patch fix for #962

* include X-CID for dial outbound if the call-session is outbound

* include X-CID for dial outbound if the call-session is outbound

* include X-CID for dial outbound if the call-session is outbound

* include X-CID for dial outbound if the call-session is outbound

lib/http-routes/api/create-call.js

@@ -293,6 +293,8 @@ router.post('/',
           },
           cbProvisional: (prov) => {
             const callStatus = prov.body ? CallStatus.EarlyMedia : CallStatus.Ringing;
+            // Update call-id for sbc outbound INVITE
+            cs.callInfo.sbcCallid = prov.get('X-CID');
             if ([180, 183].includes(prov.status) && prov.body) connectStream(prov.body);
             restDial.emit('callStatus', prov.status, !!prov.body);
             cs.emit('callStatusChange', {callStatus, sipStatus: prov.status});

lib/middleware.js

@@ -385,7 +385,7 @@ module.exports = function(srf, logger) {
     const {rootSpan, siprec, application:app} = req.locals;
     let span;
     try {
-      if (app.tasks && !JAMBONES_MYSQL_REFRESH_TTL) {
+      if (app.tasks && app.tasks?.length > 0 && !JAMBONES_MYSQL_REFRESH_TTL) {
         app.tasks = normalizeJambones(logger, app.tasks).map((tdata) => makeTask(logger, tdata));
         if (0 === app.tasks.length) throw new Error('no application provided');
         return next();

lib/session/call-info.js

@@ -32,6 +32,7 @@ class CallInfo {
       this.sipStatus = 100;
       this.sipReason = 'Trying';
       this.callStatus = CallStatus.Trying;
+      this.sbcCallid = req.get('X-CID');
       this.originatingSipIp = req.get('X-Forwarded-For');
       this.originatingSipTrunkName = req.get('X-Originating-Carrier');
       const {siprec} = req.locals;
@@ -129,6 +130,7 @@ class CallInfo {
       from: this.from,
       to: this.to,
       callId: this.callId,
+      sbcCallid: this.sbcCallid,
       sipStatus: this.sipStatus,
       sipReason: this.sipReason,
       callStatus: this.callStatus,

lib/session/call-session.js

@@ -1590,17 +1590,29 @@ Duration=${duration} `
   }
 
   _lccToolOutput(tool_call_id, opts, callSid) {
-    // this whole thing requires us to be in a Dial verb
+    // only valid if we are in an LLM verb
     const task = this.currentTask;
     if (!task || !task.name.startsWith('Llm')) {
       return this.logger.info('CallSession:_lccToolOutput - invalid command since we are not in an llm');
     }
 
-    task.processToolOutput(tool_call_id, opts)
+    task.processToolOutput(tool_call_id, opts, callSid)
       .catch((err) => this.logger.error(err, 'CallSession:_lccToolOutput'));
   }
 
 
+  _lccLlmUpdate(opts, callSid) {
+    // only valid if we are in an LLM verb
+    const task = this.currentTask;
+    if (!task || !task.name.startsWith('Llm')) {
+      return this.logger.info('CallSession:_lccLlmUpdate - invalid command since we are not in an llm');
+    }
+
+    task.processLlmUpdate(opts, callSid)
+      .catch((err) => this.logger.error(err, 'CallSession:_lccLlmUpdate'));
+  }
+
+
   /**
    * perform call hangup by jambonz
    */
@@ -1660,6 +1672,12 @@ Duration=${duration} `
     else if (opts.boostAudioSignal) {
       return this._lccBoostAudioSignal(opts, callSid);
     }
+    else if (opts.llm_tool_output) {
+      return this._lccToolOutput(opts.tool_call_id, opts.llm_tool_output, callSid);
+    }
+    else if (opts.llm_update) {
+      return this._lccLlmUpdate(opts.llm_update, callSid);
+    }
 
     // whisper may be the only thing we are asked to do, or it may that
     // we are doing a whisper after having muted, paused recording etc..
@@ -1961,6 +1979,10 @@ Duration=${duration} `
         this._lccToolOutput(tool_call_id, data, call_sid);
         break;
 
+      case 'llm:update':
+        this._lccLlmUpdate(data, call_sid);
+        break;
+
       default:
         this.logger.info(`CallSession:_onCommand - invalid command ${command}`);
     }

lib/session/inbound-call-session.js

@@ -72,6 +72,8 @@ class InboundCallSession extends CallSession {
 
   _jambonzHangup() {
     this.dlg?.destroy();
+    // kill current task or wakeup the call session.
+    this._callReleased();
   }
 
   _hangup(terminatedBy = 'jambonz') {

lib/tasks/dial.js

@@ -340,15 +340,17 @@ class TaskDial extends Task {
 
         const to = parseUri(req.getParsedHeader('Refer-To').uri);
         const by = parseUri(req.getParsedHeader('Referred-By').uri);
+        const referredBy = req.get('Referred-By');
+        const userAgent = req.get('User-Agent');
         this.logger.info({to}, 'refer to parsed');
         const json = await cs.requestor.request('verb:hook', this.referHook, {
           ...(callInfo.toJSON()),
           refer_details: {
             sip_refer_to: req.get('Refer-To'),
-            sip_referred_by: req.get('Referred-By'),
-            sip_user_agent: req.get('User-Agent'),
             refer_to_user: to.scheme === 'tel' ? to.number : to.user,
-            referred_by_user: by.scheme === 'tel' ? by.number : by.user,
+            ...(referredBy && {sip_referred_by: referredBy}),
+            ...(userAgent && {sip_user_agent: userAgent}),
+            ...(by && {referred_by_user: by.scheme === 'tel' ? by.number : by.user}),
             referring_call_sid,
             referred_call_sid
           }
@@ -379,6 +381,7 @@ class TaskDial extends Task {
         res.send(202);
         this.logger.info('DialTask:handleRefer - sent 202 Accepted');
       } catch (err) {
+        this.logger.info({err}, 'DialTask:handleRefer - error processing incoming REFER');
         res.send(err.statusCode || 501);
       }
     }
@@ -484,7 +487,7 @@ class TaskDial extends Task {
   }
 
   async _attemptCalls(cs) {
-    const {req, srf} = cs;
+    const {req, callInfo, direction, srf} = cs;
     const {getSBC} = srf.locals;
     const {lookupTeamsByAccount, lookupAccountBySid} = srf.locals.dbHelpers;
     const {lookupCarrier, lookupCarrierByPhoneNumber} = dbUtils(this.logger, cs.srf);
@@ -496,8 +499,11 @@ class TaskDial extends Task {
     this.headers = {
       'X-Account-Sid': cs.accountSid,
       ...(req && req.has('X-CID') && {'X-CID': req.get('X-CID')}),
-      ...(req && req.has('P-Asserted-Identity') && !JAMBONZ_DISABLE_DIAL_PAI_HEADER &&
-        {'P-Asserted-Identity': req.get('P-Asserted-Identity')}),
+      ...(direction === 'outbound' && callInfo.sbcCallid && {'X-CID': callInfo.sbcCallid}),
+      ...(!JAMBONZ_DISABLE_DIAL_PAI_HEADER && req && {
+        ...(req.has('P-Asserted-Identity') && {'P-Asserted-Identity': req.get('P-Asserted-Identity')}),
+        ...(req.has('Privacy') && {'Privacy': req.get('Privacy')}),
+      }),
       ...(req && req.has('X-Voip-Carrier-Sid') && {'X-Voip-Carrier-Sid': req.get('X-Voip-Carrier-Sid')}),
       // Put headers at the end to make sure opt.headers override all default behavior.
       ...this.headers
@@ -612,6 +618,7 @@ class TaskDial extends Task {
                 dialCallStatus: obj.callStatus,
                 dialSipStatus: obj.sipStatus,
                 dialCallSid: sd.callSid,
+                dialSbcCallid: sd.callInfo.sbcCallid
               });
             }
             switch (obj.callStatus) {

lib/tasks/gather.js

@@ -924,7 +924,7 @@ class TaskGather extends SttTask {
         }
       }
       // If transcription received, reset timeout timer.
-      if (this._timeoutTimer) {
+      if (this._timeoutTimer && !emptyTranscript) {
         this._startTimer();
       }
       /* restart asr timer if we get a partial transcript (only if the asr timer is already running) */

lib/tasks/llm/index.js

@@ -79,7 +79,18 @@ class TaskLlm extends Task {
     this.llm.processToolOutput(this.ep, tool_call_id, data);
   }
 
-
+  async processLlmUpdate(data, callSid) {
+    if (this.ep.connected) {
+      if (typeof this.llm.processLlmUpdate === 'function') {
+        this.llm.processLlmUpdate(this.ep, data, callSid);
+      }
+      else {
+        const {vendor, model} = this.llm;
+        this.logger.info({data, callSid},
+          `TaskLlm:_processLlmUpdate: LLM ${vendor}:${model} does not support llm:update`);
+      }
+    }
+  }
 }
 
 module.exports = TaskLlm;

lib/tasks/llm/llms/openai_s2s.js

@@ -2,6 +2,7 @@ const Task = require('../../task');
 const TaskName = 'Llm_OpenAI_s2s';
 const {LlmEvents_OpenAI} = require('../../../utils/constants');
 const ClientEvent = 'client.event';
+const SessionDelete = 'session.delete';
 
 const openai_server_events = [
   'error',
@@ -125,6 +126,13 @@ class TaskLlmOpenAI_S2S extends Task {
     }
   }
 
+  async _api(ep, args) {
+    const res = await ep.api('uuid_openai_s2s', `^^|${args.join('|')}`);
+    if (!res.body?.startsWith('+OK')) {
+      throw new Error({args}, `Error calling uuid_openai_s2s: ${res.body}`);
+    }
+  }
+
   async exec(cs, {ep}) {
     await super.exec(cs);
 
@@ -140,26 +148,57 @@ class TaskLlmOpenAI_S2S extends Task {
 
   async kill(cs) {
     super.kill(cs);
+
+    this._api(cs.ep, [cs.ep.uuid, SessionDelete])
+      .catch((err) => this.logger.info({err}, 'TaskLlmOpenAI_S2S:kill - error deleting session'));
+
     this.notifyTaskDone();
   }
 
+  /**
+   * Send function call output to the OpenAI server in the form of conversation.item.create
+   * per https://platform.openai.com/docs/guides/realtime/function-calls
+   */
   async processToolOutput(ep, tool_call_id, data) {
     try {
       this.logger.debug({tool_call_id, data}, 'TaskLlmOpenAI_S2S:processToolOutput');
 
-      await this._api(ep, [ep.uuid, ClientEvent, JSON.stringify(data)]);
+      if (!data.type || data.type !== 'conversation.item.create') {
+        this.logger.info({data},
+          'TaskLlmOpenAI_S2S:processToolOutput - invalid tool output, must be conversation.item.create');
+      }
+      else {
+        await this._api(ep, [ep.uuid, ClientEvent, JSON.stringify(data)]);
 
-      // send immediate response.create per https://platform.openai.com/docs/guides/realtime/function-calls
-      await this._api(ep, [ep.uuid, ClientEvent, JSON.stringify({type: 'response.create'})]);
+        // spec also recommends to send immediate response.create
+        await this._api(ep, [ep.uuid, ClientEvent, JSON.stringify({type: 'response.create'})]);
+      }
     } catch (err) {
       this.logger.info({err}, 'TaskLlmOpenAI_S2S:processToolOutput');
     }
   }
 
-  async _api(ep, args) {
-    const res = await ep.api('uuid_openai_s2s', `^^|${args.join('|')}`);
-    if (!res.body?.startsWith('+OK')) {
-      throw new Error({args}, `Error calling uuid_openai_s2s: ${res.body}`);
+  /**
+   * Send a session.update to the OpenAI server
+   * Note: creating and deleting conversation items also supported as well as interrupting the assistant
+   */
+  async processLlmUpdate(ep, data, _callSid) {
+    try {
+      this.logger.debug({data, _callSid}, 'TaskLlmOpenAI_S2S:processLlmUpdate');
+
+      if (!data.type || ![
+        'session.update',
+        'conversation.item.create',
+        'conversation.item.delete',
+        'response.cancel'
+      ].includes(data.type)) {
+        this.logger.info({data}, 'TaskLlmOpenAI_S2S:processLlmUpdate - invalid mid-call request');
+      }
+      else {
+        await this._api(ep, [ep.uuid, ClientEvent, JSON.stringify(data)]);
+      }
+    } catch (err) {
+      this.logger.info({err}, 'TaskLlmOpenAI_S2S:processLlmUpdate');
     }
   }
 

lib/tasks/say.js

@@ -177,32 +177,22 @@ class TaskSay extends TtsTask {
                 account_sid,
                 alert_type: AlertType.TTS_FAILURE,
                 vendor,
-                detail: evt.variable_tts_error
+                detail: evt.variable_tts_error,
+                target_sid
               }).catch((err) => this.logger.info({err}, 'Error generating alert for no tts'));
             }
-            else {
-              this.logger.debug({evt}, 'Say got playback-stop');
-              if (evt.variable_tts_error) {
-                writeAlerts({
-                  account_sid,
-                  alert_type: AlertType.TTS_FAILURE,
-                  vendor,
-                  detail: evt.variable_tts_error,
-                  target_sid
-                }).catch((err) => this.logger.info({err}, 'Error generating alert for no tts'));
-              }
-              if (evt.variable_tts_cache_filename && !this.killed) {
-                const text = parseTextFromSayString(this.text[segment]);
-                addFileToCache(evt.variable_tts_cache_filename, {
-                  account_sid,
-                  vendor,
-                  language,
-                  voice,
-                  engine,
-                  text
-                }).catch((err) => this.logger.info({err}, 'Error adding file to cache'));
-              }
+            if (evt.variable_tts_cache_filename && !this.killed) {
+              const text = parseTextFromSayString(this.text[segment]);
+              addFileToCache(evt.variable_tts_cache_filename, {
+                account_sid,
+                vendor,
+                language,
+                voice,
+                engine,
+                text
+              }).catch((err) => this.logger.info({err}, 'Error adding file to cache'));
             }
+
             if (this._playResolve) {
               evt.variable_tts_error ? this._playReject(new Error(evt.variable_tts_error)) : this._playResolve();
             }

lib/utils/place-outdial.js

@@ -213,6 +213,8 @@ class SingleDialer extends Emitter {
         },
         cbProvisional: (prov) => {
           const status = {sipStatus: prov.status, sipReason: prov.reason};
+          // Update call-id for sbc outbound INVITE
+          this.callInfo.sbcCallid = prov.get('X-CID');
           if ([180, 183].includes(prov.status) && prov.body) {
             if (status.callStatus !== CallStatus.EarlyMedia) {
               status.callStatus = CallStatus.EarlyMedia;

lib/utils/transcription-utils.js

@@ -161,7 +161,12 @@ const selectDefaultDeepgramModel = (task, language) => {
 
 const optimalGoogleModels = {
   'v1' : {
-    'en-IN':['telephony', 'latest_long']
+    'en-IN':['telephony', 'telephony'],
+    'es-DO':['default', 'default'],
+    'es-MX':['default', 'default'],
+    'en-AU':['telephony', 'telephony'],
+    'en-GB':['telephony', 'telephony'],
+    'en-NZ':['telephony', 'telephony']
   },
   'v2' : {
     'en-IN':['telephony', 'long']

package-lock.json

@@ -15,7 +15,7 @@
         "@jambonz/http-health-check": "^0.0.1",
         "@jambonz/mw-registrar": "^0.2.7",
         "@jambonz/realtimedb-helpers": "^0.8.8",
-        "@jambonz/speech-utils": "^0.1.18",
+        "@jambonz/speech-utils": "^0.1.20",
         "@jambonz/stats-collector": "^0.1.10",
         "@jambonz/time-series": "^0.2.9",
         "@jambonz/verb-specifications": "^0.0.83",
@@ -1539,10 +1539,9 @@
       }
     },
     "node_modules/@jambonz/speech-utils": {
-      "version": "0.1.18",
-      "resolved": "https://registry.npmjs.org/@jambonz/speech-utils/-/speech-utils-0.1.18.tgz",
-      "integrity": "sha512-GlcPvUIKcyiiH4cfUPXyYZtP1HIIdFbrqYUmeTmeBaOuZUrJ0xW+TAp/pbysh54vgPnAfcS43Y3ciULx0S3IjQ==",
-      "license": "MIT",
+      "version": "0.1.20",
+      "resolved": "https://registry.npmjs.org/@jambonz/speech-utils/-/speech-utils-0.1.20.tgz",
+      "integrity": "sha512-3Ff9zLcFoVZhrI4jBKyjgWpv/fEMx1BpJP85daRwZNC2S0BFshULJ54fQxc8S63IsdgFf6g1cLD5VxqPaqCfbQ==",
       "dependencies": {
         "@aws-sdk/client-polly": "^3.496.0",
         "@aws-sdk/client-sts": "^3.496.0",
@@ -10637,9 +10636,9 @@
       }
     },
     "@jambonz/speech-utils": {
-      "version": "0.1.18",
-      "resolved": "https://registry.npmjs.org/@jambonz/speech-utils/-/speech-utils-0.1.18.tgz",
-      "integrity": "sha512-GlcPvUIKcyiiH4cfUPXyYZtP1HIIdFbrqYUmeTmeBaOuZUrJ0xW+TAp/pbysh54vgPnAfcS43Y3ciULx0S3IjQ==",
+      "version": "0.1.20",
+      "resolved": "https://registry.npmjs.org/@jambonz/speech-utils/-/speech-utils-0.1.20.tgz",
+      "integrity": "sha512-3Ff9zLcFoVZhrI4jBKyjgWpv/fEMx1BpJP85daRwZNC2S0BFshULJ54fQxc8S63IsdgFf6g1cLD5VxqPaqCfbQ==",
       "requires": {
         "@aws-sdk/client-polly": "^3.496.0",
         "@aws-sdk/client-sts": "^3.496.0",

package.json

@@ -31,7 +31,7 @@
     "@jambonz/http-health-check": "^0.0.1",
     "@jambonz/mw-registrar": "^0.2.7",
     "@jambonz/realtimedb-helpers": "^0.8.8",
-    "@jambonz/speech-utils": "^0.1.18",
+    "@jambonz/speech-utils": "^0.1.20",
     "@jambonz/stats-collector": "^0.1.10",
     "@jambonz/time-series": "^0.2.9",
     "@jambonz/verb-specifications": "^0.0.83",