From ab83b219795806487f480329ed7973b6dbd024c0 Mon Sep 17 00:00:00 2001
From: Hoan Luu Huu <110280845+xquanluu@users.noreply.github.com>
Date: Fri, 27 Jun 2025 21:05:18 +0700
Subject: [PATCH] support inworld tts (#1262)

* support inworld tts

* wip

* wip
---
 lib/session/call-session.js | 7 +++++++
 lib/tasks/say.js            | 6 ++++++
 lib/tasks/tts-task.js       | 3 +++
 lib/utils/db-utils.js       | 6 ++++++
 package-lock.json           | 9 +++++----
 package.json                | 2 +-
 6 files changed, 28 insertions(+), 5 deletions(-)

diff --git a/lib/session/call-session.js b/lib/session/call-session.js
index fccc9b7d..edbf09de 100644
--- a/lib/session/call-session.js
+++ b/lib/session/call-session.js
@@ -1109,6 +1109,13 @@ class CallSession extends Emitter {
             options: credential.options
           };
         }
+        else if ('inworld' === vendor) {
+          return {
+            api_key: credential.api_key,
+            model_id: credential.model_id,
+            options: credential.options
+          };
+        }
         else if ('assemblyai' === vendor) {
           return {
             speech_credential_sid: credential.speech_credential_sid,
diff --git a/lib/tasks/say.js b/lib/tasks/say.js
index 587c98a9..b05043c8 100644
--- a/lib/tasks/say.js
+++ b/lib/tasks/say.js
@@ -397,6 +397,7 @@ class TaskSay extends TtsTask {
           .replace('playht_', 'playht.')
           .replace('cartesia_', 'cartesia.')
           .replace('rimelabs_', 'rimelabs.')
+          .replace('inworld_', 'inworld.')
           .replace('verbio_', 'verbio.')
           .replace('elevenlabs_', 'elevenlabs.');
         if (spanMapping[newKey]) newKey = spanMapping[newKey];
@@ -461,6 +462,11 @@ const spanMapping = {
   'rimelabs.name_lookup_time_ms': 'name_lookup_ms',
   'rimelabs.connect_time_ms': 'connect_ms',
   'rimelabs.final_response_time_ms': 'final_response_ms',
+  // inworld
+  'inworld.name_lookup_time_ms': 'name_lookup_ms',
+  'inworld.connect_time_ms': 'connect_ms',
+  'inworld.final_response_time_ms': 'final_response_ms',
+  'inworld.x_envoy_upstream_service_time': 'upstream_service_time',
   // verbio
   'verbio.name_lookup_time_ms': 'name_lookup_ms',
   'verbio.connect_time_ms': 'connect_ms',
diff --git a/lib/tasks/tts-task.js b/lib/tasks/tts-task.js
index f6d081fa..97d5ab3b 100644
--- a/lib/tasks/tts-task.js
+++ b/lib/tasks/tts-task.js
@@ -185,6 +185,9 @@ class TtsTask extends Task {
     } else if (vendor === 'rimelabs') {
       credentials = credentials || {};
       credentials.model_id = this.options.model_id || credentials.model_id;
+    } else if (vendor === 'inworld') {
+      credentials = credentials || {};
+      credentials.model_id = this.options.model_id || credentials.model_id;
     } else if (vendor === 'whisper') {
       credentials = credentials || {};
       credentials.model_id = this.options.model_id || credentials.model_id;
diff --git a/lib/utils/db-utils.js b/lib/utils/db-utils.js
index 73bc504b..a5b7889f 100644
--- a/lib/utils/db-utils.js
+++ b/lib/utils/db-utils.js
@@ -120,6 +120,12 @@ const speechMapper = (cred) => {
       obj.model_id = o.model_id;
       obj.options = o.options;
     }
+    else if ('inworld' === obj.vendor) {
+      const o = JSON.parse(decrypt(credential));
+      obj.api_key = o.api_key;
+      obj.model_id = o.model_id;
+      obj.options = o.options;
+    }
     else if ('assemblyai' === obj.vendor) {
       const o = JSON.parse(decrypt(credential));
       obj.api_key = o.api_key;
diff --git a/package-lock.json b/package-lock.json
index 868f2d8b..48345657 100644
--- a/package-lock.json
+++ b/package-lock.json
@@ -15,7 +15,7 @@
         "@jambonz/http-health-check": "^0.0.1",
         "@jambonz/mw-registrar": "^0.2.7",
         "@jambonz/realtimedb-helpers": "^0.8.13",
-        "@jambonz/speech-utils": "^0.2.12",
+        "@jambonz/speech-utils": "^0.2.13",
         "@jambonz/stats-collector": "^0.1.10",
         "@jambonz/time-series": "^0.2.13",
         "@jambonz/verb-specifications": "^0.0.105",
@@ -1466,9 +1466,10 @@
       }
     },
     "node_modules/@jambonz/speech-utils": {
-      "version": "0.2.12",
-      "resolved": "https://registry.npmjs.org/@jambonz/speech-utils/-/speech-utils-0.2.12.tgz",
-      "integrity": "sha512-1xik/ZRUtPE2SOztxweGI+RTXUbiUXRShJ8G/l7VJJBkSWbfKKerYIRfHicAPumHicaUrbqSzZ6hr0eghv80KA==",
+      "version": "0.2.13",
+      "resolved": "https://registry.npmjs.org/@jambonz/speech-utils/-/speech-utils-0.2.13.tgz",
+      "integrity": "sha512-8ISTWTfz3fWtPmzPDsZG8zgnf6pTjLA1WasMAF/d/ktGswqVsbhoPcDh5ZyZ7BsEqOMLMIv2Hn0ESmrBuMn5kw==",
+      "license": "MIT",
       "dependencies": {
         "23": "^0.0.0",
         "@aws-sdk/client-polly": "^3.496.0",
diff --git a/package.json b/package.json
index 343559b5..2b535e5b 100644
--- a/package.json
+++ b/package.json
@@ -31,7 +31,7 @@
     "@jambonz/http-health-check": "^0.0.1",
     "@jambonz/mw-registrar": "^0.2.7",
     "@jambonz/realtimedb-helpers": "^0.8.13",
-    "@jambonz/speech-utils": "^0.2.12",
+    "@jambonz/speech-utils": "^0.2.13",
     "@jambonz/stats-collector": "^0.1.10",
     "@jambonz/time-series": "^0.2.13",
     "@jambonz/verb-specifications": "^0.0.105",