feat(video): add capability-driven video tools contract

app/(chat)/api/models/route.ts

@@ -5,6 +5,7 @@ import {
   getEnabledAiModelsWithProvider,
   getUserById,
 } from "@/lib/db/queries";
+import { buildVideoToolCapability } from "@/lib/ai/video-tool-capability";
 import { ChatSDKError } from "@/lib/errors";
 import { isAiModelAgentAuto } from "@/lib/db/ai-model-flags";
 
@@ -91,6 +92,11 @@ export async function GET() {
       isSubModel: model.isSubModel,
       parentModelId: model.parentModelId,
       capabilities: model.capabilities, // 🆕 返回子模型的能力配置
+      videoToolCapability: buildVideoToolCapability({
+        providerType: provider.type,
+        modelId: model.modelId,
+        rawCapabilities: model.capabilities,
+      }),
       imageGenConfig: model.imageGenConfig,
       sortOrder: model.sortOrder ?? 0,
     }));

app/(chat)/api/video-model-capabilities/route.ts

@@ -0,0 +1,47 @@
+import { auth } from "@/app/(auth)/auth";
+import { getAiModelById, getAiProviderById } from "@/lib/db/queries";
+import { buildVideoToolCapability } from "@/lib/ai/video-tool-capability";
+import { ChatSDKError } from "@/lib/errors";
+
+export async function GET(request: Request) {
+  try {
+    const session = await auth();
+    if (!session?.user?.id) {
+      return new ChatSDKError("unauthorized:chat").toResponse();
+    }
+
+    const { searchParams } = new URL(request.url);
+    const modelDbId = searchParams.get("modelId");
+    if (!modelDbId) {
+      return new ChatSDKError("bad_request:api", "modelId is required").toResponse();
+    }
+
+    const model = await getAiModelById({ id: modelDbId });
+    if (!model || !model.isEnabled) {
+      return new ChatSDKError("bad_request:api", "Model not found or disabled").toResponse();
+    }
+
+    const provider = await getAiProviderById({ id: model.providerId });
+    if (!provider || !provider.isEnabled) {
+      return new ChatSDKError("bad_request:api", "Provider not found or disabled").toResponse();
+    }
+
+    const capability = buildVideoToolCapability({
+      providerType: provider.type,
+      modelId: model.modelId,
+      rawCapabilities: model.capabilities,
+    });
+
+    return Response.json({
+      modelId: model.id,
+      modelName: model.displayName,
+      providerType: provider.type,
+      capability,
+    });
+  } catch (error) {
+    if (error instanceof ChatSDKError) {
+      return error.toResponse();
+    }
+    return new ChatSDKError("offline:chat").toResponse();
+  }
+}

components/canvas/CanvasChatPanel.tsx

@@ -57,6 +57,7 @@ import { getCanvasSkillIconComponent } from "@/components/canvas/canvas-skill-ic
 import { ImageParamsBar, type ImageGenerationParams } from "@/components/image-params-bar";
 import { VideoParamsBar, type VideoGenerationParams } from "@/components/video-params-bar";
 import { ImageGenerationLoading } from "@/components/image-generation-loading";
+import type { VideoToolCapabilityContract } from "@/types/video-tool-capability";
 
 // 标签悬停预览信息
 interface TagPreviewInfo {
@@ -760,8 +761,70 @@ function extractCanvasImageAnalysisSection(content: string): CanvasAnalysisSecti
 
 function normalizeVideoParamsForProvider(
   params: VideoGenerationParams | null | undefined,
-  providerType?: string
+  providerType?: string,
+  capability?: VideoToolCapabilityContract | null
 ): VideoGenerationParams {
+  if (capability) {
+    const modeOptions = capability.modeOptions || [];
+    const ratioOptions = capability.ratioOptions || [];
+    const durationOptions = capability.durationOptions || [];
+    const resolutionOptions = capability.resolutionOptions || [];
+    const modeSet = new Set(modeOptions.map((item) => item.mode));
+    const ratioSet = new Set(ratioOptions);
+    const durationSet = new Set(durationOptions);
+    const resolutionSet = new Set(resolutionOptions);
+
+    const fallbackMode = capability.defaults?.mode || modeOptions[0]?.mode || "text-to-video";
+    const fallbackRatio = capability.defaults?.ratio || ratioOptions[0] || "16:9";
+    const fallbackDuration = capability.defaults?.duration || durationOptions[0] || 5;
+    const fallbackResolution = capability.defaults?.resolution || resolutionOptions[0] || "720p";
+
+    const merged = {
+      mode: fallbackMode,
+      resolution: fallbackResolution as VideoGenerationParams["resolution"],
+      ratio: fallbackRatio as VideoGenerationParams["ratio"],
+      duration: Number(fallbackDuration) || 5,
+      generateAudio: capability.defaults?.generateAudio ?? false,
+      watermark: false,
+      cameraFixed: false,
+      cameraMove: "none" as VideoGenerationParams["cameraMove"],
+      fps: capability.defaults?.fps || 24,
+      ...(params || {}),
+    } as VideoGenerationParams;
+
+    if (!modeSet.has(merged.mode)) {
+      merged.mode = fallbackMode as VideoGenerationParams["mode"];
+    }
+    if (!ratioSet.has(merged.ratio)) {
+      merged.ratio = fallbackRatio as VideoGenerationParams["ratio"];
+    }
+    if (!durationSet.has(Number(merged.duration))) {
+      merged.duration = Number(fallbackDuration) || 5;
+    } else {
+      merged.duration = Number(merged.duration) || Number(fallbackDuration) || 5;
+    }
+    if (resolutionOptions.length > 0 && !resolutionSet.has(merged.resolution)) {
+      merged.resolution = fallbackResolution as VideoGenerationParams["resolution"];
+    }
+    if (!capability.ui.showGenerateAudio) {
+      merged.generateAudio = false;
+    }
+    if (!capability.ui.showFps) {
+      merged.fps = capability.defaults?.fps || 24;
+    }
+    if (!capability.ui.showCameraMove) {
+      merged.cameraMove = "none";
+      merged.cameraFixed = false;
+    }
+    if (!capability.ui.showReturnLastFrame) {
+      merged.returnLastFrame = false;
+    }
+    if (!capability.ui.showSeed) {
+      merged.seed = undefined;
+    }
+    return merged;
+  }
+
   const isFalVideo = providerType === "fal-video";
   const isRuxaVideo = providerType === "ruxa-video";
   const base: VideoGenerationParams = {
@@ -814,11 +877,18 @@ function normalizeVideoParamsForProvider(
 
 function getVideoImageLimit(
   params: VideoGenerationParams | null | undefined,
-  providerType?: string
+  providerType?: string,
+  capability?: VideoToolCapabilityContract | null
 ): { min: number; max: number } {
   if (!params) {
     return { min: 0, max: 0 };
   }
+  if (capability?.modeOptions?.length) {
+    const modeOption = capability.modeOptions.find((item) => item.mode === params.mode);
+    if (modeOption) {
+      return { min: modeOption.imageMin, max: modeOption.imageMax };
+    }
+  }
   const isFalVideo = providerType === "fal-video";
   const isRuxaVideo = providerType === "ruxa-video";
   if (isRuxaVideo) {
@@ -2253,7 +2323,8 @@ export function CanvasChatPanel({ chatId, initialMessages = [], onClose, onOpenF
       providerType: string;
       supportsVision?: boolean;
       supportsImageGen?: boolean;
-      capabilities?: string[] | null;
+      capabilities?: any;
+      videoToolCapability?: VideoToolCapabilityContract | null;
       imageGenConfig?: {
         maxImages?: number;
         supportsSize?: boolean;
@@ -2501,6 +2572,7 @@ export function CanvasChatPanel({ chatId, initialMessages = [], onClose, onOpenF
     sortOrder: -999,
     providerType: "auto",
     supportsImageGen: true,
+    videoToolCapability: null,
     imageGenConfig: modelsData?.agentModel?.imageGenConfig || undefined,
   };
   
@@ -2768,11 +2840,15 @@ export function CanvasChatPanel({ chatId, initialMessages = [], onClose, onOpenF
 
   useEffect(() => {
     if (!isVideoModel) return;
-    const normalized = normalizeVideoParamsForProvider(videoParams, currentModel?.providerType);
+    const normalized = normalizeVideoParamsForProvider(
+      videoParams,
+      currentModel?.providerType,
+      currentModel?.videoToolCapability
+    );
     if (!videoParams || JSON.stringify(videoParams) !== JSON.stringify(normalized)) {
       setVideoParams(normalized);
     }
-  }, [isVideoModel, currentModel?.providerType, videoParams]);
+  }, [isVideoModel, currentModel?.providerType, currentModel?.videoToolCapability, videoParams]);
 
   useEffect(() => {
     if (!isVideoModel || videoFrameSlots.length === 0) return;
@@ -5581,7 +5657,7 @@ export function CanvasChatPanel({ chatId, initialMessages = [], onClose, onOpenF
     
     let effectiveAttachments = [...attachments];
     const videoImageLimit = isVideoModel
-      ? getVideoImageLimit(videoParams, currentModel?.providerType)
+      ? getVideoImageLimit(videoParams, currentModel?.providerType, currentModel?.videoToolCapability)
       : null;
 
     if (videoImageLimit) {
@@ -6157,7 +6233,7 @@ export function CanvasChatPanel({ chatId, initialMessages = [], onClose, onOpenF
     ? detailValidationIssues.length === 0 && !isDetailGenerating
     : hasInputOrAttachments;
   const currentVideoImageLimit = isVideoModel
-    ? getVideoImageLimit(videoParams, currentModel?.providerType)
+    ? getVideoImageLimit(videoParams, currentModel?.providerType, currentModel?.videoToolCapability)
     : null;
   const showVideoAttachmentButton = Boolean(
     isVideoModel &&
@@ -8134,6 +8210,7 @@ export function CanvasChatPanel({ chatId, initialMessages = [], onClose, onOpenF
                       onParamsChange={handleVideoParamsChange}
                       inline={true}
                       providerType={currentModel?.providerType}
+                      capabilities={currentModel?.videoToolCapability}
                     />
                   </div>
                 </div>

components/chat.tsx

@@ -37,6 +37,7 @@ import { toast } from "sonner";
 import type { VisibilityType } from "@/components/visibility-selector";
 import type { VideoGenerationParams } from "@/components/video-params-bar";
 import type { ImageGenerationParams } from "@/components/image-params-bar";
+import type { VideoToolCapabilityContract } from "@/types/video-tool-capability";
 
 type ImageGenerationPhase = "idle" | "queued" | "processing" | "finalizing" | "failed";
 
@@ -68,8 +69,72 @@ function normalizeImageParamsForProvider(
 
 function normalizeVideoParamsForProvider(
   params: VideoGenerationParams | null | undefined,
-  providerType?: string
+  providerType?: string,
+  capability?: VideoToolCapabilityContract | null
 ): VideoGenerationParams {
+  if (capability) {
+    const modeOptions = capability.modeOptions || [];
+    const ratioOptions = capability.ratioOptions || [];
+    const durationOptions = capability.durationOptions || [];
+    const resolutionOptions = capability.resolutionOptions || [];
+    const modeSet = new Set(modeOptions.map((item) => item.mode));
+    const ratioSet = new Set(ratioOptions);
+    const durationSet = new Set(durationOptions);
+    const resolutionSet = new Set(resolutionOptions);
+
+    const fallbackMode = capability.defaults?.mode || modeOptions[0]?.mode || "text-to-video";
+    const fallbackRatio = capability.defaults?.ratio || ratioOptions[0] || "16:9";
+    const fallbackDuration = capability.defaults?.duration || durationOptions[0] || 5;
+    const fallbackResolution = capability.defaults?.resolution || resolutionOptions[0] || "720p";
+
+    const merged = {
+      mode: fallbackMode,
+      resolution: fallbackResolution as VideoGenerationParams["resolution"],
+      ratio: fallbackRatio as VideoGenerationParams["ratio"],
+      duration: Number(fallbackDuration) || 5,
+      generateAudio: capability.defaults?.generateAudio ?? false,
+      watermark: false,
+      cameraFixed: false,
+      cameraMove: "none" as VideoGenerationParams["cameraMove"],
+      fps: capability.defaults?.fps || 24,
+      ...(params || {}),
+    } as VideoGenerationParams;
+
+    if (!modeSet.has(merged.mode)) {
+      merged.mode = fallbackMode as VideoGenerationParams["mode"];
+    }
+    if (!ratioSet.has(merged.ratio)) {
+      merged.ratio = fallbackRatio as VideoGenerationParams["ratio"];
+    }
+    if (!durationSet.has(Number(merged.duration))) {
+      merged.duration = Number(fallbackDuration) || 5;
+    } else {
+      merged.duration = Number(merged.duration) || Number(fallbackDuration) || 5;
+    }
+    if (!resolutionSet.has(merged.resolution)) {
+      merged.resolution = fallbackResolution as VideoGenerationParams["resolution"];
+    }
+
+    if (!capability.ui.showGenerateAudio) {
+      merged.generateAudio = false;
+    }
+    if (!capability.ui.showFps) {
+      merged.fps = capability.defaults?.fps || 24;
+    }
+    if (!capability.ui.showCameraMove) {
+      merged.cameraMove = "none";
+      merged.cameraFixed = false;
+    }
+    if (!capability.ui.showReturnLastFrame) {
+      merged.returnLastFrame = false;
+    }
+    if (!capability.ui.showSeed) {
+      merged.seed = undefined;
+    }
+
+    return merged;
+  }
+
   const isFalVideo = providerType === "fal-video";
   const isRuxaVideo = providerType === "ruxa-video";
   const base: VideoGenerationParams = {
@@ -411,6 +476,7 @@ export function Chat({
       providerType: string;
       supportsVision?: boolean;
       supportsImageGen?: boolean;
+      videoToolCapability?: VideoToolCapabilityContract | null;
       imageGenConfig?: {
         maxImages?: number;
         supportsSize?: boolean;
@@ -473,12 +539,16 @@ export function Chat({
       return;
     }
 
-    const normalized = normalizeVideoParamsForProvider(videoParams, currentModel?.providerType);
+    const normalized = normalizeVideoParamsForProvider(
+      videoParams,
+      currentModel?.providerType,
+      currentModel?.videoToolCapability
+    );
     if (!videoParams || JSON.stringify(videoParams) !== JSON.stringify(normalized)) {
       setVideoParams(normalized);
       console.log("[Chat] 视频参数已按当前模型规范化:", normalized);
     }
-  }, [isVideoModel, videoParams, currentModel?.providerType]);
+  }, [isVideoModel, videoParams, currentModel?.providerType, currentModel?.videoToolCapability]);
 
   // 🆕 当切换到图片模型时，强制更新 imageParams（确保组件重新渲染）
   useEffect(() => {
@@ -1984,6 +2054,7 @@ export function Chat({
               videoModelName={currentModel?.displayName}
               videoParams={videoParams}
               videoProviderType={currentModel?.providerType}
+              videoToolCapability={currentModel?.videoToolCapability}
               onVideoParamsChange={(params) => {
                 setVideoParams(params);
                 console.log("[Chat] Video params updated:", params);

components/multimodal-input.tsx

@@ -19,6 +19,7 @@ type DynamicModel = {
   isSubModel?: boolean; // 🆕 是否为子模型
   parentModelId?: string | null; // 🆕 父模型ID
   capabilities?: any; // 🆕 模型能力配置
+  videoToolCapability?: VideoToolCapabilityContract | null;
   imageGenConfig?: {
     maxImages?: number;
     supportsSize?: boolean;
@@ -128,6 +129,7 @@ import { Button } from "./ui/button";
 import type { VisibilityType } from "./visibility-selector";
 import { VideoParamsBar, type VideoGenerationParams } from "./video-params-bar";
 import { ImageParamsBar, type ImageGenerationParams } from "./image-params-bar";
+import type { VideoToolCapabilityContract } from "@/types/video-tool-capability";
 
 function PureMultimodalInput({
   chatId,
@@ -149,6 +151,7 @@ function PureMultimodalInput({
   videoModelName,
   videoParams,
   videoProviderType,
+  videoToolCapability,
   onVideoParamsChange,
   isImageModel,
   imageModelName,
@@ -176,6 +179,7 @@ function PureMultimodalInput({
   videoModelName?: string;
   videoParams?: VideoGenerationParams | null;
   videoProviderType?: string;
+  videoToolCapability?: VideoToolCapabilityContract | null;
   onVideoParamsChange?: (params: VideoGenerationParams) => void;
   isImageModel?: boolean;
   imageModelName?: string;
@@ -691,6 +695,26 @@ function PureMultimodalInput({
     if (!isVideoModel || !videoParams) {
       return { max: Number.POSITIVE_INFINITY, label: "" as const };
     }
+    if (videoToolCapability?.modeOptions?.length) {
+      const modeOption = videoToolCapability.modeOptions.find(
+        (item) => item.mode === videoParams.mode
+      );
+      if (modeOption) {
+        if (modeOption.imageMax === 0) {
+          return { max: 0, label: `${modeOption.label}模式不支持图片` as const };
+        }
+        if (modeOption.imageMin === modeOption.imageMax) {
+          return {
+            max: modeOption.imageMax,
+            label: `${modeOption.label}模式仅支持 ${modeOption.imageMax} 张图` as const,
+          };
+        }
+        return {
+          max: modeOption.imageMax,
+          label: `${modeOption.label}模式最多 ${modeOption.imageMax} 张图` as const,
+        };
+      }
+    }
     if (videoProviderType === "fal-video" || videoProviderType === "ruxa-video") {
       const providerLabel = videoProviderType === "ruxa-video" ? "RUXA" : "FAL";
       switch (videoParams.mode) {
@@ -713,7 +737,7 @@ function PureMultimodalInput({
       default:
         return { max: Number.POSITIVE_INFINITY, label: "" as const };
     }
-  }, [isVideoModel, videoParams, videoProviderType]);
+  }, [isVideoModel, videoParams, videoProviderType, videoToolCapability]);
 
   const enforceVideoImageLimit = useCallback(
     (incomingCount: number) => {
@@ -1257,6 +1281,7 @@ function PureMultimodalInput({
                     onParamsChange={onVideoParamsChange}
                     inline={true}
                     providerType={videoProviderType}
+                    capabilities={videoToolCapability}
                   />
                 )}
                 <ModelSelectorCompact
@@ -1379,6 +1404,9 @@ export const MultimodalInput = memo(
     if (prevProps.videoProviderType !== nextProps.videoProviderType) {
       return false;
     }
+    if (!equal(prevProps.videoToolCapability, nextProps.videoToolCapability)) {
+      return false;
+    }
     if (prevProps.isImageModel !== nextProps.isImageModel) {
       return false;
     }

components/video-params-bar.tsx

@@ -41,6 +41,7 @@ import {
   VolumeX,
 } from "lucide-react";
 import { cn } from "@/lib/utils";
+import type { VideoToolCapabilityContract } from "@/types/video-tool-capability";
 
 // 视频生成模式类型
 type VideoGenerationMode =
@@ -86,6 +87,7 @@ interface VideoParamsBarProps {
   onParamsChange?: (params: VideoGenerationParams) => void;
   inline?: boolean; // 内联模式：只显示参数按钮，不显示外层容器
   providerType?: string;
+  capabilities?: VideoToolCapabilityContract | null;
 }
 
 const MODE_LABELS: Record<VideoGenerationMode, string> = {
@@ -165,8 +167,69 @@ const RUXA_RATIO_SET = new Set(["16:9", "9:16", "1:1"]);
 
 function normalizeVideoParamsByProvider(
   source: Partial<VideoGenerationParams> | undefined,
-  providerType?: string
+  providerType?: string,
+  capability?: VideoToolCapabilityContract | null
 ): VideoGenerationParams {
+  if (capability) {
+    const modes = capability.modeOptions.map((item) => item.mode);
+    const ratios = capability.ratioOptions;
+    const durations = capability.durationOptions;
+    const resolutions = capability.resolutionOptions;
+
+    const fallbackMode = capability.defaults.mode || modes[0] || "text-to-video";
+    const fallbackRatio = capability.defaults.ratio || ratios[0] || "16:9";
+    const fallbackDuration = capability.defaults.duration || durations[0] || 5;
+    const fallbackResolution =
+      capability.defaults.resolution || resolutions[0] || "720p";
+
+    const merged = {
+      mode: fallbackMode,
+      resolution: fallbackResolution as VideoGenerationParams["resolution"],
+      ratio: fallbackRatio as VideoGenerationParams["ratio"],
+      duration: Number(fallbackDuration) || 5,
+      generateAudio: capability.defaults.generateAudio ?? false,
+      watermark: false,
+      cameraFixed: false,
+      cameraMove: "none" as const,
+      fps: capability.defaults.fps || 24,
+      ...source,
+    } as VideoGenerationParams;
+
+    if (!modes.includes(merged.mode)) {
+      merged.mode = fallbackMode as VideoGenerationParams["mode"];
+    }
+    if (!ratios.includes(merged.ratio)) {
+      merged.ratio = fallbackRatio as VideoGenerationParams["ratio"];
+    }
+    if (!durations.includes(Number(merged.duration))) {
+      merged.duration = Number(fallbackDuration) || 5;
+    } else {
+      merged.duration = Number(merged.duration) || Number(fallbackDuration) || 5;
+    }
+    if (resolutions.length > 0 && !resolutions.includes(merged.resolution)) {
+      merged.resolution = fallbackResolution as VideoGenerationParams["resolution"];
+    }
+
+    if (!capability.ui.showGenerateAudio) {
+      merged.generateAudio = false;
+    }
+    if (!capability.ui.showFps) {
+      merged.fps = capability.defaults.fps || 24;
+    }
+    if (!capability.ui.showCameraMove) {
+      merged.cameraMove = "none";
+      merged.cameraFixed = false;
+    }
+    if (!capability.ui.showReturnLastFrame) {
+      merged.returnLastFrame = false;
+    }
+    if (!capability.ui.showSeed) {
+      merged.seed = undefined;
+    }
+
+    return merged;
+  }
+
   const isFalVideo = providerType === "fal-video";
   const isRuxaVideo = providerType === "ruxa-video";
   const base: VideoGenerationParams = {
@@ -223,40 +286,70 @@ export function VideoParamsBar({
   onParamsChange,
   inline = false,
   providerType = "volcano-video",
+  capabilities,
 }: VideoParamsBarProps) {
   const isFalVideo = providerType === "fal-video";
   const isRuxaVideo = providerType === "ruxa-video";
-  const modeOptions = isFalVideo
+  const capabilityModeOptions = capabilities?.modeOptions?.length
+    ? capabilities.modeOptions.map((item) => ({
+        value: item.mode,
+        label: item.label || MODE_LABELS[item.mode],
+      }))
+    : null;
+  const modeOptions = capabilityModeOptions || (isFalVideo
     ? FAL_MODE_OPTIONS
     : isRuxaVideo
       ? RUXA_MODE_OPTIONS
-      : STANDARD_MODE_OPTIONS;
-  const ratioOptions = isFalVideo
+      : STANDARD_MODE_OPTIONS);
+  const capabilityRatioOptions = capabilities?.ratioOptions?.length
+    ? capabilities.ratioOptions.map((value) => ({ value, label: value }))
+    : null;
+  const ratioOptions = capabilityRatioOptions || (isFalVideo
     ? FAL_RATIO_OPTIONS
     : isRuxaVideo
       ? RUXA_RATIO_OPTIONS
-      : RATIO_OPTIONS;
-  const durationOptions = isFalVideo
+      : RATIO_OPTIONS);
+  const durationOptions = capabilities?.durationOptions?.length
+    ? capabilities.durationOptions
+    : (isFalVideo
     ? FAL_DURATION_OPTIONS
     : isRuxaVideo
       ? RUXA_DURATION_OPTIONS
-      : STANDARD_DURATION_OPTIONS;
+      : STANDARD_DURATION_OPTIONS);
+  const resolutionOptions = capabilities?.resolutionOptions?.length
+    ? capabilities.resolutionOptions.map((value) => ({
+        value,
+        label: value.toUpperCase(),
+      }))
+    : RESOLUTION_OPTIONS;
+  const showResolutionControl = capabilities
+    ? capabilities.ui.showResolution && resolutionOptions.length > 0
+    : !isFalVideo && !isRuxaVideo;
+  const showGenerateAudioControl = capabilities
+    ? capabilities.ui.showGenerateAudio
+    : isFalVideo;
+  const showCameraMoveControl = capabilities
+    ? capabilities.ui.showCameraMove
+    : !isFalVideo && !isRuxaVideo;
+  const showFpsControl = capabilities ? capabilities.ui.showFps : !isFalVideo && !isRuxaVideo;
+  const showSeedControl = capabilities ? capabilities.ui.showSeed : true;
+  const showReturnLastFrameControl = capabilities ? capabilities.ui.showReturnLastFrame : true;
 
   const [params, setParams] = useState<VideoGenerationParams>(() =>
-    normalizeVideoParamsByProvider(initialParams, providerType)
+    normalizeVideoParamsByProvider(initialParams, providerType, capabilities)
   );
 
   const [collapsed, setCollapsed] = useState(false);
 
   useEffect(() => {
-    const normalized = normalizeVideoParamsByProvider(initialParams, providerType);
+    const normalized = normalizeVideoParamsByProvider(initialParams, providerType, capabilities);
     setParams((prev) => {
       if (JSON.stringify(prev) === JSON.stringify(normalized)) {
         return prev;
       }
       return normalized;
     });
-  }, [initialParams, providerType]);
+  }, [initialParams, providerType, capabilities]);
 
   // 获取当前运镜模式的标签
   const getCurrentCameraLabel = () => {
@@ -269,6 +362,23 @@ export function VideoParamsBar({
 
   // 🆕 根据模式获取图片上传配置（根据火山引擎官方文档）
   const getImageUploadConfig = () => {
+    if (capabilities?.modeOptions?.length) {
+      const modeOption = capabilities.modeOptions.find((item) => item.mode === params.mode);
+      if (!modeOption) {
+        return { min: 0, max: 0, label: "" };
+      }
+      const label =
+        modeOption.imageMax === 0
+          ? `${modeOption.label}（无需图片）`
+          : modeOption.imageMin === modeOption.imageMax
+            ? `${modeOption.label}（${modeOption.imageMax} 张）`
+            : `${modeOption.label}（${modeOption.imageMin}-${modeOption.imageMax} 张）`;
+      return {
+        min: modeOption.imageMin,
+        max: modeOption.imageMax,
+        label,
+      };
+    }
     if (isRuxaVideo) {
       return { min: 1, max: 1, label: "图生视频（1 张起始图）" };
     }
@@ -373,7 +483,7 @@ export function VideoParamsBar({
                           ? "border-white bg-white text-[#1f2937] shadow-[0_1px_5px_rgba(15,23,42,0.08)]"
                           : "border-transparent text-[#5f6b7a] hover:bg-white/70"
                       )}
-                      onClick={() => updateParam("ratio", option.value)}
+                      onClick={() => updateParam("ratio", option.value as VideoGenerationParams["ratio"])}
                     >
                       {option.label}
                     </button>
@@ -382,11 +492,11 @@ export function VideoParamsBar({
               </div>
 
               {/* 分辨率（非 fal） */}
-              {!isFalVideo && !isRuxaVideo && (
+              {showResolutionControl && (
                 <div>
                   <p className="text-[11px] font-medium text-[#8a94a3]">选择分辨率</p>
                   <div className="mt-1.5 grid grid-cols-3 gap-1.5 rounded-[12px] bg-[#eceff2]/90 p-1">
-                    {RESOLUTION_OPTIONS.map((option) => (
+                    {resolutionOptions.map((option) => (
                       <button
                         key={option.value}
                         type="button"
@@ -451,8 +561,7 @@ export function VideoParamsBar({
             className="w-[280px] max-w-[calc(100vw-24px)] rounded-[14px] border border-[#d8dde4] bg-[#f3f4f6]/95 p-2 shadow-[0_10px_24px_rgba(15,23,42,0.14)]"
           >
             <div className="space-y-2.5">
-              {isFalVideo ? (
-                /* fal-video：音频开关 */
+              {showGenerateAudioControl ? (
                 <div>
                   <p className="text-[11px] font-medium text-[#8a94a3]">音频</p>
                   <div className="mt-1.5 grid grid-cols-2 gap-1.5 rounded-[12px] bg-[#eceff2]/90 p-1.5">
@@ -484,71 +593,75 @@ export function VideoParamsBar({
                     </button>
                   </div>
                 </div>
-              ) : isRuxaVideo ? (
+              ) : !showFpsControl && !showCameraMoveControl ? (
                 <div className="rounded-[12px] bg-[#eceff2]/90 p-2 text-[11px] text-[#7b8796]">
                   当前模型支持图生视频（1 张起始图），已自动按官方参数约束处理。
                 </div>
               ) : (
                 <>
-                  {/* 帧率 */}
-                  <div>
-                    <p className="text-[11px] font-medium text-[#8a94a3]">帧率</p>
-                    <div className="mt-1.5 grid grid-cols-3 gap-1.5 rounded-[12px] bg-[#eceff2]/90 p-1.5">
-                      {FPS_OPTIONS.map((option) => (
-                        <button
-                          key={option.value}
-                          type="button"
-                          className={cn(
-                            "flex h-8 items-center justify-center rounded-[9px] border text-[12px] font-semibold transition",
-                            (params.fps || 24) === option.value
-                              ? "border-white bg-white text-[#1f2937] shadow-[0_1px_5px_rgba(15,23,42,0.08)]"
-                              : "border-transparent text-[#5f6b7a] hover:bg-white/70"
-                          )}
-                          onClick={() => updateParam("fps", option.value as VideoGenerationParams["fps"])}
-                        >
-                          {option.label}
-                        </button>
-                      ))}
+                  {showFpsControl && (
+                    <div>
+                      <p className="text-[11px] font-medium text-[#8a94a3]">帧率</p>
+                      <div className="mt-1.5 grid grid-cols-3 gap-1.5 rounded-[12px] bg-[#eceff2]/90 p-1.5">
+                        {FPS_OPTIONS.map((option) => (
+                          <button
+                            key={option.value}
+                            type="button"
+                            className={cn(
+                              "flex h-8 items-center justify-center rounded-[9px] border text-[12px] font-semibold transition",
+                              (params.fps || 24) === option.value
+                                ? "border-white bg-white text-[#1f2937] shadow-[0_1px_5px_rgba(15,23,42,0.08)]"
+                                : "border-transparent text-[#5f6b7a] hover:bg-white/70"
+                            )}
+                            onClick={() => updateParam("fps", option.value as VideoGenerationParams["fps"])}
+                          >
+                            {option.label}
+                          </button>
+                        ))}
+                      </div>
                     </div>
-                  </div>
+                  )}
 
-                  {/* 运镜 */}
-                  <div>
-                    <p className="text-[11px] font-medium text-[#8a94a3]">运镜模式</p>
-                    <div className="mt-1.5 grid grid-cols-3 gap-1.5 rounded-[12px] bg-[#eceff2]/90 p-1.5">
-                      {CAMERA_MOVE_OPTIONS.map((option) => (
-                        <button
-                          key={option.value}
-                          type="button"
-                          className={cn(
-                            "flex h-8 items-center justify-center rounded-[9px] border text-[11px] font-semibold transition",
-                            params.cameraMove === option.value
-                              ? "border-white bg-white text-[#1f2937] shadow-[0_1px_5px_rgba(15,23,42,0.08)]"
-                              : "border-transparent text-[#5f6b7a] hover:bg-white/70"
-                          )}
-                          onClick={() => {
-                            updateParam("cameraMove", option.value);
-                            updateParam("cameraFixed", option.value === "static");
-                          }}
-                        >
-                          {option.label}
-                        </button>
-                      ))}
+                  {showCameraMoveControl && (
+                    <div>
+                      <p className="text-[11px] font-medium text-[#8a94a3]">运镜模式</p>
+                      <div className="mt-1.5 grid grid-cols-3 gap-1.5 rounded-[12px] bg-[#eceff2]/90 p-1.5">
+                        {CAMERA_MOVE_OPTIONS.map((option) => (
+                          <button
+                            key={option.value}
+                            type="button"
+                            className={cn(
+                              "flex h-8 items-center justify-center rounded-[9px] border text-[11px] font-semibold transition",
+                              params.cameraMove === option.value
+                                ? "border-white bg-white text-[#1f2937] shadow-[0_1px_5px_rgba(15,23,42,0.08)]"
+                                : "border-transparent text-[#5f6b7a] hover:bg-white/70"
+                            )}
+                            onClick={() => {
+                              updateParam("cameraMove", option.value);
+                              updateParam("cameraFixed", option.value === "static");
+                            }}
+                          >
+                            {option.label}
+                          </button>
+                        ))}
+                      </div>
                     </div>
-                  </div>
+                  )}
                 </>
               )}
 
               {/* 高级选项 */}
               <div className="space-y-2.5 border-t border-[#dce1e8] pt-2.5">
-                <div className="flex items-center justify-between">
-                  <span className="text-[11px] font-medium text-[#8a94a3]">返回尾帧</span>
-                  <Switch
-                    checked={params.returnLastFrame || false}
-                    onCheckedChange={(checked) => updateParam("returnLastFrame", checked)}
-                  />
-                </div>
-                {isFalVideo && (
+                {showReturnLastFrameControl && (
+                  <div className="flex items-center justify-between">
+                    <span className="text-[11px] font-medium text-[#8a94a3]">返回尾帧</span>
+                    <Switch
+                      checked={params.returnLastFrame || false}
+                      onCheckedChange={(checked) => updateParam("returnLastFrame", checked)}
+                    />
+                  </div>
+                )}
+                {showGenerateAudioControl && (
                   <div className="flex items-center justify-between">
                     <span className="text-[11px] font-medium text-[#8a94a3]">生成音频</span>
                     <Switch
@@ -557,16 +670,18 @@ export function VideoParamsBar({
                     />
                   </div>
                 )}
-                <div className="space-y-1">
-                  <span className="text-[11px] font-medium text-[#8a94a3]">随机种子</span>
-                  <Input
-                    type="number"
-                    placeholder="留空则随机"
-                    value={params.seed || ""}
-                    onChange={(e) => updateParam("seed", Number(e.target.value) || undefined)}
-                    className="h-7 rounded-[8px] border-[#dce1e8] bg-[#eef1f4]/90 text-[12px] text-[#2f3b48] placeholder:text-[#b0b8c4]"
-                  />
-                </div>
+                {showSeedControl && (
+                  <div className="space-y-1">
+                    <span className="text-[11px] font-medium text-[#8a94a3]">随机种子</span>
+                    <Input
+                      type="number"
+                      placeholder="留空则随机"
+                      value={params.seed || ""}
+                      onChange={(e) => updateParam("seed", Number(e.target.value) || undefined)}
+                      className="h-7 rounded-[8px] border-[#dce1e8] bg-[#eef1f4]/90 text-[12px] text-[#2f3b48] placeholder:text-[#b0b8c4]"
+                    />
+                  </div>
+                )}
               </div>
             </div>
           </PopoverContent>
@@ -689,7 +804,7 @@ export function VideoParamsBar({
                     variant={params.ratio === option.value ? "secondary" : "ghost"}
                     size="sm"
                     className="h-7 text-xs"
-                    onClick={() => updateParam("ratio", option.value)}
+                    onClick={() => updateParam("ratio", option.value as VideoGenerationParams["ratio"])}
                   >
                     {option.label}
                   </Button>
@@ -698,7 +813,7 @@ export function VideoParamsBar({
             </DropdownMenuContent>
           </DropdownMenu>
 
-          {!isFalVideo && !isRuxaVideo && (
+          {showResolutionControl && (
             <DropdownMenu>
               <DropdownMenuTrigger asChild>
                 <Button variant="ghost" size="sm" className="gap-1 px-2 h-8 focus-visible:ring-0 focus-visible:ring-offset-0">
@@ -706,7 +821,7 @@ export function VideoParamsBar({
                 </Button>
               </DropdownMenuTrigger>
               <DropdownMenuContent align="center" className="min-w-[100px]">
-                {RESOLUTION_OPTIONS.map((option) => (
+                {resolutionOptions.map((option) => (
                   <DropdownMenuItem key={option.value} onClick={() => updateParam("resolution", option.value)}>
                     <Check className={cn("mr-2 h-4 w-4", params.resolution === option.value ? "opacity-100" : "opacity-0")} />
                     {option.label}
@@ -734,7 +849,7 @@ export function VideoParamsBar({
             </DropdownMenuContent>
           </DropdownMenu>
 
-          {isFalVideo ? (
+          {showGenerateAudioControl ? (
             <Button
               variant="ghost"
               size="sm"
@@ -744,7 +859,7 @@ export function VideoParamsBar({
               <Sparkles size={14} />
               {params.generateAudio === false ? "静音" : "带音频"}
             </Button>
-          ) : isRuxaVideo ? (
+          ) : !showFpsControl && !showCameraMoveControl ? (
             <Button
               variant="ghost"
               size="sm"
@@ -756,54 +871,58 @@ export function VideoParamsBar({
             </Button>
           ) : (
             <>
-              <DropdownMenu>
-                <DropdownMenuTrigger asChild>
-                  <Button variant="ghost" size="sm" className="gap-1 px-2 h-8 focus-visible:ring-0 focus-visible:ring-offset-0">
-                    {params.fps || 24}fps
-                  </Button>
-                </DropdownMenuTrigger>
-                <DropdownMenuContent align="center" className="min-w-[100px]">
-                  {FPS_OPTIONS.map((option) => (
-                    <DropdownMenuItem
-                      key={option.value}
-                      onClick={() => updateParam("fps", option.value as VideoGenerationParams["fps"])}
-                    >
-                      <Check
-                        className={cn(
-                          "mr-2 h-4 w-4",
-                          (params.fps || 24) === option.value ? "opacity-100" : "opacity-0"
-                        )}
-                      />
-                      {option.label}
-                    </DropdownMenuItem>
-                  ))}
-                </DropdownMenuContent>
-              </DropdownMenu>
+              {showFpsControl && (
+                <DropdownMenu>
+                  <DropdownMenuTrigger asChild>
+                    <Button variant="ghost" size="sm" className="gap-1 px-2 h-8 focus-visible:ring-0 focus-visible:ring-offset-0">
+                      {params.fps || 24}fps
+                    </Button>
+                  </DropdownMenuTrigger>
+                  <DropdownMenuContent align="center" className="min-w-[100px]">
+                    {FPS_OPTIONS.map((option) => (
+                      <DropdownMenuItem
+                        key={option.value}
+                        onClick={() => updateParam("fps", option.value as VideoGenerationParams["fps"])}
+                      >
+                        <Check
+                          className={cn(
+                            "mr-2 h-4 w-4",
+                            (params.fps || 24) === option.value ? "opacity-100" : "opacity-0"
+                          )}
+                        />
+                        {option.label}
+                      </DropdownMenuItem>
+                    ))}
+                  </DropdownMenuContent>
+                </DropdownMenu>
+              )}
 
-              <DropdownMenu>
-                <DropdownMenuTrigger asChild>
-                  <Button variant="ghost" size="sm" className="gap-1 px-2 h-8 focus-visible:ring-0 focus-visible:ring-offset-0">
-                    <Camera size={14} />
-                    {getCurrentCameraLabel()}
-                  </Button>
-                </DropdownMenuTrigger>
-                <DropdownMenuContent align="center" className="min-w-[140px]">
-                  {CAMERA_MOVE_OPTIONS.map((option) => (
-                    <DropdownMenuItem
-                      key={option.value}
-                      onSelect={() => {
-                        updateParam("cameraMove", option.value);
-                        updateParam("cameraFixed", option.value === "static");
-                      }}
-                    >
-                      <Check
-                        className={cn("mr-2 h-4 w-4", params.cameraMove === option.value ? "opacity-100" : "opacity-0")}
-                      />
-                      {option.label}
-                    </DropdownMenuItem>
-                  ))}
-                </DropdownMenuContent>
-              </DropdownMenu>
+              {showCameraMoveControl && (
+                <DropdownMenu>
+                  <DropdownMenuTrigger asChild>
+                    <Button variant="ghost" size="sm" className="gap-1 px-2 h-8 focus-visible:ring-0 focus-visible:ring-offset-0">
+                      <Camera size={14} />
+                      {getCurrentCameraLabel()}
+                    </Button>
+                  </DropdownMenuTrigger>
+                  <DropdownMenuContent align="center" className="min-w-[140px]">
+                    {CAMERA_MOVE_OPTIONS.map((option) => (
+                      <DropdownMenuItem
+                        key={option.value}
+                        onSelect={() => {
+                          updateParam("cameraMove", option.value);
+                          updateParam("cameraFixed", option.value === "static");
+                        }}
+                      >
+                        <Check
+                          className={cn("mr-2 h-4 w-4", params.cameraMove === option.value ? "opacity-100" : "opacity-0")}
+                        />
+                        {option.label}
+                      </DropdownMenuItem>
+                    ))}
+                  </DropdownMenuContent>
+                </DropdownMenu>
+              )}
             </>
           )}
         </div>
@@ -820,15 +939,17 @@ export function VideoParamsBar({
               <DialogTitle className="text-sm">高级设置</DialogTitle>
             </DialogHeader>
             <div className="space-y-4 py-2">
-              <div className="flex items-center justify-between">
-                <Label className="text-xs">返回尾帧</Label>
-                <Switch
-                  checked={params.returnLastFrame || false}
-                  onCheckedChange={(checked) => updateParam("returnLastFrame", checked)}
-                  className="scale-75"
-                />
-              </div>
-              {isFalVideo && (
+              {showReturnLastFrameControl && (
+                <div className="flex items-center justify-between">
+                  <Label className="text-xs">返回尾帧</Label>
+                  <Switch
+                    checked={params.returnLastFrame || false}
+                    onCheckedChange={(checked) => updateParam("returnLastFrame", checked)}
+                    className="scale-75"
+                  />
+                </div>
+              )}
+              {showGenerateAudioControl && (
                 <div className="flex items-center justify-between">
                   <Label className="text-xs">生成音频</Label>
                   <Switch
@@ -838,18 +959,20 @@ export function VideoParamsBar({
                   />
                 </div>
               )}
-              <div className="space-y-1.5">
-                <Label className="text-xs">随机种子</Label>
-                <Input
-                  type="number"
-                  placeholder="留空则随机"
-                  value={params.seed || ""}
-                  onChange={(e) =>
-                    updateParam("seed", Number(e.target.value) || undefined)
-                  }
-                  className="h-8 text-xs"
-                />
-              </div>
+              {showSeedControl && (
+                <div className="space-y-1.5">
+                  <Label className="text-xs">随机种子</Label>
+                  <Input
+                    type="number"
+                    placeholder="留空则随机"
+                    value={params.seed || ""}
+                    onChange={(e) =>
+                      updateParam("seed", Number(e.target.value) || undefined)
+                    }
+                    className="h-8 text-xs"
+                  />
+                </div>
+              )}
             </div>
           </DialogContent>
         </Dialog>

lib/ai/video-tool-capability.ts

@@ -0,0 +1,322 @@
+import type { VideoModelCapabilities } from "@/types/ai";
+import type {
+  VideoToolCapabilityContract,
+  VideoToolMode,
+  VideoToolModeOption,
+  VideoToolResolution,
+} from "@/types/video-tool-capability";
+
+const MODE_LABELS: Record<VideoToolMode, string> = {
+  "text-to-video": "文生视频",
+  "first-frame": "主体参考",
+  "first-last-frame": "首尾帧",
+  "reference-images": "智能多帧",
+};
+
+const MODE_ORDER: VideoToolMode[] = [
+  "text-to-video",
+  "first-last-frame",
+  "reference-images",
+  "first-frame",
+];
+
+const STANDARD_RATIOS = ["21:9", "16:9", "4:3", "1:1", "3:4", "9:16"];
+const PORTRAIT_RATIOS = ["16:9", "9:16", "1:1"];
+const STANDARD_RESOLUTIONS: VideoToolResolution[] = ["480p", "720p", "1080p"];
+const STANDARD_DURATIONS = [5, 10];
+
+function asObject(value: unknown) {
+  return value && typeof value === "object" ? (value as Record<string, any>) : null;
+}
+
+function normalizeDurations(min?: number, max?: number) {
+  const safeMin = Number.isFinite(min) ? Math.max(1, Math.round(Number(min))) : undefined;
+  const safeMax = Number.isFinite(max) ? Math.max(1, Math.round(Number(max))) : undefined;
+  if (!safeMin || !safeMax || safeMax < safeMin) {
+    return [...STANDARD_DURATIONS];
+  }
+
+  if (safeMax - safeMin <= 12) {
+    return Array.from({ length: safeMax - safeMin + 1 }, (_, index) => safeMin + index);
+  }
+  return [safeMin, safeMax];
+}
+
+function normalizeModeOptions(
+  modes: VideoToolMode[],
+  maxReferenceImages = 4
+): VideoToolModeOption[] {
+  return MODE_ORDER.filter((mode) => modes.includes(mode)).map((mode) => {
+    if (mode === "text-to-video") {
+      return { mode, label: MODE_LABELS[mode], imageMin: 0, imageMax: 0 };
+    }
+    if (mode === "first-frame") {
+      return { mode, label: MODE_LABELS[mode], imageMin: 1, imageMax: 1 };
+    }
+    if (mode === "first-last-frame") {
+      return { mode, label: MODE_LABELS[mode], imageMin: 2, imageMax: 2 };
+    }
+    return {
+      mode,
+      label: MODE_LABELS[mode],
+      imageMin: 1,
+      imageMax: Math.max(1, Math.min(8, maxReferenceImages)),
+    };
+  });
+}
+
+function normalizeVideoCapabilities(rawCapabilities: unknown): Partial<VideoModelCapabilities> {
+  const raw = asObject(rawCapabilities);
+  if (!raw) return {};
+
+  const supportedModes = Array.isArray(raw.supportedModes)
+    ? raw.supportedModes.filter((item): item is VideoToolMode =>
+        item === "text-to-video" ||
+        item === "first-frame" ||
+        item === "first-last-frame" ||
+        item === "reference-images"
+      )
+    : undefined;
+
+  const supportedRatios = Array.isArray(raw.supportedRatios)
+    ? raw.supportedRatios
+        .map((item) => String(item || "").trim())
+        .filter(Boolean)
+    : undefined;
+
+  const supportedResolutions = Array.isArray(raw.supportedResolutions)
+    ? raw.supportedResolutions.filter(
+        (item): item is VideoToolResolution => item === "480p" || item === "720p" || item === "1080p"
+      )
+    : undefined;
+
+  const durationRange = asObject(raw.durationRange);
+  const min = Number(durationRange?.min);
+  const max = Number(durationRange?.max);
+
+  return {
+    supportedModes,
+    supportedRatios: supportedRatios as any,
+    supportedResolutions: supportedResolutions as any,
+    durationRange:
+      Number.isFinite(min) && Number.isFinite(max)
+        ? { min: Math.round(min), max: Math.round(max) }
+        : undefined,
+    maxReferenceImages: Number.isFinite(Number(raw.maxReferenceImages))
+      ? Math.round(Number(raw.maxReferenceImages))
+      : undefined,
+    supportReturnLastFrame:
+      typeof raw.supportReturnLastFrame === "boolean" ? raw.supportReturnLastFrame : undefined,
+    supportSeed: typeof raw.supportSeed === "boolean" ? raw.supportSeed : undefined,
+  };
+}
+
+function createProviderDefaultCapability(providerType: string): VideoToolCapabilityContract {
+  if (providerType === "fal-video") {
+    return {
+      schemaVersion: "v1",
+      source: "provider-default",
+      providerType,
+      modeOptions: normalizeModeOptions(["first-frame", "first-last-frame"]),
+      ratioOptions: [...PORTRAIT_RATIOS],
+      durationOptions: Array.from({ length: 13 }, (_, index) => index + 3),
+      resolutionOptions: [],
+      ui: {
+        showResolution: false,
+        showFps: false,
+        showCameraMove: false,
+        showGenerateAudio: true,
+        showReturnLastFrame: true,
+        showSeed: true,
+      },
+      defaults: {
+        mode: "first-frame",
+        ratio: "16:9",
+        duration: 5,
+        generateAudio: true,
+        fps: 24,
+      },
+    };
+  }
+
+  if (providerType === "ruxa-video") {
+    return {
+      schemaVersion: "v1",
+      source: "provider-default",
+      providerType,
+      modeOptions: normalizeModeOptions(["first-frame"]),
+      ratioOptions: [...PORTRAIT_RATIOS],
+      durationOptions: [5, 10],
+      resolutionOptions: [],
+      ui: {
+        showResolution: false,
+        showFps: false,
+        showCameraMove: false,
+        showGenerateAudio: false,
+        showReturnLastFrame: false,
+        showSeed: false,
+      },
+      defaults: {
+        mode: "first-frame",
+        ratio: "16:9",
+        duration: 10,
+        generateAudio: false,
+        fps: 24,
+      },
+    };
+  }
+
+  if (providerType === "qianwen-video") {
+    return {
+      schemaVersion: "v1",
+      source: "provider-default",
+      providerType,
+      modeOptions: normalizeModeOptions(["first-frame"]),
+      ratioOptions: ["16:9"],
+      durationOptions: [5],
+      resolutionOptions: [],
+      ui: {
+        showResolution: false,
+        showFps: false,
+        showCameraMove: false,
+        showGenerateAudio: false,
+        showReturnLastFrame: false,
+        showSeed: false,
+      },
+      defaults: {
+        mode: "first-frame",
+        ratio: "16:9",
+        duration: 5,
+        generateAudio: false,
+        fps: 24,
+      },
+    };
+  }
+
+  return {
+    schemaVersion: "v1",
+    source: "provider-default",
+    providerType,
+    modeOptions: normalizeModeOptions([
+      "text-to-video",
+      "first-last-frame",
+      "reference-images",
+      "first-frame",
+    ]),
+    ratioOptions: [...STANDARD_RATIOS],
+    durationOptions: [...STANDARD_DURATIONS],
+    resolutionOptions: [...STANDARD_RESOLUTIONS],
+    ui: {
+      showResolution: true,
+      showFps: true,
+      showCameraMove: true,
+      showGenerateAudio: false,
+      showReturnLastFrame: true,
+      showSeed: true,
+    },
+    defaults: {
+      mode: "text-to-video",
+      ratio: "16:9",
+      duration: 5,
+      resolution: "720p",
+      generateAudio: false,
+      fps: 24,
+    },
+  };
+}
+
+function pickPreferredRatio(ratioOptions: string[]) {
+  if (ratioOptions.includes("16:9")) return "16:9";
+  return ratioOptions[0] || "16:9";
+}
+
+function pickPreferredDuration(durationOptions: number[]) {
+  if (durationOptions.includes(5)) return 5;
+  if (durationOptions.includes(10)) return 10;
+  return durationOptions[0] || 5;
+}
+
+function pickPreferredMode(modeOptions: VideoToolModeOption[], fallback: VideoToolMode) {
+  const modeSet = new Set(modeOptions.map((item) => item.mode));
+  if (modeSet.has(fallback)) return fallback;
+  if (modeSet.has("first-frame")) return "first-frame";
+  if (modeSet.has("text-to-video")) return "text-to-video";
+  return modeOptions[0]?.mode || fallback;
+}
+
+export function buildVideoToolCapability(params: {
+  providerType?: string | null;
+  modelId?: string | null;
+  rawCapabilities?: unknown;
+}): VideoToolCapabilityContract | null {
+  const providerType = String(params.providerType || "").trim();
+  if (!providerType) return null;
+  const base = createProviderDefaultCapability(providerType);
+
+  if (providerType === "fal-video" || providerType === "qianwen-video" || providerType === "ruxa-video") {
+    return {
+      ...base,
+      modelId: params.modelId || undefined,
+    };
+  }
+
+  const parsed = normalizeVideoCapabilities(params.rawCapabilities);
+  const hasModelCapability =
+    Boolean(parsed.supportedModes?.length) ||
+    Boolean(parsed.supportedRatios?.length) ||
+    Boolean(parsed.supportedResolutions?.length) ||
+    Boolean(parsed.durationRange);
+
+  if (!hasModelCapability) {
+    return {
+      ...base,
+      modelId: params.modelId || undefined,
+    };
+  }
+
+  const supportedModes =
+    parsed.supportedModes && parsed.supportedModes.length > 0
+      ? parsed.supportedModes
+      : base.modeOptions.map((item) => item.mode);
+
+  const modeOptions = normalizeModeOptions(supportedModes, parsed.maxReferenceImages || 4);
+  const ratioOptions =
+    parsed.supportedRatios && parsed.supportedRatios.length > 0
+      ? Array.from(new Set(parsed.supportedRatios))
+      : base.ratioOptions;
+  const resolutionOptions =
+    parsed.supportedResolutions && parsed.supportedResolutions.length > 0
+      ? parsed.supportedResolutions
+      : base.resolutionOptions;
+  const durationOptions = parsed.durationRange
+    ? normalizeDurations(parsed.durationRange.min, parsed.durationRange.max)
+    : base.durationOptions;
+
+  return {
+    ...base,
+    source: "model-capabilities",
+    modelId: params.modelId || undefined,
+    modeOptions,
+    ratioOptions,
+    resolutionOptions,
+    durationOptions,
+    ui: {
+      ...base.ui,
+      showReturnLastFrame:
+        typeof parsed.supportReturnLastFrame === "boolean"
+          ? parsed.supportReturnLastFrame
+          : base.ui.showReturnLastFrame,
+      showSeed:
+        typeof parsed.supportSeed === "boolean" ? parsed.supportSeed : base.ui.showSeed,
+    },
+    defaults: {
+      ...base.defaults,
+      mode: pickPreferredMode(modeOptions, base.defaults.mode),
+      ratio: pickPreferredRatio(ratioOptions),
+      duration: pickPreferredDuration(durationOptions),
+      resolution: resolutionOptions.includes("720p")
+        ? "720p"
+        : resolutionOptions[0] || undefined,
+    },
+  };
+}

types/video-tool-capability.ts

@@ -0,0 +1,45 @@
+export type VideoToolMode =
+  | "text-to-video"
+  | "first-frame"
+  | "first-last-frame"
+  | "reference-images";
+
+export type VideoToolResolution = "480p" | "720p" | "1080p";
+
+export interface VideoToolModeOption {
+  mode: VideoToolMode;
+  label: string;
+  imageMin: number;
+  imageMax: number;
+}
+
+export interface VideoToolUiToggles {
+  showResolution: boolean;
+  showFps: boolean;
+  showCameraMove: boolean;
+  showGenerateAudio: boolean;
+  showReturnLastFrame: boolean;
+  showSeed: boolean;
+}
+
+export interface VideoToolDefaults {
+  mode: VideoToolMode;
+  ratio: string;
+  duration: number;
+  resolution?: VideoToolResolution;
+  generateAudio?: boolean;
+  fps?: number;
+}
+
+export interface VideoToolCapabilityContract {
+  schemaVersion: "v1";
+  source: "provider-default" | "model-capabilities";
+  providerType: string;
+  modelId?: string;
+  modeOptions: VideoToolModeOption[];
+  ratioOptions: string[];
+  durationOptions: number[];
+  resolutionOptions: VideoToolResolution[];
+  ui: VideoToolUiToggles;
+  defaults: VideoToolDefaults;
+}