@shop-os/modelkit 0.12.0 → 0.14.0

package/dist/index.d.cts

@@ -132,7 +132,7 @@ interface Model3dConstraints {
     };
 }
 /** What a `video-tool` row does to its source clip. */
-type VideoToolKind = "upscale" | "interpolate" | "hdr" | "bg-remove" | "subtitles";
+type VideoToolKind = "upscale" | "interpolate" | "hdr" | "bg-remove" | "subtitles" | "depth";
 /** Container/codec a video-tool row can write. */
 /** `exr`: a ZIP of 16-bit EXR frames in `VideoToolResult.frames`, with an MP4 preview as `video`. */
 type VideoToolOutputFormat = "h265" | "prores" | "exr";
@@ -1351,6 +1351,7 @@ declare const CatalogModelSchemaBase: z.ZodObject<{
                 interpolate: "interpolate";
                 hdr: "hdr";
                 subtitles: "subtitles";
+                depth: "depth";
             }>;
             upscaleFactor: z.ZodOptional<z.ZodObject<{
                 min: z.ZodNumber;

package/dist/index.d.ts

@@ -132,7 +132,7 @@ interface Model3dConstraints {
     };
 }
 /** What a `video-tool` row does to its source clip. */
-type VideoToolKind = "upscale" | "interpolate" | "hdr" | "bg-remove" | "subtitles";
+type VideoToolKind = "upscale" | "interpolate" | "hdr" | "bg-remove" | "subtitles" | "depth";
 /** Container/codec a video-tool row can write. */
 /** `exr`: a ZIP of 16-bit EXR frames in `VideoToolResult.frames`, with an MP4 preview as `video`. */
 type VideoToolOutputFormat = "h265" | "prores" | "exr";
@@ -1351,6 +1351,7 @@ declare const CatalogModelSchemaBase: z.ZodObject<{
                 interpolate: "interpolate";
                 hdr: "hdr";
                 subtitles: "subtitles";
+                depth: "depth";
             }>;
             upscaleFactor: z.ZodOptional<z.ZodObject<{
                 min: z.ZodNumber;

package/dist/index.js

@@ -6569,6 +6569,33 @@ 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: [],
@@ -7786,7 +7813,7 @@ var LanguageOptionSchema = z3.looseObject({
   name: z3.string().min(1)
 });
 var VideoToolConstraintsSchema = z3.looseObject({
-  kind: z3.enum(["upscale", "interpolate", "hdr", "bg-remove", "subtitles"]),
+  kind: z3.enum(["upscale", "interpolate", "hdr", "bg-remove", "subtitles", "depth"]),
   upscaleFactor: RangeSchema.optional(),
   maxOutputLongEdge: z3.number().positive().optional(),
   prompt: z3.boolean().optional(),
@@ -9556,7 +9583,7 @@ function checkedQueueUrl(ctx, url, what) {
 import { z as z12 } from "zod";
 var RangeSchema2 = z12.object({ min: z12.number(), max: z12.number() });
 var ToolSchema = z12.object({
-  kind: z12.enum(["upscale", "interpolate", "hdr", "bg-remove", "subtitles"]),
+  kind: z12.enum(["upscale", "interpolate", "hdr", "bg-remove", "subtitles", "depth"]),
   upscaleFactor: RangeSchema2.optional(),
   maxOutputLongEdge: z12.number().positive().optional(),
   prompt: z12.boolean().optional(),
@@ -9694,6 +9721,25 @@ 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(
@@ -9790,12 +9836,15 @@ 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;
@@ -18336,12 +18385,12 @@ function keysFromEnv(env, prefix = "") {
   set("moonshot", pick2("MOONSHOT_API_KEY"));
   set("minimax", pick2("MINIMAX_API_KEY"));
   set("ark", pick2("BYTEDANCE_API_KEY"));
-  set("fal", pick2("FAL_KEY", "FAL_API_KEY"));
+  set("fal", pick2("FAL_KEY"));
   set("replicate", pick2("REPLICATE_API_TOKEN"));
   set("prodia", pick2("PRODIA_TOKEN"));
   set("higgsfield", pick2("HIGGSFIELD_API_KEY"));
   set("higgsfieldSecret", pick2("HIGGSFIELD_API_SECRET"));
-  set("luma", pick2("LUMA_API_KEY", "LUMA_AGENTS_API_KEY"));
+  set("luma", pick2("LUMA_API_KEY"));
   set("fish", pick2("FISH_AUDIO_API_KEY"));
   set("elevenlabs", pick2("ELEVENLABS_API_KEY"));
   set("cloudflare", pick2("CF_AIG_TOKEN"));