fix(wp7-02): normalize generated image dimensions

apps/worker/package.json

@@ -9,7 +9,8 @@
   },
   "dependencies": {
     "better-sqlite3": "13.0.1",
-    "drizzle-orm": "0.45.2"
+    "drizzle-orm": "0.45.2",
+    "sharp": "0.35.3"
   },
   "devDependencies": {
     "@types/better-sqlite3": "7.6.13",

apps/worker/src/ai-adapter-gemini-flash.ts

@@ -1,4 +1,5 @@
 import type { GenerationAdapterRequest, NormalizedGenerationOutput } from "./ai-adapter-contract.js";
+import { normalizeImageOutputToRatio } from "./image-output-normalizer.mjs";
 import {
   AdapterContractError, type AdapterStartResult, type AdapterTransport, balanceSignalFromResponse,
   classifyAdapterError, dimensionsForRatio, mockPngBytes, validateAdapterRequest, validateMockContract,
@@ -41,7 +42,8 @@ export class GeminiFlashAdapter implements ModelAdapter {
         const classified = this.classifyError("error" in response ? response.error : undefined, (request.configSnapshot.error_mapping_profile as Record<string, string> | undefined) ?? {});
         return { ...classified, status: "failed" };
       }
-      return { outputs: [this.normalizeOutput(response)], status: "completed" };
+      const output = this.normalizeOutput(response);
+      return { outputs: [await normalizeImageOutputToRatio({ ...output, ratio: request.ratio })], status: "completed" };
     } catch (error) {
       const classified = error instanceof AdapterContractError
         ? { category: "gateway_contract_invalid" as const, sourceCategory: error.sourceCategory }

apps/worker/src/ai-adapter-gemini-pro.ts

@@ -1,4 +1,5 @@
 import type { GenerationAdapterRequest, NormalizedGenerationOutput } from "./ai-adapter-contract.js";
+import { normalizeImageOutputToRatio } from "./image-output-normalizer.mjs";
 import {
   AdapterContractError, type AdapterStartResult, type AdapterTransport, balanceSignalFromResponse,
   classifyAdapterError, dimensionsForRatio, mockPngBytes, validateAdapterRequest, validateMockContract,
@@ -37,7 +38,7 @@ export class GeminiProAdapter implements ModelAdapter {
     try {
       validateAdapterRequest(request, this.modelId);
       const response = await this.transport.start({ operation: "start", modelId: this.modelId, prompt: request.prompt, ratio: request.ratio, referenceAssetIds: request.referenceAssetIds });
-      return this.interpret(response, request.configSnapshot.error_mapping_profile as Record<string, string> ?? {});
+      return await this.interpret(response, request.configSnapshot.error_mapping_profile as Record<string, string> ?? {}, request.ratio);
     } catch (error) {
       const classified = error instanceof AdapterContractError
         ? { category: "gateway_contract_invalid" as const, sourceCategory: error.sourceCategory }
@@ -49,7 +50,7 @@ export class GeminiProAdapter implements ModelAdapter {
   async poll(upstreamJobReference: string): Promise<AdapterStartResult> {
     try {
       const response = await this.transport.poll({ operation: "poll", modelId: this.modelId, upstreamJobReference });
-      return this.interpret(response, {});
+      return await this.interpret(response, {});
     } catch (error) {
       const classified = error instanceof AdapterContractError
         ? { category: "gateway_contract_invalid" as const, sourceCategory: error.sourceCategory }
@@ -58,7 +59,7 @@ export class GeminiProAdapter implements ModelAdapter {
     }
   }
 
-  private interpret(response: unknown, mappingProfile: Readonly<Record<string, string>>): AdapterStartResult {
+  private async interpret(response: unknown, mappingProfile: Readonly<Record<string, string>>, ratio?: GenerationAdapterRequest["ratio"]): Promise<AdapterStartResult> {
     if (!response || typeof response !== "object" || !("operation" in response) || !response.operation || typeof response.operation !== "object") {
       return { category: "gateway_contract_invalid", sourceCategory: "response_shape_invalid", status: "failed" };
     }
@@ -72,7 +73,8 @@ export class GeminiProAdapter implements ModelAdapter {
       return reference ? { status: "pending", upstreamJobReference: reference } : { category: "gateway_contract_invalid", sourceCategory: "upstream_reference_missing", status: "failed" };
     }
     try {
-      return { outputs: [this.normalizeOutput("response" in operation ? operation.response : undefined)], status: "completed" };
+      const output = this.normalizeOutput("response" in operation ? operation.response : undefined);
+      return { outputs: [ratio ? await normalizeImageOutputToRatio({ ...output, ratio }) : output], status: "completed" };
     } catch (error) {
       return { category: "gateway_contract_invalid", sourceCategory: error instanceof AdapterContractError ? error.sourceCategory : "response_shape_invalid", status: "failed" };
     }

apps/worker/src/ai-adapter-gpt-image.ts

@@ -1,4 +1,5 @@
 import type { GenerationAdapterRequest, NormalizedGenerationOutput } from "./ai-adapter-contract.js";
+import { normalizeImageOutputToRatio } from "./image-output-normalizer.mjs";
 import {
   AdapterContractError, type AdapterStartResult, type AdapterTransport, balanceSignalFromResponse,
   classifyAdapterError, dimensionsForRatio, mockPngBytes, validateAdapterRequest, validateMockContract,
@@ -41,7 +42,8 @@ export class GptImageAdapter implements ModelAdapter {
         const classified = this.classifyError("error" in response ? response.error : undefined, (request.configSnapshot.error_mapping_profile as Record<string, string> | undefined) ?? {});
         return { ...classified, status: "failed" };
       }
-      return { outputs: [this.normalizeOutput(response)], status: "completed" };
+      const output = this.normalizeOutput(response);
+      return { outputs: [await normalizeImageOutputToRatio({ ...output, ratio: request.ratio })], status: "completed" };
     } catch (error) {
       const classified = error instanceof AdapterContractError
         ? { category: "gateway_contract_invalid" as const, sourceCategory: error.sourceCategory }

apps/worker/src/image-output-normalizer.mjs

@@ -0,0 +1,74 @@
+import sharp from "sharp";
+
+const productDimensions = Object.freeze({
+  "3:4": Object.freeze({ pixelHeight: 1440, pixelWidth: 1080 }),
+  "1:1": Object.freeze({ pixelHeight: 1080, pixelWidth: 1080 }),
+  "4:3": Object.freeze({ pixelHeight: 1080, pixelWidth: 1440 }),
+  "9:16": Object.freeze({ pixelHeight: 1920, pixelWidth: 1080 }),
+});
+
+const gptImageRequestSizes = Object.freeze({
+  "3:4": "1056x1408",
+  "1:1": "1088x1088",
+  "4:3": "1408x1056",
+  "9:16": "1008x1792",
+});
+
+const allowedMimeTypes = new Set(["image/jpeg", "image/png", "image/webp"]);
+const maximumInputBytes = 20 * 1024 * 1024;
+
+function assertRatio(ratio) {
+  if (!(ratio in productDimensions)) throw new Error("image_output_ratio_unsupported");
+  return ratio;
+}
+
+export function productDimensionsForRatio(ratio) {
+  return { ...productDimensions[assertRatio(ratio)] };
+}
+
+export function gptImageRequestSizeForRatio(ratio) {
+  return gptImageRequestSizes[assertRatio(ratio)];
+}
+
+export async function normalizeImageOutputToRatio(input) {
+  const ratio = assertRatio(input?.ratio);
+  if (!Buffer.isBuffer(input?.bytes) || input.bytes.length === 0 || input.bytes.length > maximumInputBytes
+    || !allowedMimeTypes.has(input?.mimeType)) {
+    throw new Error("image_output_media_invalid");
+  }
+  const target = productDimensions[ratio];
+  if (input.pixelWidth === target.pixelWidth && input.pixelHeight === target.pixelHeight) {
+    return {
+      bytes: Buffer.from(input.bytes),
+      mimeType: input.mimeType,
+      normalized: false,
+      ...target,
+      upstreamPixelHeight: input.pixelHeight,
+      upstreamPixelWidth: input.pixelWidth,
+    };
+  }
+
+  const image = sharp(input.bytes, { failOn: "error", limitInputPixels: 40_000_000 });
+  const metadata = await image.metadata();
+  if (!metadata.width || !metadata.height) throw new Error("image_output_dimensions_missing");
+  const requestedRatio = target.pixelWidth / target.pixelHeight;
+  const upstreamRatio = metadata.width / metadata.height;
+  if (Math.abs(upstreamRatio - requestedRatio) / requestedRatio > 0.02) {
+    throw new Error("image_output_aspect_ratio_mismatch");
+  }
+  const { data, info } = await image
+    .resize(target.pixelWidth, target.pixelHeight, { fit: "fill", kernel: sharp.kernel.lanczos3 })
+    .png({ compressionLevel: 9 })
+    .toBuffer({ resolveWithObject: true });
+  if (info.width !== target.pixelWidth || info.height !== target.pixelHeight || info.format !== "png") {
+    throw new Error("image_output_normalization_failed");
+  }
+  return {
+    bytes: data,
+    mimeType: "image/png",
+    normalized: true,
+    ...target,
+    upstreamPixelHeight: metadata.height,
+    upstreamPixelWidth: metadata.width,
+  };
+}

apps/worker/tsconfig.json

@@ -1,6 +1,7 @@
 {
   "extends": "../../tsconfig.base.json",
   "compilerOptions": {
+    "allowJs": true,
     "module": "NodeNext",
     "moduleResolution": "NodeNext",
     "lib": ["ES2024"],