update drachtio-srf (#1515)

* escape tag data in listen

* deep copy call data for listen

lib/session/call-session.js

@@ -2028,7 +2028,7 @@ Duration=${duration} `
         return this._lccDub(opts.dub, callSid);
       }
       else if (opts.boostAudioSignal) {
-        return this._lccBoostAudioSignal(opts, callSid);
+        return this._lccBoostAudioSignal(opts.boostAudioSignal, callSid);
       }
       else if (opts.media_path) {
         return this._lccMediaPath(opts.media_path, callSid);

lib/tasks/dial.js

@@ -195,6 +195,9 @@ class TaskDial extends Task {
   async exec(cs) {
     await super.exec(cs);
 
+    /* capture whether A leg was already answered before this dial task started */
+    this._aLegAlreadyAnswered = !!cs.dlg;
+
     if (this.data.anchorMedia && this.data.exitMediaPath) {
       this.logger.info('Dial:exec - incompatible anchorMedia and exitMediaPath are both set, will obey anchorMedia');
       delete this.data.exitMediaPath;
@@ -550,7 +553,7 @@ class TaskDial extends Task {
     let sbcAddress = this.proxy || getSBC();
     const teamsInfo = {};
     let fqdn;
-    const forwardPAI = this.forwardPAI ?? JAMBONZ_DIAL_PAI_HEADER; // dial verb overides env var
+    const forwardPAI = this.forwardPAI ?? !JAMBONZ_DIAL_PAI_HEADER; // dial verb overides env var
     this.logger.debug(forwardPAI, 'forwardPAI value');
     if (!sbcAddress) throw new Error('no SBC found for outbound call');
     this.headers = {
@@ -872,8 +875,12 @@ class TaskDial extends Task {
     this.sd = sd;
     this.callSid = sd.callSid;
     if (this.earlyMedia) {
-      debug('Dial:_selectSingleDial propagating answer supervision on A leg now that B is connected');
-      await cs.propagateAnswer();
+      if (this._aLegAlreadyAnswered) {
+        debug('Dial:_selectSingleDial A leg was already answered, skipping propagateAnswer');
+      } else {
+        debug('Dial:_selectSingleDial propagating answer supervision on A leg now that B is connected');
+        await cs.propagateAnswer();
+      }
     }
     if (this.timeLimit) {
       this.timerMaxCallDuration = setTimeout(this._onMaxCallDuration.bind(this, cs), this.timeLimit * 1000);

lib/tasks/listen.js

@@ -152,9 +152,17 @@ class TaskListen extends Task {
 
   async _startListening(cs, ep) {
     this._initListeners(ep);
-    const ci = this.nested ? this.parentTask.sd.callInfo : cs.callInfo.toJSON();
+    const tempci = this.nested ? this.parentTask.sd.callInfo : cs.callInfo.toJSON();
+    const ci = structuredClone(tempci);
     if (this._ignoreCustomerData) {
       delete ci.customerData;
+    } else {
+      for (const key in ci.customerData) {
+        if (ci.customerData.hasOwnProperty(key)) {
+          const value = ci.customerData[key];
+          ci.customerData[key] = typeof value === 'string' ? escapeString(value) : value;
+        }
+      }
     }
     const metadata = Object.assign(
       {sampleRate: this.sampleRate, mixType: this.mixType},

lib/tasks/llm/llms/google_s2s.js

@@ -36,6 +36,9 @@ class TaskLlmGoogle_S2S extends Task {
     this.model = this.parent.model || 'models/gemini-2.0-flash-live-001';
     this.auth = this.parent.auth;
     this.connectionOptions = this.parent.connectOptions;
+    const {host, version} = this.connectionOptions || {};
+    this.host = host;
+    this.version = version;
 
     const {apiKey} = this.auth || {};
     if (!apiKey) throw new Error('auth.apiKey is required for Google S2S');
@@ -46,7 +49,7 @@ class TaskLlmGoogle_S2S extends Task {
     this.eventHook = this.data.eventHook;
     this.toolHook = this.data.toolHook;
 
-    const {setup} = this.data.llmOptions;
+    const {setup, sessionResumption} = this.data.llmOptions;
 
     if (typeof setup !== 'object') {
       throw new Error('llmOptions with an initial setup is required for Google S2S');
@@ -54,6 +57,7 @@ class TaskLlmGoogle_S2S extends Task {
     this.setup = {
       ...setup,
       model: this.model,
+      ...(sessionResumption && {sessionResumption}),
       // make sure output is always audio
       generationConfig: {
         ...(setup.generationConfig || {}),
@@ -138,6 +142,10 @@ class TaskLlmGoogle_S2S extends Task {
 
     try {
       const args = [ep.uuid, 'session.create', this.apiKey];
+      if (this.host) {
+        args.push(this.host);
+        if (this.version) args.push(this.version);
+      }
       await this._api(ep, args);
     } catch (err) {
       this.logger.error({err}, 'TaskLlmGoogle_S2S:_startListening');

lib/tasks/tts-task.js

@@ -157,6 +157,13 @@ class TtsTask extends Task {
           ...(reduceLatency && {RIMELABS_TTS_STREAMING_REDUCE_LATENCY: reduceLatency})
         };
         break;
+      case 'google':
+        obj = {
+          GOOGLE_TTS_LANGUAGE_CODE: language,
+          GOOGLE_TTS_VOICE_NAME: voice,
+          GOOGLE_APPLICATION_CREDENTIALS: JSON.stringify(credentials.credentials)
+        };
+        break;
       default:
         if (vendor.startsWith('custom:')) {
           const use_tls = custom_tts_streaming_url.startsWith('wss://');

lib/utils/constants.json

@@ -311,6 +311,11 @@
     "ConnectFailure": "deepgram_tts_streaming::connect_failed",
     "Connect": "deepgram_tts_streaming::connect"
   },
+  "GoogleTtsStreamingEvents": {
+    "Empty": "google_tts_streaming::empty",
+    "ConnectFailure": "google_tts_streaming::connect_failed",
+    "Connect": "google_tts_streaming::connect"
+  },
   "CartesiaTtsStreamingEvents": {
     "Empty": "cartesia_tts_streaming::empty",
     "ConnectFailure": "cartesia_tts_streaming::connect_failed",

lib/utils/transcription-utils.js

@@ -1310,6 +1310,9 @@ module.exports = (logger) => {
           ...(openaiOptions.turn_detection.silence_duration_ms && {
             OPENAI_TURN_DETECTION_SILENCE_DURATION_MS: openaiOptions.turn_detection.silence_duration_ms
           }),
+          ...(openaiOptions.turn_detection.eagerness && {
+            OPENAI_TURN_DETECTION_EAGERNESS: openaiOptions.turn_detection.eagerness
+          })
         };
       }
     }
@@ -1375,7 +1378,9 @@ module.exports = (logger) => {
             speechmaticsOptions.transcription_config.audio_filtering_config.volume_threshold}),
         ...(speechmaticsOptions.transcription_config?.transcript_filtering_config?.remove_disfluencies &&
           {SPEECHMATICS_REMOVE_DISFLUENCIES:
-            speechmaticsOptions.transcription_config.transcript_filtering_config.remove_disfluencies})
+            speechmaticsOptions.transcription_config.transcript_filtering_config.remove_disfluencies}),
+        SPEECHMATICS_END_OF_UTTERANCE_SILENCE_TRIGGER:
+          speechmaticsOptions.transcription_config?.conversation_config?.end_of_utterance_silence_trigger || 0.5
       };
     }
     else if (vendor.startsWith('custom:')) {

lib/utils/tts-streaming-buffer.js

@@ -421,6 +421,7 @@ class TtsStreamingBuffer extends Emitter {
       'cartesia',
       'elevenlabs',
       'rimelabs',
+      'google',
       'custom'
     ].forEach((vendor) => {
       const eventClassName = `${vendor.charAt(0).toUpperCase() + vendor.slice(1)}TtsStreamingEvents`;

package.json

@@ -49,7 +49,7 @@
     "debug": "^4.3.4",
     "deepcopy": "^2.1.0",
     "drachtio-fsmrf": "^4.1.2",
-    "drachtio-srf": "^5.0.14",
+    "drachtio-srf": "^5.0.18",
     "express": "^4.19.2",
     "express-validator": "^7.0.1",
     "moment": "^2.30.1",