fix(models): 模型目录 - 保留人工能力并按声明发送视频分辨率

backend/internal/service/model_capability.go

@@ -67,6 +67,7 @@ type VideoCapabilityConfig struct {
 
 type VideoReferenceConfig struct {
 	PromptMaxChars   int   `json:"promptMaxChars"`
+	MinImages        int   `json:"minImages"`
 	MaxImages        int   `json:"maxImages"`
 	MaxImageBytes    int64 `json:"maxImageBytes"`
 	MaxVideos        int   `json:"maxVideos"`
@@ -154,7 +155,7 @@ func defaultImageSizeValues() []string {
 
 func DefaultModelCapabilityConfigForModel(protocol string, modelName string) *ModelCapabilityConfig {
 	video := &VideoCapabilityConfig{
-		References:        VideoReferenceConfig{PromptMaxChars: 1000, MaxImages: 9, MaxImageBytes: 30 * 1024 * 1024, MaxVideos: 0, MaxVideoBytes: 0, MaxVideoDuration: 0, MaxAudios: 0, MaxAudioBytes: 0, MaxAudioDuration: 0},
+		References:        VideoReferenceConfig{PromptMaxChars: 1000, MinImages: 0, MaxImages: 9, MaxImageBytes: 30 * 1024 * 1024, MaxVideos: 0, MaxVideoBytes: 0, MaxVideoDuration: 0, MaxAudios: 0, MaxAudioBytes: 0, MaxAudioDuration: 0},
 		Duration:          VideoDurationConfig{Selection: "range", Min: 1, Max: 15, Step: 1, Default: 6},
 		Ratios:            []string{"16:9", "9:16", "1:1", "4:3", "3:4", "21:9"},
 		DefaultRatio:      "16:9",
@@ -277,11 +278,14 @@ func validateVideoCapabilityConfig(value *VideoCapabilityConfig) error {
 	if value.References.PromptMaxChars < 1 || value.References.PromptMaxChars > 1000000 {
 		return BadAuthRequest("提示词最大字符数必须在 1-1000000 之间")
 	}
-	for name, number := range map[string]int{"最大图片引用数": value.References.MaxImages, "最大视频引用数": value.References.MaxVideos, "最大音频引用数": value.References.MaxAudios} {
+	for name, number := range map[string]int{"最少图片引用数": value.References.MinImages, "最大图片引用数": value.References.MaxImages, "最大视频引用数": value.References.MaxVideos, "最大音频引用数": value.References.MaxAudios} {
 		if number < 0 || number > 100 {
 			return BadAuthRequest(name + "必须在 0-100 之间")
 		}
 	}
+	if value.References.MinImages > value.References.MaxImages {
+		return BadAuthRequest("最少图片引用数不能超过最大图片引用数")
+	}
 	if value.References.MaxImageBytes < 0 || value.References.MaxVideoBytes < 0 || value.References.MaxAudioBytes < 0 || value.References.MaxVideoDuration < 0 || value.References.MaxAudioDuration < 0 {
 		return BadAuthRequest("引用素材限制不能小于 0")
 	}
@@ -291,8 +295,12 @@ func validateVideoCapabilityConfig(value *VideoCapabilityConfig) error {
 	if len(value.Ratios) == 0 || strings.TrimSpace(value.DefaultRatio) == "" || !containsCapabilityString(value.Ratios, value.DefaultRatio) {
 		return BadAuthRequest("请至少配置一个画面比例，并选择默认比例")
 	}
-	if len(value.Resolutions) == 0 || strings.TrimSpace(value.DefaultResolution) == "" || !containsCapabilityString(value.Resolutions, value.DefaultResolution) {
-		return BadAuthRequest("请至少配置一个输出分辨率，并选择默认分辨率")
+	if len(value.Resolutions) == 0 {
+		if strings.TrimSpace(value.DefaultResolution) != "" {
+			return BadAuthRequest("未配置输出分辨率时不能设置默认分辨率")
+		}
+	} else if strings.TrimSpace(value.DefaultResolution) == "" || !containsCapabilityString(value.Resolutions, value.DefaultResolution) {
+		return BadAuthRequest("默认输出分辨率必须属于支持值")
 	}
 	if len(value.Operations) == 0 || strings.TrimSpace(value.DefaultOperation) == "" || !containsCapabilityString(value.Operations, value.DefaultOperation) {
 		return BadAuthRequest("请至少配置一个生成模式，并选择默认模式")
@@ -377,6 +385,9 @@ func validateVideoTask(profile *VideoCapabilityConfig, input canvasGenerationInp
 	if len(input.ReferenceImages) > profile.References.MaxImages || len(input.ReferenceVideos) > profile.References.MaxVideos || len(input.ReferenceAudios) > profile.References.MaxAudios {
 		return BadAuthRequest("参考素材数量超过当前模型限制")
 	}
+	if len(input.ReferenceImages) < profile.References.MinImages {
+		return BadAuthRequest(fmt.Sprintf("当前视频模型至少需要 %d 张参考图", profile.References.MinImages))
+	}
 	for _, media := range input.ReferenceImages {
 		if profile.References.MaxImageBytes > 0 && media.Bytes > profile.References.MaxImageBytes {
 			return BadAuthRequest("参考图片文件超过当前模型大小限制")

backend/internal/service/model_capability_test.go

@@ -116,3 +116,32 @@ func TestValidateVideoTaskIgnoresGlobalResolutionWhenCatalogDeclaresNone(t *test
 		t.Fatalf("validateVideoTask() error = %v", err)
 	}
 }
+
+func TestNormalizeVideoCapabilityAllowsOmittedResolution(t *testing.T) {
+	profile := DefaultModelCapabilityConfigForModel("newapi-channel-2", "endpoint-video").Video
+	profile.Resolutions = nil
+	profile.DefaultResolution = ""
+
+	result, err := NormalizeModelCapabilityConfig("video", "newapi-channel-2", &ModelCapabilityConfig{Version: 1, Video: profile})
+	if err != nil {
+		t.Fatalf("NormalizeModelCapabilityConfig() error = %v", err)
+	}
+	if result.Video == nil || len(result.Video.Resolutions) != 0 || result.Video.DefaultResolution != "" {
+		t.Fatalf("normalized video resolution = %#v", result.Video)
+	}
+}
+
+func TestValidateVideoTaskRequiresDeclaredMinimumImages(t *testing.T) {
+	profile := DefaultModelCapabilityConfigForModel("newapi-channel-2", "image-required-video").Video
+	profile.References.MinImages = 1
+	profile.References.MaxImages = 2
+	profile.Operations = []string{"image_to_video"}
+	profile.DefaultOperation = "image_to_video"
+
+	err := validateVideoTask(profile, canvasGenerationInput{
+		Config: providerConfig{Model: "image-required-video", VideoSeconds: "6", Size: "16:9", VQuality: "720"},
+	})
+	if err == nil || !strings.Contains(err.Error(), "至少需要 1 张参考图") {
+		t.Fatalf("validateVideoTask() error = %v", err)
+	}
+}

backend/internal/service/provider.go

@@ -1973,7 +1973,10 @@ func newAPIChannel2VideoRequestBody(input canvasGenerationInput) (newAPIVideoReq
 		seconds = 6
 	}
 	ratio := normalizeNewAPIChannel2Ratio(input.Config.Size, modelName)
-	resolution := normalizeNewAPIChannel2Resolution(input.Config.VQuality, modelName)
+	resolution := videoResolutionNameRequest(input.VideoCapability, input.Config.VQuality)
+	if modelName == "grok-video-1.5-1080p" {
+		resolution = "1080p"
+	}
 	body := newAPIVideoRequest{
 		Model:       input.Config.Model,
 		Prompt:      strings.TrimSpace(input.Prompt),

backend/internal/service/provider_request_types.go

@@ -19,7 +19,7 @@ type newAPIVideoRequest struct {
 	Prompt        string   `json:"prompt"`
 	Seconds       string   `json:"seconds"`
 	AspectRatio   string   `json:"aspect_ratio"`
-	Resolution    string   `json:"resolution"`
+	Resolution    string   `json:"resolution,omitempty"`
 	GenerateAudio *bool    `json:"generate_audio,omitempty"`
 	ImageURLs     []string `json:"image_urls,omitempty"`
 	VideoURLs     []string `json:"video_urls,omitempty"`

backend/internal/service/provider_test.go

@@ -1278,9 +1278,12 @@ func TestRunNewAPIChannel2VideoTaskDownloadsTemporaryResult(t *testing.T) {
 	}))
 	defer server.Close()
 
+	profile := DefaultModelCapabilityConfigForModel("newapi-channel-2", "grok-image-video").Video
+	profile.Resolutions = []string{"720p"}
+	profile.DefaultResolution = "720p"
 	result, err := runVideoTask(context.Background(), canvasGenerationInput{
 		Prompt: "make it move",
-		Config: providerConfig{BaseURL: server.URL, APIKey: "test-key", Model: "grok-image-video", InterfaceType: "newapi-channel-2", VideoSeconds: "15", Size: "720x1280", VQuality: "high"},
+		Config: providerConfig{BaseURL: server.URL, APIKey: "test-key", Model: "grok-image-video", InterfaceType: "newapi-channel-2", VideoSeconds: "15", Size: "720x1280", VQuality: "720"},
 		ReferenceImages: []providerMedia{
 			{ID: "image-1", DataURL: testReferenceImageDataURL},
 			{ID: "image-2", DataURL: testReferenceImageDataURL},
@@ -1288,6 +1291,7 @@ func TestRunNewAPIChannel2VideoTaskDownloadsTemporaryResult(t *testing.T) {
 		ReferenceVideos: []providerMedia{{ID: "video-1", URL: server.URL + "/reference.mp4"}},
 		ReferenceAudios: []providerMedia{{ID: "audio-1", URL: server.URL + "/reference.mp3"}},
 		Metadata:        map[string]interface{}{"videoEditOperation": "image_to_video"},
+		VideoCapability: profile,
 	})
 	if err != nil {
 		t.Fatalf("runVideoTask() error = %v", err)
@@ -1354,6 +1358,50 @@ func TestNewAPIChannel2SingleImageModelsRequireOneReference(t *testing.T) {
 	}
 }
 
+func TestNewAPIChannel2SendsOnlyDeclaredResolution(t *testing.T) {
+	tests := []struct {
+		name        string
+		model       string
+		quality     string
+		resolutions []string
+		want        string
+	}{
+		{name: "catalog omits resolution", model: "endpoint-video", quality: "720"},
+		{name: "declared 2K alias", model: "declared-video", quality: "2K", resolutions: []string{"1440p"}, want: "1440p"},
+		{name: "fixed grok 1080p model", model: "grok-video-1.5-1080p", quality: "720", want: "1080p"},
+	}
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			profile := DefaultModelCapabilityConfigForModel("newapi-channel-2", test.model).Video
+			profile.Resolutions = test.resolutions
+			profile.DefaultResolution = ""
+			input := canvasGenerationInput{
+				Config:          providerConfig{Model: test.model, VideoSeconds: "6", Size: "16:9", VQuality: test.quality},
+				VideoCapability: profile,
+			}
+			if strings.HasPrefix(test.model, "grok-video-1.5") {
+				input.ReferenceImages = []providerMedia{{ID: "image-1", DataURL: testReferenceImageDataURL}}
+			}
+
+			body, err := newAPIChannel2VideoRequestBody(input)
+			if err != nil {
+				t.Fatalf("newAPIChannel2VideoRequestBody() error = %v", err)
+			}
+			if body.Resolution != test.want {
+				t.Fatalf("resolution = %q, want %q", body.Resolution, test.want)
+			}
+			mapped, err := requestAsMap(body)
+			if err != nil {
+				t.Fatalf("requestAsMap() error = %v", err)
+			}
+			_, hasResolution := mapped["resolution"]
+			if hasResolution != (test.want != "") {
+				t.Fatalf("resolution presence = %v, want %v; body = %#v", hasResolution, test.want != "", mapped)
+			}
+		})
+	}
+}
+
 func TestNewAPIChannel2RejectsAudioWithoutReferenceVideo(t *testing.T) {
 	_, err := newAPIChannel2VideoRequestBody(canvasGenerationInput{
 		Config:          providerConfig{Model: "grok-image-video", VideoSeconds: "6"},