@shop-os/modelkit 0.13.0 → 0.15.0

package/dist/index.cjs

@@ -136,6 +136,7 @@ __export(index_exports, {
   toGeneratedImage: () => toGeneratedImage,
   validateInputSpec: () => validateInputSpec,
   videoLengthViolation: () => videoLengthViolation,
+  videoSizeViolation: () => videoSizeViolation,
   videoToolConstraints: () => videoToolConstraints,
   withRetry: () => withRetry,
   xaiVideoAdapter: () => xaiVideoAdapter
@@ -5235,6 +5236,7 @@ function klingMotionControl(tier) {
         videos: 1,
         audios: 0,
         videoSeconds: { min: 3, max: 30.05 },
+        videoPixels: { minWidth: 340, minHeight: 340 },
         min: { images: 1, videos: 1 }
       },
       requires: ["references"],
@@ -5301,6 +5303,7 @@ var VIDEO_FAMILY_MODELS = [
       references: { ...NO_REFERENCES, images: 4 },
       requires: ["sourceVideo"],
       sourceVideoSeconds: { min: 3, max: 15.05 },
+      sourceVideoPixels: { minWidth: 720, minHeight: 720 },
       // generateAudio keeps (true) or drops (false) the source soundtrack.
       supports: ["generateAudio"],
       prompt: KLING_PROMPT,
@@ -5341,7 +5344,8 @@ var VIDEO_FAMILY_MODELS = [
         ...NO_REFERENCES,
         images: 10,
         videos: 5,
-        videoSeconds: { min: 3, max: 30.05, total: 30 }
+        videoSeconds: { min: 3, max: 30.05, total: 30 },
+        videoPixels: { minWidth: 700, minHeight: 700 }
       },
       supports: ["lastFrame"],
       prompt: KLING_V4_PROMPT,
@@ -5375,6 +5379,7 @@ var VIDEO_FAMILY_MODELS = [
       references: { ...NO_REFERENCES, images: 10 },
       requires: ["sourceVideo"],
       sourceVideoSeconds: { min: 3, max: 30.05 },
+      sourceVideoPixels: { minWidth: 700, minHeight: 700 },
       supports: [],
       prompt: KLING_V4_PROMPT,
       audio: KLING_V4_AUDIO
@@ -6711,6 +6716,33 @@ var VIDEO_TOOL_MODELS = [
       description: "Topaz Hyperion 2.5 SDR to HDR: a 10-bit HDR master in H.265 (HDR10) or ProRes 422 HQ, at the source's size and frame rate."
     }
   },
+  {
+    // https://fal.ai/models/fal-ai/depth-anything-video/api: greyscale depth (white = near)
+    // at the source's frame rate, short edge capped at 1080 px (a 2160x3840 source came
+    // back 1080x1920), at most 2400 frames; one fal price of $0.04 per second for every
+    // model size. The adapter always sends VDA-Small, the only Apache-2.0 size. 100 s is
+    // 2400 frames at 24 fps; the frame check refuses faster clips sooner.
+    id: "depth-anything-video",
+    aliases: [],
+    label: "Video Depth Anything",
+    type: "video-tool",
+    adapter: "fal",
+    providerModelId: "fal-ai/depth-anything-video",
+    lifecycle: "active",
+    exposure: ["cowork"],
+    constraints: {
+      supports: [],
+      sourceVideoSeconds: { max: 100 },
+      tool: { kind: "depth", maxOutputShortEdge: 1080, maxOutputFrames: 2400 }
+    },
+    defaults: {},
+    usage: SECONDS_USAGE,
+    pricingKind: "video_per_second",
+    display: {
+      provider: "depth-anything",
+      description: "Video Depth Anything: a greyscale depth map of the clip (near is white), frame for frame, for motion and camera control."
+    }
+  },
   {
     id: "flux-video-upscale-precise",
     aliases: [],
@@ -7892,6 +7924,10 @@ var VideoSecondsLimitsSchema = import_zod3.z.looseObject({
   total: import_zod3.z.number().optional(),
   editMin: import_zod3.z.number().optional()
 });
+var VideoPixelLimitsSchema = import_zod3.z.looseObject({
+  minWidth: import_zod3.z.number().int().positive().optional(),
+  minHeight: import_zod3.z.number().int().positive().optional()
+});
 var ReferenceLimitsSchema = import_zod3.z.looseObject({
   images: import_zod3.z.number().optional(),
   imagesWithVideo: import_zod3.z.number().optional(),
@@ -7900,6 +7936,7 @@ var ReferenceLimitsSchema = import_zod3.z.looseObject({
   total: import_zod3.z.number().optional(),
   audioOnly: import_zod3.z.boolean().optional(),
   videoSeconds: VideoSecondsLimitsSchema.optional(),
+  videoPixels: VideoPixelLimitsSchema.optional(),
   audioSeconds: import_zod3.z.looseObject({
     min: import_zod3.z.number().optional(),
     max: import_zod3.z.number().optional(),
@@ -7928,7 +7965,7 @@ var LanguageOptionSchema = import_zod3.z.looseObject({
   name: import_zod3.z.string().min(1)
 });
 var VideoToolConstraintsSchema = import_zod3.z.looseObject({
-  kind: import_zod3.z.enum(["upscale", "interpolate", "hdr", "bg-remove", "subtitles"]),
+  kind: import_zod3.z.enum(["upscale", "interpolate", "hdr", "bg-remove", "subtitles", "depth"]),
   upscaleFactor: RangeSchema.optional(),
   maxOutputLongEdge: import_zod3.z.number().positive().optional(),
   prompt: import_zod3.z.boolean().optional(),
@@ -7989,6 +8026,7 @@ var ModelConstraintsSchema = import_zod3.z.looseObject({
   }).optional(),
   languages: import_zod3.z.array(LanguageOptionSchema).optional(),
   outputFormats: import_zod3.z.array(import_zod3.z.string().min(1)).min(1).optional(),
+  sourceVideoPixels: VideoPixelLimitsSchema.optional(),
   sourceVideoSeconds: VideoSecondsLimitsSchema.omit({ total: true }).extend({ inReferenceTotal: import_zod3.z.boolean().optional() }).optional(),
   // segmentSeconds/maxTotalSeconds stay required: a kit before 0.3.0 drops a row whose
   // extension lacks them, so every extension block keeps them.
@@ -9698,7 +9736,7 @@ function checkedQueueUrl(ctx, url, what) {
 var import_zod12 = require("zod");
 var RangeSchema2 = import_zod12.z.object({ min: import_zod12.z.number(), max: import_zod12.z.number() });
 var ToolSchema = import_zod12.z.object({
-  kind: import_zod12.z.enum(["upscale", "interpolate", "hdr", "bg-remove", "subtitles"]),
+  kind: import_zod12.z.enum(["upscale", "interpolate", "hdr", "bg-remove", "subtitles", "depth"]),
   upscaleFactor: RangeSchema2.optional(),
   maxOutputLongEdge: import_zod12.z.number().positive().optional(),
   prompt: import_zod12.z.boolean().optional(),
@@ -9836,6 +9874,25 @@ function planVideoTool(row2, req) {
       }
     };
   }
+  if (tool.kind === "depth") {
+    const limit = tool.maxOutputFrames;
+    if (limit !== void 0) {
+      if (fps === void 0) return fail("The source video's frame rate is required");
+      if (durationSeconds * fps > limit) {
+        return fail(`${row2.label} renders up to ${limit} frames (${seconds(limit, fps)}).`);
+      }
+    }
+    const out2 = shrinkToShortEdge(width, height, tool.maxOutputShortEdge);
+    return {
+      ok: true,
+      output: {
+        ...base,
+        ...out2,
+        resolution: resolutionTier(out2.width, out2.height),
+        format: "h264"
+      }
+    };
+  }
   if (tool.kind === "subtitles") {
     const resolution = resolutionTier(width, height);
     const problem = subtitleProblem(
@@ -9932,12 +9989,15 @@ function toolInput(ctx, req, output) {
     input["output_codec"] = "vp9";
   } else if (tool?.kind === "subtitles") {
     Object.assign(input, subtitleFields(tool, req.subtitles));
+  } else if (tool?.kind === "depth") {
+    input["model"] = "VDA-Small";
+    input["colormap"] = "grayscale";
   } else if (output.format === "exr") {
     const prompt = tool?.prompt === true ? req.prompt?.trim() : void 0;
     input["prompt"] = prompt || EXR_DEFAULT_PROMPT;
     if (output.fps !== void 0) input["frames_per_second"] = output.fps;
     if (output.frames !== void 0) input["num_frames"] = output.frames;
-  } else {
+  } else if (tool?.kind === "hdr") {
     input["output_format"] = output.format === "prores" ? "prores" : "mp4";
   }
   return input;
@@ -17707,6 +17767,28 @@ function videoLengthViolation(model, input) {
     return `${model.label} takes a source video of at least ${String(sourceMin)} s; this one is ${seconds2(source2, sourceMin)} s.`;
   return void 0;
 }
+function undersizedSide(frame, limits) {
+  if (!frame || !limits) return void 0;
+  const { width, height } = frame;
+  if (limits.minWidth !== void 0 && width !== void 0 && width > 0 && width < limits.minWidth)
+    return { side: "wide", min: limits.minWidth, actual: width };
+  if (limits.minHeight !== void 0 && height !== void 0 && height > 0 && height < limits.minHeight)
+    return { side: "tall", min: limits.minHeight, actual: height };
+  return void 0;
+}
+function videoSizeViolation(model, input) {
+  const refLimits = model.constraints.references?.videoPixels;
+  const refs = input.referenceVideos ?? [];
+  for (let i = 0; i < refs.length; i++) {
+    const under2 = undersizedSide(refs[i], refLimits);
+    if (under2)
+      return `${model.label} takes reference videos at least ${String(under2.min)} px ${under2.side}; video ${i + 1} is ${String(under2.actual)} px.`;
+  }
+  const under = undersizedSide(input.sourceVideo, model.constraints.sourceVideoPixels);
+  if (under)
+    return `${model.label} takes a source video at least ${String(under.min)} px ${under.side}; this one is ${String(under.actual)} px.`;
+  return void 0;
+}
 function resolveImageOptions(model, request, onUnsupported = "error") {
   const warnings = [];
   const c = model.constraints;
@@ -19820,6 +19902,7 @@ var describe = (modelId) => createDefaultModelKit().describe(modelId);
   toGeneratedImage,
   validateInputSpec,
   videoLengthViolation,
+  videoSizeViolation,
   videoToolConstraints,
   withRetry,
   xaiVideoAdapter,