ci: Bump version to 0.7.2 [skip ci]

package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "flyer-crawler",
-  "version": "0.7.0",
+  "version": "0.7.2",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "flyer-crawler",
-      "version": "0.7.0",
+      "version": "0.7.2",
       "dependencies": {
         "@bull-board/api": "^6.14.2",
         "@bull-board/express": "^6.14.2",

package.json

@@ -1,7 +1,7 @@
 {
   "name": "flyer-crawler",
   "private": true,
-  "version": "0.7.0",
+  "version": "0.7.2",
   "type": "module",
   "scripts": {
     "dev": "concurrently \"npm:start:dev\" \"vite\"",

src/services/aiService.server.ts

@@ -43,7 +43,7 @@ const ExtractedFlyerItemSchema = z.object({
 });
 
 export const AiFlyerDataSchema = z.object({
-  store_name: requiredString('Store name cannot be empty'),
+  store_name: z.string().nullable(),
   valid_from: z.string().nullable(),
   valid_to: z.string().nullable(),
   store_address: z.string().nullable(),
@@ -507,7 +507,7 @@ export class AIService {
     userProfileAddress?: string,
     logger: Logger = this.logger,
   ): Promise<{
-    store_name: string;
+    store_name: string | null;
     valid_from: string | null;
     valid_to: string | null;
     store_address: string | null;

src/services/flyerAiProcessor.server.test.ts

@@ -127,4 +127,98 @@ describe('FlyerAiProcessor', () => {
     expect(result.needsReview).toBe(true);
     expect(logger.warn).toHaveBeenCalledWith(expect.any(Object), expect.stringContaining('contains no items. The flyer will be saved with an item_count of 0. Flagging for review.'));
   });
+
+  describe('Batching Logic', () => {
+    it('should process images in batches and merge the results correctly', async () => {
+      // Arrange
+      const jobData = createMockJobData({});
+      // 5 images, with BATCH_SIZE = 4, should result in 2 batches.
+      const imagePaths = [
+        { path: 'page1.jpg', mimetype: 'image/jpeg' },
+        { path: 'page2.jpg', mimetype: 'image/jpeg' },
+        { path: 'page3.jpg', mimetype: 'image/jpeg' },
+        { path: 'page4.jpg', mimetype: 'image/jpeg' },
+        { path: 'page5.jpg', mimetype: 'image/jpeg' },
+      ];
+
+      const mockAiResponseBatch1 = {
+        store_name: 'Batch 1 Store',
+        valid_from: '2025-01-01',
+        valid_to: '2025-01-07',
+        store_address: '123 Batch St',
+        items: [
+          { item: 'Item A', price_display: '$1', price_in_cents: 100, quantity: '1', category_name: 'Cat A', master_item_id: 1 },
+          { item: 'Item B', price_display: '$2', price_in_cents: 200, quantity: '1', category_name: 'Cat B', master_item_id: 2 },
+        ],
+      };
+
+      const mockAiResponseBatch2 = {
+        store_name: 'Batch 2 Store', // This should be ignored in the merge
+        valid_from: null,
+        valid_to: null,
+        store_address: null,
+        items: [
+          { item: 'Item C', price_display: '$3', price_in_cents: 300, quantity: '1', category_name: 'Cat C', master_item_id: 3 },
+        ],
+      };
+
+      // Mock the AI service to return different results for each batch call
+      vi.mocked(mockAiService.extractCoreDataFromFlyerImage)
+        .mockResolvedValueOnce(mockAiResponseBatch1)
+        .mockResolvedValueOnce(mockAiResponseBatch2);
+
+      // Act
+      const result = await service.extractAndValidateData(imagePaths, jobData, logger);
+
+      // Assert
+      // 1. AI service was called twice (for 2 batches)
+      expect(mockAiService.extractCoreDataFromFlyerImage).toHaveBeenCalledTimes(2);
+
+      // 2. Check the arguments for each call
+      expect(mockAiService.extractCoreDataFromFlyerImage).toHaveBeenNthCalledWith(1, imagePaths.slice(0, 4), [], undefined, undefined, logger);
+      expect(mockAiService.extractCoreDataFromFlyerImage).toHaveBeenNthCalledWith(2, imagePaths.slice(4, 5), [], undefined, undefined, logger);
+
+      // 3. Check the merged data
+      expect(result.data.store_name).toBe('Batch 1 Store'); // Metadata from the first batch
+      expect(result.data.valid_from).toBe('2025-01-01');
+      expect(result.data.valid_to).toBe('2025-01-07');
+      expect(result.data.store_address).toBe('123 Batch St');
+
+      // 4. Check that items from both batches are merged
+      expect(result.data.items).toHaveLength(3);
+      expect(result.data.items).toEqual(expect.arrayContaining([
+        expect.objectContaining({ item: 'Item A' }),
+        expect.objectContaining({ item: 'Item B' }),
+        expect.objectContaining({ item: 'Item C' }),
+      ]));
+
+      // 5. Check that the job is not flagged for review
+      expect(result.needsReview).toBe(false);
+    });
+
+    it('should fill in missing metadata from subsequent batches', async () => {
+      // Arrange
+      const jobData = createMockJobData({});
+      const imagePaths = [
+        { path: 'page1.jpg', mimetype: 'image/jpeg' }, { path: 'page2.jpg', mimetype: 'image/jpeg' }, { path: 'page3.jpg', mimetype: 'image/jpeg' }, { path: 'page4.jpg', mimetype: 'image/jpeg' }, { path: 'page5.jpg', mimetype: 'image/jpeg' },
+      ];
+
+      const mockAiResponseBatch1 = { store_name: null, valid_from: '2025-01-01', valid_to: '2025-01-07', store_address: null, items: [{ item: 'Item A', price_display: '$1', price_in_cents: 100, quantity: '1', category_name: 'Cat A', master_item_id: 1 }] };
+      const mockAiResponseBatch2 = { store_name: 'Batch 2 Store', valid_from: '2025-01-02', valid_to: null, store_address: '456 Subsequent St', items: [{ item: 'Item C', price_display: '$3', price_in_cents: 300, quantity: '1', category_name: 'Cat C', master_item_id: 3 }] };
+
+      vi.mocked(mockAiService.extractCoreDataFromFlyerImage)
+        .mockResolvedValueOnce(mockAiResponseBatch1)
+        .mockResolvedValueOnce(mockAiResponseBatch2);
+
+      // Act
+      const result = await service.extractAndValidateData(imagePaths, jobData, logger);
+
+      // Assert
+      expect(result.data.store_name).toBe('Batch 2 Store'); // Filled from batch 2
+      expect(result.data.valid_from).toBe('2025-01-01'); // Kept from batch 1
+      expect(result.data.valid_to).toBe('2025-01-07'); // Kept from batch 1
+      expect(result.data.store_address).toBe('456 Subsequent St'); // Filled from batch 2
+      expect(result.data.items).toHaveLength(2);
+    });
+  });
 });

src/services/flyerAiProcessor.server.ts

@@ -94,19 +94,64 @@ export class FlyerAiProcessor {
     jobData: FlyerJobData,
     logger: Logger,
   ): Promise<AiProcessorResult> {
-    logger.info(`Starting AI data extraction.`);
+    logger.info(`Starting AI data extraction for ${imagePaths.length} pages.`);
     const { submitterIp, userProfileAddress } = jobData;
     const masterItems = await this.personalizationRepo.getAllMasterItems(logger);
     logger.debug(`Retrieved ${masterItems.length} master items for AI matching.`);
 
-    const extractedData = await this.ai.extractCoreDataFromFlyerImage(
-      imagePaths,
-      masterItems,
-      submitterIp,
-      userProfileAddress,
-      logger,
-    );
+    // BATCHING LOGIC: Process images in chunks to avoid hitting AI payload/token limits.
+    const BATCH_SIZE = 4;
+    const batches = [];
+    for (let i = 0; i < imagePaths.length; i += BATCH_SIZE) {
+      batches.push(imagePaths.slice(i, i + BATCH_SIZE));
+    }
 
-    return this._validateAiData(extractedData, logger);
+    // Initialize container for merged data
+    const mergedData: ValidatedAiDataType = {
+      store_name: null,
+      valid_from: null,
+      valid_to: null,
+      store_address: null,
+      items: [],
+    };
+
+    logger.info(`Processing ${imagePaths.length} pages in ${batches.length} batches (Batch Size: ${BATCH_SIZE}).`);
+
+    for (const [index, batch] of batches.entries()) {
+      logger.info(`Processing batch ${index + 1}/${batches.length} (${batch.length} pages)...`);
+
+      // The AI service handles rate limiting internally (e.g., max 5 RPM).
+      // Processing these sequentially ensures we respect that limit.
+      const batchResult = await this.ai.extractCoreDataFromFlyerImage(
+        batch,
+        masterItems,
+        submitterIp,
+        userProfileAddress,
+        logger,
+      );
+
+      // MERGE LOGIC:
+      // 1. Metadata (Store Name, Dates): Prioritize the first batch (usually the cover page).
+      //    If subsequent batches have data and the current is null, fill it in.
+      if (index === 0) {
+        mergedData.store_name = batchResult.store_name;
+        mergedData.valid_from = batchResult.valid_from;
+        mergedData.valid_to = batchResult.valid_to;
+        mergedData.store_address = batchResult.store_address;
+      } else {
+        if (!mergedData.store_name && batchResult.store_name) mergedData.store_name = batchResult.store_name;
+        if (!mergedData.valid_from && batchResult.valid_from) mergedData.valid_from = batchResult.valid_from;
+        if (!mergedData.valid_to && batchResult.valid_to) mergedData.valid_to = batchResult.valid_to;
+        if (!mergedData.store_address && batchResult.store_address) mergedData.store_address = batchResult.store_address;
+      }
+
+      // 2. Items: Append all found items to the master list.
+      mergedData.items.push(...batchResult.items);
+    }
+
+    logger.info(`Batch processing complete. Total items extracted: ${mergedData.items.length}`);
+
+    // Validate the final merged dataset
+    return this._validateAiData(mergedData, logger);
   }
 }

src/services/workers.server.ts

@@ -1,3 +1,4 @@
+// src/services/workers.server.ts
 import { Worker, Job, UnrecoverableError } from 'bullmq';
 import fsPromises from 'node:fs/promises';
 import { exec } from 'child_process';