@shop-os/modelkit 0.1.10 → 0.2.0

package/dist/catalog.js

@@ -93,11 +93,18 @@ var Catalog = class _Catalog {
       ...rows
     ]);
   }
+  /** Atomically swap every row. Builds the maps first so a bad batch throws before anything changes. */
+  replace(rows) {
+    const next = new _Catalog(rows);
+    this.byId.clear();
+    this.byAlias.clear();
+    for (const [id, model] of next.byId) this.byId.set(id, model);
+    for (const [alias, id] of next.byAlias) this.byAlias.set(alias, id);
+  }
 };
 
 // src/catalog/image.ts
-var IMAGE_USAGE = { unit: "images", modality: "IMAGE" };
-var IMAGE_GENERATE_EDIT = { imageGenerate: true, imageEdit: true };
+var IMAGE_USAGE = { unit: "images" };
 var GEMINI_RATIOS = [
   "1:1",
   "2:3",
@@ -129,8 +136,7 @@ function geminiImage(id, label, aliases, sizes, display, exposure) {
     defaults: { size: "1K", aspectRatio: "1:1" },
     usage: { ...IMAGE_USAGE, slug: "gemini" },
     pricingKind: "image_size",
-    display,
-    capabilities: { ...IMAGE_GENERATE_EDIT }
+    display
   };
   return model;
 }
@@ -217,8 +223,7 @@ var IMAGE_MODELS = [
       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.
@@ -252,8 +257,7 @@ var IMAGE_MODELS = [
     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.
@@ -287,8 +291,7 @@ var IMAGE_MODELS = [
     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).
@@ -326,8 +329,7 @@ var IMAGE_MODELS = [
       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",
@@ -353,8 +355,7 @@ var IMAGE_MODELS = [
       provider: "bytedance",
       logo: "https://d31vvfpk2ppdyq.cloudfront.net/assets/image/3bba5d81-4e57-4e71-911b-cdfa05acca2c_compressed.png",
       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",
@@ -391,8 +392,7 @@ var IMAGE_MODELS = [
     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",
@@ -414,9 +414,9 @@ var IMAGE_MODELS = [
     pricingKind: "image_size",
     display: {
       provider: "black-forest-labs",
+      logo: "https://cdn.jsdelivr.net/npm/@lobehub/icons-static-png@1.97.0/light/flux.png",
       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
@@ -442,8 +442,7 @@ var IMAGE_MODELS = [
     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,
@@ -468,8 +467,7 @@ var IMAGE_MODELS = [
     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
@@ -490,13 +488,12 @@ var IMAGE_MODELS = [
       supports: []
     },
     defaults: { size: "1K", aspectRatio: "1:1" },
-    usage: { unit: "compute_seconds", modality: "IMAGE" },
+    usage: { unit: "compute_seconds" },
     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
@@ -518,13 +515,12 @@ var IMAGE_MODELS = [
       supports: []
     },
     defaults: { aspectRatio: "1:1" },
-    usage: { unit: "compute_seconds", modality: "IMAGE" },
+    usage: { unit: "compute_seconds" },
     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
@@ -546,13 +542,12 @@ var IMAGE_MODELS = [
       supports: ["outputFormat"]
     },
     defaults: { outputFormat: "png" },
-    usage: { unit: "compute_seconds", modality: "IMAGE" },
+    usage: { unit: "compute_seconds" },
     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 }
+    }
   }
 ];
 
@@ -574,7 +569,7 @@ var OTHER_MODELS = [
       supports: []
     },
     defaults: {},
-    usage: { unit: "images", modality: "AUDIO" },
+    usage: { unit: "images" },
     pricingKind: "flat",
     display: {
       provider: "google",
@@ -597,7 +592,7 @@ var OTHER_MODELS = [
       supports: []
     },
     defaults: {},
-    usage: { unit: "images", modality: "AUDIO" },
+    usage: { unit: "images" },
     pricingKind: "flat",
     display: {
       provider: "google",
@@ -620,7 +615,7 @@ var OTHER_MODELS = [
       supports: ["seed", "negativePrompt"]
     },
     defaults: {},
-    usage: { unit: "compute_seconds", modality: "AUDIO" },
+    usage: { unit: "compute_seconds" },
     pricingKind: "flat",
     display: {
       provider: "google",
@@ -643,7 +638,7 @@ var OTHER_MODELS = [
       supports: ["seed", "negativePrompt"]
     },
     defaults: { durationSeconds: 8 },
-    usage: { unit: "compute_seconds", modality: "AUDIO" },
+    usage: { unit: "compute_seconds" },
     pricingKind: "flat",
     display: {
       provider: "mmaudio",
@@ -666,7 +661,7 @@ var OTHER_MODELS = [
     },
     defaults: { durationSeconds: 15, instrumental: true },
     // Replicate's run() does not report predict_time; billed by generated length.
-    usage: { unit: "seconds", modality: "AUDIO" },
+    usage: { unit: "seconds" },
     pricingKind: "flat",
     display: {
       provider: "elevenlabs",
@@ -689,13 +684,13 @@ var OTHER_MODELS = [
       supports: ["scaleFactor", "outputFormat", "upscale"]
     },
     defaults: { scaleFactor: 2, outputFormat: "png" },
-    usage: { unit: "compute_seconds", modality: "IMAGE" },
+    usage: { unit: "compute_seconds" },
     pricingKind: "flat",
     display: {
       provider: "topaz",
+      logo: "https://cdn.jsdelivr.net/npm/@lobehub/icons-static-png@1.97.0/light/topazlabs.png",
       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",
@@ -712,19 +707,30 @@ var OTHER_MODELS = [
       supports: []
     },
     defaults: {},
-    usage: { unit: "images", modality: "IMAGE" },
+    usage: { unit: "images" },
     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" };
+var TEXT_USAGE = { unit: "tokens" };
+var CDN = "https://d31vvfpk2ppdyq.cloudfront.net/assets/image";
+var PROVIDER_LOGOS = {
+  anthropic: `${CDN}/25fad392-188c-4829-a3d0-c4fa02d1c8cd.png`,
+  openai: `${CDN}/1204b494-8962-4f7f-af5e-699a8bdc98e3.png`,
+  google: `${CDN}/192485d2-a76a-4657-9c06-930b2c70badd.png`,
+  deepseek: `${CDN}/9b7406d2-a696-4558-ab79-90a37e3f4574.png`,
+  xai: `${CDN}/858b7fcc-9bd5-411d-aea1-a7b35ecfffab.png`,
+  moonshot: `${CDN}/5a1e2eae-3474-4fac-9c74-a93f41ee6f67.png`,
+  minimax: `${CDN}/f4c2e906-28b7-4a4a-b491-df8fef7808ed.png`
+};
 function text(id, label, providerModelId, extra) {
+  const display = extra?.display;
+  const logo = display && !display.logo ? PROVIDER_LOGOS[display.provider] : void 0;
   return {
     id,
     aliases: [],
@@ -738,7 +744,8 @@ function text(id, label, providerModelId, extra) {
     defaults: {},
     usage: { ...TEXT_USAGE },
     pricingKind: "token",
-    ...extra
+    ...extra,
+    ...display && logo ? { display: { ...display, logo } } : {}
   };
 }
 var ANTHROPIC_THINKING = {
@@ -757,7 +764,7 @@ var TEXT_MODELS = [
       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 }
+    capabilities: { vision: true, contextWindow: 128e3 }
   }),
   text("gpt-4o-mini", "GPT-4o mini", "gpt-4o-mini", {
     exposure: "internal",
@@ -766,7 +773,7 @@ var TEXT_MODELS = [
       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 }
+    capabilities: { vision: true, contextWindow: 128e3 }
   }),
   text("gpt-5.1", "GPT-5.1", "gpt-5.1", {
     exposure: "catalog",
@@ -775,7 +782,7 @@ var TEXT_MODELS = [
       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 }
+    capabilities: { vision: true, contextWindow: 4e5 }
   }),
   text("gpt-5.2", "GPT-5.2", "gpt-5.2", {
     exposure: "catalog",
@@ -784,7 +791,7 @@ var TEXT_MODELS = [
       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 }
+    capabilities: { vision: 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", {
@@ -796,8 +803,6 @@ var TEXT_MODELS = [
     },
     capabilities: {
       vision: true,
-      tools: true,
-      structuredOutput: true,
       contextWindow: 105e4
     }
   }),
@@ -810,7 +815,7 @@ var TEXT_MODELS = [
       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 }
+    capabilities: { vision: true, contextWindow: 4e5 }
   }),
   text("gpt-5.5", "GPT-5.5", "gpt-5.5", {
     exposure: "catalog",
@@ -821,8 +826,6 @@ var TEXT_MODELS = [
     },
     capabilities: {
       vision: true,
-      tools: true,
-      structuredOutput: true,
       contextWindow: 105e4
     }
   }),
@@ -837,8 +840,6 @@ var TEXT_MODELS = [
     capabilities: {
       reasoning: true,
       vision: true,
-      tools: true,
-      structuredOutput: true,
       contextWindow: 105e4
     }
   }),
@@ -853,8 +854,6 @@ var TEXT_MODELS = [
     capabilities: {
       reasoning: true,
       vision: true,
-      tools: true,
-      structuredOutput: true,
       contextWindow: 105e4
     }
   }),
@@ -869,8 +868,6 @@ var TEXT_MODELS = [
     capabilities: {
       reasoning: true,
       vision: true,
-      tools: true,
-      structuredOutput: true,
       contextWindow: 105e4
     }
   }),
@@ -883,7 +880,7 @@ var TEXT_MODELS = [
       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 }
+    capabilities: { vision: 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", {
@@ -903,8 +900,6 @@ var TEXT_MODELS = [
     capabilities: {
       reasoning: true,
       vision: true,
-      tools: true,
-      structuredOutput: true,
       contextWindow: 1e6
     }
   }),
@@ -919,8 +914,6 @@ var TEXT_MODELS = [
     capabilities: {
       reasoning: true,
       vision: true,
-      tools: true,
-      structuredOutput: true,
       contextWindow: 1e6
     }
   }),
@@ -936,8 +929,6 @@ var TEXT_MODELS = [
     capabilities: {
       reasoning: true,
       vision: true,
-      tools: true,
-      structuredOutput: true,
       contextWindow: 1e6
     }
   }),
@@ -953,8 +944,6 @@ var TEXT_MODELS = [
     capabilities: {
       reasoning: true,
       vision: true,
-      tools: true,
-      structuredOutput: true,
       contextWindow: 2e5
     }
   }),
@@ -968,8 +957,6 @@ var TEXT_MODELS = [
     },
     capabilities: {
       vision: true,
-      tools: true,
-      structuredOutput: true,
       contextWindow: 1e6
     }
   }),
@@ -982,8 +969,6 @@ var TEXT_MODELS = [
     },
     capabilities: {
       vision: true,
-      tools: true,
-      structuredOutput: true,
       contextWindow: 1e6
     }
   }),
@@ -996,8 +981,6 @@ var TEXT_MODELS = [
     },
     capabilities: {
       vision: true,
-      tools: true,
-      structuredOutput: true,
       contextWindow: 1048576
     }
   }),
@@ -1010,8 +993,6 @@ var TEXT_MODELS = [
     },
     capabilities: {
       vision: true,
-      tools: true,
-      structuredOutput: true,
       contextWindow: 1048576
     }
   }),
@@ -1024,8 +1005,6 @@ var TEXT_MODELS = [
     },
     capabilities: {
       vision: true,
-      tools: true,
-      structuredOutput: true,
       contextWindow: 1048576
     }
   }),
@@ -1038,8 +1017,6 @@ var TEXT_MODELS = [
     },
     capabilities: {
       vision: true,
-      tools: true,
-      structuredOutput: true,
       contextWindow: 1048576
     }
   }),
@@ -1052,8 +1029,6 @@ var TEXT_MODELS = [
     },
     capabilities: {
       vision: true,
-      tools: true,
-      structuredOutput: true,
       contextWindow: 1048576
     }
   }),
@@ -1067,8 +1042,6 @@ var TEXT_MODELS = [
     },
     capabilities: {
       vision: true,
-      tools: true,
-      structuredOutput: true,
       contextWindow: 1048576
     }
   }),
@@ -1081,8 +1054,6 @@ var TEXT_MODELS = [
     },
     capabilities: {
       vision: true,
-      tools: true,
-      structuredOutput: true,
       contextWindow: 1048576
     }
   }),
@@ -1101,8 +1072,6 @@ var TEXT_MODELS = [
     },
     capabilities: {
       vision: true,
-      tools: true,
-      structuredOutput: true,
       contextWindow: 1048576
     }
   }),
@@ -1115,8 +1084,6 @@ var TEXT_MODELS = [
     },
     capabilities: {
       vision: true,
-      tools: false,
-      structuredOutput: true,
       contextWindow: 128e3
     }
   }),
@@ -1129,8 +1096,6 @@ var TEXT_MODELS = [
     },
     capabilities: {
       vision: true,
-      tools: false,
-      structuredOutput: true,
       contextWindow: 2e5
     }
   }),
@@ -1143,8 +1108,6 @@ var TEXT_MODELS = [
     },
     capabilities: {
       vision: true,
-      tools: false,
-      structuredOutput: true,
       contextWindow: 128e3
     }
   }),
@@ -1164,8 +1127,6 @@ var TEXT_MODELS = [
     capabilities: {
       reasoning: true,
       vision: true,
-      tools: true,
-      structuredOutput: true,
       contextWindow: 1e6
     }
   }),
@@ -1180,8 +1141,6 @@ var TEXT_MODELS = [
     capabilities: {
       reasoning: true,
       vision: false,
-      tools: true,
-      structuredOutput: true,
       contextWindow: 1e6
     }
   }),
@@ -1195,14 +1154,14 @@ var TEXT_MODELS = [
     },
     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", {
-    exposure: "catalog",
+    lifecycle: "deprecated",
+    replacedBy: "kimi-k3",
+    exposure: "internal",
     display: {
       chat: { temperature: 1 },
       provider: "moonshot",
@@ -1222,8 +1181,6 @@ var TEXT_MODELS = [
     capabilities: {
       reasoning: true,
       vision: true,
-      tools: true,
-      structuredOutput: true,
       contextWindow: 1e6
     }
   }),
@@ -1237,12 +1194,12 @@ var TEXT_MODELS = [
       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 }
+    capabilities: { vision: true, contextWindow: 1e6 }
   })
 ];
 
 // src/catalog/video.ts
-var VIDEO_USAGE = { unit: "seconds", modality: "VIDEO" };
+var VIDEO_USAGE = { unit: "seconds" };
 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"];
@@ -1416,7 +1373,7 @@ var VIDEO_MODELS = [
     constraints: {
       aspectRatios: [...SEEDANCE_RATIOS],
       resolutions: ["480p", "720p", "1080p", "4k"],
-      durations: [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
+      durations: { min: 4, max: 15 },
       durationPresets: [4, 6, 8, 10, 12, 15],
       references: { images: 9, videos: 3, audios: 3, audioOnly: false },
       ratioIgnoredWith: ["firstFrame"],
@@ -1449,7 +1406,7 @@ var VIDEO_MODELS = [
     constraints: {
       aspectRatios: [...SEEDANCE_RATIOS],
       resolutions: ["480p", "720p"],
-      durations: [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
+      durations: { min: 4, max: 15 },
       references: { images: 9, videos: 3, audios: 3, audioOnly: false },
       ratioIgnoredWith: ["firstFrame"],
       supports: ["generateAudio", "seed", "lastFrame", "watermark"]
@@ -1522,7 +1479,7 @@ var VIDEO_MODELS = [
     },
     // No resolution default: xAI picks its own unless the caller asks.
     defaults: { durationSeconds: 8, aspectRatio: "16:9" },
-    usage: { unit: "reported_cost_usd", modality: "VIDEO", slug: "grok" },
+    usage: { unit: "reported_cost_usd", slug: "grok" },
     pricingKind: "video_per_second",
     display: {
       provider: "xai",
@@ -1556,6 +1513,7 @@ var VIDEO_MODELS = [
     pricingKind: "video_per_second",
     display: {
       provider: "kuaishou",
+      logo: "https://cdn.jsdelivr.net/npm/@lobehub/icons-static-png@1.97.0/light/kling-color.png",
       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 }
@@ -1586,6 +1544,7 @@ var VIDEO_MODELS = [
     pricingKind: "video_per_second",
     display: {
       provider: "kuaishou",
+      logo: "https://cdn.jsdelivr.net/npm/@lobehub/icons-static-png@1.97.0/light/kling-color.png",
       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 }
@@ -1619,10 +1578,11 @@ var VIDEO_MODELS = [
       supports: ["generateAudio", "lastFrame"]
     },
     defaults: { durationSeconds: 5, aspectRatio: "16:9", generateAudio: false, mode: "pro" },
-    usage: { unit: "compute_seconds", modality: "VIDEO" },
+    usage: { unit: "compute_seconds" },
     pricingKind: "video_per_second",
     display: {
       provider: "kuaishou",
+      logo: "https://cdn.jsdelivr.net/npm/@lobehub/icons-static-png@1.97.0/light/kling-color.png",
       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: {}
@@ -1652,6 +1612,7 @@ var VIDEO_MODELS = [
     pricingKind: "video_per_second",
     display: {
       provider: "minimax",
+      logo: "https://d31vvfpk2ppdyq.cloudfront.net/assets/image/f4c2e906-28b7-4a4a-b491-df8fef7808ed.png",
       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 }
@@ -1681,6 +1642,7 @@ var VIDEO_MODELS = [
     pricingKind: "video_per_second",
     display: {
       provider: "minimax",
+      logo: "https://d31vvfpk2ppdyq.cloudfront.net/assets/image/f4c2e906-28b7-4a4a-b491-df8fef7808ed.png",
       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 }
@@ -1711,6 +1673,7 @@ var VIDEO_MODELS = [
     pricingKind: "video_per_second",
     display: {
       provider: "minimax",
+      logo: "https://d31vvfpk2ppdyq.cloudfront.net/assets/image/f4c2e906-28b7-4a4a-b491-df8fef7808ed.png",
       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: {}
@@ -1741,6 +1704,7 @@ var VIDEO_MODELS = [
     pricingKind: "video_per_second",
     display: {
       provider: "minimax",
+      logo: "https://d31vvfpk2ppdyq.cloudfront.net/assets/image/f4c2e906-28b7-4a4a-b491-df8fef7808ed.png",
       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: {}
@@ -1766,6 +1730,7 @@ var VIDEO_MODELS = [
     pricingKind: "video_per_second",
     display: {
       provider: "kuaishou",
+      logo: "https://cdn.jsdelivr.net/npm/@lobehub/icons-static-png@1.97.0/light/kling-color.png",
       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 }
@@ -1785,10 +1750,11 @@ var VIDEO_MODELS = [
       supports: ["lastFrame"]
     },
     defaults: { durationSeconds: 6, resolution: "768p" },
-    usage: { unit: "compute_seconds", modality: "VIDEO" },
+    usage: { unit: "compute_seconds" },
     pricingKind: "video_per_second",
     display: {
       provider: "minimax",
+      logo: "https://d31vvfpk2ppdyq.cloudfront.net/assets/image/f4c2e906-28b7-4a4a-b491-df8fef7808ed.png",
       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: {}
@@ -1825,6 +1791,387 @@ var MODELS = [
   ...OTHER_MODELS
 ];
 
+// src/catalog/schema.ts
+import { z } from "zod";
+var ADAPTER_IDS = [
+  "ai-sdk-text",
+  "google-image",
+  "openai-image",
+  "google-genai",
+  "ark",
+  "fal",
+  "replicate",
+  "prodia",
+  "higgsfield",
+  "xai-video"
+];
+var ModelTypeSchema = z.enum(["text", "image", "video", "audio", "upscale", "bg-remove"]);
+var LifecycleSchema = z.enum(["active", "deprecated", "retired"]);
+var AdapterSchema = z.enum(ADAPTER_IDS);
+var AspectRatioSchema = z.enum([
+  "1:1",
+  "16:9",
+  "9:16",
+  "4:3",
+  "3:4",
+  "3:2",
+  "2:3",
+  "4:5",
+  "5:4",
+  "21:9",
+  "9:21",
+  "97:30",
+  "adaptive"
+]);
+var ImageSizeSchema = z.enum(["512", "1K", "2K", "4K"]);
+var VideoResolutionSchema = z.enum(["360p", "480p", "720p", "768p", "1080p", "2k", "4k"]);
+var ImageQualitySchema = z.enum(["auto", "low", "medium", "high", "xhigh", "max"]);
+var OutputFormatSchema = z.enum(["png", "jpeg", "webp"]);
+var RequiredInputSchema = z.enum(["firstFrame", "sourceVideo", "references"]);
+var SupportedOptionSchema = z.enum([
+  "seed",
+  "quality",
+  "negativePrompt",
+  "cfgScale",
+  "generateAudio",
+  "promptExpansion",
+  "lastFrame",
+  "count",
+  "outputFormat",
+  "scaleFactor",
+  "safetySettings",
+  "temperature",
+  "upscale",
+  "watermark",
+  "safetyChecker",
+  "instrumental"
+]);
+var UsageUnitSchema = z.enum([
+  "tokens",
+  "images",
+  "seconds",
+  "reported_cost_usd",
+  "compute_seconds"
+]);
+var PricingKindSchema = z.enum(["token", "image_size", "video_per_second", "flat"]);
+var ProviderOptionValueSchema = z.lazy(
+  () => z.union([
+    z.string(),
+    z.number(),
+    z.boolean(),
+    z.array(ProviderOptionValueSchema),
+    z.record(z.string(), ProviderOptionValueSchema)
+  ])
+);
+var ProviderOptionsSchema = z.record(z.string(), z.record(z.string(), ProviderOptionValueSchema));
+var ReferenceLimitsSchema = z.looseObject({
+  images: z.number().optional(),
+  videos: z.number().optional(),
+  audios: z.number().optional(),
+  total: z.number().optional(),
+  audioOnly: z.boolean().optional()
+});
+var ModelConstraintsSchema = z.looseObject({
+  aspectRatios: z.array(AspectRatioSchema).optional(),
+  sizes: z.array(ImageSizeSchema).optional(),
+  resolutions: z.array(VideoResolutionSchema).optional(),
+  durations: z.union([z.looseObject({ min: z.number(), max: z.number() }), z.array(z.number())]).optional(),
+  durationPresets: z.array(z.number()).optional(),
+  qualityOptions: z.array(z.string()).optional(),
+  exactSizes: z.record(z.string(), z.string()).optional(),
+  references: ReferenceLimitsSchema.optional(),
+  requires: z.array(RequiredInputSchema).optional(),
+  supports: z.array(SupportedOptionSchema).optional(),
+  ratioIgnoredWith: z.array(z.enum(["firstFrame", "sourceVideo"])).optional()
+});
+var MediaRefSchema = z.union([
+  z.looseObject({ url: z.string(), label: z.string().optional() }),
+  z.looseObject({
+    bytes: z.instanceof(Uint8Array),
+    mime: z.string(),
+    label: z.string().optional()
+  })
+]);
+var ImageSafetySettingSchema = z.looseObject({ category: z.string(), threshold: z.string() });
+var UpscaleOptionsSchema = z.looseObject({
+  enhanceModel: z.string().optional(),
+  subjectDetection: z.string().optional(),
+  faceEnhancement: z.boolean().optional(),
+  faceEnhancementCreativity: z.number().optional(),
+  faceEnhancementStrength: z.number().optional()
+});
+var DefaultsSchema = z.looseObject({
+  references: z.array(MediaRefSchema).optional(),
+  aspectRatio: AspectRatioSchema.optional(),
+  size: ImageSizeSchema.optional(),
+  exactSize: z.string().optional(),
+  quality: ImageQualitySchema.optional(),
+  seed: z.number().optional(),
+  count: z.number().optional(),
+  outputFormat: OutputFormatSchema.optional(),
+  scaleFactor: z.number().optional(),
+  safetySettings: z.array(ImageSafetySettingSchema).optional(),
+  temperature: z.number().optional(),
+  upscale: UpscaleOptionsSchema.optional(),
+  firstFrame: MediaRefSchema.optional(),
+  lastFrame: MediaRefSchema.optional(),
+  sourceVideo: MediaRefSchema.optional(),
+  referenceImages: z.array(MediaRefSchema).optional(),
+  referenceVideos: z.array(MediaRefSchema).optional(),
+  referenceAudios: z.array(MediaRefSchema).optional(),
+  resolution: VideoResolutionSchema.optional(),
+  durationSeconds: z.number().optional(),
+  generateAudio: z.boolean().optional(),
+  negativePrompt: z.string().optional(),
+  cfgScale: z.number().optional(),
+  promptExpansion: z.enum(["fast", "balanced", "quality"]).optional(),
+  watermark: z.boolean().optional(),
+  safetyChecker: z.boolean().optional(),
+  mode: z.enum(["standard", "pro"]).optional(),
+  instrumental: z.boolean().optional()
+});
+var CatalogDisplaySchema = z.looseObject({
+  provider: z.string(),
+  description: z.string(),
+  logo: z.string().optional(),
+  costTier: z.union([z.literal(1), z.literal(2), z.literal(3)]).optional(),
+  recommended: z.boolean().optional(),
+  sortOrder: z.number().optional(),
+  chat: z.looseObject({
+    providerOptions: ProviderOptionsSchema.optional(),
+    temperature: z.number().optional()
+  }).optional()
+});
+var CatalogCapabilitiesSchema = z.looseObject({
+  acceptsSourceVideo: z.boolean().optional(),
+  framesExcludeReferences: z.boolean().optional(),
+  reasoning: z.boolean().optional(),
+  vision: z.boolean().optional(),
+  contextWindow: z.number().optional()
+});
+var CatalogModelSchemaBase = z.looseObject({
+  id: z.string().min(1),
+  aliases: z.array(z.string()).default([]),
+  label: z.string().min(1),
+  type: ModelTypeSchema,
+  adapter: AdapterSchema,
+  providerModelId: z.string().min(1),
+  lifecycle: LifecycleSchema,
+  replacedBy: z.string().optional(),
+  exposure: z.enum(["catalog", "internal"]),
+  constraints: ModelConstraintsSchema,
+  defaults: DefaultsSchema,
+  providerOptions: ProviderOptionsSchema.optional(),
+  usage: z.looseObject({
+    unit: UsageUnitSchema,
+    slug: z.string().optional(),
+    providerLabel: z.string().optional()
+  }),
+  pricingKind: PricingKindSchema,
+  display: CatalogDisplaySchema.optional(),
+  capabilities: CatalogCapabilitiesSchema.optional()
+}).superRefine((row, ctx) => {
+  if (row.lifecycle === "active" && !row.display) {
+    ctx.addIssue({
+      code: "custom",
+      path: ["display"],
+      message: "active rows need display"
+    });
+  }
+  if (row.lifecycle === "active" && row.type === "text" && !row.capabilities?.contextWindow) {
+    ctx.addIssue({
+      code: "custom",
+      path: ["capabilities", "contextWindow"],
+      message: "active text rows need contextWindow"
+    });
+  }
+  if (row.replacedBy !== void 0 && row.replacedBy === row.id) {
+    ctx.addIssue({
+      code: "custom",
+      path: ["replacedBy"],
+      message: "replacedBy must not equal id"
+    });
+  }
+});
+var CatalogModelSchema = CatalogModelSchemaBase;
+var CatalogSchema = z.array(CatalogModelSchemaBase).superRefine((rows, ctx) => {
+  const seenIds = /* @__PURE__ */ new Map();
+  rows.forEach((row, index) => {
+    if (seenIds.has(row.id)) {
+      ctx.addIssue({
+        code: "custom",
+        path: [index, "id"],
+        message: `Duplicate model id "${row.id}"`
+      });
+    } else {
+      seenIds.set(row.id, index);
+    }
+  });
+  const seenAliases = /* @__PURE__ */ new Map();
+  rows.forEach((row, index) => {
+    row.aliases.forEach((alias, aliasIndex) => {
+      const path = [index, "aliases", aliasIndex];
+      if (alias === row.id || seenIds.has(alias)) {
+        ctx.addIssue({
+          code: "custom",
+          path,
+          message: `Alias "${alias}" collides with a model id`
+        });
+        return;
+      }
+      if (seenAliases.has(alias)) {
+        ctx.addIssue({
+          code: "custom",
+          path,
+          message: `Duplicate alias "${alias}"`
+        });
+      } else {
+        seenAliases.set(alias, index);
+      }
+    });
+  });
+});
+
+// src/catalog/live.ts
+var liveCatalog = new Catalog(MODELS);
+
+// src/catalog/load.ts
+var memo;
+var timer;
+var revision;
+var origin = "snapshot";
+function rowCount() {
+  return liveCatalog.list({ lifecycle: ["active", "deprecated", "retired"] }).length;
+}
+function authHeaders(auth) {
+  return "bearer" in auth ? { Authorization: `Bearer ${auth.bearer}` } : { "x-api-key": auth.apiKey };
+}
+function describeError(err) {
+  return err instanceof Error ? err.message : String(err);
+}
+function parseEnvelope(json) {
+  if (typeof json !== "object" || json === null || !("rows" in json)) {
+    throw new Error("invalid catalog response: missing rows");
+  }
+  const rows = CatalogSchema.parse(json.rows);
+  const rev = json.revision;
+  if (typeof rev === "string") return { rows, revision: rev };
+  return { rows };
+}
+function logPrimed(logger, result) {
+  const { origin: loggedOrigin, rowCount: loggedCount, revision: loggedRevision } = result;
+  logger.info("[modelkit] catalog primed", {
+    origin: loggedOrigin,
+    rowCount: loggedCount,
+    revision: loggedRevision
+  });
+}
+async function fetchRemote(source, logger) {
+  const fetchImpl = source.fetch ?? globalThis.fetch;
+  const headers = authHeaders(source.auth);
+  if (revision !== void 0) headers["If-None-Match"] = `"${revision}"`;
+  const res = await fetchImpl(`${source.baseUrl}/models/catalog`, {
+    headers,
+    signal: AbortSignal.timeout(source.timeoutMs ?? 1e4)
+  });
+  if (res.status === 304) return { kept: true };
+  if (!res.ok) throw new Error(`catalog fetch failed with status ${res.status}`);
+  const json = await res.json();
+  const { rows, revision: rev } = parseEnvelope(json);
+  if (rev === void 0) return { rows, fromCache: false };
+  return { rows, revision: rev, fromCache: false };
+}
+async function writeCache(source, logger, rows, rev) {
+  if (!source.cache) return;
+  try {
+    await source.cache.set(JSON.stringify({ rows, revision: rev }));
+  } catch (err) {
+    logger.warn("[modelkit] catalog cache write failed", err);
+  }
+}
+async function prime(source, logger) {
+  try {
+    const fetched = await fetchRemote(source, logger);
+    if ("kept" in fetched) {
+      const result2 = { origin, rowCount: rowCount() };
+      if (revision !== void 0) result2.revision = revision;
+      logPrimed(logger, result2);
+      return result2;
+    }
+    liveCatalog.replace(fetched.rows);
+    origin = "remote";
+    revision = fetched.revision;
+    await writeCache(source, logger, fetched.rows, fetched.revision);
+    const result = { origin, rowCount: rowCount() };
+    if (fetched.revision !== void 0) result.revision = fetched.revision;
+    logPrimed(logger, result);
+    return result;
+  } catch (err) {
+    logger.warn("[modelkit] catalog remote failed, trying cache", err);
+    const error = describeError(err);
+    try {
+      const cached = await source.cache?.get();
+      if (!cached) throw new Error("no cached catalog");
+      const { rows, revision: rev } = parseEnvelope(JSON.parse(cached));
+      liveCatalog.replace(rows);
+      origin = "last-good";
+      revision = rev;
+      const result = { origin, rowCount: rowCount(), error };
+      if (rev !== void 0) result.revision = rev;
+      logPrimed(logger, result);
+      return result;
+    } catch (cacheErr) {
+      logger.warn("[modelkit] catalog cache failed, using snapshot", cacheErr);
+      liveCatalog.replace([...MODELS]);
+      origin = "snapshot";
+      revision = void 0;
+      const result = {
+        origin,
+        rowCount: rowCount(),
+        error
+      };
+      logPrimed(logger, result);
+      return result;
+    }
+  }
+}
+async function refresh(source, logger) {
+  try {
+    const fetched = await fetchRemote(source, logger);
+    if ("kept" in fetched) return;
+    liveCatalog.replace(fetched.rows);
+    origin = "remote";
+    revision = fetched.revision;
+    await writeCache(source, logger, fetched.rows, fetched.revision);
+  } catch (err) {
+    logger.warn("[modelkit] catalog refresh failed, keeping current rows", err);
+  }
+}
+function primeCatalog(source) {
+  if (memo) return memo;
+  const logger = source.logger ?? console;
+  memo = prime(source, logger).then((result) => {
+    const ttlMs = source.ttlMs ?? 3e5;
+    if (ttlMs > 0 && timer === void 0) {
+      timer = setInterval(() => void refresh(source, logger), ttlMs);
+      const maybeUnref = timer;
+      maybeUnref.unref?.();
+    }
+    return result;
+  });
+  return memo;
+}
+function resetLiveCatalog() {
+  if (timer !== void 0) {
+    clearInterval(timer);
+    timer = void 0;
+  }
+  memo = void 0;
+  revision = void 0;
+  origin = "snapshot";
+  liveCatalog.replace([...MODELS]);
+}
+
 // src/resolve-options.ts
 var SIZE_ORDER = ["512", "1K", "2K", "4K"];
 var RESOLUTION_ORDER = ["360p", "480p", "720p", "768p", "1080p", "2k", "4k"];
@@ -2207,7 +2554,10 @@ function resolveOptions(model, request, onUnsupported = "error") {
   }
 }
 export {
+  ADAPTER_IDS,
   Catalog,
+  CatalogModelSchema,
+  CatalogSchema,
   IMAGE_MODELS,
   MODELS,
   ModelKitError,
@@ -2215,7 +2565,10 @@ export {
   TEXT_MODELS,
   VIDEO_MODELS,
   isModelKitError,
+  liveCatalog,
   nearestAspectRatio,
+  primeCatalog,
+  resetLiveCatalog,
   resolveAudioOptions,
   resolveImageOptions,
   resolveOptions,