Support ASR v1.9 #2465

* Append epd param at ASR.Recognize event
nugu-developers · Mar 15, 2024 · 5b64fe7 · 5b64fe7
1 parent d97320d
commit 5b64fe7
Show file tree

Hide file tree

Showing 3 changed files with 12 additions and 2 deletions.
diff --git a/nugu-agent/src/main/java/com/skt/nugu/sdk/agent/DefaultASRAgent.kt b/nugu-agent/src/main/java/com/skt/nugu/sdk/agent/DefaultASRAgent.kt
@@ -78,7 +78,7 @@ class DefaultASRAgent(
         private const val TAG = "DefaultASRAgent"
 
         const val NAMESPACE = "ASR"
-        val VERSION = Version(1,8)
+        val VERSION = Version(1,9)
 
         const val NAME_EXPECT_SPEECH = "ExpectSpeech"
         const val NAME_RECOGNIZE = "Recognize"

diff --git a/nugu-agent/src/main/java/com/skt/nugu/sdk/agent/asr/AsrRecognizeEventPayload.kt b/nugu-agent/src/main/java/com/skt/nugu/sdk/agent/asr/AsrRecognizeEventPayload.kt
@@ -28,6 +28,7 @@ data class AsrRecognizeEventPayload(
     private val encoding: String? = null,
     private val wakeup: PayloadWakeup? = null,
     private val timeout: Timeout? = null,
+    private val epd: EndPointDetectorParam? = null,
     private val service: JsonObject? = null,
     private val requestType: RequestType? = null
 ) {
@@ -106,6 +107,14 @@ data class AsrRecognizeEventPayload(
             })
         }
 
+        epd?.let {
+            add("epd", JsonObject().apply {
+                addProperty("timeoutMilliseconds", it.timeoutInSeconds * 1000L)
+                addProperty("silenceIntervalInMilliseconds", it.pauseLengthInMilliseconds)
+                addProperty("maxSpeechDurationMilliseconds", it.maxDurationInSeconds * 1000L)
+            })
+        }
+
         service?.let {
             add("service", it)
         }
@@ -114,4 +123,4 @@ data class AsrRecognizeEventPayload(
             addProperty("requestType", it.name)
         }
     }.toString()
-}
+}
diff --git a/nugu-agent/src/main/java/com/skt/nugu/sdk/agent/asr/impl/DefaultClientSpeechRecognizer.kt b/nugu-agent/src/main/java/com/skt/nugu/sdk/agent/asr/impl/DefaultClientSpeechRecognizer.kt
@@ -160,6 +160,7 @@ class DefaultClientSpeechRecognizer(
                 endpointing = AsrRecognizeEventPayload.ENDPOINTING_CLIENT,
                 encoding = if (enablePartialResult) AsrRecognizeEventPayload.ENCODING_PARTIAL else AsrRecognizeEventPayload.ENCODING_COMPLETE,
                 wakeup = payloadWakeup,
+                epd = epdParam,
                 service = jsonService,
                 requestType = requestType
             ).toJsonString()