Adding custom models for Play/Recognize (#35328)

* Using the latest swagger

* Updating files

* Removing imports

* Moving to official swagger branch

* Fixing test

* Fixing tests

* Updating latest swagger

* Updated latest swagger and live tests

* Updating swagger location

* Adding custom models for Play/Recognize

* Fixing SSML prompt for Recognize

Author: cochi2

sdk/communication/azure-communication-callautomation/src/main/java/com/azure/communication/callautomation/CallMediaAsync.java

@@ -278,11 +278,10 @@ PlayRequest getPlayRequest(PlayOptions options) {
 
     private PlaySourceInternal getPlaySourceInternalFromFileSource(FileSource playSource) {
         FileSourceInternal fileSourceInternal = new FileSourceInternal().setUri(playSource.getUrl());
-        PlaySourceInternal playSourceInternal = new PlaySourceInternal()
+        return new PlaySourceInternal()
             .setSourceType(PlaySourceTypeInternal.FILE)
             .setFileSource(fileSourceInternal)
             .setPlaySourceId(playSource.getPlaySourceId());
-        return playSourceInternal;
     }
 
     private PlaySourceInternal getPlaySourceInternalFromTextSource(TextSource playSource) {
@@ -296,21 +295,27 @@ private PlaySourceInternal getPlaySourceInternalFromTextSource(TextSource playSo
         if (playSource.getVoiceName() != null) {
             textSourceInternal.setVoiceName(playSource.getVoiceName());
         }
+        if (playSource.getCustomVoiceEndpointId() != null) {
+            textSourceInternal.setCustomVoiceEndpointId(playSource.getCustomVoiceEndpointId());
+        }
 
-        PlaySourceInternal playSourceInternal = new PlaySourceInternal()
+        return new PlaySourceInternal()
             .setSourceType(PlaySourceTypeInternal.TEXT)
             .setTextSource(textSourceInternal)
             .setPlaySourceId(playSource.getPlaySourceId());
-        return playSourceInternal;
     }
 
     private PlaySourceInternal getPlaySourceInternalFromSsmlSource(SsmlSource playSource) {
         SsmlSourceInternal ssmlSourceInternal = new SsmlSourceInternal().setSsmlText(playSource.getSsmlText());
-        PlaySourceInternal playSourceInternal = new PlaySourceInternal()
+
+        if (playSource.getCustomVoiceEndpointId() != null) {
+            ssmlSourceInternal.setCustomVoiceEndpointId(playSource.getCustomVoiceEndpointId());
+        }
+
+        return new PlaySourceInternal()
             .setSourceType(PlaySourceTypeInternal.SSML)
             .setSsmlSource(ssmlSourceInternal)
             .setPlaySourceId(playSource.getPlaySourceId());
-        return playSourceInternal;
     }
 
     private PlaySourceInternal convertPlaySourceToPlaySourceInternal(PlaySource playSource) {
@@ -319,6 +324,8 @@ private PlaySourceInternal convertPlaySourceToPlaySourceInternal(PlaySource play
             playSourceInternal = getPlaySourceInternalFromFileSource((FileSource) playSource);
         } else if (playSource instanceof TextSource) {
             playSourceInternal = getPlaySourceInternalFromTextSource((TextSource) playSource);
+        } else if (playSource instanceof SsmlSource) {
+            playSourceInternal = getPlaySourceInternalFromSsmlSource((SsmlSource) playSource);
         }
         return playSourceInternal;
     }
@@ -398,6 +405,12 @@ private RecognizeRequest getRecognizeRequestFromChoiceConfiguration(CallMediaRec
             }
         }
 
+        if (choiceRecognizeOptions.getSpeechModelEndpointId() != null) {
+            if (!choiceRecognizeOptions.getSpeechModelEndpointId().isEmpty()) {
+                recognizeOptionsInternal.setSpeechRecognitionModelEndpointId(choiceRecognizeOptions.getSpeechModelEndpointId());
+            }
+        }
+
         PlaySourceInternal playSourceInternal = getPlaySourceInternalFromRecognizeOptions(recognizeOptions);
 
         RecognizeRequest recognizeRequest = new RecognizeRequest()
@@ -422,6 +435,12 @@ private RecognizeRequest getRecognizeRequestFromSpeechConfiguration(CallMediaRec
 
         recognizeOptionsInternal.setInitialSilenceTimeoutInSeconds((int) speechRecognizeOptions.getInitialSilenceTimeout().getSeconds());
 
+        if (speechRecognizeOptions.getSpeechModelEndpointId() != null) {
+            if (!speechRecognizeOptions.getSpeechModelEndpointId().isEmpty()) {
+                recognizeOptionsInternal.setSpeechRecognitionModelEndpointId(speechRecognizeOptions.getSpeechModelEndpointId());
+            }
+        }
+
         PlaySourceInternal playSourceInternal = getPlaySourceInternalFromRecognizeOptions(recognizeOptions);
 
         RecognizeRequest recognizeRequest = new RecognizeRequest()
@@ -459,6 +478,11 @@ private RecognizeRequest getRecognizeRequestFromSpeechOrDtmfConfiguration(CallMe
             .setInterruptPrompt(speechOrDtmfRecognizeOptions.isInterruptPrompt())
             .setTargetParticipant(CommunicationIdentifierConverter.convert(speechOrDtmfRecognizeOptions.getTargetParticipant()));
 
+        if (speechOrDtmfRecognizeOptions.getSpeechModelEndpointId() != null) {
+            if (!speechOrDtmfRecognizeOptions.getSpeechModelEndpointId().isEmpty()) {
+                recognizeOptionsInternal.setSpeechRecognitionModelEndpointId(speechOrDtmfRecognizeOptions.getSpeechModelEndpointId());
+            }
+        }
         recognizeOptionsInternal.setInitialSilenceTimeoutInSeconds((int) speechOrDtmfRecognizeOptions.getInitialSilenceTimeout().getSeconds());
 
         PlaySourceInternal playSourceInternal = getPlaySourceInternalFromRecognizeOptions(recognizeOptions);

sdk/communication/azure-communication-callautomation/src/main/java/com/azure/communication/callautomation/models/CallMediaRecognizeOptions.java

@@ -54,6 +54,12 @@ public abstract class CallMediaRecognizeOptions {
     @JsonProperty(value = "initialSilenceTimeout")
     private Duration initialSilenceTimeout;
 
+    /*
+     * Endpoint where the custom model was deployed.
+     */
+    @JsonProperty(value = "speechModelEndpointId")
+    private String speechModelEndpointId;
+
     /*
      * Target participant of DTFM tone recognition.
      */
@@ -224,4 +230,24 @@ public CallMediaRecognizeOptions setInitialSilenceTimeout(Duration initialSilenc
     public CommunicationIdentifier getTargetParticipant() {
         return this.targetParticipant;
     }
+
+    /**
+     * Get the speech model endpoint id.
+     *
+     * @return the speech model endpoint id.
+     */
+    public String getSpeechModelEndpointId() {
+        return speechModelEndpointId;
+    }
+
+    /**
+     * Set the speechModelEndpointId property: Endpoint where the custom model was deployed.
+     *
+     * @param speechModelEndpointId the initialSilenceTimeout value to set.
+     * @return the CallMediaRecognizeSpeechOrDtmfOptions object itself.
+     */
+    public CallMediaRecognizeOptions setSpeechModelEndpointId(String speechModelEndpointId) {
+        this.speechModelEndpointId = speechModelEndpointId;
+        return this;
+    }
 }

sdk/communication/azure-communication-callautomation/src/main/java/com/azure/communication/callautomation/models/SsmlSource.java

@@ -15,6 +15,12 @@ public final class SsmlSource extends PlaySource {
     @JsonProperty(value = "ssmlText", required = true)
     private String ssmlText;
 
+    /*
+     * Endpoint where the Custom Voice was deployed.
+     */
+    @JsonProperty(value = "customVoiceEndpointId")
+    private String customVoiceEndpointId;
+
     /**
      * Get the ssmlText property: Ssml string for the cognitive service to be played.
      *
@@ -34,4 +40,24 @@ public SsmlSource setSsmlText(String ssmlText) {
         this.ssmlText = ssmlText;
         return this;
     }
+
+    /**
+     * Get the customVoiceEndpointId property: Endpoint where the custom voice was deployed.
+     *
+     * @return the customVoiceEndpointId value.
+     */
+    public String getCustomVoiceEndpointId() {
+        return this.customVoiceEndpointId;
+    }
+
+    /**
+     * Set the customVoiceEndpointId property: Endpoint where the custom voice was deployed.
+     *
+     * @param customVoiceEndpointId the customVoiceEndpointId value to set.
+     * @return the TextSourceInternal object itself.
+     */
+    public SsmlSource setCustomVoiceEndpointId(String customVoiceEndpointId) {
+        this.customVoiceEndpointId = customVoiceEndpointId;
+        return this;
+    }
 }

sdk/communication/azure-communication-callautomation/src/main/java/com/azure/communication/callautomation/models/TextSource.java

@@ -33,6 +33,12 @@ public final class TextSource extends PlaySource {
     @JsonProperty(value = "voiceName")
     private String voiceName;
 
+    /*
+     * Endpoint where the Custom Voice was deployed.
+     */
+    @JsonProperty(value = "customVoiceEndpointId")
+    private String customVoiceEndpointId;
+
     /**
      * Get the text property: Text for the cognitive service to be played.
      *
@@ -112,4 +118,24 @@ public TextSource setVoiceName(String voiceName) {
         this.voiceName = voiceName;
         return this;
     }
+
+    /**
+     * Get the customVoiceEndpointId property: Endpoint where the custom voice was deployed.
+     *
+     * @return the customVoiceEndpointId value.
+     */
+    public String getCustomVoiceEndpointId() {
+        return this.customVoiceEndpointId;
+    }
+
+    /**
+     * Set the customVoiceEndpointId property: Endpoint where the custom voice was deployed.
+     *
+     * @param customVoiceEndpointId the customVoiceEndpointId value to set.
+     * @return the TextSourceInternal object itself.
+     */
+    public TextSource setCustomVoiceEndpointId(String customVoiceEndpointId) {
+        this.customVoiceEndpointId = customVoiceEndpointId;
+        return this;
+    }
 }

sdk/communication/azure-communication-callautomation/src/test/java/com/azure/communication/callautomation/CallMediaAsyncUnitTests.java

@@ -58,9 +58,11 @@ public void setup() {
         playTextSource.setVoiceGender(GenderType.MALE);
         playTextSource.setSourceLocale("en-US");
         playTextSource.setVoiceName("LULU");
+        playTextSource.setCustomVoiceEndpointId("customVoiceEndpointId");
 
         playSsmlSource = new SsmlSource();
-        playSsmlSource.setSsmlText("<speak></speak>");
+        playSsmlSource.setSsmlText("<speak><voice name=\"LULU\"></voice></speak>");
+        playSsmlSource.setCustomVoiceEndpointId("customVoiceEndpointId");
     }
 
     @Test
@@ -260,6 +262,7 @@ public void recognizeWithResponseWithFileSourceChoiceOptions() {
         recognizeOptions.setInterruptPrompt(true);
         recognizeOptions.setInitialSilenceTimeout(Duration.ofSeconds(4));
         recognizeOptions.setSpeechLanguage("en-US");
+        recognizeOptions.setSpeechModelEndpointId("customModelEndpointId");
 
         StepVerifier.create(
                 callMedia.startRecognizingWithResponse(recognizeOptions))
@@ -306,6 +309,7 @@ public void recognizeWithResponseTextSpeechOptions() {
         recognizeOptions.setOperationContext("operationContext");
         recognizeOptions.setInterruptPrompt(true);
         recognizeOptions.setInitialSilenceTimeout(Duration.ofSeconds(4));
+        recognizeOptions.setSpeechModelEndpointId("customModelEndpointId");
 
         StepVerifier.create(
                 callMedia.startRecognizingWithResponse(recognizeOptions))
@@ -325,6 +329,7 @@ public void recognizeWithResponseTextSpeechOrDtmfOptions() {
         recognizeOptions.setOperationContext("operationContext");
         recognizeOptions.setInterruptPrompt(true);
         recognizeOptions.setInitialSilenceTimeout(Duration.ofSeconds(4));
+        recognizeOptions.setSpeechModelEndpointId("customModelEndpointId");
 
         StepVerifier.create(
                 callMedia.startRecognizingWithResponse(recognizeOptions))

sdk/communication/azure-communication-callautomation/src/test/java/com/azure/communication/callautomation/CallMediaUnitTests.java

@@ -45,6 +45,7 @@ public void setup() {
         playTextSource.setVoiceGender(GenderType.MALE);
         playTextSource.setSourceLocale("en-US");
         playTextSource.setVoiceName("LULU");
+        playTextSource.setCustomVoiceEndpointId("customVoiceEndpointId");
     }
 
     @Test