@shop-os/modelkit 0.1.5 → 0.1.6

package/dist/index.js

@@ -105,7 +105,11 @@ function geminiImage(id, label, aliases, sizes) {
     adapter: "google-image",
     providerModelId: id,
     lifecycle: "active",
-    constraints: { aspectRatios: [...GEMINI_RATIOS], sizes, supports: ["safetySettings"] },
+    constraints: {
+      aspectRatios: [...GEMINI_RATIOS],
+      sizes,
+      supports: ["safetySettings", "temperature"]
+    },
     defaults: { size: "1K", aspectRatio: "1:1" },
     usage: { ...IMAGE_USAGE },
     pricingKind: "image_size"
@@ -244,8 +248,20 @@ var IMAGE_MODELS = [
     constraints: {
       // 97:30 has no Ark size spelling; the A+ flow pads inputs to 97:30 and
       // omits the ratio instead. Advertise only sizes the adapter can send.
-      aspectRatios: ["1:1", "2:3", "3:2", "4:3", "3:4", "4:5", "5:4", "9:16", "16:9", "21:9"],
-      sizes: ["1K", "2K"],
+      aspectRatios: [
+        "1:1",
+        "2:3",
+        "3:2",
+        "4:3",
+        "3:4",
+        "4:5",
+        "5:4",
+        "9:16",
+        "16:9",
+        "21:9",
+        "9:21"
+      ],
+      sizes: ["1K", "2K", "4K"],
       references: { images: 10 },
       supports: ["seed"]
     },
@@ -475,7 +491,7 @@ var OTHER_MODELS = [
     constraints: {
       references: { images: 1 },
       requires: ["references"],
-      supports: ["scaleFactor", "outputFormat"]
+      supports: ["scaleFactor", "outputFormat", "upscale"]
     },
     defaults: { scaleFactor: 2, outputFormat: "png" },
     usage: { unit: "compute_seconds", modality: "IMAGE" },
@@ -722,7 +738,7 @@ var VIDEO_MODELS = [
       durations: [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
       references: { images: 9, videos: 3, audios: 3, audioOnly: false },
       ratioIgnoredWith: ["firstFrame"],
-      supports: ["generateAudio", "seed", "lastFrame"]
+      supports: ["generateAudio", "seed", "lastFrame", "watermark"]
     },
     defaults: {
       durationSeconds: 5,
@@ -747,7 +763,7 @@ var VIDEO_MODELS = [
       durations: [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
       references: { images: 9, videos: 3, audios: 3, audioOnly: false },
       ratioIgnoredWith: ["firstFrame"],
-      supports: ["generateAudio", "seed", "lastFrame"]
+      supports: ["generateAudio", "seed", "lastFrame", "watermark"]
     },
     defaults: {
       durationSeconds: 5,
@@ -773,7 +789,7 @@ var VIDEO_MODELS = [
       durations: { min: 4, max: 30 },
       references: { images: 30, videos: 10, audios: 10 },
       ratioIgnoredWith: ["firstFrame"],
-      supports: ["generateAudio", "seed", "lastFrame"]
+      supports: ["generateAudio", "seed", "lastFrame", "watermark"]
     },
     defaults: {
       durationSeconds: 5,
@@ -800,7 +816,8 @@ var VIDEO_MODELS = [
       references: { images: 7 },
       supports: []
     },
-    defaults: { durationSeconds: 8, resolution: "720p", aspectRatio: "16:9" },
+    // No resolution default: xAI picks its own unless the caller asks.
+    defaults: { durationSeconds: 8, aspectRatio: "16:9" },
     usage: { unit: "reported_cost_usd", modality: "VIDEO" },
     pricingKind: "video_per_second"
   },
@@ -894,7 +911,7 @@ var VIDEO_MODELS = [
       aspectRatios: [...H3_RATIOS],
       resolutions: ["480p", "768p", "2k", "4k"],
       durations: { min: 5, max: 15 },
-      supports: ["seed", "promptExpansion"]
+      supports: ["seed", "promptExpansion", "lastFrame", "safetyChecker"]
     },
     defaults: {
       durationSeconds: 5,
@@ -917,7 +934,7 @@ var VIDEO_MODELS = [
       aspectRatios: [...H3_RATIOS],
       resolutions: ["480p", "768p"],
       durations: { min: 5, max: 15 },
-      supports: ["seed", "promptExpansion"]
+      supports: ["seed", "promptExpansion", "lastFrame", "safetyChecker"]
     },
     defaults: {
       durationSeconds: 5,
@@ -941,7 +958,7 @@ var VIDEO_MODELS = [
       resolutions: ["480p", "768p", "2k", "4k"],
       durations: { min: 5, max: 15 },
       references: { images: 9, videos: 3, audios: 3, total: 12, audioOnly: false },
-      supports: ["seed", "promptExpansion"]
+      supports: ["seed", "promptExpansion", "safetyChecker"]
     },
     defaults: {
       durationSeconds: 5,
@@ -965,7 +982,7 @@ var VIDEO_MODELS = [
       resolutions: ["480p", "768p"],
       durations: { min: 5, max: 15 },
       references: { images: 9, videos: 3, audios: 3, total: 12, audioOnly: false },
-      supports: ["seed", "promptExpansion"]
+      supports: ["seed", "promptExpansion", "safetyChecker"]
     },
     defaults: {
       durationSeconds: 5,
@@ -1093,19 +1110,22 @@ function resolveTier(field, code, requested, supported, order, model, onUnsuppor
   });
   return to;
 }
-function resolveDuration(requested, durations, model, warnings) {
+function resolveDuration(requested, durations, model, onUnsupported, warnings) {
   if (requested === void 0 || !durations) return requested;
-  let to = requested;
-  if (Array.isArray(durations)) {
-    if (durations.includes(requested)) return requested;
-    to = durations.reduce(
-      (best, d) => Math.abs(d - requested) < Math.abs(best - requested) ? d : best,
-      durations[0] ?? requested
-    );
-  } else {
-    to = Math.min(Math.max(requested, durations.min), durations.max);
-    if (to === requested) return requested;
+  const inSet = Array.isArray(durations) ? durations.includes(requested) : requested >= durations.min && requested <= durations.max;
+  if (inSet) return requested;
+  if (onUnsupported === "error") {
+    const supported = Array.isArray(durations) ? `Supported: ${durations.join(", ")}.` : `Supported range: ${durations.min}-${durations.max}.`;
+    throw new ModelKitError({
+      code: "unsupported_option",
+      message: `${model.id} does not support durationSeconds ${requested}. ${supported}`,
+      model: model.id
+    });
   }
+  const to = Array.isArray(durations) ? durations.reduce(
+    (best, d) => Math.abs(d - requested) < Math.abs(best - requested) ? d : best,
+    durations[0] ?? requested
+  ) : Math.min(Math.max(requested, durations.min), durations.max);
   warnings.push({
     code: "duration_coerced",
     field: "durationSeconds",
@@ -1115,6 +1135,26 @@ function resolveDuration(requested, durations, model, warnings) {
   });
   return to;
 }
+function resolveAspectRatio(requested, supported, model, onUnsupported, warnings) {
+  if (requested === void 0 || requested === "adaptive" || !supported) return requested;
+  if (supported.includes(requested)) return requested;
+  if (onUnsupported === "error") {
+    throw new ModelKitError({
+      code: "unsupported_option",
+      message: `${model.id} does not support aspectRatio "${requested}". Supported: ${supported.join(", ")}.`,
+      model: model.id
+    });
+  }
+  const to = nearestAspectRatio(requested, supported);
+  warnings.push({
+    code: "ratio_coerced",
+    field: "aspectRatio",
+    from: requested,
+    to,
+    message: `aspect ratio ${requested} coerced to ${to} for ${model.id}`
+  });
+  return to;
+}
 function compact(refs) {
   return (refs ?? []).filter((r) => Boolean(r));
 }
@@ -1133,17 +1173,14 @@ function resolveImageOptions(model, request, onUnsupported = "error") {
   const warnings = [];
   const c = model.constraints;
   const merged = { ...model.defaults, ...stripUndefined(request), model: model.id };
-  if (merged.aspectRatio !== void 0 && merged.aspectRatio !== "adaptive" && c.aspectRatios !== void 0 && !c.aspectRatios.includes(merged.aspectRatio)) {
-    const to = nearestAspectRatio(merged.aspectRatio, c.aspectRatios);
-    warnings.push({
-      code: "ratio_coerced",
-      field: "aspectRatio",
-      from: merged.aspectRatio,
-      to,
-      message: `aspect ratio ${merged.aspectRatio} coerced to ${to} for ${model.id}`
-    });
-    merged.aspectRatio = to;
-  }
+  const aspectRatio = resolveAspectRatio(
+    merged.aspectRatio,
+    c.aspectRatios,
+    model,
+    onUnsupported,
+    warnings
+  );
+  if (aspectRatio !== void 0) merged.aspectRatio = aspectRatio;
   const size = resolveTier(
     "size",
     "size_coerced",
@@ -1177,7 +1214,9 @@ function resolveImageOptions(model, request, onUnsupported = "error") {
     "count",
     "outputFormat",
     "scaleFactor",
-    "safetySettings"
+    "safetySettings",
+    "temperature",
+    "upscale"
   ]);
   return { request: merged, warnings };
 }
@@ -1207,17 +1246,14 @@ function resolveVideoOptions(model, request, onUnsupported = "error") {
       delete merged.aspectRatio;
     }
   }
-  if (merged.aspectRatio !== void 0 && merged.aspectRatio !== "adaptive" && c.aspectRatios !== void 0 && !c.aspectRatios.includes(merged.aspectRatio)) {
-    const to = nearestAspectRatio(merged.aspectRatio, c.aspectRatios);
-    warnings.push({
-      code: "ratio_coerced",
-      field: "aspectRatio",
-      from: merged.aspectRatio,
-      to,
-      message: `aspect ratio ${merged.aspectRatio} coerced to ${to} for ${model.id}`
-    });
-    merged.aspectRatio = to;
-  }
+  const aspectRatio = resolveAspectRatio(
+    merged.aspectRatio,
+    c.aspectRatios,
+    model,
+    onUnsupported,
+    warnings
+  );
+  if (aspectRatio !== void 0) merged.aspectRatio = aspectRatio;
   const resolution = resolveTier(
     "resolution",
     "resolution_coerced",
@@ -1229,7 +1265,13 @@ function resolveVideoOptions(model, request, onUnsupported = "error") {
     warnings
   );
   if (resolution !== void 0) merged.resolution = resolution;
-  const duration = resolveDuration(merged.durationSeconds, c.durations, model, warnings);
+  const duration = resolveDuration(
+    merged.durationSeconds,
+    c.durations,
+    model,
+    onUnsupported,
+    warnings
+  );
   if (duration !== void 0) merged.durationSeconds = duration;
   const limits = c.references ?? {};
   let images = trim(
@@ -1284,10 +1326,45 @@ function resolveVideoOptions(model, request, onUnsupported = "error") {
     "cfgScale",
     "generateAudio",
     "promptExpansion",
-    "lastFrame"
+    "lastFrame",
+    "watermark",
+    "safetyChecker"
   ]);
   return { request: merged, warnings };
 }
+function resolveAudioOptions(model, request, onUnsupported = "error") {
+  const warnings = [];
+  const c = model.constraints;
+  const merged = { ...model.defaults, ...stripUndefined(request), model: model.id };
+  for (const input of c.requires ?? []) {
+    const present = input === "references" ? compact(merged.references).length > 0 : input === "sourceVideo" && merged.sourceVideo !== void 0;
+    if (!present)
+      throw new ModelKitError({
+        code: "missing_input",
+        message: `${model.id} requires ${input}`,
+        model: model.id
+      });
+  }
+  const duration = resolveDuration(
+    merged.durationSeconds,
+    c.durations,
+    model,
+    onUnsupported,
+    warnings
+  );
+  if (duration !== void 0) merged.durationSeconds = duration;
+  const references = trim(
+    "references",
+    compact(merged.references),
+    c.references?.images ?? c.references?.total,
+    model,
+    warnings
+  );
+  if (references.length > 0) merged.references = references;
+  else delete merged.references;
+  dropUnsupported(merged, model, warnings, ["seed", "negativePrompt", "instrumental"]);
+  return { request: merged, warnings };
+}
 function dropUnsupported(req, model, warnings, fields) {
   const supports = model.constraints.supports ?? [];
   for (const field of fields) {
@@ -1329,6 +1406,9 @@ function hasAll(keys, names) {
     (n) => Boolean(keys[n]) && (n !== "higgsfield" || Boolean(keys.higgsfieldSecret))
   );
 }
+function missingEnterpriseKeys(config, names) {
+  return names.filter((name) => !hasAll(config.keys.enterprise, [name]));
+}
 function resolveKeys(config, names, tier = "pro") {
   if (tier === "enterprise" && hasAll(config.keys.enterprise, names))
     return config.keys.enterprise;
@@ -1361,7 +1441,6 @@ function keysFromEnv(env, prefix = "") {
   set("prodia", pick("PRODIA_API_KEY", "PRODIA_TOKEN"));
   set("higgsfield", pick("HIGGSFIELD_API_KEY"));
   set("higgsfieldSecret", pick("HIGGSFIELD_API_SECRET"));
-  set("openrouter", pick("OPENROUTER_API_KEY"));
   set("cloudflare", pick("CLOUDFLARE_AI_GATEWAY_TOKEN", "CF_AIG_TOKEN"));
   return keys;
 }
@@ -1409,7 +1488,9 @@ var textBody = z2.string();
 function classifyHttpError(status, body, provider, model) {
   const asText = textBody.safeParse(body);
   const text2 = asText.success ? asText.data : JSON.stringify(body);
-  const blocked = /content_policy|safety|blockReason|PROHIBITED_CONTENT|moderation/i.test(text2);
+  const blocked = /content_policy_violation|blockReason|PROHIBITED_CONTENT|moderation_blocked|safety(?![_A-Za-z])/i.test(
+    text2
+  );
   let code = "provider_rejected";
   if (blocked) code = "content_blocked";
   else if (status === 401 || status === 403) code = "auth";
@@ -1430,8 +1511,8 @@ function classifyHttpError(status, body, provider, model) {
 }
 function shouldRetry(err) {
   if (err instanceof ModelKitError) return err.retryable;
-  if (err.name === "AbortError" || /ECONNRESET|ETIMEDOUT|fetch failed|socket hang up/i.test(err.message))
-    return true;
+  if (err.name === "AbortError") return false;
+  if (/ECONNRESET|ETIMEDOUT|fetch failed|socket hang up/i.test(err.message)) return true;
   return false;
 }
 async function withRetry(fn, opts = {}) {
@@ -1623,6 +1704,12 @@ import { z as z4 } from "zod";
 function isUrlRef(ref) {
   return "url" in ref;
 }
+function abortedError() {
+  return new ModelKitError({ code: "timeout", message: "aborted", retryable: false });
+}
+function throwIfAborted(signal) {
+  if (signal?.aborted) throw abortedError();
+}
 function mediaRefToBase64(bytes) {
   let binary = "";
   const CHUNK = 32768;
@@ -1637,10 +1724,34 @@ function base64ToBytes(b64) {
   for (let i = 0; i < binary.length; i++) out[i] = binary.charCodeAt(i);
   return out;
 }
-async function mediaRefToBytes(ref, fetch) {
+function mediaRefToDataUrl(ref) {
+  if (isUrlRef(ref)) return ref.url;
+  return `data:${ref.mime};base64,${mediaRefToBase64(ref.bytes)}`;
+}
+var refToUrl = mediaRefToDataUrl;
+async function readBody(res) {
+  const text2 = await res.text().catch(() => "");
+  try {
+    const decoded = JsonValueSchema.safeParse(JSON.parse(text2));
+    return decoded.success ? decoded.data : text2;
+  } catch {
+    return text2;
+  }
+}
+async function mediaRefToBytes(ref, fetch, options = {}) {
   if (!isUrlRef(ref)) return { bytes: ref.bytes, mime: ref.mime };
-  const res = await fetch(ref.url);
-  if (!res.ok) throw new Error(`reference fetch failed (${res.status}): ${ref.url}`);
+  const init = {};
+  if (options.signal !== void 0) init.signal = options.signal;
+  const res = await fetch(ref.url, init);
+  if (!res.ok) {
+    const text2 = await res.text().catch(() => "");
+    throw classifyHttpError(
+      res.status,
+      text2,
+      options.provider ?? "media-reference",
+      options.model
+    );
+  }
   const mime = res.headers.get("content-type")?.split(";")[0]?.trim().toLowerCase() || "image/png";
   return { bytes: new Uint8Array(await res.arrayBuffer()), mime };
 }
@@ -1698,16 +1809,6 @@ var GATEWAY_SLUGS = {
   perplexity: "perplexity-ai",
   deepseek: "deepseek"
 };
-var OPENROUTER_VENDORS = {
-  openai: "openai",
-  anthropic: "anthropic",
-  google: "google",
-  xai: "x-ai",
-  perplexity: "perplexity",
-  deepseek: "deepseek",
-  moonshot: "moonshotai",
-  minimax: "minimax"
-};
 function detectTextProvider(modelId) {
   const id = modelId.toLowerCase();
   if (id.startsWith("claude") || id.includes("anthropic.")) return "anthropic";
@@ -1724,14 +1825,17 @@ function detectTextProvider(modelId) {
     message: `ai-sdk-text cannot detect a provider for "${modelId}". Set providerOptions.provider to one of ${TEXT_PROVIDERS.join(", ")}.`
   });
 }
-function textKeyName(model, routing) {
-  if (routing.text === "openrouter") return "openrouter";
+function textKeyName(model, _routing) {
   return textProviderOf(model);
 }
 function textProviderOf(model) {
   const explicit = model.providerOptions?.["modelkit"]?.["provider"];
   return isTextProvider(explicit) ? explicit : detectTextProvider(model.providerModelId);
 }
+function textUsesCloudflareGateway(model, routing) {
+  if (routing.text !== "cloudflare") return false;
+  return GATEWAY_SLUGS[textProviderOf(model)] !== void 0;
+}
 function nativeClientFor(provider) {
   switch (provider) {
     case "openai":
@@ -1764,16 +1868,6 @@ function resolveTextRoute(ctx) {
       retryable: false
     });
   }
-  if (mode === "openrouter") {
-    const modelId = ctx.model.providerModelId.includes("/") ? ctx.model.providerModelId : `${OPENROUTER_VENDORS[provider]}/${ctx.model.providerModelId}`;
-    return {
-      provider,
-      client: "openai-compatible",
-      modelId,
-      keyName: "openrouter",
-      baseURL: "https://openrouter.ai/api/v1"
-    };
-  }
   const compatBase = OPENAI_COMPATIBLE_BASE_URLS[provider];
   const route2 = {
     provider,
@@ -1841,7 +1935,7 @@ var aiSdkTextAdapter = {
       case "openai-compatible":
         return createOpenAI({
           ...settings,
-          name: route2.provider === "openai" ? "openai" : ctx.routing.text === "openrouter" ? "openrouter" : route2.provider
+          name: route2.provider
         }).chat(route2.modelId);
       case "openai":
         return createOpenAI(settings)(route2.modelId);
@@ -1854,8 +1948,9 @@ function requiredKeys(model, config) {
   if (model.adapter === "fal" && config.routing?.fal === "cloudflare-passthrough")
     return ["cloudflare"];
   if (model.adapter === "ai-sdk-text") {
-    const names = [textKeyName(model, config.routing ?? {})];
-    if (config.routing?.text === "cloudflare") names.push("cloudflare");
+    const routing = config.routing ?? {};
+    const names = [textKeyName(model, routing)];
+    if (textUsesCloudflareGateway(model, routing)) names.push("cloudflare");
     return names;
   }
   return ADAPTER_KEYS[model.adapter];
@@ -1929,6 +2024,11 @@ function pushIgnored(warnings, field, from, modelId) {
     message: `${modelId} provider instance does not support ${field}; ignored`
   });
 }
+function withoutModelkit(options) {
+  if (!options) return {};
+  const { modelkit: _modelkit, ...rest } = options;
+  return rest;
+}
 function logCoercions(logger, modelId, warnings) {
   if (warnings.length === 0) return;
   logger.warn(`[modelkit] ${modelId} request coerced`, { model: modelId, warnings });
@@ -1945,7 +2045,16 @@ function createModelKit(config, adapters) {
         message: `No adapter registered for ${model.adapter}`,
         model: model.id
       });
-    const keys = resolveKeys(config, requiredKeys(model, config), call.tier);
+    const names = requiredKeys(model, config);
+    if (call.tier === "enterprise") {
+      const missing = missingEnterpriseKeys(config, names);
+      if (missing.length > 0)
+        logger.warn(
+          `[modelkit] enterprise keys incomplete for ${model.adapter}; using pro`,
+          { missing }
+        );
+    }
+    const keys = resolveKeys(config, names, call.tier);
     const ctx = {
       model,
       keys,
@@ -2266,8 +2375,15 @@ function createModelKit(config, adapters) {
           model: model.id
         });
       }
-      const warnings = [];
-      const { adapter, ctx } = context(model, call, warnings);
+      const { abortSignal: _signal, ...request } = req;
+      const { request: resolved, warnings } = resolveAudioOptions(
+        model,
+        request,
+        call.onUnsupported ?? "error"
+      );
+      logCoercions(logger, model.id, warnings);
+      const signal = combineSignals(req.abortSignal, call.signal);
+      const { adapter, ctx } = context(model, signal ? { ...call, signal } : call, warnings);
       if (!adapter.generateAudio) {
         throw new ModelKitError({
           code: "unsupported_option",
@@ -2275,11 +2391,7 @@ function createModelKit(config, adapters) {
           model: model.id
         });
       }
-      const result = await adapter.generateAudio(
-        { ...model.defaults, ...req, model: model.id },
-        ctx
-      );
-      logCoercions(logger, model.id, warnings);
+      const result = await adapter.generateAudio(resolved, ctx);
       await emit({ ...result.usage, model: model.id, unit: model.usage.unit });
       return result;
     },
@@ -2300,7 +2412,7 @@ function createModelKit(config, adapters) {
         });
       return {
         model,
-        providerOptions: entry.providerOptions ?? {},
+        providerOptions: withoutModelkit(entry.providerOptions),
         catalog: entry
       };
     },
@@ -2336,7 +2448,9 @@ function createModelKit(config, adapters) {
         ...textAbortOptions(options.abortSignal, call.signal),
         model: target.languageModel,
         providerOptions: {},
-        prepareStep: steps.wrap(options.prepareStep),
+        prepareStep: steps.wrap(
+          options.prepareStep ?? options.experimental_prepareStep
+        ),
         onFinish: async (event) => {
           for (const usage2 of steps.usage(event.steps)) await emit(withRowUnit(usage2));
           await callerOnFinish?.(event);
@@ -2393,6 +2507,7 @@ var SEEDREAM_SIZES = {
     "3:2": "1248x832",
     "2:3": "832x1248",
     "21:9": "1568x672",
+    "9:21": "672x1568",
     "4:5": "864x1080",
     "5:4": "1080x864"
   },
@@ -2405,16 +2520,34 @@ var SEEDREAM_SIZES = {
     "3:2": "2496x1664",
     "2:3": "1664x2496",
     "21:9": "3136x1344",
+    "9:21": "1344x3136",
     "4:5": "1728x2160",
     "5:4": "2160x1728"
+  },
+  // 4K keeps the long edge at Ark's 4096 cap (a doubled 2K entry would exceed
+  // it); sides are multiples of 16.
+  "4K": {
+    "1:1": "4096x4096",
+    "4:3": "4096x3072",
+    "3:4": "3072x4096",
+    "16:9": "4096x2304",
+    "9:16": "2304x4096",
+    "3:2": "4096x2736",
+    "2:3": "2736x4096",
+    "21:9": "4096x1760",
+    "9:21": "1760x4096",
+    "4:5": "3280x4096",
+    "5:4": "4096x3280"
   }
 };
 var arkAdapter = {
   id: "ark",
   async generateImage(req, ctx) {
+    throwIfAborted(ctx.signal);
     return generateSeedreamImage(req, ctx);
   },
   async generateVideo(req, ctx) {
+    throwIfAborted(ctx.signal);
     return generateSeedanceVideo(req, ctx);
   }
 };
@@ -2434,7 +2567,7 @@ function arkKey(ctx) {
 function throwArkError(status, body, model) {
   const parsed = z5.string().safeParse(body);
   const text2 = parsed.success ? parsed.data : JSON.stringify(body ?? "");
-  if (/sensitive.?content|inputimage|human.?face/i.test(text2)) {
+  if (/sensitive.?content|human.?face/i.test(text2)) {
     const init = {
       code: "content_blocked",
       message: `ark rejected the request: ${text2.slice(0, 500)}`,
@@ -2459,7 +2592,7 @@ function throwArkError(status, body, model) {
   throw classifyHttpError(status, body, "ark", model);
 }
 function seedreamSize(aspect, size) {
-  const tier = size === "1K" ? "1K" : "2K";
+  const tier = size === "4K" || size === "1K" ? size : "2K";
   const table = SEEDREAM_SIZES[tier];
   return table[aspect ?? ""] ?? table["1:1"] ?? "2048x2048";
 }
@@ -2470,10 +2603,6 @@ function sniffedImageMime(bytes) {
     return "image/webp";
   return "image/jpeg";
 }
-function arkUrl(ref) {
-  if ("url" in ref) return ref.url;
-  return `data:${ref.mime};base64,${Buffer.from(ref.bytes).toString("base64")}`;
-}
 var SeedreamImageSchema = z5.object({
   url: z5.string().optional(),
   b64_json: z5.string().optional()
@@ -2500,7 +2629,7 @@ async function generateSeedreamImage(req, ctx) {
   const key = arkKey(ctx);
   const modelId = ctx.model.id;
   const size = seedreamSize(req.aspectRatio, req.size);
-  const refs = (req.references ?? []).map(arkUrl);
+  const refs = (req.references ?? []).map(refToUrl);
   const body = {
     model: ctx.model.providerModelId,
     prompt: req.prompt,
@@ -2552,10 +2681,18 @@ async function generateSeedreamImage(req, ctx) {
   if (req.quality !== void 0) usage2.imageQuality = req.quality;
   return { images: [{ bytes, mime: sniffedImageMime(bytes) }], usage: usage2, raw: data };
 }
+var SeedanceUsageSchema = z5.object({
+  total_tokens: z5.number().optional(),
+  completion_tokens: z5.number().optional()
+});
 var SeedanceResponseSchema = z5.object({
   id: z5.string().optional(),
   status: z5.string().optional(),
   content: z5.object({ video_url: z5.string().optional() }).optional(),
+  usage: SeedanceUsageSchema.optional(),
+  duration: z5.number().optional(),
+  framespersecond: z5.number().optional(),
+  resolution: z5.string().optional(),
   error: z5.union([z5.string(), z5.object({ message: z5.string().optional() })]).optional()
 });
 var SeedanceErrorMessageSchema = z5.object({ message: z5.string().optional() });
@@ -2587,42 +2724,42 @@ async function generateSeedanceVideo(req, ctx) {
   if (req.firstFrame) {
     content.push({
       type: "image_url",
-      image_url: { url: arkUrl(req.firstFrame) },
+      image_url: { url: refToUrl(req.firstFrame) },
       role: "first_frame"
     });
   }
   if (req.lastFrame) {
     content.push({
       type: "image_url",
-      image_url: { url: arkUrl(req.lastFrame) },
+      image_url: { url: refToUrl(req.lastFrame) },
       role: "last_frame"
     });
   }
   for (const ref of req.referenceImages ?? []) {
     content.push({
       type: "image_url",
-      image_url: { url: arkUrl(ref) },
+      image_url: { url: refToUrl(ref) },
       role: "reference_image"
     });
   }
   if (req.sourceVideo) {
     content.push({
       type: "video_url",
-      video_url: { url: arkUrl(req.sourceVideo) },
+      video_url: { url: refToUrl(req.sourceVideo) },
       role: "reference_video"
     });
   }
   for (const ref of req.referenceVideos ?? []) {
     content.push({
       type: "video_url",
-      video_url: { url: arkUrl(ref) },
+      video_url: { url: refToUrl(ref) },
       role: "reference_video"
     });
   }
   for (const ref of req.referenceAudios ?? []) {
     content.push({
       type: "audio_url",
-      audio_url: { url: arkUrl(ref) },
+      audio_url: { url: refToUrl(ref) },
       role: "reference_audio"
     });
   }
@@ -2636,6 +2773,7 @@ async function generateSeedanceVideo(req, ctx) {
   if (!hasFrameAnchor && req.aspectRatio !== void 0) body.ratio = req.aspectRatio;
   if (req.generateAudio !== void 0) body.generate_audio = req.generateAudio;
   if (req.seed !== void 0) body.seed = req.seed;
+  if (req.watermark !== void 0) body.watermark = req.watermark;
   const submit = await withRetry(async () => {
     const res = await arkFetch(ctx, `${arkBaseUrl(ctx)}${SUBMIT_PATH}`, {
       method: "POST",
@@ -2699,6 +2837,11 @@ async function generateSeedanceVideo(req, ctx) {
               "Seedance task succeeded but returned no video_url",
               modelId
             );
+          const tokens = polled.usage;
+          if (tokens?.total_tokens !== void 0 && tokens.completion_tokens !== void 0) {
+            usage2.inputTokens = tokens.total_tokens - tokens.completion_tokens;
+            usage2.outputTokens = tokens.completion_tokens;
+          }
           return {
             video: { url, expiresAt: new Date(Date.now() + VIDEO_URL_TTL_MS) },
             usage: usage2,
@@ -2762,24 +2905,17 @@ var defaultDeps = { createFalClient: defaultCreateFalClient };
 function createFalAdapter(deps = defaultDeps) {
   return {
     id: "fal",
-    generateImage: (req, ctx) => generateMuseImage(req, ctx, deps),
-    generateVideo: (req, ctx) => generateFalVideo(req, ctx, deps)
+    generateImage: (req, ctx) => {
+      throwIfAborted(ctx.signal);
+      return generateMuseImage(req, ctx, deps);
+    },
+    generateVideo: (req, ctx) => {
+      throwIfAborted(ctx.signal);
+      return generateFalVideo(req, ctx, deps);
+    }
   };
 }
 var falAdapter = createFalAdapter();
-function refToUrl(ref) {
-  if ("url" in ref) return ref.url;
-  return `data:${ref.mime};base64,${Buffer.from(ref.bytes).toString("base64")}`;
-}
-async function readBody(res) {
-  const text2 = await res.text().catch(() => "");
-  try {
-    const decoded = JsonValueSchema.safeParse(JSON.parse(text2));
-    return decoded.success ? decoded.data : text2;
-  } catch {
-    return text2;
-  }
-}
 var falDetailSchema = z6.object({ detail: JsonValueSchema });
 function liftDetail(body) {
   const parsed = falDetailSchema.safeParse(body);
@@ -3016,8 +3152,12 @@ async function generateMuseImage(req, ctx, deps) {
     raw: payload
   };
 }
+var H3_MAX_MODEL_IDS = /* @__PURE__ */ new Set([
+  "minimax/h3-max",
+  "minimax/h3-max/reference-to-video"
+]);
 function minimaxFamily(providerModelId) {
-  return providerModelId.toLowerCase().includes("h3-max") ? "h3-max" : "h3";
+  return H3_MAX_MODEL_IDS.has(providerModelId.toLowerCase()) ? "h3-max" : "h3";
 }
 function minimaxBase(req, family) {
   const resolution = (family === "h3-max" ? req.resolution ?? "768p" : req.resolution ?? "2k").toUpperCase();
@@ -3025,14 +3165,17 @@ function minimaxBase(req, family) {
     prompt: req.prompt.trim(),
     duration: req.durationSeconds ?? 5,
     resolution,
-    prompt_expansion_mode: req.promptExpansion ?? "balanced",
-    enable_safety_checker: true
+    prompt_expansion_mode: req.promptExpansion ?? "balanced"
   };
+  if (req.safetyChecker !== void 0) input.enable_safety_checker = req.safetyChecker;
   if (req.seed !== void 0) input.seed = req.seed;
   return input;
 }
+var KLING_MODEL_IDS = /* @__PURE__ */ new Set([
+  "fal-ai/kling-video/o1/video-to-video/reference"
+]);
 function isKling(providerModelId) {
-  return providerModelId.toLowerCase().includes("kling");
+  return KLING_MODEL_IDS.has(providerModelId.toLowerCase());
 }
 function klingO1Input(req) {
   const source = req.sourceVideo ? refToUrl(req.sourceVideo) : void 0;
@@ -3105,7 +3248,9 @@ async function generateFalVideo(req, ctx, deps) {
     if (lastFrame) base.end_image_url = lastFrame;
   } else {
     endpoint = `minimax/${family}/text-to-video`;
-    base.aspect_ratio = req.aspectRatio ?? "adaptive";
+    if (req.aspectRatio !== void 0 && req.aspectRatio !== "adaptive") {
+      base.aspect_ratio = req.aspectRatio;
+    }
   }
   if (ctx.routing.fal === "cloudflare-passthrough") {
     return await passthroughVideoJob(ctx, req, endpoint, base);
@@ -3593,6 +3738,14 @@ async function generateVeoVideo(req, ctx, createGenAIClient) {
       model: modelId
     });
   }
+  if (req.lastFrame !== void 0 && req.firstFrame === void 0) {
+    throw new ModelKitError({
+      code: "missing_input",
+      message: `${modelId} requires firstFrame when lastFrame is set`,
+      provider: "google-genai",
+      model: modelId
+    });
+  }
   const duration = req.durationSeconds ?? 8;
   const personGeneration = req.firstFrame ?? refs.length > 0 ? "allow_adult" : "allow_all";
   let image;
@@ -3659,7 +3812,7 @@ async function generateVeoVideo(req, ctx, createGenAIClient) {
       throwSdkError(cause, modelId);
     }
   };
-  const jobId = operation.name ?? "veo-operation";
+  const jobId = operation.name ?? crypto.randomUUID();
   const usage2 = videoUsage2(ctx, req, duration);
   return {
     jobId,
@@ -3798,6 +3951,7 @@ async function generateOmniVideo(req, ctx, createGenAIClient) {
       });
     }
     for (const ref of videos) {
+      if (ref.label !== void 0) parts.push({ type: "text", text: ref.label });
       const media = await resolveMedia(ref, ctx.fetch, ctx.signal, "video/mp4", modelId);
       parts.push({
         type: "video",
@@ -3806,6 +3960,8 @@ async function generateOmniVideo(req, ctx, createGenAIClient) {
       });
     }
     if (req.firstFrame) {
+      if (req.firstFrame.label !== void 0)
+        parts.push({ type: "text", text: req.firstFrame.label });
       const media = await resolveMedia(
         req.firstFrame,
         ctx.fetch,
@@ -3824,6 +3980,7 @@ async function generateOmniVideo(req, ctx, createGenAIClient) {
       }
     }
     for (const ref of images) {
+      if (ref.label !== void 0) parts.push({ type: "text", text: ref.label });
       const media = await resolveMedia(ref, ctx.fetch, ctx.signal, "image/png", modelId);
       if (media.bytes.length > OMNI_INLINE_LIMIT_BYTES) {
         parts.push({
@@ -3867,7 +4024,7 @@ async function generateOmniVideo(req, ctx, createGenAIClient) {
   } finally {
     await cleanupUploads();
   }
-  const jobId = interaction.id ?? "gemini-omni-interaction";
+  const jobId = interaction.id ?? crypto.randomUUID();
   const usage2 = videoUsage2(ctx, req, duration);
   const getInteraction = async () => {
     try {
@@ -4093,11 +4250,11 @@ function createGoogleImageAdapter(deps = defaultGoogleImageDeps) {
       const model = google(ctx.model.providerModelId);
       const content = [];
       for (const ref of req.references ?? []) {
+        if (ref.label !== void 0) content.push({ type: "text", text: ref.label });
         if ("url" in ref) content.push({ type: "image", image: ref.url });
         else content.push({ type: "image", image: ref.bytes, mediaType: ref.mime });
       }
       content.push({ type: "text", text: req.prompt });
-      void ctx.call.postCropToAspect;
       const imageConfig = {};
       if (req.aspectRatio !== void 0 && req.aspectRatio !== "adaptive")
         imageConfig.aspectRatio = req.aspectRatio;
@@ -4113,7 +4270,7 @@ function createGoogleImageAdapter(deps = defaultGoogleImageDeps) {
         const timeout = AbortSignal.timeout(GEMINI_IMAGE_TIMEOUT_MS);
         const signal = ctx.signal !== void 0 ? AbortSignal.any([ctx.signal, timeout]) : timeout;
         try {
-          return await deps.runTextGeneration({
+          const genOptions = {
             model,
             messages: [{ role: "user", content }],
             maxRetries: 0,
@@ -4121,7 +4278,9 @@ function createGoogleImageAdapter(deps = defaultGoogleImageDeps) {
               google: googleOptions
             },
             abortSignal: signal
-          });
+          };
+          if (req.temperature !== void 0) genOptions.temperature = req.temperature;
+          return await deps.runTextGeneration(genOptions);
         } catch (cause) {
           throw toProviderError(cause, ctx);
         }
@@ -4325,26 +4484,14 @@ function headers(ctx) {
   const { key, secret } = credentials(ctx);
   return { "Content-Type": "application/json", "hf-api-key": key, "hf-secret": secret };
 }
-function refToUrl2(ref) {
-  if ("url" in ref) return ref.url;
-  return `data:${ref.mime};base64,${Buffer.from(ref.bytes).toString("base64")}`;
-}
-async function readBody2(res) {
-  const text2 = await res.text().catch(() => "");
-  try {
-    const decoded = JsonValueSchema.safeParse(JSON.parse(text2));
-    return decoded.success ? decoded.data : text2;
-  } catch {
-    return text2;
-  }
-}
 function klingEndpoint(providerModelId) {
   if (providerModelId.startsWith("http")) return providerModelId;
   const path = providerModelId.startsWith("/") ? providerModelId.slice(1) : providerModelId;
   return `${PLATFORM}/${path}`;
 }
-function isDop(providerModelId) {
-  return providerModelId.toLowerCase().includes("dop");
+var DOP_MODEL_IDS = /* @__PURE__ */ new Set(["dop", "dop-preview"]);
+function higgsfieldVideoKind(providerModelId) {
+  return DOP_MODEL_IDS.has(providerModelId.toLowerCase()) ? "dop" : "kling";
 }
 function videoUsage3(ctx, req, duration) {
   const event = {
@@ -4388,12 +4535,12 @@ async function createRequest(ctx, url, init) {
     throw classifyThrown("higgsfield", cause, ctx.model.id);
   }
   if (!res.ok) {
-    throw classifyHttpError(res.status, await readBody2(res), "higgsfield", ctx.model.id);
+    throw classifyHttpError(res.status, await readBody(res), "higgsfield", ctx.model.id);
   }
   return res;
 }
 async function generateKlingVideo(req, ctx) {
-  const firstFrame = req.firstFrame ? refToUrl2(req.firstFrame) : void 0;
+  const firstFrame = req.firstFrame ? refToUrl(req.firstFrame) : void 0;
   if (!firstFrame) {
     throw new ModelKitError({
       code: "missing_input",
@@ -4412,7 +4559,7 @@ async function generateKlingVideo(req, ctx) {
   if (req.durationSeconds !== void 0) body.duration = req.durationSeconds;
   if (req.aspectRatio !== void 0 && req.aspectRatio !== "adaptive")
     body.aspect_ratio = req.aspectRatio;
-  if (req.lastFrame) body.last_image_url = refToUrl2(req.lastFrame);
+  if (req.lastFrame) body.last_image_url = refToUrl(req.lastFrame);
   const endpoint = klingEndpoint(ctx.model.providerModelId);
   const retryOptions = {};
   if (ctx.signal !== void 0) retryOptions.signal = ctx.signal;
@@ -4438,9 +4585,9 @@ async function generateKlingVideo(req, ctx) {
     } catch {
       return void 0;
     }
-    if (res.status >= 500) return void 0;
+    if (res.status === 429 || res.status >= 500) return void 0;
     if (!res.ok) {
-      throw classifyHttpError(res.status, await readBody2(res), "higgsfield", ctx.model.id);
+      throw classifyHttpError(res.status, await readBody(res), "higgsfield", ctx.model.id);
     }
     const payload = JsonValueSchema.parse(await res.json());
     const parsed = klingStatusSchema.safeParse(payload);
@@ -4547,7 +4694,7 @@ function dopUrl(body) {
 var dopCreateSchema = z10.object({ id: z10.string() });
 var dopStatusSchema = z10.object({ status: z10.string().optional() });
 async function generateDopVideo(req, ctx) {
-  const firstFrame = req.firstFrame ? refToUrl2(req.firstFrame) : void 0;
+  const firstFrame = req.firstFrame ? refToUrl(req.firstFrame) : void 0;
   if (!firstFrame) {
     throw new ModelKitError({
       code: "missing_input",
@@ -4567,7 +4714,7 @@ async function generateDopVideo(req, ctx) {
     webhook: null
   };
   if (req.lastFrame) {
-    params.input_images_end = [{ type: "image_url", image_url: refToUrl2(req.lastFrame) }];
+    params.input_images_end = [{ type: "image_url", image_url: refToUrl(req.lastFrame) }];
   }
   if (req.seed !== void 0) params.seed = req.seed;
   const retryOptions = {};
@@ -4657,7 +4804,10 @@ async function generateDopVideo(req, ctx) {
 }
 var higgsfieldAdapter = {
   id: "higgsfield",
-  generateVideo: (req, ctx) => isDop(ctx.model.providerModelId) ? generateDopVideo(req, ctx) : generateKlingVideo(req, ctx)
+  generateVideo: (req, ctx) => {
+    throwIfAborted(ctx.signal);
+    return higgsfieldVideoKind(ctx.model.providerModelId) === "dop" ? generateDopVideo(req, ctx) : generateKlingVideo(req, ctx);
+  }
 };
 
 // src/adapters/openai-image.ts
@@ -4752,6 +4902,7 @@ var openaiUsageSchema = z11.object({
 var openaiImageAdapter = {
   id: "openai-image",
   async generateImage(req, ctx) {
+    throwIfAborted(ctx.signal);
     const apiKey = ctx.keys.openai;
     if (!apiKey)
       throw new ModelKitError({
@@ -4761,6 +4912,13 @@ var openaiImageAdapter = {
         model: ctx.model.id,
         retryable: false
       });
+    if ((req.references ?? []).some((ref) => ref.label !== void 0)) {
+      ctx.warnings.push({
+        code: "option_ignored",
+        field: "references.label",
+        message: "openai-image cannot express reference labels; ignoring them"
+      });
+    }
     const client2 = new OpenAI({ apiKey, fetch: ctx.fetch, maxRetries: 0 });
     const size = req.exactSize !== void 0 ? req.exactSize : req.aspectRatio === void 0 || req.aspectRatio === "adaptive" ? void 0 : aspectRatioToSize(req.aspectRatio, req.size ?? "1K");
     const refs = req.references ?? [];
@@ -4774,24 +4932,34 @@ var openaiImageAdapter = {
       try {
         if (!isEdit) {
           if (size !== void 0) {
-            return await client2.images.generate({
+            return await client2.images.generate(
+              {
+                model: ctx.model.providerModelId,
+                prompt: req.prompt,
+                size,
+                quality: sdkQuality,
+                ...count
+              },
+              { signal: ctx.signal }
+            );
+          }
+          return await client2.images.generate(
+            {
               model: ctx.model.providerModelId,
               prompt: req.prompt,
-              size,
               quality: sdkQuality,
               ...count
-            });
-          }
-          return await client2.images.generate({
-            model: ctx.model.providerModelId,
-            prompt: req.prompt,
-            quality: sdkQuality,
-            ...count
-          });
+            },
+            { signal: ctx.signal }
+          );
         }
         const files = await Promise.all(
           refs.slice(0, MAX_EDIT_REFERENCES).map(async (ref, i) => {
-            const { bytes, mime } = await mediaRefToBytes(ref, ctx.fetch);
+            const { bytes, mime } = await mediaRefToBytes(ref, ctx.fetch, {
+              signal: ctx.signal,
+              provider: "openai-image",
+              model: ctx.model.id
+            });
             const ext = mime.includes("jpeg") ? "jpg" : mime.includes("webp") ? "webp" : "png";
             return new File([bytes], `ref-${i}.${ext}`, { type: mime });
           })
@@ -4799,22 +4967,28 @@ var openaiImageAdapter = {
         const single = files.length === 1 ? files[0] : void 0;
         const imageInput = single ?? files;
         if (size !== void 0) {
-          return await client2.images.edit({
+          return await client2.images.edit(
+            {
+              model: ctx.model.providerModelId,
+              image: imageInput,
+              prompt: req.prompt,
+              size,
+              quality: sdkQuality,
+              ...count
+            },
+            { signal: ctx.signal }
+          );
+        }
+        return await client2.images.edit(
+          {
             model: ctx.model.providerModelId,
             image: imageInput,
             prompt: req.prompt,
-            size,
             quality: sdkQuality,
             ...count
-          });
-        }
-        return await client2.images.edit({
-          model: ctx.model.providerModelId,
-          image: imageInput,
-          prompt: req.prompt,
-          quality: sdkQuality,
-          ...count
-        });
+          },
+          { signal: ctx.signal }
+        );
       } catch (cause) {
         throw toProviderError2(cause, ctx);
       }
@@ -4896,7 +5070,7 @@ async function refToBytes(ctx, ref) {
   if ("bytes" in ref) return { bytes: ref.bytes, mime: ref.mime };
   let res;
   try {
-    res = await ctx.fetch(ref.url);
+    res = await ctx.fetch(ref.url, ctx.signal ? { signal: ctx.signal } : {});
   } catch (cause) {
     throw classifyThrown("prodia", cause, ctx.model.id);
   }
@@ -4942,6 +5116,12 @@ function emptyImage() {
     retryable: false
   });
 }
+function inputExtension(mime) {
+  if (mime.includes("jpeg") || mime.includes("jpg")) return "jpg";
+  if (mime.includes("webp")) return "webp";
+  if (mime.includes("gif")) return "gif";
+  return "png";
+}
 function isBgRemove(providerModelId) {
   const id = providerModelId.toLowerCase();
   return id.includes("remove-background") || id.includes("bg-remove");
@@ -4950,6 +5130,7 @@ function isKlein(providerModelId) {
   return providerModelId.toLowerCase().includes("klein");
 }
 async function generateProdiaImage(req, ctx) {
+  throwIfAborted(ctx.signal);
   const key = ctx.keys.prodia;
   if (!key) {
     throw new ModelKitError({
@@ -5048,7 +5229,9 @@ async function generateProdiaImage(req, ctx) {
       "input",
       // Uint8Array<ArrayBufferLike> is not a valid BlobPart under the
       // strict lib types; copy into a fresh ArrayBuffer-backed view.
-      new File([new Uint8Array(input.bytes)], `input-${i}.jpg`, { type: "image/jpeg" })
+      new File([new Uint8Array(input.bytes)], `input-${i}.${inputExtension(input.mime)}`, {
+        type: input.mime
+      })
     );
   });
   const retryOpts = {
@@ -5063,11 +5246,13 @@ async function generateProdiaImage(req, ctx) {
     res = await withRetry(async () => {
       let attemptRes;
       try {
-        attemptRes = await ctx.fetch(PRODIA_JOB_URL, {
+        const init = {
           method: "POST",
           headers: { Authorization: `Bearer ${key}`, Accept: accept },
           body: form
-        });
+        };
+        if (ctx.signal !== void 0) init.signal = ctx.signal;
+        attemptRes = await ctx.fetch(PRODIA_JOB_URL, init);
       } catch (err) {
         throw classifyThrown("prodia", err, ctx.model.id);
       }
@@ -5128,11 +5313,7 @@ function client(ctx, deps) {
       retryable: false
     });
   }
-  return deps.createReplicateClient({ auth: token });
-}
-function refToUrl3(ref) {
-  if ("url" in ref) return ref.url;
-  return `data:${ref.mime};base64,${Buffer.from(ref.bytes).toString("base64")}`;
+  return deps.createReplicateClient({ auth: token, fetch: ctx.fetch });
 }
 function route(providerModelId) {
   if (providerModelId.includes(":")) {
@@ -5152,8 +5333,9 @@ function replicateError(cause, ctx) {
   const status = info.status ?? (nested.success ? nested.data.response?.status : void 0);
   return classifyThrown("replicate", cause, ctx.model.id, { status, body: info.body });
 }
-function isLower(id) {
-  return id.toLowerCase();
+function baseModelId(providerModelId) {
+  const colon = providerModelId.indexOf(":");
+  return colon === -1 ? providerModelId : providerModelId.slice(0, colon);
 }
 function predictionErrorBody(cause) {
   const parsed = JsonValueSchema.safeParse(cause);
@@ -5201,6 +5383,7 @@ function noOutput() {
   });
 }
 async function runPrediction(ctx, input, deps) {
+  throwIfAborted(ctx.signal);
   const api = client(ctx, deps);
   const createOptions = { ...route(ctx.model.providerModelId), input };
   let prediction;
@@ -5208,12 +5391,14 @@ async function runPrediction(ctx, input, deps) {
     prediction = await withRetry(
       async () => {
         try {
-          return await api.predictions.create(createOptions);
+          return await api.predictions.create(
+            ctx.signal ? { ...createOptions, signal: ctx.signal } : createOptions
+          );
         } catch (cause) {
           throw replicateError(cause, ctx);
         }
       },
-      ctx.signal ? { signal: ctx.signal } : {}
+      { attempts: 1 }
     );
   } catch (cause) {
     throw replicateError(cause, ctx);
@@ -5226,7 +5411,10 @@ async function runPrediction(ctx, input, deps) {
   const done = await pollUntil(async () => {
     let current;
     try {
-      current = await api.predictions.get(prediction.id);
+      current = await api.predictions.get(
+        prediction.id,
+        ctx.signal ? { signal: ctx.signal } : void 0
+      );
     } catch (cause) {
       throw replicateError(cause, ctx);
     }
@@ -5293,21 +5481,32 @@ async function download(ctx, url) {
   const mime = res.headers.get("content-type") ?? "image/png";
   return { bytes: new Uint8Array(await res.arrayBuffer()), mime };
 }
-function isUpscaler(providerModelId) {
-  const id = isLower(providerModelId);
-  return id.includes("upscal") || id.includes("crystal") || id.includes("topaz");
-}
-function isTopazUpscaler(providerModelId) {
-  return isLower(providerModelId).includes("topazlabs/image-upscale");
-}
-function isQwen(providerModelId) {
-  return isLower(providerModelId).includes("qwen");
-}
-function isWanImage(providerModelId) {
-  return isLower(providerModelId).includes("wan-2.7");
+var REPLICATE_IMAGE_KINDS = {
+  "topazlabs/image-upscale": "topaz",
+  "philz1337x/crystal-upscaler": "upscaler",
+  "wan-video/wan-2.7-image-pro": "wan",
+  "qwen/qwen-image-2-pro": "qwen-image",
+  "qwen-edit-apps/qwen-image-edit-plus-lora-skin": "qwen-edit"
+};
+function imageKind(providerModelId) {
+  return REPLICATE_IMAGE_KINDS[baseModelId(providerModelId)] ?? "generic";
+}
+var REPLICATE_VIDEO_KINDS = {
+  "kling-video/v2.5-turbo/pro/image-to-video": "kling",
+  "kling-video/v3.0/pro/image-to-video": "kling",
+  "kwaivgi/kling-v3-video": "kling",
+  "kwaivgi/kling-v3-omni-video": "kling-omni",
+  "minimax/hailuo-02": "hailuo"
+};
+function videoKind(providerModelId) {
+  return REPLICATE_VIDEO_KINDS[baseModelId(providerModelId)] ?? "wan";
 }
-function isQwenEditPlus(providerModelId) {
-  return isLower(providerModelId).includes("edit");
+var REPLICATE_AUDIO_KINDS = {
+  "zsxkib/mmaudio": "mmaudio",
+  "elevenlabs/music": "elevenlabs"
+};
+function audioKind(providerModelId) {
+  return REPLICATE_AUDIO_KINDS[baseModelId(providerModelId)] ?? "lyria";
 }
 var QWEN_IMAGE_RATIOS = /* @__PURE__ */ new Set([
   "1:1",
@@ -5350,87 +5549,113 @@ function wanImageSize(aspectRatio, size) {
 async function generateReplicateImage(req, ctx, deps) {
   const id = ctx.model.providerModelId;
   let input;
-  if (isTopazUpscaler(id)) {
-    const first = (req.references ?? [])[0];
-    if (!first) {
-      throw new ModelKitError({
-        code: "missing_input",
-        message: `${ctx.model.id} requires a reference image to upscale`,
-        provider: "replicate",
-        model: ctx.model.id,
-        retryable: false
-      });
+  switch (imageKind(id)) {
+    case "topaz": {
+      const first = (req.references ?? [])[0];
+      if (!first) {
+        throw new ModelKitError({
+          code: "missing_input",
+          message: `${ctx.model.id} requires a reference image to upscale`,
+          provider: "replicate",
+          model: ctx.model.id,
+          retryable: false
+        });
+      }
+      let format = req.outputFormat === "jpeg" ? "jpg" : req.outputFormat ?? "png";
+      if (format !== "png" && format !== "jpg") {
+        ctx.warnings.push({
+          code: "option_ignored",
+          field: "outputFormat",
+          from: format,
+          message: "topazlabs/image-upscale supports only png and jpg; using png"
+        });
+        format = "png";
+      }
+      const upscale = req.upscale;
+      const topaz = {
+        image: refToUrl(first),
+        enhance_model: upscale?.enhanceModel ?? "Standard V2",
+        output_format: format,
+        upscale_factor: `${req.scaleFactor ?? 2}x`,
+        face_enhancement: upscale?.faceEnhancement ?? false
+      };
+      if (upscale?.subjectDetection !== void 0)
+        topaz.subject_detection = upscale.subjectDetection;
+      if (upscale?.faceEnhancementCreativity !== void 0)
+        topaz.face_enhancement_creativity = upscale.faceEnhancementCreativity;
+      if (upscale?.faceEnhancementStrength !== void 0)
+        topaz.face_enhancement_strength = upscale.faceEnhancementStrength;
+      input = topaz;
+      break;
     }
-    let format = req.outputFormat === "jpeg" ? "jpg" : req.outputFormat ?? "png";
-    if (format !== "png" && format !== "jpg") {
-      ctx.warnings.push({
-        code: "option_ignored",
-        field: "outputFormat",
-        from: format,
-        message: "topazlabs/image-upscale supports only png and jpg; using png"
-      });
-      format = "png";
+    case "upscaler": {
+      const first = (req.references ?? [])[0];
+      if (!first) {
+        throw new ModelKitError({
+          code: "missing_input",
+          message: `${ctx.model.id} requires a reference image to upscale`,
+          provider: "replicate",
+          model: ctx.model.id,
+          retryable: false
+        });
+      }
+      const imageInput = { prompt: req.prompt };
+      imageInput.image = refToUrl(first);
+      imageInput.creativity = 0;
+      imageInput.scale_factor = 2;
+      imageInput.output_format = "png";
+      input = imageInput;
+      break;
     }
-    input = {
-      image: refToUrl3(first),
-      enhance_model: "Standard V2",
-      output_format: format,
-      upscale_factor: `${req.scaleFactor ?? 2}x`,
-      face_enhancement: true
-    };
-  } else if (isUpscaler(id)) {
-    const refs = req.references ?? [];
-    const first = refs[0];
-    if (!first) {
-      throw new ModelKitError({
-        code: "missing_input",
-        message: `${ctx.model.id} requires a reference image to upscale`,
-        provider: "replicate",
-        model: ctx.model.id,
-        retryable: false
-      });
+    case "wan": {
+      const refs = (req.references ?? []).map(refToUrl);
+      const imageInput = { prompt: req.prompt };
+      if (refs.length > 0) imageInput.images = refs;
+      imageInput.size = wanImageSize(req.aspectRatio, req.size);
+      input = imageInput;
+      break;
     }
-    const imageInput = { prompt: req.prompt };
-    imageInput.image = refToUrl3(first);
-    imageInput.creativity = 0;
-    imageInput.scale_factor = 2;
-    imageInput.output_format = "png";
-    input = imageInput;
-  } else if (isWanImage(id)) {
-    const refs = (req.references ?? []).map(refToUrl3);
-    const imageInput = { prompt: req.prompt };
-    if (refs.length > 0) imageInput.images = refs;
-    imageInput.size = wanImageSize(req.aspectRatio, req.size);
-    input = imageInput;
-  } else if (isQwen(id)) {
-    const refs = (req.references ?? []).map(refToUrl3);
-    const firstUrl = refs[0];
-    const imageInput = { prompt: req.prompt };
-    if (firstUrl !== void 0) {
-      imageInput.image = firstUrl;
-      if (!isQwenEditPlus(id)) imageInput.match_input_image = true;
+    case "qwen-image": {
+      const refs = (req.references ?? []).map(refToUrl);
+      const firstUrl = refs[0];
+      const imageInput = { prompt: req.prompt };
+      if (firstUrl !== void 0) {
+        imageInput.image = firstUrl;
+        imageInput.match_input_image = true;
+      } else if (req.aspectRatio !== void 0 && QWEN_IMAGE_RATIOS.has(req.aspectRatio)) {
+        imageInput.aspect_ratio = req.aspectRatio;
+      }
+      input = imageInput;
+      break;
     }
-    const ratios = isQwenEditPlus(id) ? QWEN_EDIT_RATIOS : QWEN_IMAGE_RATIOS;
-    if (req.aspectRatio !== void 0) {
-      if (ratios.has(req.aspectRatio)) imageInput.aspect_ratio = req.aspectRatio;
-      else
-        ctx.warnings.push({
-          code: "option_ignored",
-          field: "aspectRatio",
-          from: req.aspectRatio,
-          message: `${id} has no ${req.aspectRatio} aspect_ratio; using the provider default`
-        });
+    case "qwen-edit": {
+      const refs = (req.references ?? []).map(refToUrl);
+      const firstUrl = refs[0];
+      const imageInput = { prompt: req.prompt };
+      if (firstUrl !== void 0) imageInput.image = firstUrl;
+      if (req.aspectRatio !== void 0) {
+        if (QWEN_EDIT_RATIOS.has(req.aspectRatio))
+          imageInput.aspect_ratio = req.aspectRatio;
+        else
+          ctx.warnings.push({
+            code: "option_ignored",
+            field: "aspectRatio",
+            from: req.aspectRatio,
+            message: `${id} has no ${req.aspectRatio} aspect_ratio; using the provider default`
+          });
+      }
+      if (req.outputFormat !== void 0) {
+        imageInput.output_format = req.outputFormat === "jpeg" ? "jpg" : req.outputFormat;
+      }
+      input = imageInput;
+      break;
     }
-    if (isQwenEditPlus(id) && req.outputFormat !== void 0) {
-      imageInput.output_format = req.outputFormat === "jpeg" ? "jpg" : req.outputFormat;
+    default: {
+      const first = (req.references ?? [])[0];
+      const imageInput = { prompt: req.prompt };
+      if (first) imageInput.image = refToUrl(first);
+      input = imageInput;
     }
-    input = imageInput;
-  } else {
-    const refs = req.references ?? [];
-    const first = refs[0];
-    const imageInput = { prompt: req.prompt };
-    if (first) imageInput.image = refToUrl3(first);
-    input = imageInput;
   }
   const { url, predictTime, raw } = await runPrediction(ctx, input, deps);
   const file = await download(ctx, url);
@@ -5447,8 +5672,8 @@ function wanSize(aspectRatio) {
   return w > h ? "2048*1152" : "1536*2048";
 }
 function klingInput(req) {
-  const images = (req.firstFrame ? [req.firstFrame] : []).map(refToUrl3);
-  if (req.lastFrame) images.push(refToUrl3(req.lastFrame));
+  const images = (req.firstFrame ? [req.firstFrame] : []).map(refToUrl);
+  if (req.lastFrame) images.push(refToUrl(req.lastFrame));
   const input = {
     prompt: req.prompt,
     duration: req.durationSeconds ?? 10,
@@ -5466,13 +5691,18 @@ function klingInput(req) {
   if (req.negativePrompt !== void 0) input.negative_prompt = req.negativePrompt;
   return input;
 }
+function klingOmniMode(req) {
+  if (req.resolution === "1080p") return "pro";
+  if (req.resolution === "720p") return "standard";
+  return req.mode ?? "standard";
+}
 function klingOmniInput(req) {
-  const images = (req.firstFrame ? [req.firstFrame] : []).map(refToUrl3);
-  if (req.lastFrame) images.push(refToUrl3(req.lastFrame));
+  const images = (req.firstFrame ? [req.firstFrame] : []).map(refToUrl);
+  if (req.lastFrame) images.push(refToUrl(req.lastFrame));
   const input = {
     prompt: req.prompt,
     duration: req.durationSeconds ?? 10,
-    mode: req.mode ?? "standard",
+    mode: klingOmniMode(req),
     generate_audio: req.generateAudio ?? false
   };
   const startImage = images[0];
@@ -5483,7 +5713,7 @@ function klingOmniInput(req) {
   } else if (req.aspectRatio !== "adaptive") {
     input.aspect_ratio = req.aspectRatio ?? "16:9";
   }
-  const refs = (req.referenceImages ?? []).map(refToUrl3);
+  const refs = (req.referenceImages ?? []).map(refToUrl);
   if (refs.length > 0) input.reference_images = refs;
   return input;
 }
@@ -5494,8 +5724,8 @@ function hailuoInput(req) {
     resolution: req.resolution ?? "768p",
     prompt_optimizer: true
   };
-  if (req.firstFrame) input.first_frame_image = refToUrl3(req.firstFrame);
-  if (req.lastFrame) input.last_frame_image = refToUrl3(req.lastFrame);
+  if (req.firstFrame) input.first_frame_image = refToUrl(req.firstFrame);
+  if (req.lastFrame) input.last_frame_image = refToUrl(req.lastFrame);
   return input;
 }
 function wanInput(req) {
@@ -5504,12 +5734,13 @@ function wanInput(req) {
     duration: req.durationSeconds ?? 5,
     size: wanSize(req.aspectRatio)
   };
-  if (req.firstFrame) input.image = refToUrl3(req.firstFrame);
+  if (req.firstFrame) input.image = refToUrl(req.firstFrame);
   return input;
 }
 async function generateReplicateVideo(req, ctx, deps) {
-  const id = isLower(ctx.model.providerModelId);
-  const input = id.includes("kling") ? id.includes("omni") ? klingOmniInput(req) : klingInput(req) : id.includes("hailuo") ? hailuoInput(req) : wanInput(req);
+  throwIfAborted(ctx.signal);
+  const kind = videoKind(ctx.model.providerModelId);
+  const input = kind === "kling" ? klingInput(req) : kind === "kling-omni" ? klingOmniInput(req) : kind === "hailuo" ? hailuoInput(req) : wanInput(req);
   const api = client(ctx, deps);
   const createOptions = { ...route(ctx.model.providerModelId), input };
   let prediction;
@@ -5517,12 +5748,14 @@ async function generateReplicateVideo(req, ctx, deps) {
     prediction = await withRetry(
       async () => {
         try {
-          return await api.predictions.create(createOptions);
+          return await api.predictions.create(
+            ctx.signal ? { ...createOptions, signal: ctx.signal } : createOptions
+          );
         } catch (cause) {
           throw replicateError(cause, ctx);
         }
       },
-      ctx.signal ? { signal: ctx.signal } : {}
+      { attempts: 1 }
     );
   } catch (cause) {
     throw replicateError(cause, ctx);
@@ -5531,7 +5764,10 @@ async function generateReplicateVideo(req, ctx, deps) {
   const check = async () => {
     let current;
     try {
-      current = await api.predictions.get(jobId);
+      current = await api.predictions.get(
+        jobId,
+        ctx.signal ? { signal: ctx.signal } : void 0
+      );
     } catch (cause) {
       throw replicateError(cause, ctx);
     }
@@ -5632,7 +5868,7 @@ function mmaudioInput(req, ctx) {
     });
   }
   const input = {
-    video: refToUrl3(video),
+    video: refToUrl(video),
     duration: req.durationSeconds ?? 8,
     seed: req.seed ?? -1
   };
@@ -5651,8 +5887,8 @@ function predictionRaw(prediction) {
   return raw;
 }
 async function generateReplicateAudio(req, ctx, deps) {
-  const id = isLower(ctx.model.providerModelId);
-  const input = id.includes("mmaudio") ? mmaudioInput(req, ctx) : id.includes("elevenlabs") ? elevenLabsMusicInput(req, ctx) : lyriaAudioInput(req, ctx);
+  const kind = audioKind(ctx.model.providerModelId);
+  const input = kind === "mmaudio" ? mmaudioInput(req, ctx) : kind === "elevenlabs" ? elevenLabsMusicInput(req, ctx) : lyriaAudioInput(req, ctx);
   const { url, predictTime, raw } = await runPrediction(ctx, input, deps);
   const usage2 = audioUsage(ctx, req);
   if (ctx.model.usage.unit === "seconds") {
@@ -5681,10 +5917,6 @@ function createXaiVideoAdapter(deps = defaultXaiDeps) {
   };
 }
 var xaiVideoAdapter = createXaiVideoAdapter();
-function refToUrl4(ref) {
-  if ("url" in ref) return ref.url;
-  return `data:${ref.mime};base64,${Buffer.from(ref.bytes).toString("base64")}`;
-}
 var xaiResponseStatusSchema = z13.object({
   response: z13.object({ status: z13.number() }).partial()
 });
@@ -5716,13 +5948,14 @@ async function generateXaiVideo(req, ctx, deps) {
       retryable: false
     });
   }
-  const refUrls = (req.referenceImages ?? []).map(refToUrl4).slice(0, 7);
-  const resolution = req.resolution === "480p" ? "480p" : "720p";
+  const refUrls = (req.referenceImages ?? []).map(refToUrl).slice(0, 7);
   const xaiOptions = {
-    resolution,
     pollIntervalMs: POLL_INTERVAL_MS2,
     pollTimeoutMs: POLL_TIMEOUT_MS2
   };
+  if (req.resolution === "480p" || req.resolution === "720p" || req.resolution === "1080p") {
+    xaiOptions.resolution = req.resolution;
+  }
   if (refUrls.length > 0) {
     xaiOptions.mode = "reference-to-video";
     xaiOptions.referenceImageUrls = refUrls;
@@ -5742,7 +5975,7 @@ async function generateXaiVideo(req, ctx, deps) {
     providerOptions: { xai: xaiOptions }
   };
   if (req.firstFrame) {
-    callBody.frameImages = [{ image: refToUrl4(req.firstFrame), frameType: "first_frame" }];
+    callBody.frameImages = [{ image: refToUrl(req.firstFrame), frameType: "first_frame" }];
   }
   if (req.aspectRatio !== void 0 && req.aspectRatio !== "adaptive") {
     callBody.aspectRatio = req.aspectRatio;
@@ -5778,7 +6011,8 @@ async function generateXaiVideo(req, ctx, deps) {
     return settled;
   };
   return {
-    jobId: `${ctx.model.id}-grok`,
+    // xAI exposes no request id on the synchronous handle, so each job gets a fresh one.
+    jobId: crypto.randomUUID(),
     async poll() {
       if (failed !== void 0) return { status: "failed" };
       return settled ? { status: "succeeded" } : { status: "pending" };
@@ -5912,6 +6146,7 @@ export {
   prodiaAdapter,
   readErrorBody,
   replicateAdapter,
+  resolveAudioOptions,
   resolveImageOptions,
   resolveKeys,
   resolveVideoOptions,