@shop-os/modelkit 0.1.7 → 0.1.9

package/dist/catalog.js

@@ -0,0 +1,2216 @@
+// src/errors.ts
+var ModelKitError = class extends Error {
+  code;
+  provider;
+  model;
+  status;
+  body;
+  retryable;
+  constructor(init) {
+    super(init.message, { cause: init.cause });
+    this.name = "ModelKitError";
+    this.code = init.code;
+    this.provider = init.provider;
+    this.model = init.model;
+    this.status = init.status;
+    this.body = init.body;
+    this.retryable = init.retryable ?? defaultRetryable(init.code);
+  }
+};
+function defaultRetryable(code) {
+  return code === "timeout" || code === "network" || code === "rate_limited";
+}
+function isModelKitError(err) {
+  return err instanceof ModelKitError;
+}
+
+// src/catalog/registry.ts
+var Catalog = class _Catalog {
+  byId = /* @__PURE__ */ new Map();
+  byAlias = /* @__PURE__ */ new Map();
+  constructor(models) {
+    for (const model of models) this.add(model);
+  }
+  add(model) {
+    if (this.byId.has(model.id)) {
+      throw new Error(`Duplicate model id "${model.id}"`);
+    }
+    if (this.byAlias.has(model.id)) {
+      throw new Error(
+        `Model id "${model.id}" collides with an alias of "${this.byAlias.get(model.id)}"`
+      );
+    }
+    for (const alias of model.aliases) {
+      if (alias === model.id) {
+        throw new Error(`Alias "${alias}" collides with its own model id "${model.id}"`);
+      }
+      const existing = this.byAlias.get(alias);
+      if (existing !== void 0) {
+        throw new Error(`Duplicate alias "${alias}" (already maps to "${existing}")`);
+      }
+    }
+    this.byId.set(model.id, model);
+    this.byAlias.set(model.id, model.id);
+    for (const alias of model.aliases) this.byAlias.set(alias, model.id);
+  }
+  /** Resolves a canonical id or any historical alias. */
+  resolveId(idOrAlias) {
+    return this.byAlias.get(idOrAlias);
+  }
+  get(idOrAlias) {
+    const id = this.resolveId(idOrAlias);
+    return id ? this.byId.get(id) : void 0;
+  }
+  require(idOrAlias) {
+    const model = this.get(idOrAlias);
+    if (!model) {
+      throw new ModelKitError({
+        code: "unknown_model",
+        message: `Unknown model "${idOrAlias}"`,
+        model: idOrAlias
+      });
+    }
+    if (model.lifecycle === "retired") {
+      const hint = model.replacedBy ? ` Use "${model.replacedBy}".` : "";
+      throw new ModelKitError({
+        code: "unknown_model",
+        message: `Model "${model.id}" is retired.${hint}`,
+        model: model.id
+      });
+    }
+    return model;
+  }
+  list(filter = {}) {
+    const lifecycles = filter.lifecycle ?? ["active"];
+    return [...this.byId.values()].filter(
+      (m) => (filter.type ? m.type === filter.type : true) && lifecycles.includes(m.lifecycle)
+    );
+  }
+  /** Merge rows by id. Used for catalogOverride. */
+  withOverrides(rows) {
+    return new _Catalog([
+      ...[...this.byId.values()].filter((m) => !rows.some((r) => r.id === m.id)),
+      ...rows
+    ]);
+  }
+};
+
+// src/catalog/image.ts
+var IMAGE_USAGE = { unit: "images", modality: "IMAGE" };
+var IMAGE_GENERATE_EDIT = { imageGenerate: true, imageEdit: true };
+var GEMINI_RATIOS = [
+  "1:1",
+  "2:3",
+  "3:2",
+  "16:9",
+  "9:16",
+  "4:3",
+  "3:4",
+  "21:9",
+  "4:5",
+  "5:4",
+  "adaptive"
+];
+function geminiImage(id, label, aliases, sizes, display, exposure) {
+  const model = {
+    id,
+    aliases,
+    label,
+    type: "image",
+    adapter: "google-image",
+    providerModelId: id,
+    lifecycle: "active",
+    exposure,
+    constraints: {
+      aspectRatios: [...GEMINI_RATIOS],
+      sizes,
+      supports: ["safetySettings", "temperature"]
+    },
+    defaults: { size: "1K", aspectRatio: "1:1" },
+    usage: { ...IMAGE_USAGE, slug: "gemini" },
+    pricingKind: "image_size",
+    display,
+    capabilities: { ...IMAGE_GENERATE_EDIT }
+  };
+  return model;
+}
+var IMAGE_MODELS = [
+  geminiImage(
+    "gemini-3-pro-image-preview",
+    "Nano Banana Pro",
+    ["nb-pro"],
+    ["1K", "2K", "4K"],
+    {
+      provider: "google",
+      description: "Create stunning, production-quality visuals with exceptional realism, detail, and typography. Best for campaigns, e-commerce, product photography, and hero creatives.",
+      logo: "https://d31vvfpk2ppdyq.cloudfront.net/assets/image/192485d2-a76a-4657-9c06-930b2c70badd_compressed.png"
+    },
+    "catalog"
+  ),
+  geminiImage(
+    "gemini-3.1-flash-image-preview",
+    "Nano Banana 2",
+    ["nb-2"],
+    // Flash 3.1 is the only Gemini image model with the 512 tier.
+    ["512", "1K", "2K", "4K"],
+    {
+      provider: "google",
+      description: "Generate high-quality images at speed. Perfect for brainstorming, moodboards, concept testing, creative exploration, and large batch workflows.",
+      logo: "https://d31vvfpk2ppdyq.cloudfront.net/assets/image/192485d2-a76a-4657-9c06-930b2c70badd_compressed.png"
+    },
+    "catalog"
+  ),
+  geminiImage(
+    "gemini-3.1-flash-lite-image",
+    "Nano Banana Lite",
+    ["nb-lite"],
+    ["1K"],
+    {
+      provider: "google",
+      description: "The fastest and most affordable image model. Ideal for rapid drafts, bulk exploration, and iterative concept testing at 1K resolution.",
+      logo: "https://d31vvfpk2ppdyq.cloudfront.net/assets/image/192485d2-a76a-4657-9c06-930b2c70badd_compressed.png"
+    },
+    "catalog"
+  ),
+  // trigger-workflows validates 1K only; pricing seeds 2K/4K rows so allow all three.
+  geminiImage(
+    "gemini-2.5-flash-image",
+    "Gemini 2.5 Flash Image",
+    [],
+    ["1K", "2K", "4K"],
+    {
+      provider: "google",
+      description: "Nano Banana (legacy Gemini 2.5 Flash Image) native image generation and editing model for high-volume, low-latency creative workflows."
+    },
+    "internal"
+  ),
+  {
+    // Cowork "auto" lets OpenAI pick the size; the canonical equivalent is adaptive.
+    id: "gpt-image-2",
+    aliases: [],
+    label: "GPT-image-2",
+    type: "image",
+    adapter: "openai-image",
+    providerModelId: "gpt-image-2",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      exactSizes: {
+        "1:1": "1024x1024",
+        "3:2": "1536x1024",
+        "2:3": "1024x1536",
+        "16:9": "2048x1152",
+        "9:16": "1152x2048",
+        "4:5": "1024x1280",
+        "5:4": "1280x1024"
+      },
+      aspectRatios: ["1:1", "3:2", "2:3", "16:9", "9:16", "4:5", "5:4", "adaptive"],
+      sizes: ["1K", "2K", "4K"],
+      qualityOptions: ["auto", "low", "medium", "high"],
+      references: { images: 16 },
+      supports: ["quality", "count"]
+    },
+    defaults: { size: "1K", aspectRatio: "adaptive", quality: "auto" },
+    usage: { ...IMAGE_USAGE, slug: "gpt" },
+    pricingKind: "image_size",
+    display: {
+      provider: "openai",
+      description: "The most capable model for image editing and complex creative instructions. Excels at text rendering, visual consistency, design generation, and iterative refinement.",
+      logo: "https://d31vvfpk2ppdyq.cloudfront.net/assets/image/1204b494-8962-4f7f-af5e-699a8bdc98e3_compressed.png"
+    },
+    capabilities: { ...IMAGE_GENERATE_EDIT }
+  },
+  {
+    // Same Images API surface as gpt-image-2 plus the "xhigh" and "max" quality tiers.
+    id: "gpt-image-2.5-sunburst",
+    aliases: [],
+    label: "GPT Image 2.5 Sunburst",
+    type: "image",
+    adapter: "openai-image",
+    providerModelId: "gpt-image-2.5-sunburst",
+    lifecycle: "active",
+    exposure: "internal",
+    constraints: {
+      exactSizes: {
+        "1:1": "1024x1024",
+        "3:2": "1536x1024",
+        "2:3": "1024x1536",
+        "16:9": "2048x1152",
+        "9:16": "1152x2048",
+        "4:5": "1024x1280",
+        "5:4": "1280x1024"
+      },
+      aspectRatios: ["1:1", "3:2", "2:3", "16:9", "9:16", "4:5", "5:4", "adaptive"],
+      sizes: ["1K", "2K", "4K"],
+      qualityOptions: ["auto", "low", "medium", "high"],
+      references: { images: 16 },
+      supports: ["quality", "count"]
+    },
+    defaults: { size: "1K", aspectRatio: "adaptive", quality: "auto" },
+    usage: { ...IMAGE_USAGE },
+    pricingKind: "image_size",
+    display: {
+      provider: "openai",
+      description: "OpenAI's most capable GPT Image 2.5 tier, built for editing precision; takes text and image input and outputs images in low, medium, high, xhigh, or max quality."
+    },
+    capabilities: { ...IMAGE_GENERATE_EDIT }
+  },
+  {
+    // Same Images API surface as gpt-image-2 plus the "xhigh" and "max" quality tiers.
+    id: "gpt-image-2.5-flare",
+    aliases: [],
+    label: "GPT Image 2.5 Flare",
+    type: "image",
+    adapter: "openai-image",
+    providerModelId: "gpt-image-2.5-flare",
+    lifecycle: "active",
+    exposure: "internal",
+    constraints: {
+      exactSizes: {
+        "1:1": "1024x1024",
+        "3:2": "1536x1024",
+        "2:3": "1024x1536",
+        "16:9": "2048x1152",
+        "9:16": "1152x2048",
+        "4:5": "1024x1280",
+        "5:4": "1280x1024"
+      },
+      aspectRatios: ["1:1", "3:2", "2:3", "16:9", "9:16", "4:5", "5:4", "adaptive"],
+      sizes: ["1K", "2K", "4K"],
+      qualityOptions: ["auto", "low", "medium", "high"],
+      references: { images: 16 },
+      supports: ["quality", "count"]
+    },
+    defaults: { size: "1K", aspectRatio: "adaptive", quality: "auto" },
+    usage: { ...IMAGE_USAGE },
+    pricingKind: "image_size",
+    display: {
+      provider: "openai",
+      description: "OpenAI's fastest GPT Image 2.5 tier for high-quality everyday image generation and editing; takes text and image input and outputs images."
+    },
+    capabilities: { ...IMAGE_GENERATE_EDIT }
+  },
+  {
+    // Cowork allows 7 ratios; trigger-workflows allows 9 (adds 3:2, 9:21, drops 4:5/5:4).
+    id: "muse-image-1.0",
+    aliases: ["meta-muse"],
+    label: "Meta Muse",
+    type: "image",
+    adapter: "fal",
+    providerModelId: "muse-image-1.0",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      aspectRatios: [
+        "21:9",
+        "16:9",
+        "4:3",
+        "3:2",
+        "1:1",
+        "2:3",
+        "3:4",
+        "9:16",
+        "4:5",
+        "5:4",
+        "9:21",
+        "adaptive"
+      ],
+      sizes: ["1K"],
+      references: { images: 10 },
+      supports: ["count", "outputFormat"]
+    },
+    defaults: { size: "1K", aspectRatio: "1:1", outputFormat: "png", count: 1 },
+    usage: { ...IMAGE_USAGE, slug: "muse" },
+    pricingKind: "image_size",
+    display: {
+      provider: "meta",
+      description: "Meta's image model with built-in world knowledge and strong prompt adherence. Great for concept-rich scenes, realistic compositions, and reference-guided edits.",
+      logo: "https://d31vvfpk2ppdyq.cloudfront.net/assets/characters/1787909065794-k9a7zz.png"
+    },
+    capabilities: { ...IMAGE_GENERATE_EDIT }
+  },
+  {
+    id: "seedream-5-0-pro",
+    aliases: ["seedream-4-5"],
+    label: "Seedream 5 Pro",
+    type: "image",
+    adapter: "ark",
+    providerModelId: "dola-seedream-5-0-pro-260628",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      aspectRatios: ["1:1", "4:3", "3:4", "16:9", "9:16", "3:2", "2:3", "21:9", "4:5", "5:4"],
+      sizes: ["1K", "2K"],
+      references: { images: 10 },
+      // Ark's Images API has no output-format, quality or count field; only
+      // the seed is forwarded, so anything else must warn, not vanish.
+      supports: ["seed"]
+    },
+    defaults: { size: "1K", aspectRatio: "1:1" },
+    usage: { ...IMAGE_USAGE },
+    pricingKind: "image_size",
+    display: {
+      provider: "bytedance",
+      description: "ByteDance Seedream 5.0 Pro text-to-image and multi-reference image-to-image (up to 10 inputs) with region editing; outputs 1K, 1.5K, or 2K."
+    },
+    capabilities: { ...IMAGE_GENERATE_EDIT }
+  },
+  {
+    id: "seedream-4.5",
+    aliases: [],
+    label: "Seedream 4.5",
+    type: "image",
+    adapter: "ark",
+    providerModelId: "seedream-4-5-251128",
+    lifecycle: "active",
+    exposure: "internal",
+    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",
+        "9:21"
+      ],
+      sizes: ["1K", "2K", "4K"],
+      references: { images: 10 },
+      supports: ["seed"]
+    },
+    defaults: { size: "1K", aspectRatio: "1:1" },
+    usage: { ...IMAGE_USAGE },
+    pricingKind: "image_size",
+    display: {
+      provider: "bytedance",
+      description: "ByteDance Seedream 4.5 text-to-image and image-to-image with up to 14 reference images and batch image-set generation; outputs 2K or 4K."
+    },
+    capabilities: { ...IMAGE_GENERATE_EDIT }
+  },
+  {
+    id: "flux-2-pro",
+    aliases: [],
+    label: "Flux 2 Pro",
+    type: "image",
+    adapter: "prodia",
+    providerModelId: "flux-2-pro",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      aspectRatios: ["1:1", "2:3", "3:2", "16:9", "9:16", "4:3", "3:4", "4:5", "5:4", "21:9"],
+      sizes: ["1K"],
+      // flux-2-pro's job type always returns image/png; only the seed is forwarded.
+      supports: ["seed"]
+    },
+    defaults: { size: "1K", aspectRatio: "1:1" },
+    usage: { ...IMAGE_USAGE },
+    pricingKind: "image_size",
+    display: {
+      provider: "black-forest-labs",
+      description: "Black Forest Labs FLUX.2 [pro] text-to-image and multi-reference editing model served via Prodia; takes a prompt and up to 8 input images and outputs up to 4096x4096."
+    },
+    capabilities: { ...IMAGE_GENERATE_EDIT }
+  },
+  {
+    // Prodia's gemini-3-pro jobs take aspect_ratio + resolution (1K/2K/4K), not
+    // width/height; 97:30 is outside its aspect_ratio enum.
+    // https://docs.prodia.com/models/gemini-3/
+    id: "prodia-nano-banana-pro",
+    aliases: [],
+    label: "Nano Banana Pro via Prodia",
+    type: "image",
+    adapter: "prodia",
+    providerModelId: "inference.gemini-3-pro.img2img.v1",
+    lifecycle: "active",
+    exposure: "internal",
+    constraints: {
+      aspectRatios: ["1:1", "2:3", "3:2", "4:3", "3:4", "4:5", "5:4", "9:16", "16:9", "21:9"],
+      sizes: ["1K", "2K", "4K"],
+      requires: ["references"],
+      supports: ["seed"]
+    },
+    defaults: { size: "1K", aspectRatio: "1:1" },
+    usage: { ...IMAGE_USAGE },
+    pricingKind: "image_size",
+    display: {
+      provider: "google",
+      description: "Google Gemini 3 Pro image generation and editing served via Prodia; takes a prompt plus up to 3 reference images and outputs 1K, 2K, or 4K."
+    },
+    capabilities: { ...IMAGE_GENERATE_EDIT }
+  },
+  {
+    // Nano Banana (img2img.v2) is flat-rate with aspect_ratio only; 3:2, 4:5,
+    // 5:4 and 97:30 are not in its enum. https://docs.prodia.com/models/nano-banana/
+    id: "prodia-nano-banana",
+    aliases: [],
+    label: "Nano Banana via Prodia",
+    type: "image",
+    adapter: "prodia",
+    providerModelId: "inference.nano-banana.img2img.v2",
+    lifecycle: "active",
+    exposure: "internal",
+    constraints: {
+      aspectRatios: ["1:1", "2:3", "3:4", "4:3", "9:16", "16:9", "21:9"],
+      sizes: ["1K"],
+      requires: ["references"],
+      supports: ["seed"]
+    },
+    defaults: { size: "1K", aspectRatio: "1:1" },
+    usage: { ...IMAGE_USAGE },
+    pricingKind: "image_size",
+    display: {
+      provider: "google",
+      description: "Google Nano Banana image generation and editing served via Prodia; takes a prompt plus 1-3 input images and matches the first input's aspect ratio in roughly 8 seconds."
+    },
+    capabilities: { ...IMAGE_GENERATE_EDIT }
+  },
+  {
+    // WAN 2.7 takes `images` (up to 9) and a `size` enum whose exact values
+    // cover 1:1/16:9/9:16/4:3/3:4 only; other ratios have no spelling.
+    // https://replicate.com/wan-video/wan-2.7-image-pro (input schema)
+    id: "wan-2.7-image-pro",
+    aliases: [],
+    label: "WAN 2.7 Image Pro",
+    type: "image",
+    adapter: "replicate",
+    providerModelId: "wan-video/wan-2.7-image-pro",
+    lifecycle: "active",
+    exposure: "internal",
+    constraints: {
+      aspectRatios: ["1:1", "16:9", "9:16", "4:3", "3:4"],
+      sizes: ["1K", "2K", "4K"],
+      references: { images: 9 },
+      supports: []
+    },
+    defaults: { size: "1K", aspectRatio: "1:1" },
+    usage: { unit: "compute_seconds", modality: "IMAGE" },
+    pricingKind: "image_size",
+    display: {
+      provider: "alibaba",
+      description: "Alibaba Wan 2.7 Image Pro text-to-image and image-editing model; takes a prompt plus up to 9 reference images and generates up to 4K (4096x4096)."
+    },
+    capabilities: { ...IMAGE_GENERATE_EDIT }
+  },
+  {
+    // Qwen Image 2 Pro has aspect_ratio but no size input; resolution follows
+    // the reference image (match_input_image) or the provider default.
+    // https://replicate.com/qwen/qwen-image-2-pro (input schema)
+    id: "qwen-image-2-pro",
+    aliases: [],
+    label: "Qwen Image 2 Pro",
+    type: "image",
+    adapter: "replicate",
+    providerModelId: "qwen/qwen-image-2-pro",
+    lifecycle: "active",
+    exposure: "internal",
+    constraints: {
+      aspectRatios: ["1:1", "16:9", "9:16", "4:3", "3:4", "3:2", "2:3"],
+      sizes: [],
+      references: { images: 1 },
+      requires: ["references"],
+      supports: []
+    },
+    defaults: { aspectRatio: "1:1" },
+    usage: { unit: "compute_seconds", modality: "IMAGE" },
+    pricingKind: "image_size",
+    display: {
+      provider: "alibaba",
+      description: "Alibaba Qwen Image 2 Pro unified text-to-image and editing model; takes a prompt plus an optional single reference image and generates native 2K output."
+    },
+    capabilities: { ...IMAGE_GENERATE_EDIT }
+  },
+  {
+    // Qwen Image Edit Plus takes aspect_ratio (or match_input_image) and
+    // output_format, but no size; no default ratio so edits keep following
+    // the input image. https://replicate.com/qwen-edit-apps/qwen-image-edit-plus-lora-skin
+    id: "qwen-image-edit-plus-lora-skin",
+    aliases: [],
+    label: "Qwen Image Edit Plus",
+    type: "image",
+    adapter: "replicate",
+    providerModelId: "qwen-edit-apps/qwen-image-edit-plus-lora-skin",
+    lifecycle: "active",
+    exposure: "internal",
+    constraints: {
+      aspectRatios: ["1:1", "16:9", "9:16", "4:3", "3:4"],
+      sizes: [],
+      references: { images: 1 },
+      requires: ["references"],
+      supports: ["outputFormat"]
+    },
+    defaults: { outputFormat: "png" },
+    usage: { unit: "compute_seconds", modality: "IMAGE" },
+    pricingKind: "image_size",
+    display: {
+      provider: "alibaba",
+      description: "Alibaba Qwen Image Edit Plus with Lightning and a skin-retouching LoRA; takes one or more reference images plus an edit prompt and outputs webp, jpg, or png."
+    },
+    capabilities: { ...IMAGE_GENERATE_EDIT }
+  }
+];
+
+// src/catalog/other.ts
+var OTHER_MODELS = [
+  {
+    id: "lyria-3-clip-preview",
+    aliases: [],
+    label: "Lyria 3 Clip",
+    type: "audio",
+    adapter: "google-genai",
+    providerModelId: "lyria-3-clip-preview",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      durations: [30],
+      references: { images: 10 },
+      supports: []
+    },
+    defaults: {},
+    usage: { unit: "images", modality: "AUDIO" },
+    pricingKind: "flat",
+    display: {
+      provider: "google",
+      description: "Google Lyria 3 Clip short music clips, loops, and previews; takes text and up to 10 image prompts and returns a fixed 30-second MP3 plus lyrics."
+    },
+    capabilities: {}
+  },
+  {
+    id: "lyria-3-pro-preview",
+    aliases: [],
+    label: "Lyria 3 Pro",
+    type: "audio",
+    adapter: "google-genai",
+    providerModelId: "lyria-3-pro-preview",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      references: { images: 10 },
+      supports: []
+    },
+    defaults: {},
+    usage: { unit: "images", modality: "AUDIO" },
+    pricingKind: "flat",
+    display: {
+      provider: "google",
+      description: "Google Lyria 3 Pro flagship music model for full-length songs with verses, choruses, and bridges; takes text and up to 10 image prompts and returns MP3 plus lyrics."
+    },
+    capabilities: {}
+  },
+  {
+    id: "google/lyria-2",
+    aliases: [],
+    label: "Lyria 2 via Replicate",
+    type: "audio",
+    adapter: "replicate",
+    providerModelId: "google/lyria-2",
+    lifecycle: "active",
+    exposure: "internal",
+    constraints: {
+      durations: [30],
+      supports: ["seed", "negativePrompt"]
+    },
+    defaults: {},
+    usage: { unit: "compute_seconds", modality: "AUDIO" },
+    pricingKind: "flat",
+    display: {
+      provider: "google",
+      description: "Lyria 2 music generation on Replicate; produces 30 seconds of 48 kHz stereo audio from a text prompt."
+    },
+    capabilities: {}
+  },
+  {
+    id: "mmaudio",
+    aliases: ["zsxkib/mmaudio", "video-to-audio"],
+    label: "MMAudio video-to-audio",
+    type: "audio",
+    adapter: "replicate",
+    providerModelId: "zsxkib/mmaudio:62871fb59889b2d7c13777f08deb3b36bdff88f7e1d53a50ad7694548a41b484",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      durations: { min: 1, max: 30 },
+      requires: ["sourceVideo"],
+      supports: ["seed", "negativePrompt"]
+    },
+    defaults: { durationSeconds: 8 },
+    usage: { unit: "compute_seconds", modality: "AUDIO" },
+    pricingKind: "flat",
+    display: {
+      provider: "mmaudio",
+      description: "MMAudio V2 video-to-audio model on Replicate; takes a source video and prompt and synthesizes synchronized environmental, action, and speech-like audio."
+    },
+    capabilities: {}
+  },
+  {
+    id: "elevenlabs-music",
+    aliases: ["elevenlabs/music"],
+    label: "ElevenLabs Music via Replicate",
+    type: "audio",
+    adapter: "replicate",
+    providerModelId: "elevenlabs/music",
+    lifecycle: "active",
+    exposure: "internal",
+    constraints: {
+      durations: { min: 5, max: 300 },
+      supports: ["instrumental"]
+    },
+    defaults: { durationSeconds: 15, instrumental: true },
+    // Replicate's run() does not report predict_time; billed by generated length.
+    usage: { unit: "seconds", modality: "AUDIO" },
+    pricingKind: "flat",
+    display: {
+      provider: "elevenlabs",
+      description: "ElevenLabs Music text-to-music on Replicate; generates full songs, instrumental or with vocals, up to 5 minutes long at up to 44.1 kHz."
+    },
+    capabilities: {}
+  },
+  {
+    id: "topaz",
+    aliases: ["replicate-upscaler", "upscale", "topazlabs/image-upscale"],
+    label: "Topaz Upscale",
+    type: "upscale",
+    adapter: "replicate",
+    providerModelId: "topazlabs/image-upscale",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      references: { images: 1 },
+      requires: ["references"],
+      supports: ["scaleFactor", "outputFormat", "upscale"]
+    },
+    defaults: { scaleFactor: 2, outputFormat: "png" },
+    usage: { unit: "compute_seconds", modality: "IMAGE" },
+    pricingKind: "flat",
+    display: {
+      provider: "topaz",
+      description: "Topaz Labs AI image upscaling on Replicate; takes one input image and upscales up to 6x with selectable enhancement models and optional face enhancement."
+    },
+    capabilities: { imageEdit: true }
+  },
+  {
+    id: "bg-remove",
+    aliases: [],
+    label: "Background Remover",
+    type: "bg-remove",
+    adapter: "prodia",
+    providerModelId: "inference.remove-background.v1",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      references: { images: 1 },
+      requires: ["references"],
+      supports: []
+    },
+    defaults: {},
+    usage: { unit: "images", modality: "IMAGE" },
+    pricingKind: "flat",
+    display: {
+      provider: "prodia",
+      description: "Prodia background removal; takes one input image and returns a PNG cut-out of the subject plus an optional black-and-white mask."
+    },
+    capabilities: { imageEdit: true }
+  }
+];
+
+// src/catalog/text.ts
+var TEXT_USAGE = { unit: "tokens", modality: "TEXT" };
+function text(id, label, providerModelId, extra) {
+  return {
+    id,
+    aliases: [],
+    label,
+    type: "text",
+    adapter: "ai-sdk-text",
+    providerModelId,
+    lifecycle: "active",
+    exposure: "internal",
+    constraints: { supports: [] },
+    defaults: {},
+    usage: { ...TEXT_USAGE },
+    pricingKind: "token",
+    ...extra
+  };
+}
+var ANTHROPIC_THINKING = {
+  anthropic: { thinking: { type: "enabled", budgetTokens: 4e3 }, sendReasoning: true }
+};
+var ANTHROPIC_ADAPTIVE = {
+  anthropic: { thinking: { type: "adaptive", display: "summarized" }, sendReasoning: true }
+};
+var OPENAI_REASONING = { openai: { reasoningEffort: "medium", reasoningSummary: "auto" } };
+var DEEPSEEK_THINKING = { deepseek: { thinking: { type: "enabled" } } };
+var TEXT_MODELS = [
+  text("gpt-4o", "GPT-4o", "gpt-4o", {
+    exposure: "internal",
+    display: {
+      provider: "openai",
+      costTier: 2,
+      description: "OpenAI GPT-4o flagship multimodal model; text and image input, 128K-token context."
+    },
+    capabilities: { vision: true, tools: true, structuredOutput: true, contextWindow: 128e3 }
+  }),
+  text("gpt-4o-mini", "GPT-4o mini", "gpt-4o-mini", {
+    exposure: "internal",
+    display: {
+      provider: "openai",
+      costTier: 1,
+      description: "OpenAI GPT-4o mini fast, low-cost multimodal model; text and image input, 128K-token context."
+    },
+    capabilities: { vision: true, tools: true, structuredOutput: true, contextWindow: 128e3 }
+  }),
+  text("gpt-5.1", "GPT-5.1", "gpt-5.1", {
+    exposure: "catalog",
+    display: {
+      provider: "openai",
+      costTier: 3,
+      description: "OpenAI GPT-5.1 reasoning model for coding and agentic tasks; text and image input, 400K-token context."
+    },
+    capabilities: { vision: true, tools: true, structuredOutput: true, contextWindow: 4e5 }
+  }),
+  text("gpt-5.2", "GPT-5.2", "gpt-5.2", {
+    exposure: "catalog",
+    display: {
+      provider: "openai",
+      costTier: 3,
+      description: "OpenAI GPT-5.2 reasoning model for complex professional work; text and image input, 400K-token context."
+    },
+    capabilities: { vision: true, tools: true, structuredOutput: true, contextWindow: 4e5 }
+  }),
+  // 1.05M-token context: https://platform.openai.com/docs/models/gpt-5.4
+  text("gpt-5.4", "GPT-5.4", "gpt-5.4", {
+    exposure: "internal",
+    display: {
+      provider: "openai",
+      costTier: 3,
+      description: "OpenAI GPT-5.4 flagship reasoning model; text and image input, 1.05M-token context."
+    },
+    capabilities: {
+      vision: true,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 105e4
+    }
+  }),
+  // OpenAI never shipped gpt-5.1-mini or gpt-5.2-mini; callers that still send them get 5.4 mini.
+  text("gpt-5.4-mini", "GPT-5.4 mini", "gpt-5.4-mini", {
+    exposure: "catalog",
+    aliases: ["gpt-5.1-mini", "gpt-5.2-mini"],
+    display: {
+      provider: "openai",
+      costTier: 1,
+      description: "OpenAI GPT-5.4 mini efficient reasoning model for high-volume workloads; text and image input, 400K-token context."
+    },
+    capabilities: { vision: true, tools: true, structuredOutput: true, contextWindow: 4e5 }
+  }),
+  text("gpt-5.5", "GPT-5.5", "gpt-5.5", {
+    exposure: "catalog",
+    display: {
+      provider: "openai",
+      costTier: 3,
+      description: "OpenAI GPT-5.5 flagship reasoning model for the most complex professional and coding work; text and image input, 1.05M-token context."
+    },
+    capabilities: {
+      vision: true,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 105e4
+    }
+  }),
+  text("gpt-5.6-luna", "GPT-5.6 Luna", "gpt-5.6-luna", {
+    providerOptions: OPENAI_REASONING,
+    exposure: "catalog",
+    display: {
+      provider: "openai",
+      costTier: 1,
+      description: "OpenAI GPT-5.6 Luna cost-optimized tier of the GPT-5.6 reasoning family; text and image input, 1.05M-token context."
+    },
+    capabilities: {
+      reasoning: true,
+      vision: true,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 105e4
+    }
+  }),
+  text("gpt-5.6-terra", "GPT-5.6 Terra", "gpt-5.6-terra", {
+    providerOptions: OPENAI_REASONING,
+    exposure: "catalog",
+    display: {
+      provider: "openai",
+      costTier: 2,
+      description: "OpenAI GPT-5.6 Terra balanced tier of the GPT-5.6 reasoning family; text and image input, 1.05M-token context."
+    },
+    capabilities: {
+      reasoning: true,
+      vision: true,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 105e4
+    }
+  }),
+  text("gpt-5.6-sol", "GPT-5.6 Sol", "gpt-5.6-sol", {
+    providerOptions: OPENAI_REASONING,
+    exposure: "catalog",
+    display: {
+      provider: "openai",
+      costTier: 3,
+      description: "OpenAI GPT-5.6 Sol flagship tier of the GPT-5.6 reasoning family; text and image input, 1.05M-token context."
+    },
+    capabilities: {
+      reasoning: true,
+      vision: true,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 105e4
+    }
+  }),
+  // Library key "claude-sonnet-4.5" maps to the SDK id "claude-sonnet-4-5".
+  text("claude-sonnet-4-5", "Claude Sonnet 4.5", "claude-sonnet-4-5", {
+    exposure: "internal",
+    aliases: ["claude-sonnet-4.5", "claude-sonnet-4-5-20250929"],
+    display: {
+      provider: "anthropic",
+      costTier: 2,
+      description: "Anthropic Claude Sonnet 4.5 (legacy) balanced model; text and image input, 200K-token context."
+    },
+    capabilities: { vision: true, tools: true, structuredOutput: true, contextWindow: 2e5 }
+  }),
+  // Anthropic retired the dated 4.0 ids; callers that still send them get 4.6.
+  text("claude-sonnet-4-6", "Claude Sonnet 4.6", "claude-sonnet-4-6", {
+    providerOptions: ANTHROPIC_THINKING,
+    exposure: "catalog",
+    aliases: [
+      "claude-sonnet-4.6",
+      "claude-sonnet-4",
+      "claude-sonnet-4.0",
+      "claude-sonnet-4-20250514"
+    ],
+    display: {
+      provider: "anthropic",
+      costTier: 2,
+      description: "Anthropic Claude Sonnet 4.6 (legacy) balanced model; text and image input, 1M-token context."
+    },
+    capabilities: {
+      reasoning: true,
+      vision: true,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 1e6
+    }
+  }),
+  text("claude-sonnet-5", "Claude Sonnet 5", "claude-sonnet-5", {
+    providerOptions: ANTHROPIC_ADAPTIVE,
+    exposure: "catalog",
+    display: {
+      provider: "anthropic",
+      costTier: 3,
+      description: "Anthropic Claude Sonnet 5 current balanced model; text and image input, 1M-token context."
+    },
+    capabilities: {
+      reasoning: true,
+      vision: true,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 1e6
+    }
+  }),
+  text("claude-opus-4-8", "Claude Opus 4.8", "claude-opus-4-8", {
+    providerOptions: ANTHROPIC_ADAPTIVE,
+    exposure: "catalog",
+    aliases: ["claude-opus-4.8", "claude-opus-4", "claude-opus-4.0", "claude-opus-4-20250514"],
+    display: {
+      provider: "anthropic",
+      costTier: 3,
+      description: "Anthropic Claude Opus 4.8 (legacy) frontier model; text and image input, 1M-token context."
+    },
+    capabilities: {
+      reasoning: true,
+      vision: true,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 1e6
+    }
+  }),
+  text("claude-haiku-4-5", "Claude Haiku 4.5", "claude-haiku-4-5", {
+    providerOptions: ANTHROPIC_THINKING,
+    exposure: "catalog",
+    aliases: ["claude-haiku-4.5", "claude-haiku-4-5-20251001", "claude-haiku-4"],
+    display: {
+      provider: "anthropic",
+      costTier: 1,
+      description: "Anthropic Claude Haiku 4.5 fast model with near-frontier intelligence; text and image input, 200K-token context."
+    },
+    capabilities: {
+      reasoning: true,
+      vision: true,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 2e5
+    }
+  }),
+  text("claude-opus-5", "Claude Opus 5", "claude-opus-5", {
+    exposure: "internal",
+    aliases: ["claude-opus-5.0"],
+    display: {
+      provider: "anthropic",
+      costTier: 3,
+      description: "Anthropic Claude Opus 5 current frontier model for agentic coding and enterprise work; text and image input, 1M-token context."
+    },
+    capabilities: {
+      vision: true,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 1e6
+    }
+  }),
+  text("claude-fable-5", "Claude Fable 5", "claude-fable-5", {
+    exposure: "internal",
+    display: {
+      provider: "anthropic",
+      costTier: 3,
+      description: "Anthropic Claude Fable 5 (legacy) model for demanding reasoning and long-horizon agentic work; text and image input, 1M-token context."
+    },
+    capabilities: {
+      vision: true,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 1e6
+    }
+  }),
+  text("gemini-2.5-flash", "Gemini 2.5 Flash", "gemini-2.5-flash", {
+    exposure: "internal",
+    display: {
+      provider: "google",
+      costTier: 1,
+      description: "Google Gemini 2.5 Flash low-latency multimodal model for high-volume reasoning over text, images, video, and audio; 1M-token context."
+    },
+    capabilities: {
+      vision: true,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 1048576
+    }
+  }),
+  text("gemini-2.5-pro", "Gemini 2.5 Pro", "gemini-2.5-pro", {
+    exposure: "internal",
+    display: {
+      provider: "google",
+      costTier: 2,
+      description: "Google Gemini 2.5 Pro thinking model for complex reasoning over code, math, and documents; text, image, video, audio, and PDF input, 1M-token context."
+    },
+    capabilities: {
+      vision: true,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 1048576
+    }
+  }),
+  text("gemini-3.1-pro-preview", "Gemini 3.1 Pro (preview)", "gemini-3.1-pro-preview", {
+    exposure: "catalog",
+    display: {
+      provider: "google",
+      costTier: 2,
+      description: "Google Gemini 3.1 Pro preview reasoning model for software engineering, tool use, and reliable agentic workflows; 1M-token context."
+    },
+    capabilities: {
+      vision: true,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 1048576
+    }
+  }),
+  text("gemini-3.5-flash", "Gemini 3.5 Flash", "gemini-3.5-flash", {
+    exposure: "catalog",
+    display: {
+      provider: "google",
+      costTier: 1,
+      description: "Google Gemini 3.5 Flash high-throughput model for frontier-level agentic tasks; text, image, video, audio, and PDF input, 1M-token context."
+    },
+    capabilities: {
+      vision: true,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 1048576
+    }
+  }),
+  text("gemini-3.6-flash", "Gemini 3.6 Flash", "gemini-3.6-flash", {
+    exposure: "catalog",
+    display: {
+      provider: "google",
+      costTier: 1,
+      description: "Google Gemini 3.6 Flash general-purpose multimodal agentic model balancing speed and capability; 1M-token context."
+    },
+    capabilities: {
+      vision: true,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 1048576
+    }
+  }),
+  text("gemini-3.7-flash", "Gemini 3.7 Flash", "gemini-3.7-flash", {
+    exposure: "catalog",
+    display: {
+      provider: "google",
+      costTier: 1,
+      recommended: true,
+      description: "Google Gemini 3.7 Flash natively multimodal reasoning model for coding, agentic workflows, and multi-step execution; 1M-token context."
+    },
+    capabilities: {
+      vision: true,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 1048576
+    }
+  }),
+  text("gemini-3.8-flash", "Gemini 3.8 Flash", "gemini-3.8-flash", {
+    exposure: "internal",
+    display: {
+      provider: "google",
+      costTier: 1,
+      description: "Google Gemini 3.8 Flash frontier model for long-horizon software engineering, autonomous agents, and enterprise workflows; 1M-token context."
+    },
+    capabilities: {
+      vision: true,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 1048576
+    }
+  }),
+  // Renamed to gemini-3.6-flash in vibe-shared model-migration.ts.
+  text("gemini-3-flash-preview", "Gemini 3 Flash (preview)", "gemini-3-flash-preview", {
+    exposure: "internal",
+    lifecycle: "deprecated",
+    replacedBy: "gemini-3.6-flash"
+  }),
+  text("gemini-3.1-flash-lite", "Gemini 3.1 Flash Lite", "gemini-3.1-flash-lite", {
+    exposure: "internal",
+    display: {
+      provider: "google",
+      costTier: 1,
+      description: "Google Gemini 3.1 Flash-Lite low-latency multimodal model for high-volume agentic, translation, and data-extraction tasks; 1M-token context."
+    },
+    capabilities: {
+      vision: true,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 1048576
+    }
+  }),
+  text("sonar", "Sonar", "sonar", {
+    exposure: "internal",
+    display: {
+      provider: "perplexity",
+      costTier: 1,
+      description: "Perplexity Sonar lightweight web-grounded search model returning cited answers; 128K-token context."
+    },
+    capabilities: {
+      vision: true,
+      tools: false,
+      structuredOutput: true,
+      contextWindow: 128e3
+    }
+  }),
+  text("sonar-pro", "Sonar Pro", "sonar-pro", {
+    exposure: "internal",
+    display: {
+      provider: "perplexity",
+      costTier: 2,
+      description: "Perplexity Sonar Pro advanced web-grounded search model for complex multi-step queries; 200K-token context."
+    },
+    capabilities: {
+      vision: true,
+      tools: false,
+      structuredOutput: true,
+      contextWindow: 2e5
+    }
+  }),
+  text("sonar-reasoning-pro", "Sonar Reasoning Pro", "sonar-reasoning-pro", {
+    exposure: "internal",
+    display: {
+      provider: "perplexity",
+      costTier: 2,
+      description: "Perplexity Sonar Reasoning Pro chain-of-thought web-grounded search model for complex problem-solving; 128K-token context."
+    },
+    capabilities: {
+      vision: true,
+      tools: false,
+      structuredOutput: true,
+      contextWindow: 128e3
+    }
+  }),
+  // DeepSeek retired V4 Flash upstream and routes the old id to V4.1 Flash at
+  // Flash rates (https://api-docs.deepseek.com/news/news260910/); the old id
+  // stays an alias so stored selections keep resolving.
+  text("deepseek-flash", "DeepSeek V4.1 Flash", "deepseek-flash", {
+    providerOptions: DEEPSEEK_THINKING,
+    aliases: ["deepseek-v4-flash"],
+    exposure: "catalog",
+    display: {
+      provider: "deepseek",
+      costTier: 1,
+      recommended: true,
+      description: "DeepSeek V4.1 Flash thinking/non-thinking model with vision, tool calls, and JSON output; 1M-token context and up to 384K output tokens."
+    },
+    capabilities: {
+      reasoning: true,
+      vision: true,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 1e6
+    }
+  }),
+  text("deepseek-v4-pro", "DeepSeek V4 Pro", "deepseek-v4-pro", {
+    providerOptions: DEEPSEEK_THINKING,
+    exposure: "catalog",
+    display: {
+      provider: "deepseek",
+      costTier: 2,
+      description: "DeepSeek V4 Pro thinking/non-thinking text model with tool calls and JSON output (no vision); 1M-token context and up to 384K output tokens."
+    },
+    capabilities: {
+      reasoning: true,
+      vision: false,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 1e6
+    }
+  }),
+  // 1M-token context: https://docs.x.ai/docs/models/grok-4.3
+  text("grok-4.3", "Grok 4.3", "grok-4.3", {
+    exposure: "catalog",
+    display: {
+      provider: "xai",
+      costTier: 2,
+      description: "xAI Grok 4.3 reasoning model with tool calling and structured outputs; text and image input, 1M-token context."
+    },
+    capabilities: {
+      vision: true,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 1e6
+    }
+  }),
+  // Moonshot retired kimi-k2.5 on 2026-08-31 and no longer publishes its specs.
+  text("kimi-k2.5", "Kimi K2.5", "kimi-k2.5", {
+    defaults: { temperature: 1 },
+    exposure: "catalog",
+    display: {
+      provider: "moonshot",
+      costTier: 2,
+      description: "Moonshot Kimi K2.5; the provider retired the model on 2026-08-31 and calls now return a model-not-found error."
+    },
+    capabilities: {}
+  }),
+  text("kimi-k3", "Kimi K3", "kimi-k3", {
+    defaults: { temperature: 1 },
+    exposure: "catalog",
+    display: {
+      provider: "moonshot",
+      costTier: 2,
+      description: "Moonshot Kimi K3 flagship model with native visual understanding and always-on thinking; text, image, and video input, 1M-token context."
+    },
+    capabilities: {
+      reasoning: true,
+      vision: true,
+      tools: true,
+      structuredOutput: true,
+      contextWindow: 1e6
+    }
+  }),
+  // Cowork calls the provider "MiniMax-M3"; the pricing key is lowercase.
+  // 1M-token context: https://platform.minimax.io/docs/api-reference/text-openai-api
+  text("minimax-m3", "MiniMax M3", "MiniMax-M3", {
+    exposure: "catalog",
+    aliases: ["MiniMax-M3"],
+    display: {
+      provider: "minimax",
+      costTier: 2,
+      description: "MiniMax M3 frontier multimodal model for agentic reasoning, tool use, and coding; text, image, and video input, 1M-token context."
+    },
+    capabilities: { vision: true, tools: true, contextWindow: 1e6 }
+  })
+];
+
+// src/catalog/video.ts
+var VIDEO_USAGE = { unit: "seconds", modality: "VIDEO" };
+var SEEDANCE_RATIOS = ["21:9", "16:9", "4:3", "1:1", "3:4", "9:16", "adaptive"];
+var KLING_RATIOS = ["16:9", "9:16", "1:1", "adaptive"];
+var OMNI_RATIOS = ["16:9", "9:16"];
+var GROK_RATIOS = ["1:1", "16:9", "9:16", "4:3", "3:4", "3:2", "2:3"];
+var H3_RATIOS = ["21:9", "16:9", "4:3", "1:1", "3:4", "9:16"];
+var H3_REFERENCE_RATIOS = ["adaptive", ...H3_RATIOS];
+var NEGATIVE_PROMPT = "blur, distort, and low quality";
+var FRAMES_EXCLUDE_REFERENCES = { framesExcludeReferences: true };
+var GOOGLE_LOGO = "https://d31vvfpk2ppdyq.cloudfront.net/assets/image/192485d2-a76a-4657-9c06-930b2c70badd_compressed.png";
+var VIDEO_MODELS = [
+  {
+    id: "veo-3.1",
+    aliases: [
+      "veo-3.1-generate-001",
+      "veo-3.1-generate-preview",
+      "veo3",
+      "veo-3.1-frame-specific"
+    ],
+    label: "Veo 3.1",
+    type: "video",
+    adapter: "google-genai",
+    providerModelId: "veo-3.1-generate-preview",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      aspectRatios: [...OMNI_RATIOS],
+      resolutions: ["720p", "1080p", "4k"],
+      durations: [8],
+      references: { images: 3 },
+      supports: ["lastFrame"]
+    },
+    defaults: { durationSeconds: 8, resolution: "720p", aspectRatio: "16:9" },
+    usage: { ...VIDEO_USAGE, slug: "veo" },
+    pricingKind: "video_per_second",
+    display: {
+      provider: "google",
+      description: "Google's highest-quality video generation model. Supports up to 4K resolution and 8-second clips with reference image conditioning.",
+      logo: GOOGLE_LOGO
+    },
+    capabilities: { ...FRAMES_EXCLUDE_REFERENCES }
+  },
+  {
+    id: "veo-3.1-fast",
+    aliases: [
+      "veo-3.1-fast-generate-001",
+      "veo-3.1-fast-generate-preview",
+      "veo-3.1-fast-generate-001-frame-specific"
+    ],
+    label: "Veo 3.1 Fast",
+    type: "video",
+    adapter: "google-genai",
+    providerModelId: "veo-3.1-fast-generate-preview",
+    lifecycle: "active",
+    exposure: "internal",
+    constraints: {
+      aspectRatios: [...OMNI_RATIOS],
+      resolutions: ["720p", "1080p", "4k"],
+      durations: [4, 8],
+      references: { images: 3 },
+      supports: ["lastFrame"]
+    },
+    defaults: { durationSeconds: 8, resolution: "720p", aspectRatio: "16:9" },
+    usage: { ...VIDEO_USAGE },
+    pricingKind: "video_per_second",
+    display: {
+      provider: "google",
+      description: "Faster variant of Google Veo 3.1; text- and image-to-video with natively generated audio at 4 or 8 seconds."
+    },
+    capabilities: { ...FRAMES_EXCLUDE_REFERENCES }
+  },
+  {
+    id: "veo-3.1-lite",
+    aliases: [
+      "veo-3.1-lite-generate-001",
+      "veo-3.1-lite-generate-preview",
+      "veo-3.1-lite-generate-001-frame-specific"
+    ],
+    label: "Veo 3.1 Lite",
+    type: "video",
+    adapter: "google-genai",
+    providerModelId: "veo-3.1-lite-generate-preview",
+    lifecycle: "active",
+    exposure: "internal",
+    constraints: {
+      aspectRatios: [...OMNI_RATIOS],
+      resolutions: ["720p", "1080p"],
+      durations: [4, 8],
+      references: { images: 3 },
+      supports: ["lastFrame"]
+    },
+    defaults: { durationSeconds: 8, resolution: "720p", aspectRatio: "16:9" },
+    usage: { ...VIDEO_USAGE },
+    pricingKind: "video_per_second",
+    display: {
+      provider: "google",
+      description: "Efficiency-focused Google Veo 3.1 Lite for high-volume video generation and editing; text and image input, with no 4K output or video extension."
+    },
+    capabilities: { ...FRAMES_EXCLUDE_REFERENCES }
+  },
+  {
+    id: "gemini-omni-flash",
+    aliases: [
+      "gemini-omni-flash-preview",
+      "geminiOmniFlash",
+      "gemini-omni-video",
+      "gemini-omni-reference",
+      "gemini-omni-edit"
+    ],
+    label: "Gemini Omni Flash",
+    type: "video",
+    adapter: "google-genai",
+    providerModelId: "gemini-omni-flash-preview",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      // 1.0 rejects `resolution`; it renders 720p. No defaults: the adapter
+      // sends ratio and duration only when the caller set them, so a
+      // reference clip keeps its own framing.
+      aspectRatios: [...OMNI_RATIOS],
+      durations: [3, 4, 5, 6, 7, 8, 9, 10],
+      references: { images: 6, videos: 1 },
+      supports: []
+    },
+    defaults: {},
+    usage: { ...VIDEO_USAGE, slug: "gemini_omni" },
+    pricingKind: "video_per_second",
+    display: {
+      provider: "google",
+      description: "Google's conversational video model with native audio, human support, and natural language editing. Aspect ratio only \u2014 resolution and duration are model-determined.",
+      logo: GOOGLE_LOGO
+    },
+    capabilities: { acceptsSourceVideo: true }
+  },
+  {
+    id: "gemini-omni-1.1-flash",
+    aliases: ["gemini-omni-1.1-video", "gemini-omni-1.1-reference", "gemini-omni-1.1-edit"],
+    label: "Gemini Omni 1.1 Flash",
+    type: "video",
+    adapter: "google-genai",
+    providerModelId: "gemini-omni-1.1-flash",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      aspectRatios: [...OMNI_RATIOS],
+      resolutions: ["360p", "720p", "1080p", "4k"],
+      durations: [3, 4, 5, 6, 7, 8, 9, 10],
+      references: { images: 6, videos: 1 },
+      supports: []
+    },
+    defaults: {},
+    usage: { ...VIDEO_USAGE },
+    pricingKind: "video_per_second",
+    display: {
+      provider: "google",
+      description: "Google Gemini Omni Flash 1.1 conversational video generation and editing; text, image, and up to 10-second video input, 3-10 second output at up to 4K and 24 FPS."
+    },
+    capabilities: { acceptsSourceVideo: true }
+  },
+  {
+    // vibe-triggers and Space Builder drop the ratio when a frame is wired; Cowork always sends it.
+    id: "seedance-2.0",
+    aliases: ["seedance2", "seedance-2-0", "dreamina-seedance-2-0-260128"],
+    label: "Seedance 2.0",
+    type: "video",
+    adapter: "ark",
+    providerModelId: "dreamina-seedance-2-0-260128",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      aspectRatios: [...SEEDANCE_RATIOS],
+      resolutions: ["480p", "720p", "1080p", "4k"],
+      durations: [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
+      durationPresets: [4, 6, 8, 10, 12, 15],
+      references: { images: 9, videos: 3, audios: 3, audioOnly: false },
+      ratioIgnoredWith: ["firstFrame"],
+      supports: ["generateAudio", "seed", "lastFrame", "watermark"]
+    },
+    defaults: {
+      durationSeconds: 5,
+      resolution: "720p",
+      aspectRatio: "16:9",
+      generateAudio: true
+    },
+    usage: { ...VIDEO_USAGE, slug: "seedance2", providerLabel: "byteplus" },
+    pricingKind: "video_per_second",
+    display: {
+      provider: "bytedance",
+      description: "High-quality AI video generation with reference image conditioning. Supports 4-15 second clips with human preprocessing.",
+      logo: "https://d31vvfpk2ppdyq.cloudfront.net/assets/image/3bba5d81-4e57-4e71-911b-cdfa05acca2c_compressed.png"
+    },
+    capabilities: { acceptsSourceVideo: true, ...FRAMES_EXCLUDE_REFERENCES }
+  },
+  {
+    id: "seedance-2.0-fast",
+    aliases: ["seedance-2-0-fast", "dreamina-seedance-2-0-fast-260128"],
+    label: "Seedance 2.0 Fast",
+    type: "video",
+    adapter: "ark",
+    providerModelId: "dreamina-seedance-2-0-fast-260128",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      aspectRatios: [...SEEDANCE_RATIOS],
+      resolutions: ["480p", "720p"],
+      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", "watermark"]
+    },
+    defaults: {
+      durationSeconds: 5,
+      resolution: "720p",
+      aspectRatio: "16:9",
+      generateAudio: true
+    },
+    usage: { ...VIDEO_USAGE },
+    pricingKind: "video_per_second",
+    display: {
+      provider: "bytedance",
+      description: "Speed- and cost-optimized ByteDance Seedance 2.0 Fast; text plus image, video, and audio references to generate or edit 4-15 second clips."
+    },
+    capabilities: { acceptsSourceVideo: true, ...FRAMES_EXCLUDE_REFERENCES }
+  },
+  {
+    // generateAudio defaults to false here (trigger-workflows); Space Builder defaults it to true.
+    id: "seedance-2.5",
+    aliases: ["seedance-2-5", "dreamina-seedance-2-5-260628"],
+    label: "Seedance 2.5",
+    type: "video",
+    adapter: "ark",
+    providerModelId: "dreamina-seedance-2-5-260628",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      aspectRatios: [...SEEDANCE_RATIOS],
+      resolutions: ["480p", "720p", "1080p"],
+      durations: { min: 4, max: 30 },
+      references: { images: 30, videos: 10, audios: 10 },
+      ratioIgnoredWith: ["firstFrame"],
+      supports: ["generateAudio", "seed", "lastFrame", "watermark"]
+    },
+    defaults: {
+      durationSeconds: 5,
+      resolution: "720p",
+      aspectRatio: "16:9",
+      generateAudio: false
+    },
+    usage: { ...VIDEO_USAGE },
+    pricingKind: "video_per_second",
+    display: {
+      provider: "bytedance",
+      description: "ByteDance Seedance 2.5 for long-form storytelling and omni-reference generation/editing; up to 50 reference assets and 4-30 second clips at up to 1080p."
+    },
+    capabilities: { acceptsSourceVideo: true, ...FRAMES_EXCLUDE_REFERENCES }
+  },
+  {
+    // Cowork allows 7 reference images and durations 4/6/8/10; Space Builder allows 5 and 1-15s.
+    id: "grok-imagine",
+    aliases: ["grok-imagine-video", "grokImagine"],
+    label: "Grok Imagine",
+    type: "video",
+    adapter: "xai-video",
+    providerModelId: "grok-imagine-video",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      aspectRatios: [...GROK_RATIOS],
+      resolutions: ["480p", "720p"],
+      durations: { min: 1, max: 15 },
+      durationPresets: [4, 6, 8, 10],
+      references: { images: 7 },
+      supports: []
+    },
+    // 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",
+    display: {
+      provider: "xai",
+      description: "xAI's reference-to-video model. Generates short clips conditioned on up to 7 reference images with flexible aspect ratios.",
+      logo: "https://d31vvfpk2ppdyq.cloudfront.net/assets/image/858b7fcc-9bd5-411d-aea1-a7b35ecfffab_compressed.png"
+    },
+    capabilities: { ...FRAMES_EXCLUDE_REFERENCES }
+  },
+  {
+    id: "kling-2.5",
+    aliases: ["kling-v2.5-turbo-pro", "kling-2.5-turbo-pro", "kling-v2.5"],
+    label: "Kling 2.5",
+    type: "video",
+    adapter: "higgsfield",
+    providerModelId: "kling-video/v2.5-turbo/pro/image-to-video",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      aspectRatios: [...KLING_RATIOS],
+      durations: [5, 10],
+      requires: ["firstFrame"],
+      supports: ["cfgScale", "negativePrompt", "lastFrame"]
+    },
+    defaults: {
+      durationSeconds: 5,
+      aspectRatio: "16:9",
+      cfgScale: 0.5,
+      negativePrompt: NEGATIVE_PROMPT
+    },
+    usage: { ...VIDEO_USAGE, slug: "grok" },
+    pricingKind: "video_per_second",
+    display: {
+      provider: "kuaishou",
+      description: "Kuaishou Kling 2.5 Turbo Pro image-to-video; animates a single input image from a text prompt with an optional end image, 5- or 10-second clips."
+    },
+    capabilities: { ...FRAMES_EXCLUDE_REFERENCES }
+  },
+  {
+    // trigger-workflows validates 5/10 only; Space Builder allows 3-15s (Kling30Input).
+    id: "kling-3.0",
+    aliases: ["kling-v3.0-pro", "kling-v3.0", "kling-3.0-pro"],
+    label: "Kling 3.0",
+    type: "video",
+    adapter: "higgsfield",
+    providerModelId: "kling-video/v3.0/pro/image-to-video",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      aspectRatios: [...KLING_RATIOS],
+      durations: { min: 3, max: 15 },
+      requires: ["firstFrame"],
+      supports: ["cfgScale", "negativePrompt", "lastFrame"]
+    },
+    defaults: {
+      durationSeconds: 5,
+      aspectRatio: "16:9",
+      cfgScale: 0.5,
+      negativePrompt: NEGATIVE_PROMPT
+    },
+    usage: { ...VIDEO_USAGE },
+    pricingKind: "video_per_second",
+    display: {
+      provider: "kuaishou",
+      description: "Kuaishou Kling 3.0 Pro image-to-video; a required start image plus a text prompt (or multi-shot prompts) with optional end image and native audio, 3-15 second clips."
+    },
+    capabilities: { ...FRAMES_EXCLUDE_REFERENCES }
+  },
+  {
+    // mode standard/pro selects 720p/1080p; the endpoint takes no separate resolution.
+    id: "kling-v3-omni-video",
+    aliases: ["kwaivgi/kling-v3-omni-video", "kling-v3-omni"],
+    label: "Kling V3 Omni",
+    type: "video",
+    adapter: "replicate",
+    providerModelId: "kwaivgi/kling-v3-omni-video",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      aspectRatios: [
+        "1:1",
+        "16:9",
+        "9:16",
+        "4:3",
+        "3:4",
+        "3:2",
+        "2:3",
+        "21:9",
+        "4:5",
+        "adaptive"
+      ],
+      resolutions: ["720p", "1080p"],
+      durations: { min: 3, max: 15 },
+      references: { images: 7 },
+      supports: ["generateAudio", "lastFrame"]
+    },
+    defaults: { durationSeconds: 5, aspectRatio: "16:9", generateAudio: false, mode: "pro" },
+    usage: { unit: "compute_seconds", modality: "VIDEO" },
+    pricingKind: "video_per_second",
+    display: {
+      provider: "kuaishou",
+      description: "Kuaishou Kling Video 3.0 Omni unified generation and editing from text, start/end images, up to 7 reference images, and reference video; up to 15 seconds at up to 1080p."
+    },
+    capabilities: {}
+  },
+  {
+    id: "minimax-h3",
+    aliases: [],
+    label: "MiniMax H3",
+    type: "video",
+    adapter: "fal",
+    providerModelId: "minimax/h3",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      aspectRatios: [...H3_RATIOS],
+      resolutions: ["480p", "768p", "2k", "4k"],
+      durations: { min: 5, max: 15 },
+      supports: ["seed", "promptExpansion", "lastFrame", "safetyChecker"]
+    },
+    defaults: {
+      durationSeconds: 5,
+      resolution: "2k",
+      aspectRatio: "16:9",
+      promptExpansion: "balanced"
+    },
+    usage: { ...VIDEO_USAGE },
+    pricingKind: "video_per_second",
+    display: {
+      provider: "minimax",
+      description: "MiniMax H3 multimodal video model for text-to-video, image-to-video (first/last frame), and reference-to-video with native audio, up to 2K."
+    },
+    capabilities: { ...FRAMES_EXCLUDE_REFERENCES }
+  },
+  {
+    id: "minimax-h3-max",
+    aliases: [],
+    label: "MiniMax H3 Max",
+    type: "video",
+    adapter: "fal",
+    providerModelId: "minimax/h3-max",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      aspectRatios: [...H3_RATIOS],
+      resolutions: ["480p", "768p"],
+      durations: { min: 5, max: 15 },
+      supports: ["seed", "promptExpansion", "lastFrame", "safetyChecker"]
+    },
+    defaults: {
+      durationSeconds: 5,
+      resolution: "768p",
+      aspectRatio: "16:9",
+      promptExpansion: "balanced"
+    },
+    usage: { ...VIDEO_USAGE },
+    pricingKind: "video_per_second",
+    display: {
+      provider: "minimax",
+      description: "High-speed MiniMax H3 Max variant for text-to-video and image-to-video (first/last frame) only, up to 768p."
+    },
+    capabilities: { ...FRAMES_EXCLUDE_REFERENCES }
+  },
+  {
+    id: "minimax-h3-reference",
+    aliases: [],
+    label: "MiniMax H3 Reference",
+    type: "video",
+    adapter: "fal",
+    providerModelId: "minimax/h3/reference-to-video",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      aspectRatios: [...H3_REFERENCE_RATIOS],
+      resolutions: ["480p", "768p", "2k", "4k"],
+      durations: { min: 5, max: 15 },
+      references: { images: 9, videos: 3, audios: 3, total: 12, audioOnly: false },
+      supports: ["seed", "promptExpansion", "safetyChecker"]
+    },
+    defaults: {
+      durationSeconds: 5,
+      resolution: "2k",
+      aspectRatio: "adaptive",
+      promptExpansion: "balanced"
+    },
+    usage: { ...VIDEO_USAGE },
+    pricingKind: "video_per_second",
+    display: {
+      provider: "minimax",
+      description: "MiniMax H3 reference-to-video mode guided by up to 9 reference images, 3 reference videos, and 3 audio clips; reference and image-to-video inputs cannot be mixed."
+    },
+    capabilities: {}
+  },
+  {
+    id: "minimax-h3-max-reference",
+    aliases: [],
+    label: "MiniMax H3 Max Reference",
+    type: "video",
+    adapter: "fal",
+    providerModelId: "minimax/h3-max/reference-to-video",
+    lifecycle: "active",
+    exposure: "catalog",
+    constraints: {
+      aspectRatios: [...H3_REFERENCE_RATIOS],
+      resolutions: ["480p", "768p"],
+      durations: { min: 5, max: 15 },
+      references: { images: 9, videos: 3, audios: 3, total: 12, audioOnly: false },
+      supports: ["seed", "promptExpansion", "safetyChecker"]
+    },
+    defaults: {
+      durationSeconds: 5,
+      resolution: "768p",
+      aspectRatio: "adaptive",
+      promptExpansion: "balanced"
+    },
+    usage: { ...VIDEO_USAGE },
+    pricingKind: "video_per_second",
+    display: {
+      provider: "minimax",
+      description: "MiniMax H3 Max reference-to-video on fal; conditions on reference images, videos, and audio. MiniMax's own docs do not list a reference mode for H3 Max."
+    },
+    capabilities: {}
+  },
+  {
+    // Duration 3-10s per trigger-workflows mapDuration; the endpoint takes no resolution.
+    id: "kling-o1-v2v",
+    aliases: [],
+    label: "Kling O1 Video to Video",
+    type: "video",
+    adapter: "fal",
+    providerModelId: "fal-ai/kling-video/o1/video-to-video/reference",
+    lifecycle: "active",
+    exposure: "internal",
+    constraints: {
+      aspectRatios: [...KLING_RATIOS],
+      durations: { min: 3, max: 10 },
+      requires: ["sourceVideo"],
+      supports: []
+    },
+    defaults: { durationSeconds: 5, aspectRatio: "16:9" },
+    usage: { ...VIDEO_USAGE },
+    pricingKind: "video_per_second",
+    display: {
+      provider: "kuaishou",
+      description: "Kuaishou Kling O1 Reference Video-to-Video; a prompt plus a 3-10 second reference video (optional reference images) generates 3-10 second clips preserving motion and camera language."
+    },
+    capabilities: { acceptsSourceVideo: true }
+  },
+  {
+    id: "hailuo-02",
+    aliases: [],
+    label: "Hailuo 02",
+    type: "video",
+    adapter: "replicate",
+    providerModelId: "minimax/hailuo-02",
+    lifecycle: "active",
+    exposure: "internal",
+    constraints: {
+      resolutions: ["768p"],
+      durations: [6, 10],
+      supports: ["lastFrame"]
+    },
+    defaults: { durationSeconds: 6, resolution: "768p" },
+    usage: { unit: "compute_seconds", modality: "VIDEO" },
+    pricingKind: "video_per_second",
+    display: {
+      provider: "minimax",
+      description: "MiniMax Hailuo 02 text-to-video and image-to-video model with realistic physics and strong prompt following; outputs 768p at 6 or 10 seconds."
+    },
+    capabilities: {}
+  },
+  {
+    id: "higgsfield-dop-preview",
+    aliases: [],
+    label: "Higgsfield DOP Preview",
+    type: "video",
+    adapter: "higgsfield",
+    providerModelId: "dop-preview",
+    lifecycle: "active",
+    exposure: "internal",
+    constraints: {
+      requires: ["firstFrame"],
+      supports: ["seed", "lastFrame"]
+    },
+    defaults: {},
+    usage: { ...VIDEO_USAGE },
+    pricingKind: "video_per_second",
+    display: {
+      provider: "higgsfield",
+      description: "Higgsfield DOP image-to-video model; animates a required first-frame image from a text prompt with an optional last-frame image and motion controls."
+    },
+    capabilities: {}
+  }
+];
+
+// src/catalog/models.ts
+var MODELS = [
+  ...TEXT_MODELS,
+  ...IMAGE_MODELS,
+  ...VIDEO_MODELS,
+  ...OTHER_MODELS
+];
+
+// src/resolve-options.ts
+var SIZE_ORDER = ["512", "1K", "2K", "4K"];
+var RESOLUTION_ORDER = ["360p", "480p", "720p", "768p", "1080p", "2k", "4k"];
+function ratioValue(r) {
+  if (r === "adaptive") return 1;
+  const [w, h] = r.split(":").map(Number);
+  return w / h;
+}
+function orientation(r) {
+  const v = ratioValue(r);
+  return v > 1 ? "landscape" : v < 1 ? "portrait" : "square";
+}
+function nearestAspectRatio(requested, supported) {
+  if (supported.includes(requested)) return requested;
+  const target = requested === "adaptive" ? "1:1" : requested;
+  const sameBucket = supported.filter(
+    (s) => s !== "adaptive" && orientation(s) === orientation(target)
+  );
+  const pool = sameBucket.length > 0 ? sameBucket : supported.filter((s) => s !== "adaptive");
+  const fallback = pool[0] ?? supported[0];
+  if (!fallback) return target;
+  const dist = (s) => Math.abs(Math.log(ratioValue(s)) - Math.log(ratioValue(target)));
+  return pool.reduce((best, s) => dist(s) < dist(best) ? s : best, fallback);
+}
+function nearestLowerTier(requested, supported, order) {
+  const idx = order.indexOf(requested);
+  for (let i = idx; i >= 0; i--) {
+    const candidate = order[i];
+    if (candidate !== void 0 && supported.includes(candidate)) return candidate;
+  }
+  for (let i = idx + 1; i < order.length; i++) {
+    const candidate = order[i];
+    if (candidate !== void 0 && supported.includes(candidate)) return candidate;
+  }
+  return requested;
+}
+function resolveTier(field, code, requested, supported, order, model, onUnsupported, warnings) {
+  if (requested === void 0 || !supported || supported.includes(requested)) return requested;
+  if (onUnsupported === "error") {
+    throw new ModelKitError({
+      code: "unsupported_option",
+      message: `${model.id} does not support ${field} "${requested}". Supported: ${supported.join(", ")}.`,
+      model: model.id
+    });
+  }
+  const to = nearestLowerTier(requested, supported, order);
+  warnings.push({
+    code,
+    field,
+    from: requested,
+    to,
+    message: `${field} ${requested} coerced to ${to} for ${model.id}`
+  });
+  return to;
+}
+function resolveDuration(requested, durations, model, onUnsupported, warnings) {
+  if (requested === void 0 || !durations) 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",
+    from: requested,
+    to,
+    message: `duration ${requested}s coerced to ${to}s for ${model.id}`
+  });
+  return to;
+}
+function resolveAspectRatio(requested, supported, model, onUnsupported, warnings) {
+  if (requested === void 0 || requested === "adaptive" || !supported) return requested;
+  if (supported.includes(requested)) return requested;
+  if (onUnsupported === "error") {
+    throw new ModelKitError({
+      code: "unsupported_option",
+      message: `${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 ${model.id}`
+  });
+  return to;
+}
+function compact(refs) {
+  return (refs ?? []).filter((r) => Boolean(r));
+}
+function trim(field, refs, limit, model, warnings) {
+  if (limit === void 0 || refs.length <= limit) return refs;
+  warnings.push({
+    code: "references_trimmed",
+    field,
+    from: refs.length,
+    to: limit,
+    message: `${field} trimmed from ${refs.length} to ${limit} for ${model.id}`
+  });
+  return refs.slice(0, limit);
+}
+function resolveImageOptions(model, request, onUnsupported = "error") {
+  const warnings = [];
+  const c = model.constraints;
+  const merged = { ...model.defaults, ...stripUndefined(request), model: model.id };
+  const aspectRatio = resolveAspectRatio(
+    merged.aspectRatio,
+    c.aspectRatios,
+    model,
+    onUnsupported,
+    warnings
+  );
+  if (aspectRatio !== void 0) merged.aspectRatio = aspectRatio;
+  const size = resolveTier(
+    "size",
+    "size_coerced",
+    merged.size,
+    c.sizes,
+    SIZE_ORDER,
+    model,
+    onUnsupported,
+    warnings
+  );
+  if (size !== void 0) merged.size = size;
+  const references = trim(
+    "references",
+    compact(merged.references),
+    c.references?.images ?? c.references?.total,
+    model,
+    warnings
+  );
+  if (references.length > 0) merged.references = references;
+  else delete merged.references;
+  if (c.requires?.includes("references") && references.length === 0) {
+    throw new ModelKitError({
+      code: "missing_input",
+      message: `${model.id} requires at least one reference image`,
+      model: model.id
+    });
+  }
+  dropUnsupported(merged, model, warnings, [
+    "seed",
+    "quality",
+    "count",
+    "outputFormat",
+    "scaleFactor",
+    "safetySettings",
+    "temperature",
+    "upscale"
+  ]);
+  return { request: merged, warnings };
+}
+function resolveVideoOptions(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.referenceImages).length + compact(merged.referenceVideos).length + compact(merged.referenceAudios).length > 0 : merged[input] !== void 0;
+    if (!present)
+      throw new ModelKitError({
+        code: "missing_input",
+        message: `${model.id} requires ${input}`,
+        model: model.id
+      });
+  }
+  if (merged.sourceVideo !== void 0 && !model.capabilities?.acceptsSourceVideo) {
+    if (onUnsupported === "error") {
+      throw new ModelKitError({
+        code: "unsupported_option",
+        message: `${model.id} does not take a source video`,
+        model: model.id
+      });
+    }
+    warnings.push({
+      code: "option_ignored",
+      field: "sourceVideo",
+      message: `${model.id} does not take a source video; dropped`
+    });
+    delete merged.sourceVideo;
+  }
+  const hasFrames = merged.firstFrame !== void 0 || merged.lastFrame !== void 0;
+  const hasRefs = compact(merged.referenceImages).length + compact(merged.referenceVideos).length + compact(merged.referenceAudios).length > 0;
+  if (hasFrames && hasRefs && model.capabilities?.framesExcludeReferences) {
+    if (onUnsupported === "error") {
+      throw new ModelKitError({
+        code: "unsupported_option",
+        message: `${model.id} takes frames or references, not both`,
+        model: model.id
+      });
+    }
+    warnings.push({
+      code: "option_ignored",
+      field: "referenceImages",
+      message: `${model.id} takes frames or references, not both; references dropped`
+    });
+    delete merged.referenceImages;
+    delete merged.referenceVideos;
+    delete merged.referenceAudios;
+  }
+  for (const input of c.ratioIgnoredWith ?? []) {
+    if (merged[input] !== void 0 && merged.aspectRatio !== void 0) {
+      if (request.aspectRatio !== void 0) {
+        warnings.push({
+          code: "ratio_dropped",
+          field: "aspectRatio",
+          from: merged.aspectRatio,
+          message: `${model.id} ignores aspectRatio when ${input} is set`
+        });
+      }
+      delete merged.aspectRatio;
+    }
+  }
+  const aspectRatio = resolveAspectRatio(
+    merged.aspectRatio,
+    c.aspectRatios,
+    model,
+    onUnsupported,
+    warnings
+  );
+  if (aspectRatio !== void 0) merged.aspectRatio = aspectRatio;
+  const resolution = resolveTier(
+    "resolution",
+    "resolution_coerced",
+    merged.resolution,
+    c.resolutions,
+    RESOLUTION_ORDER,
+    model,
+    onUnsupported,
+    warnings
+  );
+  if (resolution !== void 0) merged.resolution = resolution;
+  const duration = resolveDuration(
+    merged.durationSeconds,
+    c.durations,
+    model,
+    onUnsupported,
+    warnings
+  );
+  if (duration !== void 0) merged.durationSeconds = duration;
+  const limits = c.references ?? {};
+  let images = trim(
+    "referenceImages",
+    compact(merged.referenceImages),
+    limits.images,
+    model,
+    warnings
+  );
+  let videos = trim(
+    "referenceVideos",
+    compact(merged.referenceVideos),
+    limits.videos,
+    model,
+    warnings
+  );
+  let audios = trim(
+    "referenceAudios",
+    compact(merged.referenceAudios),
+    limits.audios,
+    model,
+    warnings
+  );
+  if (limits.total !== void 0) {
+    let over = images.length + videos.length + audios.length - limits.total;
+    if (over > 0) {
+      const cut = Math.min(over, audios.length);
+      audios = trim("referenceAudios", audios, audios.length - cut, model, warnings);
+      over -= cut;
+    }
+    if (over > 0) {
+      const cut = Math.min(over, videos.length);
+      videos = trim("referenceVideos", videos, videos.length - cut, model, warnings);
+      over -= cut;
+    }
+    if (over > 0)
+      images = trim("referenceImages", images, images.length - over, model, warnings);
+  }
+  if (limits.audioOnly === false && audios.length > 0 && images.length === 0 && videos.length === 0 && !merged.firstFrame && !merged.sourceVideo) {
+    throw new ModelKitError({
+      code: "missing_input",
+      message: `${model.id} rejects audio-only references`,
+      model: model.id
+    });
+  }
+  setOrDelete(merged, "referenceImages", images);
+  setOrDelete(merged, "referenceVideos", videos);
+  setOrDelete(merged, "referenceAudios", audios);
+  dropUnsupported(merged, model, warnings, [
+    "seed",
+    "negativePrompt",
+    "cfgScale",
+    "generateAudio",
+    "promptExpansion",
+    "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 ${input}`,
+        model: model.id
+      });
+  }
+  const duration = resolveDuration(
+    merged.durationSeconds,
+    c.durations,
+    model,
+    onUnsupported,
+    warnings
+  );
+  if (duration !== void 0) merged.durationSeconds = duration;
+  const references = trim(
+    "references",
+    compact(merged.references),
+    c.references?.images ?? c.references?.total,
+    model,
+    warnings
+  );
+  if (references.length > 0) merged.references = references;
+  else delete merged.references;
+  dropUnsupported(merged, model, warnings, ["seed", "negativePrompt", "instrumental"]);
+  return { request: merged, warnings };
+}
+function dropUnsupported(req, model, warnings, fields) {
+  const supports = model.constraints.supports ?? [];
+  for (const field of fields) {
+    if (req[field] !== void 0 && !supports.includes(field)) {
+      warnings.push({
+        code: "option_ignored",
+        field,
+        from: req[field],
+        message: `${model.id} does not support ${field}; ignored`
+      });
+      delete req[field];
+    }
+  }
+}
+function setOrDelete(req, field, refs) {
+  if (refs.length > 0) req[field] = refs;
+  else delete req[field];
+}
+function stripUndefined(obj) {
+  return Object.fromEntries(Object.entries(obj).filter(([, v]) => v !== void 0));
+}
+
+// src/catalog-entry.ts
+function resolveOptions(model, request, onUnsupported = "error") {
+  switch (model.type) {
+    case "image":
+    case "upscale":
+    case "bg-remove":
+      return resolveImageOptions(model, request, onUnsupported);
+    case "video":
+      return resolveVideoOptions(model, request, onUnsupported);
+    case "audio":
+      return resolveAudioOptions(model, request, onUnsupported);
+    default:
+      throw new Error(`resolveOptions: unsupported model type "${model.type}"`);
+  }
+}
+export {
+  Catalog,
+  IMAGE_MODELS,
+  MODELS,
+  ModelKitError,
+  OTHER_MODELS,
+  TEXT_MODELS,
+  VIDEO_MODELS,
+  isModelKitError,
+  nearestAspectRatio,
+  resolveAudioOptions,
+  resolveImageOptions,
+  resolveOptions,
+  resolveVideoOptions
+};
+//# sourceMappingURL=catalog.js.map