feat pause resume transcribe (#438)

* feat pause resume transcribe

* wip

* fix jslint

* update fsmrf

lib/session/call-session.js

@@ -1075,6 +1075,24 @@ ${credential.label ? `, label: ${credential.label}` : ''} is chosen`);
     listenTask.updateListen(opts.listen_status);
   }
 
+  /**
+   * perform live call control -- change Transcribe status
+   * @param {object} opts
+   * @param {string} opts.transcribe_status - 'pause' or 'resume'
+  */
+  async _lccTranscribeStatus(opts) {
+    const task = this.currentTask;
+    if (!task || ![TaskName.Dial, TaskName.Transcribe].includes(task.name)) {
+      return this.logger.info(`CallSession:_lccTranscribeStatus - invalid transcribe_status in task ${task.name}`);
+    }
+    const transcribeTask = task.name === TaskName.Transcribe ? task : task.transcribeTask;
+    if (!transcribeTask) {
+      return this.logger
+        .info('CallSession:_lccTranscribeStatus - invalid transcribe_status: Dial does not have a Transcribe');
+    }
+    transcribeTask.updateTranscribe(opts.transcribe_status);
+  }
+
   async _lccMuteStatus(callSid, mute) {
     // this whole thing requires us to be in a Dial or Conference verb
     const task = this.currentTask;
@@ -1193,6 +1211,9 @@ ${credential.label ? `, label: ${credential.label}` : ''} is chosen`);
     if (opts.listen_status) {
       await this._lccListenStatus(opts);
     }
+    if (opts.transcribe_status) {
+      await this._lccTranscribeStatus(opts);
+    }
     else if (opts.mute_status) {
       await this._lccMuteStatus(callSid, opts.mute_status === 'mute');
     }
@@ -1369,6 +1390,10 @@ ${credential.label ? `, label: ${credential.label}` : ''} is chosen`);
         this._lccListenStatus(data);
         break;
 
+      case 'transcribe:status':
+        this._lccTranscribeStatus(data);
+        break;
+
       case 'whisper':
         this._lccWhisper(data, call_sid);
         break;

lib/tasks/transcribe.js

@@ -9,7 +9,8 @@ const {
   SonioxTranscriptionEvents,
   IbmTranscriptionEvents,
   NvidiaTranscriptionEvents,
-  JambonzTranscriptionEvents
+  JambonzTranscriptionEvents,
+  TranscribeStatus
 } = require('../utils/constants');
 const { normalizeJambones } = require('@jambonz/verb-specifications');
 const SttTask = require('./stt-task');
@@ -119,8 +120,7 @@ class TaskTranscribe extends SttTask {
     this.removeSpeechListeners(ep);
   }
 
-  async kill(cs) {
-    super.kill(cs);
+  async _stopTranscription() {
     let stopTranscription = false;
     if (this.ep?.connected) {
       stopTranscription = true;
@@ -132,6 +132,13 @@ class TaskTranscribe extends SttTask {
       this.ep2.stopTranscription({vendor: this.vendor})
         .catch((err) => this.logger.info(err, 'Error TaskTranscribe:kill'));
     }
+
+    return stopTranscription;
+  }
+
+  async kill(cs) {
+    super.kill(cs);
+    const stopTranscription = this._stopTranscription();
     // hangup after 1 sec if we don't get a final transcription
     if (stopTranscription) this._timer = setTimeout(() => this.notifyTaskDone(), 1500);
     else this.notifyTaskDone();
@@ -139,6 +146,23 @@ class TaskTranscribe extends SttTask {
     await this.awaitTaskDone();
   }
 
+  async updateTranscribe(status) {
+    if (!this.killed && this.ep && this.ep.connected) {
+      this.logger.info(`TaskTranscribe:updateTranscribe status ${status}`);
+      switch (status) {
+        case TranscribeStatus.Pause:
+          await this._stopTranscription();
+          break;
+        case TranscribeStatus.Resume:
+          await this._startTranscribing(this.cs, this.ep, 1);
+          if (this.separateRecognitionPerChannel && this.ep2) {
+            await this._startTranscribing(this.cs, this.ep2, 2);
+          }
+          break;
+      }
+    }
+  }
+
   async _startTranscribing(cs, ep, channel) {
     const opts = this.setChannelVarsForStt(this, this.sttCredentials, this.data.recognizer);
     switch (this.vendor) {

lib/utils/constants.json

@@ -51,6 +51,11 @@
     "Silence": "silence",
     "Resume": "resume"
   },
+  "TranscribeStatus": {
+    "Pause": "pause",
+    "Silence": "silence",
+    "Resume": "resume"
+  },
   "TaskPreconditions": {
     "None": "none",
     "Endpoint": "endpoint",

package-lock.json

@@ -30,7 +30,7 @@
         "bent": "^7.3.12",
         "debug": "^4.3.4",
         "deepcopy": "^2.1.0",
-        "drachtio-fsmrf": "^3.0.23",
+        "drachtio-fsmrf": "^3.0.24",
         "drachtio-srf": "^4.5.26",
         "express": "^4.18.2",
         "ip": "^1.1.8",
@@ -5152,9 +5152,9 @@
       }
     },
     "node_modules/drachtio-fsmrf": {
-      "version": "3.0.23",
-      "resolved": "https://registry.npmjs.org/drachtio-fsmrf/-/drachtio-fsmrf-3.0.23.tgz",
-      "integrity": "sha512-ElruNKuPzFiMOUH06PUd3dR9tElEGRhbP/gXxai58qhrqRQNLJxzCRJkbgbjrZdYWFQPHOAzy4ZQb7+qq0AUPw==",
+      "version": "3.0.24",
+      "resolved": "https://registry.npmjs.org/drachtio-fsmrf/-/drachtio-fsmrf-3.0.24.tgz",
+      "integrity": "sha512-MsTwHDOOg5yrKCcQ2buNTVSEwoMZzTcVeU5/BZ2Km0TPxBI1lErW6IiyZxA5pyoLfSdBtbqSh9ZwXrtxWTp6kA==",
       "dependencies": {
         "camel-case": "^4.1.2",
         "debug": "^2.6.9",
@@ -14610,9 +14610,9 @@
       }
     },
     "drachtio-fsmrf": {
-      "version": "3.0.23",
-      "resolved": "https://registry.npmjs.org/drachtio-fsmrf/-/drachtio-fsmrf-3.0.23.tgz",
-      "integrity": "sha512-ElruNKuPzFiMOUH06PUd3dR9tElEGRhbP/gXxai58qhrqRQNLJxzCRJkbgbjrZdYWFQPHOAzy4ZQb7+qq0AUPw==",
+      "version": "3.0.24",
+      "resolved": "https://registry.npmjs.org/drachtio-fsmrf/-/drachtio-fsmrf-3.0.24.tgz",
+      "integrity": "sha512-MsTwHDOOg5yrKCcQ2buNTVSEwoMZzTcVeU5/BZ2Km0TPxBI1lErW6IiyZxA5pyoLfSdBtbqSh9ZwXrtxWTp6kA==",
       "requires": {
         "camel-case": "^4.1.2",
         "debug": "^2.6.9",

package.json

@@ -46,7 +46,7 @@
     "bent": "^7.3.12",
     "debug": "^4.3.4",
     "deepcopy": "^2.1.0",
-    "drachtio-fsmrf": "^3.0.23",
+    "drachtio-fsmrf": "^3.0.24",
     "drachtio-srf": "^4.5.26",
     "express": "^4.18.2",
     "ip": "^1.1.8",