@wildix/wilma-voice-client 0.0.1

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

@@ -0,0 +1,805 @@
+import type { VoiceAudioEncoding, VoiceCommandType, VoiceControlAction, VoiceEventType, VoiceInputMode } from "./enums";
+/**
+ * A chunk of raw caller audio, base64-encoded in the session's negotiated input format
+ *     (inputAudio from the connect handshake). Used by web/app clients in AUDIO input mode; the
+ *     service buffers and transcribes it server-side. Not used in TRANSCRIPTION mode.
+ * @public
+ */
+export interface AudioEvent {
+    /**
+     * Discriminator for the inner payload of an `event` frame.
+     * @public
+     */
+    type: VoiceEventType;
+    data: string;
+    /**
+     * Optional monotonic chunk sequence number, for gap detection.
+     * @public
+     */
+    seq?: number | undefined;
+}
+/**
+ * Envelope of a WebSocket event frame: `\{ "type": "event", "event": \{ ... \} \}`.
+ * @public
+ */
+export interface AudioWebsocketEvent {
+    type: string;
+    /**
+     * A chunk of raw caller audio, base64-encoded in the session's negotiated input format
+     *     (inputAudio from the connect handshake). Used by web/app clients in AUDIO input mode; the
+     *     service buffers and transcribes it server-side. Not used in TRANSCRIPTION mode.
+     * @public
+     */
+    event: AudioEvent;
+}
+/**
+ * The call/conversation completed (e.g. the caller hung up or the call was transferred away).
+ * @public
+ */
+export interface CompleteEvent {
+    /**
+     * Discriminator for the inner payload of an `event` frame.
+     * @public
+     */
+    type: VoiceEventType;
+    reason: string;
+    time: number;
+}
+/**
+ * Envelope of a WebSocket event frame: `\{ "type": "event", "event": \{ ... \} \}`.
+ * @public
+ */
+export interface CompleteWebsocketEvent {
+    type: string;
+    /**
+     * The call/conversation completed (e.g. the caller hung up or the call was transferred away).
+     * @public
+     */
+    event: CompleteEvent;
+}
+/**
+ * A lightweight control frame from the client (AUDIO mode).
+ * @public
+ */
+export interface ControlEvent {
+    /**
+     * Discriminator for the inner payload of an `event` frame.
+     * @public
+     */
+    type: VoiceEventType;
+    /**
+     * A client control action on the audio input (AUDIO mode).
+     * @public
+     */
+    action: VoiceControlAction;
+}
+/**
+ * Envelope of a WebSocket event frame: `\{ "type": "event", "event": \{ ... \} \}`.
+ * @public
+ */
+export interface ControlWebsocketEvent {
+    type: string;
+    /**
+     * A lightweight control frame from the client (AUDIO mode).
+     * @public
+     */
+    event: ControlEvent;
+}
+/**
+ * @public
+ */
+export interface DescribeVoiceProtocolInput {
+}
+/**
+ * A PCM/compressed audio format description.
+ * @public
+ */
+export interface VoiceAudioFormat {
+    /**
+     * Encoding of raw audio carried over the connection (client `audio` chunks and server `play` frames).
+     * @public
+     */
+    encoding: VoiceAudioEncoding;
+    /**
+     * Samples per second, e.g. 8000, 16000 or 24000.
+     * @public
+     */
+    sampleRate: number;
+    /**
+     * Channel count. Defaults to 1 (mono).
+     * @public
+     */
+    channels?: number | undefined;
+}
+/**
+ * Parameters supplied on the WebSocket connect (upgrade) request, as query-string parameters.
+ * @public
+ */
+export interface VoiceConnectParams {
+    /**
+     * Protocol version advertised by the client (wim-voicebots `v`). 0 (or absent) is the legacy PBX protocol; >= 1 enables the sessionStarted event.
+     * @public
+     */
+    protocolVersion?: number | undefined;
+    /**
+     * Caller-supplied session id, unique per connection.
+     * @public
+     */
+    sessionId: string;
+    /**
+     * Agent to run on this session.
+     * @public
+     */
+    agentId: string;
+    /**
+     * Deployment used to resolve the published agent version. Defaults to 'production'. Mutually exclusive with agentVersion and agentDraft.
+     * @public
+     */
+    agentDeployment?: string | undefined;
+    /**
+     * Published agent version to run. Mutually exclusive with agentDeployment and agentDraft.
+     * @public
+     */
+    agentVersion?: number | undefined;
+    /**
+     * Run the agent's current draft (for testing). Mutually exclusive with agentDeployment and agentVersion.
+     * @public
+     */
+    agentDraft?: boolean | undefined;
+    /**
+     * PBX call id, when the session is bridged to a call.
+     * @public
+     */
+    callId?: string | undefined;
+    /**
+     * BCP-47 language of the conversation, e.g. 'en-US'.
+     * @public
+     */
+    language?: string | undefined;
+    callerName?: string | undefined;
+    callerNumber?: string | undefined;
+    callerEmail?: string | undefined;
+    callerCompany?: string | undefined;
+    /**
+     * Override the TTS voice for this session, as a voice URI, e.g. 'elevenlabs://<voiceId>?apiKey=...'. Defaults to the agent's configured voice.
+     * @public
+     */
+    voice?: string | undefined;
+    /**
+     * First message spoken to the caller. Overrides the agent's configured greeting.
+     * @public
+     */
+    initialMessage?: string | undefined;
+    /**
+     * Base64-encoded JSON object of agent variables for this session, e.g. base64('\{"orderId":"123"\}').
+     * @public
+     */
+    variables?: string | undefined;
+    /**
+     * How caller speech reaches the service. Defaults to TRANSCRIPTION (PBX). Web/app clients use AUDIO.
+     * @public
+     */
+    inputMode?: VoiceInputMode | undefined;
+    /**
+     * Format of client `audio` chunks in AUDIO mode.
+     * @public
+     */
+    inputAudio?: VoiceAudioFormat | undefined;
+    /**
+     * Preferred format for server-synthesized `play` frames.
+     * @public
+     */
+    outputAudio?: VoiceAudioFormat | undefined;
+}
+/**
+ * An error surfaced on the connection. Non-terminal unless followed by a SessionEndedEvent.
+ * @public
+ */
+export interface VoiceErrorEvent {
+    /**
+     * Discriminator for the inner payload of an `event` frame.
+     * @public
+     */
+    type: VoiceEventType;
+    /**
+     * Stable machine-readable error code.
+     * @public
+     */
+    code: string;
+    message: string;
+    time: number;
+}
+/**
+ * Envelope of a WebSocket event frame: `\{ "type": "event", "event": \{ ... \} \}`.
+ * @public
+ */
+export interface ErrorWebsocketEvent {
+    type: string;
+    /**
+     * An error surfaced on the connection. Non-terminal unless followed by a SessionEndedEvent.
+     * @public
+     */
+    event: VoiceErrorEvent;
+}
+/**
+ * Terminate the session and hang up the call.
+ * @public
+ */
+export interface HangupCommandData {
+    reason?: string | undefined;
+    /**
+     * Discriminator for the inner payload of a `command` frame.
+     * @public
+     */
+    type: VoiceCommandType;
+}
+/**
+ * Envelope of a WebSocket command frame: `\{ "type": "command", "command": \{ ... \} \}`.
+ * @public
+ */
+export interface HangupWebsocketCommand {
+    type: string;
+    /**
+     * Terminate the session and hang up the call.
+     * @public
+     */
+    command: HangupCommandData;
+}
+/**
+ * The caller interrupted (barged in over) a reply that was being played.
+ * @public
+ */
+export interface InterruptionEvent {
+    /**
+     * Discriminator for the inner payload of an `event` frame.
+     * @public
+     */
+    type: VoiceEventType;
+    /**
+     * Correlates a caller phrase, the agent reply it produced, and the playback/interruption of that reply.
+     * @public
+     */
+    replyId: string;
+    /**
+     * Portion of the reply that had been played before the interruption.
+     * @public
+     */
+    playedText: string;
+    /**
+     * The full reply text that was being played.
+     * @public
+     */
+    originalText: string;
+    time: number;
+}
+/**
+ * Envelope of a WebSocket event frame: `\{ "type": "event", "event": \{ ... \} \}`.
+ * @public
+ */
+export interface InterruptionWebsocketEvent {
+    type: string;
+    /**
+     * The caller interrupted (barged in over) a reply that was being played.
+     * @public
+     */
+    event: InterruptionEvent;
+}
+/**
+ * Client-reported progress of an in-flight reply playback.
+ * @public
+ */
+export interface PlaybackEvent {
+    /**
+     * Discriminator for the inner payload of an `event` frame.
+     * @public
+     */
+    type: VoiceEventType;
+    /**
+     * Correlates a caller phrase, the agent reply it produced, and the playback/interruption of that reply.
+     * @public
+     */
+    replyId: string;
+    text: string;
+    complete: boolean;
+    startTime: number;
+    endTime?: number | undefined;
+}
+/**
+ * Envelope of a WebSocket event frame: `\{ "type": "event", "event": \{ ... \} \}`.
+ * @public
+ */
+export interface PlaybackWebsocketEvent {
+    type: string;
+    /**
+     * Client-reported progress of an in-flight reply playback.
+     * @public
+     */
+    event: PlaybackEvent;
+}
+/**
+ * A chunk of synthesized reply audio to play. `data` is base64, PCM-encoded at the sample rate declared in the matching `say` command; an empty `data` marks the end of the stream.
+ * @public
+ */
+export interface PlayCommandData {
+    /**
+     * Correlates a caller phrase, the agent reply it produced, and the playback/interruption of that reply.
+     * @public
+     */
+    replyId: string;
+    /**
+     * Identifies one synthesized audio stream within a session; carried by a `say` command and the `play` frames that render it.
+     * @public
+     */
+    streamId: string;
+    data: string;
+    /**
+     * Discriminator for the inner payload of a `command` frame.
+     * @public
+     */
+    type: VoiceCommandType;
+}
+/**
+ * Envelope of a WebSocket command frame: `\{ "type": "command", "command": \{ ... \} \}`.
+ * @public
+ */
+export interface PlayWebsocketCommand {
+    type: string;
+    /**
+     * A chunk of synthesized reply audio to play. `data` is base64, PCM-encoded at the sample rate declared in the matching `say` command; an empty `data` marks the end of the stream.
+     * @public
+     */
+    command: PlayCommandData;
+}
+/**
+ * A final caller phrase that should trigger an agent turn (PBX / TRANSCRIPTION mode).
+ * @public
+ */
+export interface ReplyEvent {
+    /**
+     * Discriminator for the inner payload of an `event` frame.
+     * @public
+     */
+    type: VoiceEventType;
+    /**
+     * Correlates a caller phrase, the agent reply it produced, and the playback/interruption of that reply.
+     * @public
+     */
+    id: string;
+    text: string;
+    startTime: number;
+    endTime: number;
+}
+/**
+ * Envelope of a WebSocket event frame: `\{ "type": "event", "event": \{ ... \} \}`.
+ * @public
+ */
+export interface ReplyWebsocketEvent {
+    type: string;
+    /**
+     * A final caller phrase that should trigger an agent turn (PBX / TRANSCRIPTION mode).
+     * @public
+     */
+    event: ReplyEvent;
+}
+/**
+ * Audio-stream parameters when the server synthesizes the reply itself (TTS provider configured on the agent): the `say` is followed by a stream of `play` frames on this stream id.
+ * @public
+ */
+export interface SayCommandAudioOptions {
+    /**
+     * Identifies one synthesized audio stream within a session; carried by a `say` command and the `play` frames that render it.
+     * @public
+     */
+    streamId: string;
+    sampleRate: number;
+}
+/**
+ * Play a reply to the caller. When `audio` is absent the client performs text-to-speech from `text`; when present the server streams synthesized `play` frames for this reply.
+ * @public
+ */
+export interface SayCommandData {
+    text: string;
+    /**
+     * Correlates a caller phrase, the agent reply it produced, and the playback/interruption of that reply.
+     * @public
+     */
+    replyId: string;
+    /**
+     * Whether the caller may barge in over this reply.
+     * @public
+     */
+    interruptible?: boolean | undefined;
+    /**
+     * Audio-stream parameters when the server synthesizes the reply itself (TTS provider configured on the agent): the `say` is followed by a stream of `play` frames on this stream id.
+     * @public
+     */
+    audio?: SayCommandAudioOptions | undefined;
+    /**
+     * Discriminator for the inner payload of a `command` frame.
+     * @public
+     */
+    type: VoiceCommandType;
+}
+/**
+ * Envelope of a WebSocket command frame: `\{ "type": "command", "command": \{ ... \} \}`.
+ * @public
+ */
+export interface SayWebsocketCommand {
+    type: string;
+    /**
+     * Play a reply to the caller. When `audio` is absent the client performs text-to-speech from `text`; when present the server streams synthesized `play` frames for this reply.
+     * @public
+     */
+    command: SayCommandData;
+}
+/**
+ * Terminal server event: the session ended, optionally with an error.
+ * @public
+ */
+export interface SessionEndedEvent {
+    /**
+     * Discriminator for the inner payload of an `event` frame.
+     * @public
+     */
+    type: VoiceEventType;
+    reason?: string | undefined;
+    /**
+     * Present when the session ended because of an error.
+     * @public
+     */
+    error?: string | undefined;
+    time: number;
+}
+/**
+ * Envelope of a WebSocket event frame: `\{ "type": "event", "event": \{ ... \} \}`.
+ * @public
+ */
+export interface SessionEndedWebsocketEvent {
+    type: string;
+    /**
+     * Terminal server event: the session ended, optionally with an error.
+     * @public
+     */
+    event: SessionEndedEvent;
+}
+/**
+ * Emitted once right after the handshake to confirm the session and the agent it is bound to. Sent only when the client advertises protocol version >= 1.
+ * @public
+ */
+export interface SessionStartedEvent {
+    /**
+     * Discriminator for the inner payload of an `event` frame.
+     * @public
+     */
+    type: VoiceEventType;
+    time: number;
+    /**
+     * Caller-facing voice session identifier, unique per connection. Supplied by the client at connect (the PBX call/session id, or a client-generated id for web sessions).
+     * @public
+     */
+    sessionId: string;
+    /**
+     * Agent bound to the session.
+     * @public
+     */
+    agentId: string;
+    /**
+     * Published agent version resolved at connect; absent when running the agent draft.
+     * @public
+     */
+    agentVersion?: number | undefined;
+    /**
+     * Agent display name.
+     * @public
+     */
+    agentName?: string | undefined;
+    /**
+     * Agent category.
+     * @public
+     */
+    category?: string | undefined;
+    /**
+     * PBX group used for data ACL; empty when the agent is not assigned to a group.
+     * @public
+     */
+    groupId?: string | undefined;
+    /**
+     * Negotiated format of caller `audio` chunks the client should send (AUDIO mode).
+     * @public
+     */
+    inputAudio?: VoiceAudioFormat | undefined;
+    /**
+     * Format of the synthesized `play` frames the server will send when it performs TTS.
+     * @public
+     */
+    outputAudio?: VoiceAudioFormat | undefined;
+}
+/**
+ * Envelope of a WebSocket event frame: `\{ "type": "event", "event": \{ ... \} \}`.
+ * @public
+ */
+export interface SessionStartedWebsocketEvent {
+    type: string;
+    /**
+     * Emitted once right after the handshake to confirm the session and the agent it is bound to. Sent only when the client advertises protocol version >= 1.
+     * @public
+     */
+    event: SessionStartedEvent;
+}
+/**
+ * @public
+ */
+export interface StopCommandData {
+    /**
+     * Discriminator for the inner payload of a `command` frame.
+     * @public
+     */
+    type: VoiceCommandType;
+}
+/**
+ * Envelope of a WebSocket command frame: `\{ "type": "command", "command": \{ ... \} \}`.
+ * @public
+ */
+export interface StopWebsocketCommand {
+    type: string;
+    command: StopCommandData;
+}
+/**
+ * A transcription chunk of caller speech, provided by a client that does its own speech-to-text (PBX / TRANSCRIPTION mode), or emitted by the service (AUDIO mode) and echoed to the client.
+ * @public
+ */
+export interface TranscriptionEvent {
+    /**
+     * Discriminator for the inner payload of an `event` frame.
+     * @public
+     */
+    type: VoiceEventType;
+    /**
+     * Identifier of a transcription chunk within a session.
+     * @public
+     */
+    id: string;
+    text: string;
+    /**
+     * Transcription confidence score (accuracy level) in range from 0.0 to 1.0. Value -1.0 indicates confidence is unavailable.
+     * @public
+     */
+    confidence?: number | undefined;
+    /**
+     * True when this is the final transcript for the utterance rather than an interim hypothesis.
+     * @public
+     */
+    final?: boolean | undefined;
+    time: number;
+}
+/**
+ * Envelope of a WebSocket event frame: `\{ "type": "event", "event": \{ ... \} \}`.
+ * @public
+ */
+export interface TranscriptionWebsocketEvent {
+    type: string;
+    /**
+     * A transcription chunk of caller speech, provided by a client that does its own speech-to-text (PBX / TRANSCRIPTION mode), or emitted by the service (AUDIO mode) and echoed to the client.
+     * @public
+     */
+    event: TranscriptionEvent;
+}
+/**
+ * Transfer the call to a PBX destination and terminate the session. Honoured on PBX clients.
+ * @public
+ */
+export interface TransferCommandData {
+    context: string;
+    extension: string;
+    /**
+     * Discriminator for the inner payload of a `command` frame.
+     * @public
+     */
+    type: VoiceCommandType;
+}
+/**
+ * Envelope of a WebSocket command frame: `\{ "type": "command", "command": \{ ... \} \}`.
+ * @public
+ */
+export interface TransferWebsocketCommand {
+    type: string;
+    /**
+     * Transfer the call to a PBX destination and terminate the session. Honoured on PBX clients.
+     * @public
+     */
+    command: TransferCommandData;
+}
+/**
+ * @public
+ */
+export interface DescribeVoiceProtocolOutput {
+    /**
+     * Parameters supplied on the WebSocket connect (upgrade) request, as query-string parameters.
+     * @public
+     */
+    connectParams?: VoiceConnectParams | undefined;
+    /**
+     * Envelope of a WebSocket event frame: `\{ "type": "event", "event": \{ ... \} \}`.
+     * @public
+     */
+    audioWebsocketEvent?: AudioWebsocketEvent | undefined;
+    /**
+     * Envelope of a WebSocket event frame: `\{ "type": "event", "event": \{ ... \} \}`.
+     * @public
+     */
+    transcriptionWebsocketEvent?: TranscriptionWebsocketEvent | undefined;
+    /**
+     * Envelope of a WebSocket event frame: `\{ "type": "event", "event": \{ ... \} \}`.
+     * @public
+     */
+    replyWebsocketEvent?: ReplyWebsocketEvent | undefined;
+    /**
+     * Envelope of a WebSocket event frame: `\{ "type": "event", "event": \{ ... \} \}`.
+     * @public
+     */
+    playbackWebsocketEvent?: PlaybackWebsocketEvent | undefined;
+    /**
+     * Envelope of a WebSocket event frame: `\{ "type": "event", "event": \{ ... \} \}`.
+     * @public
+     */
+    interruptionWebsocketEvent?: InterruptionWebsocketEvent | undefined;
+    /**
+     * Envelope of a WebSocket event frame: `\{ "type": "event", "event": \{ ... \} \}`.
+     * @public
+     */
+    completeWebsocketEvent?: CompleteWebsocketEvent | undefined;
+    /**
+     * Envelope of a WebSocket event frame: `\{ "type": "event", "event": \{ ... \} \}`.
+     * @public
+     */
+    controlWebsocketEvent?: ControlWebsocketEvent | undefined;
+    /**
+     * Envelope of a WebSocket event frame: `\{ "type": "event", "event": \{ ... \} \}`.
+     * @public
+     */
+    sessionStartedWebsocketEvent?: SessionStartedWebsocketEvent | undefined;
+    /**
+     * Envelope of a WebSocket event frame: `\{ "type": "event", "event": \{ ... \} \}`.
+     * @public
+     */
+    sessionEndedWebsocketEvent?: SessionEndedWebsocketEvent | undefined;
+    /**
+     * Envelope of a WebSocket event frame: `\{ "type": "event", "event": \{ ... \} \}`.
+     * @public
+     */
+    errorWebsocketEvent?: ErrorWebsocketEvent | undefined;
+    /**
+     * Envelope of a WebSocket command frame: `\{ "type": "command", "command": \{ ... \} \}`.
+     * @public
+     */
+    sayWebsocketCommand?: SayWebsocketCommand | undefined;
+    /**
+     * Envelope of a WebSocket command frame: `\{ "type": "command", "command": \{ ... \} \}`.
+     * @public
+     */
+    playWebsocketCommand?: PlayWebsocketCommand | undefined;
+    /**
+     * Envelope of a WebSocket command frame: `\{ "type": "command", "command": \{ ... \} \}`.
+     * @public
+     */
+    stopWebsocketCommand?: StopWebsocketCommand | undefined;
+    /**
+     * Envelope of a WebSocket command frame: `\{ "type": "command", "command": \{ ... \} \}`.
+     * @public
+     */
+    hangupWebsocketCommand?: HangupWebsocketCommand | undefined;
+    /**
+     * Envelope of a WebSocket command frame: `\{ "type": "command", "command": \{ ... \} \}`.
+     * @public
+     */
+    transferWebsocketCommand?: TransferWebsocketCommand | undefined;
+}
+/**
+ * Terminate the session and hang up the call.
+ * @public
+ */
+export interface SendHangupInput {
+    reason?: string | undefined;
+    /**
+     * Caller-facing voice session identifier, unique per connection. Supplied by the client at connect (the PBX call/session id, or a client-generated id for web sessions).
+     * @public
+     */
+    sessionId: string;
+}
+/**
+ * @public
+ */
+export interface SendHangupOutput {
+}
+/**
+ * A chunk of synthesized reply audio to play. `data` is base64, PCM-encoded at the sample rate declared in the matching `say` command; an empty `data` marks the end of the stream.
+ * @public
+ */
+export interface SendPlayInput {
+    /**
+     * Correlates a caller phrase, the agent reply it produced, and the playback/interruption of that reply.
+     * @public
+     */
+    replyId: string;
+    /**
+     * Identifies one synthesized audio stream within a session; carried by a `say` command and the `play` frames that render it.
+     * @public
+     */
+    streamId: string;
+    data: string;
+    /**
+     * Caller-facing voice session identifier, unique per connection. Supplied by the client at connect (the PBX call/session id, or a client-generated id for web sessions).
+     * @public
+     */
+    sessionId: string;
+}
+/**
+ * @public
+ */
+export interface SendPlayOutput {
+}
+/**
+ * Play a reply to the caller. When `audio` is absent the client performs text-to-speech from `text`; when present the server streams synthesized `play` frames for this reply.
+ * @public
+ */
+export interface SendSayInput {
+    text: string;
+    /**
+     * Correlates a caller phrase, the agent reply it produced, and the playback/interruption of that reply.
+     * @public
+     */
+    replyId: string;
+    /**
+     * Whether the caller may barge in over this reply.
+     * @public
+     */
+    interruptible?: boolean | undefined;
+    /**
+     * Audio-stream parameters when the server synthesizes the reply itself (TTS provider configured on the agent): the `say` is followed by a stream of `play` frames on this stream id.
+     * @public
+     */
+    audio?: SayCommandAudioOptions | undefined;
+    /**
+     * Caller-facing voice session identifier, unique per connection. Supplied by the client at connect (the PBX call/session id, or a client-generated id for web sessions).
+     * @public
+     */
+    sessionId: string;
+}
+/**
+ * @public
+ */
+export interface SendSayOutput {
+}
+/**
+ * @public
+ */
+export interface SendStopInput {
+    /**
+     * Caller-facing voice session identifier, unique per connection. Supplied by the client at connect (the PBX call/session id, or a client-generated id for web sessions).
+     * @public
+     */
+    sessionId: string;
+}
+/**
+ * @public
+ */
+export interface SendStopOutput {
+}
+/**
+ * Transfer the call to a PBX destination and terminate the session. Honoured on PBX clients.
+ * @public
+ */
+export interface SendTransferInput {
+    context: string;
+    extension: string;
+    /**
+     * Caller-facing voice session identifier, unique per connection. Supplied by the client at connect (the PBX call/session id, or a client-generated id for web sessions).
+     * @public
+     */
+    sessionId: string;
+}
+/**
+ * @public
+ */
+export interface SendTransferOutput {
+}