feat: add multi-speaker voice config

PiperOrigin-RevId: 759431774

Author: MarkDaoust

live_converters.go

@@ -91,6 +91,35 @@ func TestLiveConnect(t *testing.T) {
 			},
 			wantRequestBody: `{"setup":{"generationConfig":{"temperature":0.5},"model":"models/test-model","systemInstruction":{"parts":[{"text":"test instruction"}]},"tools":[{"googleSearch":{}}]}}`,
 		},
+		{
+			desc:   "Fail if multispeaker config.",
+			client: mldevClient,
+			config: &LiveConnectConfig{
+				SpeechConfig: &SpeechConfig{
+					MultiSpeakerVoiceConfig: &MultiSpeakerVoiceConfig{
+						SpeakerVoiceConfigs: []*SpeakerVoiceConfig{
+							{
+								Speaker: "Alice",
+								VoiceConfig: &VoiceConfig{
+									PrebuiltVoiceConfig: &PrebuiltVoiceConfig{VoiceName: "kore"},
+								},
+							},
+							{
+								Speaker: "Bob",
+								VoiceConfig: &VoiceConfig{
+									PrebuiltVoiceConfig: &PrebuiltVoiceConfig{VoiceName: "puck"},
+								},
+							},
+						},
+					},
+				},
+				Temperature:       Ptr[float32](0.5),
+				SystemInstruction: &Content{Parts: []*Part{{Text: "test instruction"}}},
+				Tools:             []*Tool{{GoogleSearch: &GoogleSearch{}}},
+			},
+			wantErr:        true,
+			wantErrMessage: "multiSpeakerVoiceConfig is not supported",
+		},
 		{
 			desc:            "successful connection with http options mldev",
 			client:          mldevClient,

live_test.go

@@ -149,6 +149,61 @@ func TestModelsGenerateContentAudio(t *testing.T) {
 	}
 }
 
+func TestModelsGenerateContentMultiSpeakerVoiceConfigAudio(t *testing.T) {
+	if *mode != apiMode {
+		t.Skip("Skip. This test is only in the API mode")
+	}
+	ctx := context.Background()
+	for _, backend := range backends {
+		t.Run(backend.name, func(t *testing.T) {
+			t.Parallel()
+			if isDisabledTest(t) {
+				t.Skip("Skip: disabled test")
+			}
+			client, err := NewClient(ctx, &ClientConfig{Backend: backend.Backend})
+			if err != nil {
+				t.Fatal(err)
+			}
+			config := &GenerateContentConfig{
+				ResponseModalities: []string{"AUDIO"},
+				SpeechConfig: &SpeechConfig{
+					MultiSpeakerVoiceConfig: &MultiSpeakerVoiceConfig{
+						SpeakerVoiceConfigs: []*SpeakerVoiceConfig{
+							{
+								Speaker: "Alice",
+								VoiceConfig: &VoiceConfig{
+									PrebuiltVoiceConfig: &PrebuiltVoiceConfig{
+										VoiceName: "Aoede",
+									},
+								},
+							},
+							{
+								Speaker: "Bob",
+								VoiceConfig: &VoiceConfig{
+									PrebuiltVoiceConfig: &PrebuiltVoiceConfig{
+										VoiceName: "Kore",
+									},
+								},
+							},
+						},
+					},
+					LanguageCode: "en-US",
+				},
+			}
+			result, err := client.Models.GenerateContent(ctx, "gemini-2.0-flash", Text("say something nice to me"), config)
+			if err != nil {
+				t.Errorf("GenerateContent failed unexpectedly: %v", err)
+			}
+			if result == nil {
+				t.Fatalf("expected at least one response, got none")
+			}
+			if len(result.Candidates) == 0 {
+				t.Errorf("expected at least one candidate, got none")
+			}
+		})
+	}
+}
+
 func TestModelsGenerateVideosText2VideoPoll(t *testing.T) {
 	if *mode != apiMode {
 		t.Skip("Skip. This test is only in the API mode")

models.go

@@ -125,6 +125,19 @@ func tSpeechConfig(_ *apiClient, speechConfig any) (any, error) {
 	return speechConfig, nil
 }
 
+func tLiveSpeechConfig(_ *apiClient, speechConfig any) (any, error) {
+	switch config := speechConfig.(type) {
+	case map[string]any:
+		if _, ok := config["multiSpeakerVoiceConfig"]; ok {
+			return nil, fmt.Errorf("multiSpeakerVoiceConfig is not supported in the live API")
+		}
+		return config, nil
+	case nil:
+		return nil, nil
+	default:
+		return nil, fmt.Errorf("unsupported speechConfig type: %T", speechConfig)
+	}
+}
 func tBytes(_ *apiClient, fromImageBytes any) (any, error) {
 	// TODO(b/389133914): Remove dummy bytes converter.
 	return fromImageBytes, nil