迁移音频生成与语音合成到 gateway 并补充 simulation 测试

2026-06-07 10:26:57 +08:00
parent 78ab867a9f
commit dc14866210
22 changed files with 2475 additions and 55 deletions
@@ -2,6 +2,7 @@ package clients

 import (
 	"context"
+	"encoding/base64"
 	"encoding/json"
 	"net/http"
 	"net/http/httptest"
@@ -65,6 +66,35 @@ func TestSimulationClientReturnsVideoDemoAssets(t *testing.T) {
 	}
 }

+func TestSimulationClientReturnsAudioDemoAssets(t *testing.T) {
+	response, err := (SimulationClient{}).Run(context.Background(), Request{
+		Kind:      "speech.generations",
+		ModelType: "text_to_speech",
+		Model:     "speech-2.6-turbo",
+		Body: map[string]any{
+			"text":                 "hello from simulation",
+			"voice_id":             "female-shaonv",
+			"count":                2,
+			"simulationDurationMs": 5,
+		},
+		Candidate: store.RuntimeModelCandidate{Provider: "simulation"},
+	})
+	if err != nil {
+		t.Fatalf("run simulation audio client: %v", err)
+	}
+	data, _ := response.Result["data"].([]any)
+	if len(data) != 2 || response.Result["status"] != "success" {
+		t.Fatalf("unexpected simulated audio response: %+v", response.Result)
+	}
+	item, _ := data[0].(map[string]any)
+	if item["type"] != "audio" || item["url"] != "/static/simulation/audio.wav" || item["audio_url"] != "/static/simulation/audio.wav" {
+		t.Fatalf("unexpected simulated audio item: %+v", item)
+	}
+	if item["revised_text"] != "hello from simulation" || item["assetSource"] != "simulation" {
+		t.Fatalf("unexpected simulated audio metadata: %+v", item)
+	}
+}
+
 func TestSimulationDurationDefaultsByMediaType(t *testing.T) {
 	imageDuration := simulationDuration(Request{Kind: "images.generations"})
 	if imageDuration < 10*time.Second || imageDuration > 30*time.Second {
@@ -74,12 +104,84 @@ func TestSimulationDurationDefaultsByMediaType(t *testing.T) {
 	if videoDuration < 2*time.Minute || videoDuration > 3*time.Minute {
 		t.Fatalf("video simulation duration should default to 2-3m, got %s", videoDuration)
 	}
+	audioDuration := simulationDuration(Request{Kind: "speech.generations"})
+	if audioDuration < 2*time.Second || audioDuration > 6*time.Second {
+		t.Fatalf("audio simulation duration should default to 2-6s, got %s", audioDuration)
+	}
 	textDuration := simulationDuration(Request{Kind: "chat.completions"})
 	if textDuration < 800*time.Millisecond || textDuration > 2400*time.Millisecond {
 		t.Fatalf("text simulation duration should keep short defaults, got %s", textDuration)
 	}
 }

+func TestMinimaxClientSpeechUsesT2AV2AndNormalizesAudio(t *testing.T) {
+	var captured map[string]any
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if r.Method != http.MethodPost || r.URL.Path != "/t2a_v2" {
+			t.Fatalf("unexpected request: %s %s", r.Method, r.URL.String())
+		}
+		if got := r.Header.Get("Authorization"); got != "Bearer test-key" {
+			t.Fatalf("unexpected auth header: %q", got)
+		}
+		if err := json.NewDecoder(r.Body).Decode(&captured); err != nil {
+			t.Fatalf("decode request: %v", err)
+		}
+		w.Header().Set("x-request-id", "req-minimax-speech")
+		_ = json.NewEncoder(w).Encode(map[string]any{
+			"data":      map[string]any{"audio": "68656c6c6f"},
+			"base_resp": map[string]any{"status_code": 0},
+		})
+	}))
+	defer server.Close()
+
+	response, err := (MinimaxClient{HTTPClient: server.Client()}).Run(context.Background(), Request{
+		Kind:  "speech.generations",
+		Model: "MiniMax Speech 2.6 Turbo",
+		Body: map[string]any{
+			"text":     "hello",
+			"voice_id": "female-shaonv",
+			"speed":    1.2,
+			"vol":      0.8,
+			"pitch":    -1,
+			"emotion":  "happy",
+		},
+		Candidate: store.RuntimeModelCandidate{
+			Provider:          "minimax",
+			BaseURL:           server.URL,
+			ProviderModelName: "speech-2.6-turbo",
+			Credentials:       map[string]any{"apiKey": "test-key"},
+		},
+	})
+	if err != nil {
+		t.Fatalf("run minimax speech client: %v", err)
+	}
+	if captured["model"] != "speech-2.6-turbo" || captured["text"] != "hello" {
+		t.Fatalf("unexpected minimax speech payload: %+v", captured)
+	}
+	if _, ok := captured["voice_id"]; ok {
+		t.Fatalf("voice_id should be moved into voice_setting: %+v", captured)
+	}
+	voiceSetting, ok := captured["voice_setting"].(map[string]any)
+	if !ok {
+		t.Fatalf("missing voice_setting: %+v", captured)
+	}
+	if voiceSetting["voice_id"] != "female-shaonv" || voiceSetting["speed"] != 1.2 || voiceSetting["vol"] != 0.8 || voiceSetting["pitch"] != float64(-1) || voiceSetting["emotion"] != "happy" {
+		t.Fatalf("unexpected voice_setting: %+v", voiceSetting)
+	}
+	data, _ := response.Result["data"].([]any)
+	if len(data) != 1 {
+		t.Fatalf("unexpected minimax speech response: %+v", response.Result)
+	}
+	item, _ := data[0].(map[string]any)
+	expectedContent := "data:audio/mpeg;base64," + base64.StdEncoding.EncodeToString([]byte("hello"))
+	if item["type"] != "audio" || item["content"] != expectedContent || item["mime_type"] != "audio/mpeg" {
+		t.Fatalf("unexpected normalized audio item: %+v", item)
+	}
+	if response.RequestID != "req-minimax-speech" {
+		t.Fatalf("unexpected request id: %q", response.RequestID)
+	}
+}
+
 func TestSimulationDurationCanBeControlledByParams(t *testing.T) {
 	fixedDuration := simulationDuration(Request{Body: map[string]any{"simulationDurationSeconds": 7}})
 	if fixedDuration != 7*time.Second {