@node-webrtc-rust/bindings 0.9.19 → 0.9.21

package/index.d.ts

@@ -299,10 +299,14 @@ export interface JsLanguageIdConfig {
   allowlist?: Array<string>
   minSpeechMs?: number
   continuous?: boolean
+  timing?: string
   lidMaxClipMs?: number
   lidGateMaxWaitMs?: number
   ttsExclusion?: boolean
 }
+export interface JsReplayConfig {
+  maxAgeMs?: number
+}
 export interface JsVoiceAgentConfig {
   vad?: JsVadConfig
   events?: JsEventsConfig
@@ -311,6 +315,7 @@ export interface JsVoiceAgentConfig {
   languageId?: JsLanguageIdConfig
   postUtteranceSilenceMs?: number
   noiseSuppression?: JsNoiseSuppressionConfig
+  replay?: JsReplayConfig
 }
 export const enum JsSpeechEventType {
   UserSpeakingStart = 'user_speaking_start',
@@ -318,6 +323,7 @@ export const enum JsSpeechEventType {
   UserSpeechPartial = 'user_speech_partial',
   UserSpeechFinal = 'user_speech_final',
   UserLanguage = 'user_language',
+  LanguageIdSkipped = 'language_id_skipped',
   AgentSpeakingStart = 'agent_speaking_start',
   AgentSpeakingEnd = 'agent_speaking_end',
   VadTriggered = 'vad_triggered',
@@ -329,7 +335,9 @@ export const enum JsSpeechEventType {
   BargeIn = 'barge_in',
   Error = 'error',
   SttConfigUpdated = 'stt_config_updated',
-  TtsConfigUpdated = 'tts_config_updated'
+  TtsConfigUpdated = 'tts_config_updated',
+  SttHoldStarted = 'stt_hold_started',
+  SttHoldEnded = 'stt_hold_ended'
 }
 export interface JsSpeechEvent {
   eventType: JsSpeechEventType
@@ -343,6 +351,23 @@ export interface JsSpeechEvent {
   voice?: string
   modelPath?: string
   endpoint?: string
+  holdMode?: string
+  holdOutcome?: string
+  bufferedMs?: number
+  droppedMs?: number
+  reason?: string
+  speechMs?: number
+}
+export const enum JsSttHoldMode {
+  BufferReplay = 'buffer_replay',
+  FirstUtterance = 'first_utterance'
+}
+export interface JsBeginSttHoldOptions {
+  mode: JsSttHoldMode
+  maxBufferMs?: number
+}
+export interface JsReleaseSttHoldOptions {
+  replay: boolean
 }
 export interface JsUpdateTtsOptions {
   cancelInflight?: boolean
@@ -582,6 +607,12 @@ export declare class JsVoiceAgent {
   updateStt(config: JsSttConfig): Promise<void>
   /** Apply a new TTS config before the next synthesis job. */
   updateTts(config: JsTtsConfig, options?: JsUpdateTtsOptions | undefined | null): Promise<void>
+  /** Start a host STT hold: stop feeding/polling the current STT and buffer user PCM. */
+  beginSttHold(options: JsBeginSttHoldOptions): Promise<void>
+  /** End the hold after `update_stt`; `replay` decodes held audio through the new STT. */
+  releaseSttHold(options: JsReleaseSttHoldOptions): Promise<void>
+  /** Abort the hold: the old STT resumes and receives the buffered audio. */
+  cancelSttHold(): Promise<void>
   /** Re-transcribe the last utterance PCM on the current STT provider (after `update_stt`). */
   replayLastUtterance(): Promise<void>
 }

package/index.js

@@ -310,7 +310,7 @@ if (!nativeBinding) {
   throw new Error(`Failed to load native binding`)
 }
 
-const { JsConferenceRoom, JsConferenceServer, JsMixGraph, JsMixPlacement, JsRoomNoiseSuppression, JsMuteScope, JsRtcDataChannel, JsMediaStreamTrack, JsMediaStream, JsLocalAudioTrack, JsPeerConnection, JsRtpReceiver, JsRtpSender, JsRtpTransceiver, JsSessionAudioFormat, JsSessionRecorder, JsEventDeliveryMode, JsNoiseSuppressionProvider, JsVadSampleRate, JsSttVendor, JsTtsVendor, JsSpeechEventType, JsVoiceAgent, JsClipStatus, JsGrowingClipWriter, playClipFromPath, playClipFromBytes, playClipProgressive, getClipStatus, stopClip, takeClipFrame, preloadLanguageId, jsQuatIdentity, jsVec3Zero, version } = nativeBinding
+const { JsConferenceRoom, JsConferenceServer, JsMixGraph, JsMixPlacement, JsRoomNoiseSuppression, JsMuteScope, JsRtcDataChannel, JsMediaStreamTrack, JsMediaStream, JsLocalAudioTrack, JsPeerConnection, JsRtpReceiver, JsRtpSender, JsRtpTransceiver, JsSessionAudioFormat, JsSessionRecorder, JsEventDeliveryMode, JsNoiseSuppressionProvider, JsVadSampleRate, JsSttVendor, JsTtsVendor, JsSpeechEventType, JsSttHoldMode, JsVoiceAgent, JsClipStatus, JsGrowingClipWriter, playClipFromPath, playClipFromBytes, playClipProgressive, getClipStatus, stopClip, takeClipFrame, preloadLanguageId, jsQuatIdentity, jsVec3Zero, version } = nativeBinding
 
 module.exports.JsConferenceRoom = JsConferenceRoom
 module.exports.JsConferenceServer = JsConferenceServer
@@ -334,6 +334,7 @@ module.exports.JsVadSampleRate = JsVadSampleRate
 module.exports.JsSttVendor = JsSttVendor
 module.exports.JsTtsVendor = JsTtsVendor
 module.exports.JsSpeechEventType = JsSpeechEventType
+module.exports.JsSttHoldMode = JsSttHoldMode
 module.exports.JsVoiceAgent = JsVoiceAgent
 module.exports.JsClipStatus = JsClipStatus
 module.exports.JsGrowingClipWriter = JsGrowingClipWriter

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@node-webrtc-rust/bindings",
-  "version": "0.9.19",
+  "version": "0.9.21",
   "description": "NAPI-RS native bindings for node-webrtc-rust — WebRTC peer connections, data channels, and conference audio mixing",
   "keywords": [
     "webrtc",
@@ -58,12 +58,12 @@
     "@napi-rs/cli": "^2.18.0"
   },
   "optionalDependencies": {
-    "@node-webrtc-rust/bindings-win32-x64-msvc": "0.9.19",
-    "@node-webrtc-rust/bindings-darwin-x64": "0.9.19",
-    "@node-webrtc-rust/bindings-linux-x64-gnu": "0.9.19",
-    "@node-webrtc-rust/bindings-darwin-arm64": "0.9.19",
-    "@node-webrtc-rust/bindings-linux-arm64-gnu": "0.9.19",
-    "@node-webrtc-rust/bindings-linux-x64-musl": "0.9.19"
+    "@node-webrtc-rust/bindings-win32-x64-msvc": "0.9.21",
+    "@node-webrtc-rust/bindings-darwin-x64": "0.9.21",
+    "@node-webrtc-rust/bindings-linux-x64-gnu": "0.9.21",
+    "@node-webrtc-rust/bindings-darwin-arm64": "0.9.21",
+    "@node-webrtc-rust/bindings-linux-arm64-gnu": "0.9.21",
+    "@node-webrtc-rust/bindings-linux-x64-musl": "0.9.21"
   },
   "engines": {
     "node": ">=18"