@shop-os/modelkit 0.1.1 → 0.1.3

package/dist/index.cjs

@@ -292,7 +292,10 @@ var IMAGE_MODELS = [
     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 }
+      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 },
@@ -307,21 +310,12 @@ var IMAGE_MODELS = [
     providerModelId: "seedream-4-5-251128",
     lifecycle: "active",
     constraints: {
-      aspectRatios: [
-        "1:1",
-        "2:3",
-        "3:2",
-        "4:3",
-        "3:4",
-        "4:5",
-        "5:4",
-        "9:16",
-        "16:9",
-        "21:9",
-        "97:30"
-      ],
+      // 97:30 has no Ark size spelling; the A+ flow pads inputs to 97:30 and
+      // omits the ratio instead. Advertise only sizes the adapter can send.
+      aspectRatios: ["1:1", "2:3", "3:2", "4:3", "3:4", "4:5", "5:4", "9:16", "16:9", "21:9"],
       sizes: ["1K", "2K"],
-      references: { images: 10 }
+      references: { images: 10 },
+      supports: ["seed"]
     },
     defaults: { size: "1K", aspectRatio: "1:1" },
     usage: { ...IMAGE_USAGE },
@@ -345,6 +339,9 @@ var IMAGE_MODELS = [
     pricingKind: "image_size"
   },
   {
+    // 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",
@@ -353,19 +350,7 @@ var IMAGE_MODELS = [
     providerModelId: "inference.gemini-3-pro.img2img.v1",
     lifecycle: "active",
     constraints: {
-      aspectRatios: [
-        "1:1",
-        "2:3",
-        "3:2",
-        "4:3",
-        "3:4",
-        "4:5",
-        "5:4",
-        "9:16",
-        "16:9",
-        "21:9",
-        "97:30"
-      ],
+      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"]
     },
@@ -374,6 +359,8 @@ var IMAGE_MODELS = [
     pricingKind: "image_size"
   },
   {
+    // 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",
@@ -382,19 +369,7 @@ var IMAGE_MODELS = [
     providerModelId: "inference.nano-banana.img2img.v2",
     lifecycle: "active",
     constraints: {
-      aspectRatios: [
-        "1:1",
-        "2:3",
-        "3:2",
-        "4:3",
-        "3:4",
-        "4:5",
-        "5:4",
-        "9:16",
-        "16:9",
-        "21:9",
-        "97:30"
-      ],
+      aspectRatios: ["1:1", "2:3", "3:4", "4:3", "9:16", "16:9", "21:9"],
       sizes: ["1K"],
       requires: ["references"]
     },
@@ -403,6 +378,9 @@ var IMAGE_MODELS = [
     pricingKind: "image_size"
   },
   {
+    // 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",
@@ -411,19 +389,7 @@ var IMAGE_MODELS = [
     providerModelId: "wan-video/wan-2.7-image-pro",
     lifecycle: "active",
     constraints: {
-      aspectRatios: [
-        "1:1",
-        "2:3",
-        "3:2",
-        "4:3",
-        "3:4",
-        "4:5",
-        "5:4",
-        "9:16",
-        "16:9",
-        "21:9",
-        "97:30"
-      ],
+      aspectRatios: ["1:1", "16:9", "9:16", "4:3", "3:4"],
       sizes: ["1K", "2K", "4K"],
       references: { images: 9 }
     },
@@ -432,6 +398,9 @@ var IMAGE_MODELS = [
     pricingKind: "image_size"
   },
   {
+    // 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",
@@ -440,28 +409,19 @@ var IMAGE_MODELS = [
     providerModelId: "qwen/qwen-image-2-pro",
     lifecycle: "active",
     constraints: {
-      aspectRatios: [
-        "1:1",
-        "2:3",
-        "3:2",
-        "4:3",
-        "3:4",
-        "4:5",
-        "5:4",
-        "9:16",
-        "16:9",
-        "21:9",
-        "97:30"
-      ],
-      sizes: ["1K", "2K", "4K"],
+      aspectRatios: ["1:1", "16:9", "9:16", "4:3", "3:4", "3:2", "2:3"],
+      sizes: [],
       references: { images: 1 },
       requires: ["references"]
     },
-    defaults: { size: "1K", aspectRatio: "1:1" },
+    defaults: { aspectRatio: "1:1" },
     usage: { unit: "compute_seconds", modality: "IMAGE" },
     pricingKind: "image_size"
   },
   {
+    // 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",
@@ -470,25 +430,13 @@ var IMAGE_MODELS = [
     providerModelId: "qwen-edit-apps/qwen-image-edit-plus-lora-skin",
     lifecycle: "active",
     constraints: {
-      aspectRatios: [
-        "1:1",
-        "2:3",
-        "3:2",
-        "4:3",
-        "3:4",
-        "4:5",
-        "5:4",
-        "9:16",
-        "16:9",
-        "21:9",
-        "97:30"
-      ],
-      sizes: ["1K", "2K", "4K"],
+      aspectRatios: ["1:1", "16:9", "9:16", "4:3", "3:4"],
+      sizes: [],
       references: { images: 1 },
       requires: ["references"],
       supports: ["outputFormat"]
     },
-    defaults: { size: "1K", aspectRatio: "1:1", outputFormat: "png" },
+    defaults: { outputFormat: "png" },
     usage: { unit: "compute_seconds", modality: "IMAGE" },
     pricingKind: "image_size"
   }
@@ -826,7 +774,7 @@ var VIDEO_MODELS = [
       durations: [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
       references: { images: 9, videos: 3, audios: 3, audioOnly: false },
       ratioIgnoredWith: ["firstFrame"],
-      supports: ["generateAudio", "seed"]
+      supports: ["generateAudio", "seed", "lastFrame"]
     },
     defaults: {
       durationSeconds: 5,
@@ -851,7 +799,7 @@ var VIDEO_MODELS = [
       durations: [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
       references: { images: 9, videos: 3, audios: 3, audioOnly: false },
       ratioIgnoredWith: ["firstFrame"],
-      supports: ["generateAudio", "seed"]
+      supports: ["generateAudio", "seed", "lastFrame"]
     },
     defaults: {
       durationSeconds: 5,
@@ -877,7 +825,7 @@ var VIDEO_MODELS = [
       durations: { min: 4, max: 30 },
       references: { images: 30, videos: 10, audios: 10 },
       ratioIgnoredWith: ["firstFrame"],
-      supports: ["generateAudio", "seed"]
+      supports: ["generateAudio", "seed", "lastFrame"]
     },
     defaults: {
       durationSeconds: 5,
@@ -979,7 +927,7 @@ var VIDEO_MODELS = [
       resolutions: ["720p", "1080p"],
       durations: { min: 3, max: 15 },
       references: { images: 7 },
-      supports: ["generateAudio"]
+      supports: ["generateAudio", "lastFrame"]
     },
     defaults: { durationSeconds: 5, aspectRatio: "16:9", generateAudio: false, mode: "pro" },
     usage: { unit: "compute_seconds", modality: "VIDEO" },
@@ -1898,7 +1846,11 @@ var aiSdkTextAdapter = {
 function requiredKeys(model, config) {
   if (model.adapter === "fal" && config.routing?.fal === "cloudflare-passthrough")
     return ["cloudflare"];
-  if (model.adapter === "ai-sdk-text") return [textKeyName(model, config.routing ?? {})];
+  if (model.adapter === "ai-sdk-text") {
+    const names = [textKeyName(model, config.routing ?? {})];
+    if (config.routing?.text === "cloudflare") names.push("cloudflare");
+    return names;
+  }
   return ADAPTER_KEYS[model.adapter];
 }
 var IMAGE_OUTPUT_TYPES = /* @__PURE__ */ new Set(["image", "upscale", "bg-remove"]);
@@ -1956,6 +1908,20 @@ function sdkAspectRatio(ratio) {
   const parsed = sdkRatio.safeParse(ratio);
   return parsed.success ? parsed.data : void 0;
 }
+function sdkDataContent(ref) {
+  return "url" in ref ? ref.url : ref.bytes;
+}
+function sdkReference(ref) {
+  return "url" in ref ? ref.url : { data: ref.bytes, mediaType: ref.mime };
+}
+function pushIgnored(warnings, field, from, modelId) {
+  warnings.push({
+    code: "option_ignored",
+    field,
+    from,
+    message: `${modelId} provider instance does not support ${field}; ignored`
+  });
+}
 function createModelKit(config, adapters) {
   const catalog = new Catalog(MODELS).withOverrides(config.catalogOverride ?? []);
   const logger = config.logger ?? console;
@@ -1978,6 +1944,7 @@ function createModelKit(config, adapters) {
       call,
       warnings
     };
+    if (config.endpoints) ctx.endpoints = config.endpoints;
     if (call.signal) ctx.signal = call.signal;
     return { adapter, ctx };
   }
@@ -2017,16 +1984,42 @@ function createModelKit(config, adapters) {
     };
   }
   async function passthroughImage(req, call) {
-    const { model, prompt, n, count, size, aspectRatio, seed, providerOptions } = req;
+    const {
+      model,
+      prompt,
+      n,
+      count,
+      size,
+      aspectRatio,
+      seed,
+      providerOptions,
+      references,
+      exactSize: exactSizeField,
+      quality: _quality,
+      outputFormat: _outputFormat
+    } = req;
     const options = { model, prompt };
+    const warnings = [];
     const quantity = n ?? count;
     if (quantity !== void 0) options.n = quantity;
-    const exact = exactSize.safeParse(size);
-    if (exact.success) options.size = exact.data;
+    if (references !== void 0 && references.length > 0)
+      options.prompt = { images: references.map(sdkDataContent), text: prompt };
+    const exact = exactSizeField === void 0 ? void 0 : exactSize.safeParse(exactSizeField);
+    if (exact?.success) options.size = exact.data;
+    else if (exactSizeField !== void 0)
+      pushIgnored(warnings, "exactSize", exactSizeField, model.modelId);
+    const requested = size === void 0 ? void 0 : exactSize.safeParse(size);
+    if (requested?.success) options.size ??= requested.data;
+    else if (size !== void 0) pushIgnored(warnings, "size", size, model.modelId);
     const ratio = sdkAspectRatio(aspectRatio);
     if (ratio) options.aspectRatio = ratio;
+    else if (aspectRatio !== void 0)
+      pushIgnored(warnings, "aspectRatio", aspectRatio, model.modelId);
     if (seed !== void 0) options.seed = seed;
     if (providerOptions) options.providerOptions = providerOptions;
+    if (_quality !== void 0) pushIgnored(warnings, "quality", _quality, model.modelId);
+    if (_outputFormat !== void 0)
+      pushIgnored(warnings, "outputFormat", _outputFormat, model.modelId);
     const abortSignal = combineSignals(req.abortSignal, call.signal);
     if (abortSignal) options.abortSignal = abortSignal;
     const result = await (0, import_ai2.generateImage)(options);
@@ -2041,17 +2034,73 @@ function createModelKit(config, adapters) {
       quantity: images.length
     };
     await emit(usage2);
-    return { image: images[0], images, usage: usage2, raw: result, warnings: [] };
+    return { image: images[0], images, usage: usage2, raw: result, warnings };
   }
   function passthroughVideo(req, call) {
-    const { model, prompt, aspectRatio, duration, durationSeconds, seed, providerOptions } = req;
+    const {
+      model,
+      prompt,
+      aspectRatio,
+      duration,
+      durationSeconds,
+      seed,
+      providerOptions,
+      firstFrame,
+      lastFrame,
+      sourceVideo,
+      referenceImages,
+      referenceVideos,
+      referenceAudios,
+      resolution,
+      generateAudio: generateAudio2,
+      negativePrompt: _negativePrompt,
+      cfgScale: _cfgScale,
+      promptExpansion: _promptExpansion,
+      mode: _mode
+    } = req;
     const seconds = duration ?? durationSeconds;
     const abortSignal = combineSignals(req.abortSignal, call.signal);
     const options = { model, prompt };
-    const ratio = sdkAspectRatio(aspectRatio);
+    const warnings = [];
+    const ratio = sdkAspectRatio(aspectRatio) ?? (aspectRatio === "adaptive" ? "adaptive" : void 0);
     if (ratio) options.aspectRatio = ratio;
+    else if (aspectRatio !== void 0)
+      pushIgnored(warnings, "aspectRatio", aspectRatio, model.modelId);
     if (seconds !== void 0) options.duration = seconds;
     if (seed !== void 0) options.seed = seed;
+    if (generateAudio2 !== void 0) options.generateAudio = generateAudio2;
+    const frames = [
+      ...firstFrame ? [{ image: sdkDataContent(firstFrame), frameType: "first_frame" }] : [],
+      ...lastFrame ? [{ image: sdkDataContent(lastFrame), frameType: "last_frame" }] : []
+    ];
+    if (frames.length > 0) options.frameImages = frames;
+    const references = [
+      ...(referenceImages ?? []).map(sdkDataContent),
+      ...(referenceVideos ?? []).map(sdkReference)
+    ];
+    if (references.length > 0) {
+      if (frames.length > 0) {
+        warnings.push({
+          code: "option_ignored",
+          field: "inputReferences",
+          from: references.length,
+          message: `${model.modelId} provider instance ignores references when frames are set`
+        });
+      } else {
+        options.inputReferences = references;
+      }
+    }
+    if (sourceVideo) pushIgnored(warnings, "sourceVideo", void 0, model.modelId);
+    if (referenceAudios?.length)
+      pushIgnored(warnings, "referenceAudios", referenceAudios.length, model.modelId);
+    if (resolution !== void 0)
+      pushIgnored(warnings, "resolution", resolution, model.modelId);
+    if (_negativePrompt !== void 0)
+      pushIgnored(warnings, "negativePrompt", void 0, model.modelId);
+    if (_cfgScale !== void 0) pushIgnored(warnings, "cfgScale", _cfgScale, model.modelId);
+    if (_promptExpansion !== void 0)
+      pushIgnored(warnings, "promptExpansion", _promptExpansion, model.modelId);
+    if (_mode !== void 0) pushIgnored(warnings, "mode", _mode, model.modelId);
     if (providerOptions) options.providerOptions = providerOptions;
     if (abortSignal) options.abortSignal = abortSignal;
     const pending = (0, import_ai2.experimental_generateVideo)(options);
@@ -2079,7 +2128,7 @@ function createModelKit(config, adapters) {
     });
     return {
       jobId: `ai-sdk:${model.provider}:${model.modelId}`,
-      warnings: [],
+      warnings,
       async poll() {
         if (settled) return { status: "succeeded", raw: settled.raw };
         if (failed) return { status: "failed" };
@@ -2303,6 +2352,9 @@ var SUBMIT_PATH = "/contents/generations/tasks";
 var SEEDANCE_POLL_INTERVAL_MS = 15e3;
 var RESULT_TIMEOUT_MS = 15 * 60 * 1e3;
 var VIDEO_URL_TTL_MS = 24 * 60 * 60 * 1e3;
+function arkBaseUrl(ctx) {
+  return ctx.endpoints?.ark ?? ARK_BASE_URL;
+}
 var SEEDREAM_SIZES = {
   "1K": {
     "1:1": "1024x1024",
@@ -2313,7 +2365,8 @@ var SEEDREAM_SIZES = {
     "3:2": "1248x832",
     "2:3": "832x1248",
     "21:9": "1568x672",
-    "4:5": "864x1080"
+    "4:5": "864x1080",
+    "5:4": "1080x864"
   },
   "2K": {
     "1:1": "2048x2048",
@@ -2324,7 +2377,8 @@ var SEEDREAM_SIZES = {
     "3:2": "2496x1664",
     "2:3": "1664x2496",
     "21:9": "3136x1344",
-    "4:5": "1728x2160"
+    "4:5": "1728x2160",
+    "5:4": "2160x1728"
   }
 };
 var arkAdapter = {
@@ -2371,9 +2425,11 @@ function seedreamSize(aspect, size) {
   const table = SEEDREAM_SIZES[tier];
   return table[aspect ?? ""] ?? table["1:1"] ?? "2048x2048";
 }
-function outputMime(format) {
-  if (format === "png") return "image/png";
-  if (format === "webp") return "image/webp";
+function sniffedImageMime(bytes) {
+  if (bytes.length >= 4 && bytes[0] === 137 && bytes[1] === 80 && bytes[2] === 78 && bytes[3] === 71)
+    return "image/png";
+  if (bytes.length >= 12 && bytes[0] === 82 && bytes[1] === 73 && bytes[2] === 70 && bytes[3] === 70 && bytes[8] === 87 && bytes[9] === 69 && bytes[10] === 66 && bytes[11] === 80)
+    return "image/webp";
   return "image/jpeg";
 }
 function arkUrl(ref) {
@@ -2415,7 +2471,7 @@ async function generateSeedreamImage(req, ctx) {
   else if (refs.length > 1) body.image = refs;
   const data = await withRetry(async () => {
     const res = await ctx.fetch(
-      `${ARK_BASE_URL}/images/generations`,
+      `${arkBaseUrl(ctx)}/images/generations`,
       fetchInitFor(ctx, {
         method: "POST",
         headers: { Authorization: `Bearer ${key}`, "Content-Type": "application/json" },
@@ -2454,7 +2510,7 @@ async function generateSeedreamImage(req, ctx) {
   };
   if (req.size !== void 0) usage2.imageSize = req.size;
   if (req.quality !== void 0) usage2.imageQuality = req.quality;
-  return { images: [{ bytes, mime: outputMime(req.outputFormat) }], usage: usage2, raw: data };
+  return { images: [{ bytes, mime: sniffedImageMime(bytes) }], usage: usage2, raw: data };
 }
 var SeedanceResponseSchema = import_zod4.z.object({
   id: import_zod4.z.string().optional(),
@@ -2539,9 +2595,10 @@ async function generateSeedanceVideo(req, ctx) {
   };
   if (!hasFrameAnchor && req.aspectRatio !== void 0) body.ratio = req.aspectRatio;
   if (req.generateAudio !== void 0) body.generate_audio = req.generateAudio;
+  if (req.seed !== void 0) body.seed = req.seed;
   const submit = await withRetry(async () => {
     const res = await ctx.fetch(
-      `${ARK_BASE_URL}${SUBMIT_PATH}`,
+      `${arkBaseUrl(ctx)}${SUBMIT_PATH}`,
       fetchInitFor(ctx, {
         method: "POST",
         headers: { Authorization: `Bearer ${key}`, "Content-Type": "application/json" },
@@ -2561,7 +2618,7 @@ async function generateSeedanceVideo(req, ctx) {
   const usage2 = seedanceUsage(ctx, req, duration);
   const fetchStatus = async () => {
     const res = await ctx.fetch(
-      `${ARK_BASE_URL}${SUBMIT_PATH}/${jobId}`,
+      `${arkBaseUrl(ctx)}${SUBMIT_PATH}/${jobId}`,
       fetchInitFor(ctx, {
         headers: { Authorization: `Bearer ${key}` }
       })
@@ -3620,60 +3677,6 @@ async function generateOmniVideo(req, ctx, createGenAIClient) {
     uploadedNames.push(file.fileName);
     return file.fileUri;
   };
-  const parts = [];
-  if (req.sourceVideo) {
-    const media = await resolveMedia(
-      req.sourceVideo,
-      ctx.fetch,
-      ctx.signal,
-      "video/mp4",
-      modelId
-    );
-    parts.push({
-      type: "video",
-      uri: await uploadOne("video", media.bytes, media.mime),
-      mime_type: media.mime
-    });
-  }
-  for (const ref of videos) {
-    const media = await resolveMedia(ref, ctx.fetch, ctx.signal, "video/mp4", modelId);
-    parts.push({
-      type: "video",
-      uri: await uploadOne("video", media.bytes, media.mime),
-      mime_type: media.mime
-    });
-  }
-  if (req.firstFrame) {
-    const media = await resolveMedia(
-      req.firstFrame,
-      ctx.fetch,
-      ctx.signal,
-      "image/png",
-      modelId
-    );
-    if (media.bytes.length > OMNI_INLINE_LIMIT_BYTES) {
-      parts.push({
-        type: "image",
-        uri: await uploadOne("image", media.bytes, media.mime),
-        mime_type: media.mime
-      });
-    } else {
-      parts.push({ type: "image", data: toBase64(media.bytes), mime_type: media.mime });
-    }
-  }
-  for (const ref of images) {
-    const media = await resolveMedia(ref, ctx.fetch, ctx.signal, "image/png", modelId);
-    if (media.bytes.length > OMNI_INLINE_LIMIT_BYTES) {
-      parts.push({
-        type: "image",
-        uri: await uploadOne("image", media.bytes, media.mime),
-        mime_type: media.mime
-      });
-    } else {
-      parts.push({ type: "image", data: toBase64(media.bytes), mime_type: media.mime });
-    }
-  }
-  parts.push({ type: "text", text: req.prompt });
   const cleanupUploads = async () => {
     for (const fileName of uploadedNames) await deleteGoogleFile(fileName, apiKey, ctx.fetch);
   };
@@ -3687,6 +3690,60 @@ async function generateOmniVideo(req, ctx, createGenAIClient) {
   if (ctx.signal !== void 0) createRetryOptions.signal = ctx.signal;
   let interaction;
   try {
+    const parts = [];
+    if (req.sourceVideo) {
+      const media = await resolveMedia(
+        req.sourceVideo,
+        ctx.fetch,
+        ctx.signal,
+        "video/mp4",
+        modelId
+      );
+      parts.push({
+        type: "video",
+        uri: await uploadOne("video", media.bytes, media.mime),
+        mime_type: media.mime
+      });
+    }
+    for (const ref of videos) {
+      const media = await resolveMedia(ref, ctx.fetch, ctx.signal, "video/mp4", modelId);
+      parts.push({
+        type: "video",
+        uri: await uploadOne("video", media.bytes, media.mime),
+        mime_type: media.mime
+      });
+    }
+    if (req.firstFrame) {
+      const media = await resolveMedia(
+        req.firstFrame,
+        ctx.fetch,
+        ctx.signal,
+        "image/png",
+        modelId
+      );
+      if (media.bytes.length > OMNI_INLINE_LIMIT_BYTES) {
+        parts.push({
+          type: "image",
+          uri: await uploadOne("image", media.bytes, media.mime),
+          mime_type: media.mime
+        });
+      } else {
+        parts.push({ type: "image", data: toBase64(media.bytes), mime_type: media.mime });
+      }
+    }
+    for (const ref of images) {
+      const media = await resolveMedia(ref, ctx.fetch, ctx.signal, "image/png", modelId);
+      if (media.bytes.length > OMNI_INLINE_LIMIT_BYTES) {
+        parts.push({
+          type: "image",
+          uri: await uploadOne("image", media.bytes, media.mime),
+          mime_type: media.mime
+        });
+      } else {
+        parts.push({ type: "image", data: toBase64(media.bytes), mime_type: media.mime });
+      }
+    }
+    parts.push({ type: "text", text: req.prompt });
     try {
       interaction = await withRetry(async () => {
         try {
@@ -3937,6 +3994,7 @@ function toProviderError(cause, ctx) {
     });
   }
   const info = describeThrown(cause);
+  if (info.status === void 0) return transportError("google-image", cause, ctx.model.id);
   return classifyHttpError(info.status, info.body, "google-image", ctx.model.id);
 }
 function createGoogleImageAdapter(deps = defaultGoogleImageDeps) {
@@ -4751,7 +4809,8 @@ var RATIO_TO_DIMS = {
   "9:16": { width: 768, height: 1360 },
   "3:4": { width: 768, height: 1024 },
   "2:3": { width: 768, height: 1152 },
-  "4:5": { width: 896, height: 1120 }
+  "4:5": { width: 896, height: 1120 },
+  "5:4": { width: 1120, height: 896 }
 };
 function dimsFor(aspectRatio) {
   if (aspectRatio !== void 0) {
@@ -4760,6 +4819,27 @@ function dimsFor(aspectRatio) {
   }
   return { width: 1024, height: 1024 };
 }
+var GEMINI_JOB_RATIOS = /* @__PURE__ */ new Set([
+  "1:1",
+  "2:3",
+  "3:2",
+  "3:4",
+  "4:3",
+  "4:5",
+  "5:4",
+  "9:16",
+  "16:9",
+  "21:9"
+]);
+var NANO_BANANA_RATIOS = /* @__PURE__ */ new Set([
+  "1:1",
+  "2:3",
+  "3:4",
+  "4:3",
+  "9:16",
+  "16:9",
+  "21:9"
+]);
 async function refToBytes(ctx, ref) {
   if ("bytes" in ref) return { bytes: ref.bytes, mime: ref.mime };
   const res = await ctx.fetch(ref.url);
@@ -4859,7 +4939,7 @@ async function generateProdiaImage(req, ctx) {
     inputs = refs.length > 0 && firstRef !== void 0 ? [firstRef] : [];
     accept = req.outputFormat === "png" ? "image/png" : "image/jpeg";
     mime = accept;
-  } else if (providerModelId === "flux-2-pro" || providerModelId === FLUX_2_PRO_JOB || providerModelId === "inference.gemini-3-pro.img2img.v1" || providerModelId === "inference.nano-banana.img2img.v2") {
+  } else if (providerModelId === "flux-2-pro" || providerModelId === FLUX_2_PRO_JOB) {
     const dims = dimsFor(req.aspectRatio);
     const config = {
       prompt: req.prompt,
@@ -4868,8 +4948,30 @@ async function generateProdiaImage(req, ctx) {
       safety_tolerance: 5
     };
     if (req.seed !== void 0) config.seed = req.seed;
-    const jobType = providerModelId === "flux-2-pro" ? FLUX_2_PRO_JOB : providerModelId;
-    job = { type: jobType, config };
+    job = { type: FLUX_2_PRO_JOB, config };
+    inputs = refs;
+    accept = "image/png";
+    mime = "image/png";
+  } else if (providerModelId === "inference.gemini-3-pro.img2img.v1" || providerModelId === "inference.nano-banana.img2img.v2") {
+    const allowed = providerModelId === "inference.nano-banana.img2img.v2" ? NANO_BANANA_RATIOS : GEMINI_JOB_RATIOS;
+    const config = {
+      prompt: req.prompt,
+      safety_tolerance: 5
+    };
+    if (req.aspectRatio !== void 0 && req.aspectRatio !== "adaptive") {
+      if (allowed.has(req.aspectRatio)) config.aspect_ratio = req.aspectRatio;
+      else
+        ctx.warnings.push({
+          code: "option_ignored",
+          field: "aspectRatio",
+          from: req.aspectRatio,
+          message: `${providerModelId} has no ${req.aspectRatio} aspect_ratio; using the provider default`
+        });
+    }
+    if (providerModelId === "inference.gemini-3-pro.img2img.v1")
+      config.resolution = req.size ?? "1K";
+    if (req.seed !== void 0) config.seed = req.seed;
+    job = { type: providerModelId, config };
     inputs = refs;
     accept = "image/png";
     mime = "image/png";
@@ -5143,6 +5245,50 @@ function isUpscaler(providerModelId) {
 function isQwen(providerModelId) {
   return isLower(providerModelId).includes("qwen");
 }
+function isWanImage(providerModelId) {
+  return isLower(providerModelId).includes("wan-2.7");
+}
+function isQwenEditPlus(providerModelId) {
+  return isLower(providerModelId).includes("edit");
+}
+var QWEN_IMAGE_RATIOS = /* @__PURE__ */ new Set([
+  "1:1",
+  "16:9",
+  "9:16",
+  "4:3",
+  "3:4",
+  "3:2",
+  "2:3"
+]);
+var QWEN_EDIT_RATIOS = /* @__PURE__ */ new Set(["1:1", "16:9", "9:16", "4:3", "3:4"]);
+var WAN_IMAGE_SIZES = {
+  "1K": {
+    "1:1": "1024*1024",
+    "16:9": "1280*720",
+    "9:16": "720*1280",
+    "4:3": "1024*768",
+    "3:4": "768*1024"
+  },
+  "2K": {
+    "1:1": "2048*2048",
+    "16:9": "2048*1152",
+    "9:16": "1152*2048",
+    "4:3": "2048*1536",
+    "3:4": "1536*2048"
+  },
+  "4K": {
+    "1:1": "4096*4096",
+    "16:9": "4096*2304",
+    "9:16": "2304*4096",
+    "4:3": "4096*3072",
+    "3:4": "3072*4096"
+  }
+};
+function wanImageSize(aspectRatio, size) {
+  const tier = size === "2K" || size === "4K" ? size : "1K";
+  const table = WAN_IMAGE_SIZES[tier];
+  return table[aspectRatio ?? ""] ?? table["1:1"] ?? "1024*1024";
+}
 async function generateReplicateImage(req, ctx, deps) {
   const id = ctx.model.providerModelId;
   const input = { prompt: req.prompt };
@@ -5162,14 +5308,31 @@ async function generateReplicateImage(req, ctx, deps) {
     input.creativity = 0;
     input.scale_factor = 2;
     input.output_format = "png";
+  } else if (isWanImage(id)) {
+    const refs = (req.references ?? []).map(refToUrl3);
+    if (refs.length > 0) input.images = refs;
+    input.size = wanImageSize(req.aspectRatio, req.size);
   } else if (isQwen(id)) {
     const refs = (req.references ?? []).map(refToUrl3);
     const firstUrl = refs[0];
     if (firstUrl !== void 0) {
       input.image = firstUrl;
-      input.match_input_image = true;
+      if (!isQwenEditPlus(id)) input.match_input_image = true;
+    }
+    const ratios = isQwenEditPlus(id) ? QWEN_EDIT_RATIOS : QWEN_IMAGE_RATIOS;
+    if (req.aspectRatio !== void 0) {
+      if (ratios.has(req.aspectRatio)) input.aspect_ratio = req.aspectRatio;
+      else
+        ctx.warnings.push({
+          code: "option_ignored",
+          field: "aspectRatio",
+          from: req.aspectRatio,
+          message: `${id} has no ${req.aspectRatio} aspect_ratio; using the provider default`
+        });
+    }
+    if (isQwenEditPlus(id) && req.outputFormat !== void 0) {
+      input.output_format = req.outputFormat === "jpeg" ? "jpg" : req.outputFormat;
     }
-    if (req.outputFormat !== void 0) input.output_format = req.outputFormat;
   } else {
     const refs = req.references ?? [];
     const first = refs[0];
@@ -5398,7 +5561,13 @@ async function generateReplicateAudio(req, ctx, deps) {
   const input = id.includes("mmaudio") ? mmaudioInput(req, ctx) : id.includes("elevenlabs") ? elevenLabsMusicInput(req, ctx) : lyriaAudioInput(req, ctx);
   const { url, predictTime, raw } = await runPrediction(ctx, input, deps);
   const usage2 = audioUsage(ctx, req);
-  usage2.quantity = ctx.model.usage.unit === "seconds" ? elevenLabsMusicDuration(req.durationSeconds) : predictTime;
+  if (ctx.model.usage.unit === "seconds") {
+    const seconds = elevenLabsMusicDuration(req.durationSeconds);
+    usage2.quantity = seconds;
+    usage2.durationSeconds = seconds;
+  } else {
+    usage2.quantity = predictTime;
+  }
   return { audio: { url }, usage: usage2, raw: predictionRaw(raw) };
 }