@shop-os/modelkit 0.4.6 → 0.4.7

package/dist/index.cjs

@@ -52,6 +52,7 @@ __export(index_exports, {
   SurfaceSchema: () => SurfaceSchema,
   VIDEO_INPUT_SPEC_KEYS: () => VIDEO_INPUT_SPEC_KEYS,
   VIDEO_TOOL_MODELS: () => VIDEO_TOOL_MODELS,
+  VideoResolutionSchema: () => VideoResolutionSchema,
   adapterServesType: () => adapterServesType,
   aiSdkTextAdapter: () => aiSdkTextAdapter,
   arkAdapter: () => arkAdapter,
@@ -3932,17 +3933,19 @@ var SeedanceResponseSchema = import_zod7.z.object({
   error: import_zod7.z.union([import_zod7.z.string(), import_zod7.z.object({ message: import_zod7.z.string().optional() })]).optional()
 });
 var SeedanceErrorMessageSchema = import_zod7.z.object({ message: import_zod7.z.string().optional() });
+function seedanceResolution(req) {
+  return req.resolution ?? "720p";
+}
 function seedanceUsage(ctx, req, duration) {
-  const usage2 = {
+  return {
     model: ctx.model.id,
     provider: "ark",
     modality: "VIDEO",
     unit: "seconds",
     quantity: duration,
-    durationSeconds: duration
+    durationSeconds: duration,
+    resolution: seedanceResolution(req)
   };
-  if (req.resolution !== void 0) usage2.resolution = req.resolution;
-  return usage2;
 }
 function seedanceInitialDuration(req) {
   return req.durationSeconds ?? 5;
@@ -4007,6 +4010,10 @@ function seedanceJob(ctx, key, jobId, usage2) {
             usage2.quantity = polled.duration;
             usage2.durationSeconds = polled.duration;
           }
+          const rendered = VideoResolutionSchema.safeParse(
+            polled.resolution?.toLowerCase()
+          );
+          if (rendered.success) usage2.resolution = rendered.data;
           const tokens = polled.usage;
           if (tokens?.total_tokens !== void 0 && tokens.completion_tokens !== void 0) {
             usage2.inputTokens = tokens.total_tokens - tokens.completion_tokens;
@@ -4043,7 +4050,7 @@ async function generateSeedanceVideo(req, ctx) {
   const requestedDuration = seedanceInitialDuration(req);
   let duration = requestedDuration;
   if (taskType === "edit") duration = -1;
-  const resolution = req.resolution ?? "720p";
+  const resolution = seedanceResolution(req);
   const content = [{ type: "text", text: req.prompt }];
   if (req.firstFrame) {
     content.push({
@@ -4911,9 +4918,15 @@ function videoUsage(ctx, req, duration) {
     quantity: duration,
     durationSeconds: duration
   };
-  if (req.resolution !== void 0) event.resolution = req.resolution;
+  const resolution = falVideoResolution(ctx.model.providerModelId, req);
+  if (resolution !== void 0) event.resolution = resolution;
   return event;
 }
+function falVideoResolution(providerModelId, req) {
+  if (isWan(providerModelId)) return wanResolution(req);
+  if (isMinimax(providerModelId)) return minimaxResolution(req, minimaxFamily(providerModelId));
+  return req.resolution;
+}
 async function decodeImage(ctx, url, contentType) {
   if (url.startsWith("data:")) {
     const comma = url.indexOf(",");
@@ -5045,12 +5058,14 @@ function isFalVideoKnown(providerModelId) {
 function minimaxFamily(providerModelId) {
   return H3_MAX_MODEL_IDS.has(providerModelId.toLowerCase()) ? "h3-max" : "h3";
 }
+function minimaxResolution(req, family) {
+  return req.resolution ?? (family === "h3-max" ? "768p" : "2k");
+}
 function minimaxBase(req, family) {
-  const resolution = (family === "h3-max" ? req.resolution ?? "768p" : req.resolution ?? "2k").toUpperCase();
   const input = {
     prompt: req.prompt.trim(),
     duration: req.durationSeconds ?? 5,
-    resolution,
+    resolution: minimaxResolution(req, family).toUpperCase(),
     prompt_expansion_mode: req.promptExpansion ?? "balanced"
   };
   if (req.safetyChecker !== void 0) input.enable_safety_checker = req.safetyChecker;
@@ -5061,11 +5076,14 @@ var WAN_MODEL_IDS = /* @__PURE__ */ new Set(["alibaba/wan-3.0"]);
 function isWan(providerModelId) {
   return WAN_MODEL_IDS.has(providerModelId.toLowerCase());
 }
+function wanResolution(req) {
+  return req.resolution ?? "1080p";
+}
 function wanVideo(req, family) {
   const input = {
     prompt: req.prompt.trim(),
     duration: req.durationSeconds ?? 5,
-    resolution: req.resolution ?? "1080p",
+    resolution: wanResolution(req),
     aspect_ratio: req.aspectRatio ?? "adaptive",
     audio: req.generateAudio ?? true
   };
@@ -5664,7 +5682,7 @@ async function downloadVideo(ctx, url, apiKey, signal) {
     mime: res.headers.get("content-type")?.split(";")[0]?.trim() || "video/mp4"
   };
 }
-function videoUsage2(ctx, req, duration) {
+function videoUsage2(ctx, duration, resolution) {
   const event = {
     model: ctx.model.id,
     provider: "google-genai",
@@ -5673,7 +5691,7 @@ function videoUsage2(ctx, req, duration) {
     quantity: duration,
     durationSeconds: duration
   };
-  if (req.resolution !== void 0) event.resolution = req.resolution;
+  if (resolution !== void 0) event.resolution = resolution;
   return event;
 }
 async function generateVeoVideo(req, ctx, createGenAIClient) {
@@ -5767,7 +5785,7 @@ async function generateVeoVideo(req, ctx, createGenAIClient) {
     }
   };
   const jobId = operation.name ?? crypto.randomUUID();
-  const usage2 = videoUsage2(ctx, req, duration);
+  const usage2 = videoUsage2(ctx, duration, req.resolution ?? "720p");
   return {
     jobId,
     async poll() {
@@ -5891,10 +5909,8 @@ async function generateOmniVideo(req, ctx, createGenAIClient) {
   if (req.sourceVideo === void 0 && !extend && req.aspectRatio !== void 0 && req.aspectRatio !== "adaptive") {
     responseFormat.aspect_ratio = toOmniAspect(req.aspectRatio);
   }
-  if (!isOmniV1) {
-    const omniResolution = toOmniResolution(req.resolution);
-    if (omniResolution !== void 0) responseFormat.resolution = omniResolution;
-  }
+  const omniResolution = isOmniV1 ? void 0 : toOmniResolution(req.resolution);
+  if (omniResolution !== void 0) responseFormat.resolution = omniResolution;
   if (req.durationSeconds !== void 0) responseFormat.duration = `${req.durationSeconds}s`;
   const client2 = createGenAIClient({
     apiKey,
@@ -6012,7 +6028,7 @@ async function generateOmniVideo(req, ctx, createGenAIClient) {
     if (!(extend && !chained)) await cleanupUploads();
   }
   const jobId = interaction.id ?? crypto.randomUUID();
-  const usage2 = videoUsage2(ctx, req, duration);
+  const usage2 = videoUsage2(ctx, duration, omniResolution ?? "720p");
   const getInteraction = async () => {
     try {
       const getParams = { stream: false };
@@ -7594,7 +7610,8 @@ function videoUsage4(ctx, req) {
     unit: "compute_seconds",
     quantity: 0
   };
-  if (req.resolution !== void 0) event.resolution = req.resolution;
+  const resolution = replicateVideoResolution(ctx.model.providerModelId, req);
+  if (resolution !== void 0) event.resolution = resolution;
   if (req.durationSeconds !== void 0) event.durationSeconds = req.durationSeconds;
   return event;
 }
@@ -7831,7 +7848,7 @@ function klingInput(req) {
   const input = {
     prompt: req.prompt,
     duration: req.durationSeconds ?? 10,
-    mode: req.mode ?? "standard",
+    mode: klingTier(req).mode,
     generate_audio: req.generateAudio ?? false
   };
   const startImage = images[0];
@@ -7845,17 +7862,25 @@ function klingInput(req) {
   if (req.negativePrompt !== void 0) input.negative_prompt = req.negativePrompt;
   return input;
 }
-function klingOmniMode(req) {
-  if (req.resolution === "4k") return "4k";
-  if (req.resolution === "1080p") return "pro";
-  if (req.resolution === "720p") return "standard";
-  return req.mode ?? "standard";
+function replicateVideoResolution(providerModelId, req) {
+  const kind = videoKind(providerModelId);
+  if (kind === "kling" || kind === "kling-omni") return klingTier(req).resolution;
+  if (kind === "hailuo") return hailuoResolution(req);
+  return req.resolution;
+}
+var KLING_TIERS = [
+  { mode: "standard", resolution: "720p" },
+  { mode: "pro", resolution: "1080p" },
+  { mode: "4k", resolution: "4k" }
+];
+function klingTier(req) {
+  return KLING_TIERS.find((tier) => tier.resolution === req.resolution) ?? KLING_TIERS.find((tier) => tier.mode === req.mode) ?? KLING_TIERS[0];
 }
 function klingOmniInput(req) {
   const input = {
     prompt: req.prompt,
     duration: req.durationSeconds ?? 10,
-    mode: klingOmniMode(req),
+    mode: klingTier(req).mode,
     generate_audio: req.generateAudio ?? false
   };
   const start = req.firstFrame ?? req.lastFrame;
@@ -7869,11 +7894,14 @@ function klingOmniInput(req) {
   if (refs.length > 0) input.reference_images = refs;
   return input;
 }
+function hailuoResolution(req) {
+  return req.resolution ?? "768p";
+}
 function hailuoInput(req) {
   const input = {
     prompt: req.prompt,
     duration: req.durationSeconds ?? 6,
-    resolution: req.resolution ?? "768p",
+    resolution: hailuoResolution(req),
     prompt_optimizer: true
   };
   if (req.firstFrame) input.first_frame_image = refToUrl(req.firstFrame);
@@ -10102,6 +10130,7 @@ var describe = (modelId) => createDefaultModelKit().describe(modelId);
   SurfaceSchema,
   VIDEO_INPUT_SPEC_KEYS,
   VIDEO_TOOL_MODELS,
+  VideoResolutionSchema,
   adapterServesType,
   aiSdkTextAdapter,
   arkAdapter,