update speech-utils (#1324)

* Sending synthesized-audio notification for servedFromCache as well
https://github.com/jambonz/jambonz-feature-server/issues/1319

* Sending back the id that was set, to track the synthesized-audio
e.g if we send a say verb having 100, it's synthesized-audio event will return 100 in the data to correleate the say verb and synthesized-audio event

lib/session/call-session.js

@@ -1146,6 +1146,13 @@ class CallSession extends Emitter {
             options: credential.options
           };
         }
+        else if ('resemble' === vendor) {
+          return {
+            api_key: credential.api_key,
+            resemble_tts_use_tls: credential.resemble_tts_use_tls,
+            resemble_tts_uri: credential.resemble_tts_uri,
+          };
+        }
         else if ('inworld' === vendor) {
           return {
             api_key: credential.api_key,
@@ -1963,6 +1970,17 @@ Duration=${duration} `
       }
     } catch (err) {
       this.logger.info({err, opts, callSid}, 'CallSession:updateCall - error updating call');
+      const {writeAlerts} = this.srf.locals;
+      try {
+        writeAlerts({
+          alert_type: 'error-updating-call',
+          account_sid: this.accountSid,
+          message: err.message,
+          target_sid: callSid
+        });
+      } catch (err) {
+        this.logger.error({err}, 'Error writing error-updating-call alert');
+      }
     }
   }
 
@@ -2367,6 +2385,7 @@ Duration=${duration} `
       const ep = await this._createMediaEndpoint({
         headers: {
           'X-Jambones-Call-ID': this.callId,
+          'X-Call-Sid': this.callSid,
         },
         remoteSdp: this.req.body
       });

lib/session/confirm-call-session.js

@@ -8,7 +8,8 @@ const CallSession = require('./call-session');
 
  */
 class ConfirmCallSession extends CallSession {
-  constructor({logger, application, dlg, ep, tasks, callInfo, accountInfo, memberId, confName, rootSpan, req}) {
+  // eslint-disable-next-line max-len
+  constructor({logger, application, dlg, ep, tasks, callInfo, accountInfo, memberId, confName, rootSpan, req, tmpFiles}) {
     super({
       logger,
       application,
@@ -24,6 +25,7 @@ class ConfirmCallSession extends CallSession {
     this.dlg = dlg;
     this.ep = ep;
     this.req = req;
+    this.tmpFiles = tmpFiles;
   }
 
   /**

lib/tasks/conference.js

@@ -673,7 +673,8 @@ class Conference extends Task {
         confName: this.confName,
         tasks,
         rootSpan: cs.rootSpan,
-        req: cs.req
+        req: cs.req,
+        tmpFiles: cs.tmpFiles,
       });
       await this._playSession.exec();
       this._playSession = null;

lib/tasks/dial.js

@@ -1098,7 +1098,8 @@ class TaskDial extends Task {
               accountInfo: this.cs.accountInfo,
               tasks,
               rootSpan: this.cs.rootSpan,
-              req: this.cs.req
+              req: this.cs.req,
+              tmpFiles: this.cs.tmpFiles,
             });
             await this._onHoldSession.exec();
             this._onHoldSession = null;

lib/tasks/enqueue.js

@@ -370,7 +370,8 @@ class TaskEnqueue extends Task {
         accountInfo: cs.accountInfo,
         tasks: tasksToRun,
         rootSpan: cs.rootSpan,
-        req: cs.req
+        req: cs.req,
+        tmpFiles: cs.tmpFiles,
       });
       await this._playSession.exec();
       this._playSession = null;

lib/tasks/listen.js

@@ -72,7 +72,7 @@ class TaskListen extends Task {
     } catch (err) {
       this.logger.info(err, `TaskListen:exec - error ${this.url}`);
     }
-    if (this.transcribeTask) this.transcribeTask.kill();
+    if (this.transcribeTask) this.transcribeTask.kill(cs);
     this._removeListeners(ep);
   }
 

lib/tasks/say.js

@@ -452,6 +452,7 @@ class TaskSay extends TtsTask {
           .replace('playht_', 'playht.')
           .replace('cartesia_', 'cartesia.')
           .replace('rimelabs_', 'rimelabs.')
+          .replace('resemble_', 'resemble.')
           .replace('inworld_', 'inworld.')
           .replace('verbio_', 'verbio.')
           .replace('elevenlabs_', 'elevenlabs.');
@@ -517,6 +518,9 @@ const spanMapping = {
   'rimelabs.name_lookup_time_ms': 'name_lookup_ms',
   'rimelabs.connect_time_ms': 'connect_ms',
   'rimelabs.final_response_time_ms': 'final_response_ms',
+  // Resemble
+  'resemble.connect_time_ms': 'connect_ms',
+  'resemble.final_response_time_ms': 'final_response_ms',
   // inworld
   'inworld.name_lookup_time_ms': 'name_lookup_ms',
   'inworld.connect_time_ms': 'connect_ms',

lib/tasks/tts-task.js

@@ -293,7 +293,18 @@ class TtsTask extends Task {
                 vendor,
                 language,
                 characters: text.length,
-                elapsedTime: rtt
+                elapsedTime: rtt,
+                servedFromCache,
+                'id': this.id
+              });
+            }
+            if (servedFromCache) {
+              this.notifyStatus({
+                event: 'synthesized-audio',
+                vendor,
+                language,
+                servedFromCache,
+                'id': this.id
               });
             }
           }

lib/utils/amd-utils.js

@@ -281,13 +281,17 @@ module.exports = (logger) => {
 
     /* set stt options */
     logger.info(`starting amd for vendor ${vendor} and language ${language}`);
-    const sttOpts = amd.setChannelVarsForStt({name: TaskName.Gather}, sttCredentials, language, {
-      vendor,
-      hints,
-      enhancedModel: true,
-      altLanguages: opts.recognizer?.altLanguages || [],
-      initialSpeechTimeoutMs: opts.resolveTimeoutMs,
-    });
+    /* if opts contains recognizer object use that config for stt, otherwise use defaults */
+    const rOpts = opts.recognizer ?
+      opts.recognizer :
+      {
+        vendor,
+        hints,
+        enhancedModel: true,
+        altLanguages: opts.recognizer?.altLanguages || [],
+        initialSpeechTimeoutMs: opts.resolveTimeoutMs,
+      };
+    const sttOpts = amd.setChannelVarsForStt({name: TaskName.Gather}, sttCredentials, language, rOpts);
 
     await ep.set(sttOpts).catch((err) => logger.info(err, 'Error setting channel variables'));
 

lib/utils/background-task-manager.js

@@ -65,7 +65,7 @@ class BackgroundTaskManager extends Emitter {
       this.logger.info(`stopping background task: ${type}`);
       task.removeAllListeners();
       task.span.end();
-      task.kill();
+      task.kill(this.cs);
       // Remove task from managed List
       this.tasks.delete(type);
     }

lib/utils/db-utils.js

@@ -124,6 +124,12 @@ const speechMapper = (cred) => {
       obj.model_id = o.model_id;
       obj.options = o.options;
     }
+    else if ('resemble' === obj.vendor) {
+      const o = JSON.parse(decrypt(credential));
+      obj.api_key = o.api_key;
+      obj.resemble_tts_use_tls = o.resemble_tts_use_tls;
+      obj.resemble_tts_uri = o.resemble_tts_uri;
+    }
     else if ('inworld' === obj.vendor) {
       const o = JSON.parse(decrypt(credential));
       obj.api_key = o.api_key;

lib/utils/http-requestor.js

@@ -41,7 +41,7 @@ class HttpRequestor extends BaseRequestor {
   constructor(logger, account_sid, hook, secret) {
     super(logger, account_sid, hook, secret);
 
-    this.method = hook.method || 'POST';
+    this.method = hook.method?.toUpperCase() || 'POST';
     this.authHeader = basicAuth(hook.username, hook.password);
     this.backoffMs = 500;
 
@@ -111,7 +111,7 @@ class HttpRequestor extends BaseRequestor {
 
     const payload = params ? snakeCaseKeys(params, ['customerData', 'sip', 'env_vars', 'args']) : null;
     const url = hook.url || hook;
-    const method = hook.method || 'POST';
+    const method = hook.method?.toUpperCase() || 'POST';
     let buf = '';
     httpHeaders = {
       ...httpHeaders,
@@ -119,7 +119,7 @@ class HttpRequestor extends BaseRequestor {
     };
 
     assert.ok(url, 'HttpRequestor:request url was not provided');
-    assert.ok, (['GET', 'POST'].includes(method), `HttpRequestor:request method must be 'GET' or 'POST' not ${method}`);
+    assert.ok(['GET', 'POST'].includes(method), `HttpRequestor:request method must be 'GET' or 'POST' not ${method}`);
     const startAt = process.hrtime();
 
     /* if we have an absolute url, and it is ws then do a websocket connection */

lib/utils/place-outdial.js

@@ -401,7 +401,8 @@ class SingleDialer extends Emitter {
         accountInfo: this.accountInfo,
         tasks,
         rootSpan: this.rootSpan,
-        req: this.req
+        req: this.req,
+        tmpFiles: cs.tmpFiles,
       });
       await cs.exec();
 

lib/utils/stt-latency-calculator.js

@@ -101,8 +101,6 @@ class SttLatencyCalculator extends Emitter {
       }
       this.isRunning = false;
       this.logger.info('STT Latency Calculator stopped');
-    } else {
-      this.logger.warn('Latency calculator is not running, no VAD detection to stop');
     }
   }
 
@@ -116,7 +114,7 @@ class SttLatencyCalculator extends Emitter {
       return;
     }
     this._startVad();
-    this.logger.info('STT Latency Calculator started');
+    this.logger.debug('STT Latency Calculator started');
   }
 
   stop() {

lib/utils/transcription-utils.js

@@ -785,12 +785,15 @@ module.exports = (logger) => {
           AWS_ACCESS_KEY_ID: sttCredentials.accessKeyId,
           AWS_SECRET_ACCESS_KEY: sttCredentials.secretAccessKey,
           AWS_REGION: sttCredentials.region,
-          AWS_SECURITY_TOKEN: sttCredentials.securityToken
+          AWS_SECURITY_TOKEN: sttCredentials.securityToken,
+          AWS_SESSION_TOKEN: sttCredentials.sessionToken ? sttCredentials.sessionToken : sttCredentials.securityToken
         }),
         ...(awsOptions.accessKey && {AWS_ACCESS_KEY_ID: awsOptions.accessKey}),
         ...(awsOptions.secretKey && {AWS_SECRET_ACCESS_KEY: awsOptions.secretKey}),
         ...(awsOptions.region && {AWS_REGION: awsOptions.region}),
         ...(awsOptions.securityToken && {AWS_SECURITY_TOKEN: awsOptions.securityToken}),
+        ...(awsOptions.sessionToken && {AWS_SESSION_TOKEN: awsOptions.sessionToken ?
+          awsOptions.sessionToken : awsOptions.securityToken}),
         ...(awsOptions.languageModelName && {AWS_LANGUAGE_MODEL_NAME: awsOptions.languageModelName}),
         ...(awsOptions.piiEntityTypes?.length && {AWS_PII_ENTITY_TYPES: awsOptions.piiEntityTypes.join(',')}),
         ...(awsOptions.piiIdentifyEntities && {AWS_PII_IDENTIFY_ENTITIES: true}),
@@ -891,6 +894,14 @@ module.exports = (logger) => {
       const deepgramUri = deepgramOptions.deepgramSttUri || sttCredentials.deepgram_stt_uri;
       const useTls = deepgramOptions.deepgramSttUseTls || sttCredentials.deepgram_stt_use_tls;
 
+      // DH (2025-08-11) entity_prompt is currently limited to 100 words
+      const entityPrompt = deepgramOptions.entityPrompt ?
+        deepgramOptions.entityPrompt
+          .split(/\s+/)
+          .slice(0, 100)
+          .join(' ')
+        : undefined;
+
       /* default to a sensible model if not supplied */
       if (!model) {
         model = selectDefaultDeepgramModel(task, language);
@@ -949,7 +960,9 @@ module.exports = (logger) => {
         ...(deepgramOptions.fillerWords) &&
           {DEEPGRAM_SPEECH_FILLER_WORDS: deepgramOptions.fillerWords},
         ...((Array.isArray(deepgramOptions.keyterms) && deepgramOptions.keyterms.length > 0) &&
-          {DEEPGRAM_SPEECH_KEYTERMS: deepgramOptions.keyterms.join(',')})
+          {DEEPGRAM_SPEECH_KEYTERMS: deepgramOptions.keyterms.join(',')}),
+        ...(deepgramOptions.mipOptOut && {DEEPGRAM_SPEECH_MIP_OPT_OUT: deepgramOptions.mipOptOut}),
+        ...(entityPrompt && {DEEPGRAM_SPEECH_ENTITY_PROMPT: entityPrompt}),
       };
     }
     else if ('deepgramriver' === vendor) {

package.json

@@ -27,14 +27,14 @@
   "dependencies": {
     "@aws-sdk/client-auto-scaling": "^3.549.0",
     "@aws-sdk/client-sns": "^3.549.0",
-    "@jambonz/db-helpers": "^0.9.12",
+    "@jambonz/db-helpers": "^0.9.16",
     "@jambonz/http-health-check": "^0.0.1",
     "@jambonz/mw-registrar": "^0.2.7",
-    "@jambonz/realtimedb-helpers": "^0.8.13",
-    "@jambonz/speech-utils": "^0.2.15",
+    "@jambonz/realtimedb-helpers": "^0.8.15",
+    "@jambonz/speech-utils": "^0.2.19",
     "@jambonz/stats-collector": "^0.1.10",
     "@jambonz/time-series": "^0.2.14",
-    "@jambonz/verb-specifications": "^0.0.111",
+    "@jambonz/verb-specifications": "^0.0.113",
     "@modelcontextprotocol/sdk": "^1.9.0",
     "@opentelemetry/api": "^1.8.0",
     "@opentelemetry/exporter-jaeger": "^1.23.0",