+     * @public
+     */
+    version?: number | undefined;
+}
+/**
+ * Format of the binary PBX audio (`MediaPbxAudio`). Channel 0 is the caller; channel 1, when present, is the callee. Samples interleave.
+ * @public
+ */
+export interface MediaAudioFormat {
+    /**
+     * Sample encoding of PBX audio.
+     * @public
+     */
+    encoding: MediaAudioEncoding;
+    /**
+     * Sample rate in Hz.
+     * @public
+     */
+    sampleRate: number;
+    /**
+     * Number of channels.
+     * @public
+     */
+    channels: number;
+}
+/**
+ * Caller identity. Each member is absent, null or empty when unknown.
+ * @public
+ */
+export interface MediaStartCaller {
+    name?: string | undefined;
+    number?: string | undefined;
+    email?: string | undefined;
+    company?: string | undefined;
+}
+/**
+ * First text frame of a socket, sent within 10 s of the upgrade. Opens a run, or attaches a reconnect to it.
+ * @public
+ */
+export interface MediaStartFrame {
+    type: MediaStartFrameType;
+    /**
+     * Purpose of a media run.
+     * @public
+     */
+    purpose: MediaRunPurpose;
+    /**
+     * Id of the PBX call. Use it to correlate the call events of the run.
+     * @public
+     */
+    callId: string;
+    /**
+     * Index of this run on the call, from 0.
+     * @public
+     */
+    flowIndex: number;
+    /**
+     * Id of this run. The same on every connection of the run; the cloud attaches a reconnect and settles a takeover by it.
+     * @public
+     */
+    sessionId: string;
+    /**
+     * True when an earlier connection of this run got `started`: the cloud continues that session without a new greeting.
+     * @public
+     */
+    resume: boolean;
+    /**
+     * Run-clock time of the first audio byte of this connection: 0 on the first connection.
+     * @public
+     */
+    resumeAtMs: number;
+    /**
+     * Format of the binary PBX audio (`MediaPbxAudio`). Channel 0 is the caller; channel 1, when present, is the callee. Samples interleave.
+     * @public
+     */
+    audio: MediaAudioFormat;
+    /**
+     * Present when `purpose` is `agent`.
+     * @public
+     */
+    agent?: MediaStartAgent | undefined;
+    /**
+     * Languages to recognise when `purpose` is `transcription`. May be empty only when `autoDetect` is true. Ignored when `purpose` is `agent`: the agent settings give the languages.
+     * @public
+     */
+    languages?: string[] | undefined;
+    /**
+     * Language of the call. Required when `purpose` is `agent`: the agent starts in its configured language that matches it. With `transcription`, absent or null means the first of `languages`.
+     * @public
+     */
+    preferredLanguage?: string | undefined;
+    /**
+     * True to identify the caller's language among `languages` when `purpose` is `transcription`. Absent: false. Ignored when `purpose` is `agent`.
+     * @public
+     */
+    autoDetect?: boolean | undefined;
+    /**
+     * Caller identity. Each member is absent, null or empty when unknown.
+     * @public
+     */
+    caller: MediaStartCaller;
+    /**
+     * Number that the caller dialled. Absent, null or empty when unknown.
+     * @public
+     */
+    dialedNumber?: string | undefined;
+    /**
+     * Base64 of a JSON object of the `VOICEBOT_*` channel variables. Absent, null or empty when there are none.
+     * @public
+     */
+    variables?: string | undefined;
+}
+/**
+ * Voice activity of a party as the PBX detects it. Sent when the state changes, and every 100 ms while agent audio plays.
+ * @public
+ */
+export interface MediaVadFrame {
+    type: MediaVadFrameType;
+    /**
+     * Party whose audio the report describes.
+     * @public
+     */
+    party: MediaParty;
+    /**
+     * True while the party speaks.
+     * @public
+     */
+    speaking: boolean;
+    /**
+     * Energy of the audio frame that produced the report, in dBFS.
+     * @public
+     */
+    energyDb: number;
+    /**
+     * Run-clock time or duration in ms of PBX audio since the run started. It continues across reconnects.
+     * @public
+     */
+    atMs: number;
+}
+/**
+ * One example of each message that the PBX sends.
+ * @public
+ */
+export interface MediaPbxFrames {
+    /**
+     * First text frame of a socket, sent within 10 s of the upgrade. Opens a run, or attaches a reconnect to it.
+     * @public
+     */
+    start?: MediaStartFrame | undefined;
+    /**
+     * One binary audio message. In the describe answer it is base64; on the socket it is raw bytes.
+     * @public
+     */
+    audio?: Uint8Array | undefined;
+    /**
+     * Voice activity of a party as the PBX detects it. Sent when the state changes, and every 100 ms while agent audio plays.
+     * @public
+     */
+    vad?: MediaVadFrame | undefined;
+    /**
+     * DTMF input that the PBX collected. The cloud takes `digits` as the caller's reply.
+     * @public
+     */
+    dtmf?: MediaDtmfFrame | undefined;
+    /**
+     * An agent audio stream changed state on the PBX.
+     * @public
+     */
+    playback?: MediaPlaybackFrame | undefined;
+    /**
+     * The run ended on the PBX. The PBX sends no audio after it and waits up to 3 s for `finished`.
+     * @public
+     */
+    finish?: MediaFinishFrame | undefined;
+}
+/**
+ * One example of each message of the PBX media socket.
+ * @public
+ */
+export interface MediaProtocolFrames {
+    /**
+     * Messages that the PBX sends.
+     * @public
+     */
+    pbx?: MediaPbxFrames | undefined;
+    /**
+     * Messages that the cloud sends.
+     * @public
+     */
+    cloud?: MediaCloudFrames | undefined;
+}
 /**
  * Client-reported progress of an in-flight reply playback.
  * @public
@@ -1273,18 +1853,111 @@ export interface DescribeVoiceProtocolOutput {
      * @public
      */
     callEvent?: VoiceCallEvent | undefined;
+    /**
+     * Messages of the PBX media socket `/v2/voice/media/ws` (subprotocol `wilma-media-v2`): one example of each. The binary audio messages are base64 here and raw bytes on the socket.
+     * @public
+     */
+    media?: MediaProtocolFrames | undefined;
+}
+/**
+ * @public
+ */
+export interface GetVoiceDefaultsInput {
+}
+/**
+ * Defaults of the tuning members of `VoiceChannelSettings.interruptions`.
+ * @public
+ */
+export interface VoiceInterruptionDefaults {
+    /**
+     * Minimum number of caller words that count as an interruption.
+     * @public
+     */
+    minWords: number;
+    /**
+     * Minimum recognition confidence, from 0 to 1, of the caller words that count as an interruption.
+     * @public
+     */
+    minConfidence: number;
+    /**
+     * Caller speech in ms before the agent's reply is lowered.
+     * @public
+     */
+    duckAfterMs: number;
+    /**
+     * Volume in dB of a lowered reply.
+     * @public
+     */
+    duckVolumeDb: number;
+    /**
+     * Caller silence in ms before a lowered reply gets its volume back.
+     * @public
+     */
+    restoreAfterMs: number;
+}
+/**
+ * Defaults of `VoiceChannelSettings.turnEnd`.
+ * @public
+ */
+export interface VoiceTurnEndDefaults {
+    /**
+     * Wait in ms after the recogniser marks the end of a caller phrase, before the agent answers.
+     * @public
+     */
+    holdMs: number;
+}
+/**
+ * Defaults of `VoiceChannelSettings.voiceDetection`.
+ * @public
+ */
+export interface VoiceDetectionDefaults {
+    /**
+     * Silence in ms that ends the caller speech.
+     * @public
+     */
+    silenceMs: number;
+    /**
+     * Level in dBFS below which caller audio is silence.
+     * @public
+     */
+    silenceLevelDb: number;
+    /**
+     * Mean frame level in dBFS from which caller audio is speech.
+     * @public
+     */
+    speechLevelDb: number;
+}
+/**
+ * @public
+ */
+export interface GetVoiceDefaultsOutput {
+    /**
+     * Defaults of `VoiceChannelSettings.voiceDetection`.
+     * @public
+     */
+    voiceDetection: VoiceDetectionDefaults;
+    /**
+     * Defaults of the tuning members of `VoiceChannelSettings.interruptions`.
+     * @public
+     */
+    interruptions: VoiceInterruptionDefaults;
+    /**
+     * Defaults of `VoiceChannelSettings.turnEnd`.
+     * @public
+     */
+    turnEnd: VoiceTurnEndDefaults;
 }
 /**
  * @public
  */
 export interface ListVoicesInput {
     /**
-     * Provider whose catalog to read: 'elevenlabs', 'google', 'amazon_polly'. This is the scheme of the URIs answered on `voices`.
+     * Provider whose catalog to read: 'elevenlabs', 'google', 'amazon_polly', 'soniox'. This is the scheme of the URIs answered on `voices`.
      * @public
      */
     provider: string;
     /**
-     * Server-side filter for language-bound providers: only voices whose language SET CONTAINS this code. Ignored for ElevenLabs, whose voices are not language bound.
+     * Server-side filter for language-bound providers: only voices whose language SET CONTAINS this code. Ignored for ElevenLabs, whose voices are not language bound. For Soniox it keeps the voices whose model speaks the language.
      * @public
      */
     language?: string | undefined;