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

apps/api/internal/runner/pricing.go

@@ -112,6 +112,24 @@ func (s *Service) billings(ctx context.Context, user *auth.User, kind string, bo
 			"durationUnitCount": durationUnits,
 		})}
 	}
+	if kind == "song.generations" || kind == "music.generations" {
+		resource = "music"
+		unit = "song"
+		baseKey = "musicBase"
+		amount := float64(count) * resourcePrice(config, resource, baseKey, "basePrice") * discount
+		return []any{billingLine(candidate, resource, unit, count, roundPrice(amount), discount, simulated)}
+	}
+	if kind == "speech.generations" {
+		resource = "audio"
+		unit = "character"
+		baseKey = "audioBase"
+		quantity := len([]rune(stringFromMap(body, "text")))
+		if quantity <= 0 {
+			quantity = 1
+		}
+		amount := float64(quantity) * resourcePrice(config, resource, baseKey, "basePrice") * discount
+		return []any{billingLine(candidate, resource, unit, quantity, roundPrice(amount), discount, simulated)}
+	}
 	amount := float64(count) * resourcePrice(config, resource, baseKey, "basePrice") * resourceWeight(config, resource, "qualityWeights", stringFromMap(body, "quality")) * resourceWeight(config, resource, "sizeWeights", stringFromMap(body, "size")) * resourceWeight(config, resource, "resolutionWeights", firstNonEmptyString(stringFromMap(body, "resolution"), stringFromMap(body, "size"))) * discount
 	return []any{billingLine(candidate, resource, unit, count, roundPrice(amount), discount, simulated)}
 }

apps/api/internal/runner/pricing_test.go

@@ -84,6 +84,66 @@ func TestVideoBillingEstimateUsesFiveSecondUnitsAndDynamicWeights(t *testing.T)
 	}
 }
 
+func TestMusicBillingUsesSongResourceAndOutputCount(t *testing.T) {
+	service := &Service{}
+	candidate := store.RuntimeModelCandidate{
+		ModelName: "suno-model",
+		BaseBillingConfig: map[string]any{
+			"musicBase": 6,
+			"music":     map[string]any{"basePrice": 9},
+		},
+	}
+
+	items := service.billings(context.Background(), nil, "song.generations", map[string]any{
+		"prompt": "city lights",
+		"count":  3,
+	}, candidate, clients.Response{}, true)
+
+	line := firstBillingLine(t, items)
+	if got, want := line["resourceType"], "music"; got != want {
+		t.Fatalf("music resource type = %v, want %v", got, want)
+	}
+	if got, want := line["unit"], "song"; got != want {
+		t.Fatalf("music billing unit = %v, want %v", got, want)
+	}
+	if got, want := line["quantity"], 3; got != want {
+		t.Fatalf("music quantity = %v, want %v", got, want)
+	}
+	if got, want := floatFromAny(line["amount"]), 18.0; got != want {
+		t.Fatalf("music amount = %v, want %v", got, want)
+	}
+}
+
+func TestSpeechBillingUsesAudioCharacters(t *testing.T) {
+	service := &Service{}
+	candidate := store.RuntimeModelCandidate{
+		ModelName: "speech-model",
+		BaseBillingConfig: map[string]any{
+			"audioBase": 0.5,
+			"audio":     map[string]any{"basePrice": 0.8},
+		},
+	}
+
+	items := service.billings(context.Background(), nil, "speech.generations", map[string]any{
+		"text":     "你好abc",
+		"voice_id": "female-shaonv",
+	}, candidate, clients.Response{}, true)
+
+	line := firstBillingLine(t, items)
+	if got, want := line["resourceType"], "audio"; got != want {
+		t.Fatalf("speech resource type = %v, want %v", got, want)
+	}
+	if got, want := line["unit"], "character"; got != want {
+		t.Fatalf("speech billing unit = %v, want %v", got, want)
+	}
+	if got, want := line["quantity"], 5; got != want {
+		t.Fatalf("speech character quantity = %v, want %v", got, want)
+	}
+	if got, want := floatFromAny(line["amount"]), 2.5; got != want {
+		t.Fatalf("speech amount = %v, want %v", got, want)
+	}
+}
+
 func TestVideoBillingPrefersGeneratedDuration(t *testing.T) {
 	service := &Service{}
 	candidate := store.RuntimeModelCandidate{

apps/api/internal/runner/service.go

@@ -64,6 +64,7 @@ func New(cfg config.Config, db *store.Store, logger *slog.Logger) *Service {
 			"midjourney":            clients.MidjourneyClient{HTTPClient: httpClients.none},
 			"minimax":               clients.MinimaxClient{HTTPClient: httpClients.none},
 			"newapi":                clients.NewAPIClient{HTTPClient: httpClients.none},
+			"suno":                  clients.SunoClient{HTTPClient: httpClients.none},
 			"tencent-hunyuan-image": clients.HunyuanImageClient{HTTPClient: httpClients.none},
 			"tencent-hunyuan-video": clients.HunyuanVideoClient{HTTPClient: httpClients.none},
 			"vidu":                  clients.ViduClient{HTTPClient: httpClients.none},
@@ -957,6 +958,10 @@ func modelTypeFromKind(kind string, body map[string]any) string {
 			return "image_to_video"
 		}
 		return "video_generate"
+	case "song.generations", "music.generations":
+		return "audio_generate"
+	case "speech.generations":
+		return "text_to_speech"
 	default:
 		return "task"
 	}
@@ -979,6 +984,10 @@ func canonicalModelType(value string) string {
 		return "text_embedding"
 	case "rerank", "reranks":
 		return "text_rerank"
+	case "audio", "music", "music_generate", "song", "songs":
+		return "audio_generate"
+	case "speech", "tts":
+		return "text_to_speech"
 	default:
 		return normalized
 	}
@@ -986,7 +995,7 @@ func canonicalModelType(value string) string {
 
 func isKnownModelType(value string) bool {
 	switch value {
-	case "text_generate", "text_embedding", "text_rerank", "image_generate", "image_edit", "video_generate", "image_to_video", "text_to_video", "video_edit", "video_reference", "video_first_last_frame", "omni_video", "omni":
+	case "text_generate", "text_embedding", "text_rerank", "image_generate", "image_edit", "video_generate", "image_to_video", "text_to_video", "video_edit", "video_reference", "video_first_last_frame", "omni_video", "omni", "audio_generate", "text_to_speech":
 		return true
 	default:
 		return false
@@ -1171,6 +1180,17 @@ func validateRequest(kind string, body map[string]any) error {
 		if strings.TrimSpace(stringFromMap(body, "prompt")) == "" {
 			return errors.New("prompt is required")
 		}
+	case "song.generations", "music.generations":
+		if strings.TrimSpace(stringFromMap(body, "prompt")) == "" {
+			return errors.New("prompt is required")
+		}
+	case "speech.generations":
+		if strings.TrimSpace(stringFromMap(body, "text")) == "" && strings.TrimSpace(stringFromMap(body, "text_file_id")) == "" {
+			return errors.New("text or text_file_id is required")
+		}
+		if strings.TrimSpace(stringFromMap(body, "voice_id")) == "" {
+			return errors.New("voice_id is required")
+		}
 	}
 	return nil
 }

apps/api/internal/runner/upload.go

@@ -943,6 +943,9 @@ func mediaKindForAsset(taskKind string, item map[string]any, sourceKey string, c
 	if strings.Contains(kind, "video") {
 		return "video"
 	}
+	if strings.Contains(kind, "audio") || strings.Contains(kind, "song") || strings.Contains(kind, "music") || strings.Contains(kind, "speech") {
+		return "audio"
+	}
 	if strings.Contains(kind, "image") {
 		return "image"
 	}