feat: Add inputAudioTranscription support to Java ADK

Author: jinnigu

core/src/main/java/com/google/adk/agents/RunConfig.java

@@ -48,8 +48,12 @@ public enum StreamingMode {
 
   public abstract @Nullable AudioTranscriptionConfig outputAudioTranscription();
 
+  public abstract @Nullable AudioTranscriptionConfig inputAudioTranscription();
+
   public abstract int maxLlmCalls();
 
+  public abstract Builder toBuilder();
+
   public static Builder builder() {
     return new AutoValue_RunConfig.Builder()
         .setSaveInputBlobsAsArtifacts(false)
@@ -65,7 +69,8 @@ public static Builder builder(RunConfig runConfig) {
         .setMaxLlmCalls(runConfig.maxLlmCalls())
         .setResponseModalities(runConfig.responseModalities())
         .setSpeechConfig(runConfig.speechConfig())
-        .setOutputAudioTranscription(runConfig.outputAudioTranscription());
+        .setOutputAudioTranscription(runConfig.outputAudioTranscription())
+        .setInputAudioTranscription(runConfig.inputAudioTranscription());
   }
 
   /** Builder for {@link RunConfig}. */
@@ -88,6 +93,10 @@ public abstract static class Builder {
     public abstract Builder setOutputAudioTranscription(
         AudioTranscriptionConfig outputAudioTranscription);
 
+    @CanIgnoreReturnValue
+    public abstract Builder setInputAudioTranscription(
+        AudioTranscriptionConfig inputAudioTranscription);
+
     @CanIgnoreReturnValue
     public abstract Builder setMaxLlmCalls(int maxLlmCalls);
 

core/src/main/java/com/google/adk/flows/llmflows/Basic.java

@@ -48,6 +48,8 @@ public Single<RequestProcessor.RequestProcessingResult> processRequest(
         .ifPresent(liveConnectConfigBuilder::speechConfig);
     Optional.ofNullable(context.runConfig().outputAudioTranscription())
         .ifPresent(liveConnectConfigBuilder::outputAudioTranscription);
+    Optional.ofNullable(context.runConfig().inputAudioTranscription())
+        .ifPresent(liveConnectConfigBuilder::inputAudioTranscription);
 
     LlmRequest.Builder builder =
         request.toBuilder()

core/src/main/java/com/google/adk/runner/Runner.java

@@ -366,8 +366,9 @@ private Single<Session> emitStateDeltaEvent(
   private InvocationContext newInvocationContextForLive(
       Session session, Optional<LiveRequestQueue> liveRequestQueue, RunConfig runConfig) {
     RunConfig.Builder runConfigBuilder = RunConfig.builder(runConfig);
-    if (!CollectionUtils.isNullOrEmpty(runConfig.responseModalities())
-        && liveRequestQueue.isPresent()) {
+    if (liveRequestQueue.isPresent() && !this.agent.subAgents().isEmpty()) {
+      // Parity with Python: apply modality defaults and transcription settings
+      // only for multi-agent live scenarios.
       // Default to AUDIO modality if not specified.
       if (CollectionUtils.isNullOrEmpty(runConfig.responseModalities())) {
         runConfigBuilder.setResponseModalities(
@@ -380,6 +381,10 @@ private InvocationContext newInvocationContextForLive(
           runConfigBuilder.setOutputAudioTranscription(AudioTranscriptionConfig.builder().build());
         }
       }
+      // Need input transcription for agent transferring in live mode.
+      if (runConfig.inputAudioTranscription() == null) {
+        runConfigBuilder.setInputAudioTranscription(AudioTranscriptionConfig.builder().build());
+      }
     }
     return newInvocationContext(
         session, /* newMessage= */ Optional.empty(), liveRequestQueue, runConfigBuilder.build());

core/src/test/java/com/google/adk/agents/RunConfigTest.java

@@ -25,6 +25,7 @@ public void testBuilderWithVariousValues() {
             .setSaveInputBlobsAsArtifacts(true)
             .setStreamingMode(RunConfig.StreamingMode.SSE)
             .setOutputAudioTranscription(audioTranscriptionConfig)
+            .setInputAudioTranscription(audioTranscriptionConfig)
             .setMaxLlmCalls(10)
             .build();
 
@@ -33,6 +34,7 @@ public void testBuilderWithVariousValues() {
     assertThat(runConfig.saveInputBlobsAsArtifacts()).isTrue();
     assertThat(runConfig.streamingMode()).isEqualTo(RunConfig.StreamingMode.SSE);
     assertThat(runConfig.outputAudioTranscription()).isEqualTo(audioTranscriptionConfig);
+    assertThat(runConfig.inputAudioTranscription()).isEqualTo(audioTranscriptionConfig);
     assertThat(runConfig.maxLlmCalls()).isEqualTo(10);
   }
 
@@ -45,6 +47,7 @@ public void testBuilderDefaults() {
     assertThat(runConfig.saveInputBlobsAsArtifacts()).isFalse();
     assertThat(runConfig.streamingMode()).isEqualTo(RunConfig.StreamingMode.NONE);
     assertThat(runConfig.outputAudioTranscription()).isNull();
+    assertThat(runConfig.inputAudioTranscription()).isNull();
     assertThat(runConfig.maxLlmCalls()).isEqualTo(500);
   }
 
@@ -66,6 +69,7 @@ public void testBuilderWithDifferentValues() {
             .setSaveInputBlobsAsArtifacts(true)
             .setStreamingMode(RunConfig.StreamingMode.BIDI)
             .setOutputAudioTranscription(audioTranscriptionConfig)
+            .setInputAudioTranscription(audioTranscriptionConfig)
             .setMaxLlmCalls(20)
             .build();
 
@@ -74,6 +78,24 @@ public void testBuilderWithDifferentValues() {
     assertThat(runConfig.saveInputBlobsAsArtifacts()).isTrue();
     assertThat(runConfig.streamingMode()).isEqualTo(RunConfig.StreamingMode.BIDI);
     assertThat(runConfig.outputAudioTranscription()).isEqualTo(audioTranscriptionConfig);
+    assertThat(runConfig.inputAudioTranscription()).isEqualTo(audioTranscriptionConfig);
     assertThat(runConfig.maxLlmCalls()).isEqualTo(20);
   }
+
+  @Test
+  public void testInputAudioTranscriptionOnly() {
+    AudioTranscriptionConfig inputTranscriptionConfig = AudioTranscriptionConfig.builder().build();
+
+    RunConfig runConfig =
+        RunConfig.builder()
+            .setStreamingMode(RunConfig.StreamingMode.BIDI)
+            .setResponseModalities(ImmutableList.of(new Modality(Modality.Known.AUDIO)))
+            .setInputAudioTranscription(inputTranscriptionConfig)
+            .build();
+
+    assertThat(runConfig.inputAudioTranscription()).isEqualTo(inputTranscriptionConfig);
+    assertThat(runConfig.outputAudioTranscription()).isNull();
+    assertThat(runConfig.streamingMode()).isEqualTo(RunConfig.StreamingMode.BIDI);
+    assertThat(runConfig.responseModalities()).containsExactly(new Modality(Modality.Known.AUDIO));
+  }
 }

core/src/test/java/com/google/adk/flows/llmflows/BasicTest.java

@@ -220,13 +220,33 @@ public void processRequest_buildsLiveConnectConfigFromRunConfig_outputAudioTrans
     assertThat(result.events()).isEmpty();
   }
 
+  @Test
+  public void processRequest_buildsLiveConnectConfigFromRunConfig_inputAudioTranscription() {
+    RunConfig runConfig =
+        RunConfig.builder().setInputAudioTranscription(TEST_AUDIO_TRANSCRIPTION_CONFIG).build();
+    LlmAgent agentWithConfig = LlmAgent.builder().name("agentWithConfig").model(testLlm).build();
+    InvocationContext contextWithRunConfig = createInvocationContext(agentWithConfig, runConfig);
+
+    RequestProcessingResult result =
+        basicProcessor.processRequest(contextWithRunConfig, initialRequest).blockingGet();
+
+    LlmRequest updatedRequest = result.updatedRequest();
+    assertThat(updatedRequest.liveConnectConfig()).isNotNull();
+    assertThat(updatedRequest.liveConnectConfig().responseModalities().get()).isEmpty();
+    assertThat(updatedRequest.liveConnectConfig().speechConfig()).isEmpty();
+    assertThat(updatedRequest.liveConnectConfig().inputAudioTranscription())
+        .hasValue(TEST_AUDIO_TRANSCRIPTION_CONFIG);
+    assertThat(result.events()).isEmpty();
+  }
+
   @Test
   public void processRequest_buildsLiveConnectConfigFromRunConfig_allFields() {
     RunConfig runConfig =
         RunConfig.builder()
             .setResponseModalities(ImmutableList.of(new Modality(Modality.Known.AUDIO)))
             .setSpeechConfig(TEST_SPEECH_CONFIG)
             .setOutputAudioTranscription(TEST_AUDIO_TRANSCRIPTION_CONFIG)
+            .setInputAudioTranscription(TEST_AUDIO_TRANSCRIPTION_CONFIG)
             .build();
     LlmAgent agentWithConfig = LlmAgent.builder().name("agentWithConfig").model(testLlm).build();
     InvocationContext contextWithRunConfig = createInvocationContext(agentWithConfig, runConfig);
@@ -241,6 +261,8 @@ public void processRequest_buildsLiveConnectConfigFromRunConfig_allFields() {
     assertThat(updatedRequest.liveConnectConfig().speechConfig()).hasValue(TEST_SPEECH_CONFIG);
     assertThat(updatedRequest.liveConnectConfig().outputAudioTranscription())
         .hasValue(TEST_AUDIO_TRANSCRIPTION_CONFIG);
+    assertThat(updatedRequest.liveConnectConfig().inputAudioTranscription())
+        .hasValue(TEST_AUDIO_TRANSCRIPTION_CONFIG);
     assertThat(result.events()).isEmpty();
   }
 }