feat: getAssetOcr endpoint (#23331)

* feat: getAssetOcr endpoint

* pr feedback

server/src/controllers/asset.controller.ts

@@ -16,6 +16,7 @@ import {
   UpdateAssetDto,
 } from 'src/dtos/asset.dto';
 import { AuthDto } from 'src/dtos/auth.dto';
+import { AssetOcrResponseDto } from 'src/dtos/ocr.dto';
 import { Permission, RouteKey } from 'src/enum';
 import { Auth, Authenticated } from 'src/middleware/auth.guard';
 import { AssetService } from 'src/services/asset.service';
@@ -95,6 +96,12 @@ export class AssetController {
     return this.service.getMetadata(auth, id);
   }
 
+  @Get(':id/ocr')
+  @Authenticated({ permission: Permission.AssetRead })
+  getAssetOcr(@Auth() auth: AuthDto, @Param() { id }: UUIDParamDto): Promise<AssetOcrResponseDto[]> {
+    return this.service.getOcr(auth, id);
+  }
+
   @Put(':id/metadata')
   @Authenticated({ permission: Permission.AssetUpdate })
   updateAssetMetadata(

server/src/dtos/ocr.dto.ts

@@ -0,0 +1,42 @@
+import { ApiProperty } from '@nestjs/swagger';
+
+export class AssetOcrResponseDto {
+  @ApiProperty({ type: 'string', format: 'uuid' })
+  id!: string;
+
+  @ApiProperty({ type: 'string', format: 'uuid' })
+  assetId!: string;
+
+  @ApiProperty({ type: 'number', format: 'double', description: 'Normalized x coordinate of box corner 1 (0-1)' })
+  x1!: number;
+
+  @ApiProperty({ type: 'number', format: 'double', description: 'Normalized y coordinate of box corner 1 (0-1)' })
+  y1!: number;
+
+  @ApiProperty({ type: 'number', format: 'double', description: 'Normalized x coordinate of box corner 2 (0-1)' })
+  x2!: number;
+
+  @ApiProperty({ type: 'number', format: 'double', description: 'Normalized y coordinate of box corner 2 (0-1)' })
+  y2!: number;
+
+  @ApiProperty({ type: 'number', format: 'double', description: 'Normalized x coordinate of box corner 3 (0-1)' })
+  x3!: number;
+
+  @ApiProperty({ type: 'number', format: 'double', description: 'Normalized y coordinate of box corner 3 (0-1)' })
+  y3!: number;
+
+  @ApiProperty({ type: 'number', format: 'double', description: 'Normalized x coordinate of box corner 4 (0-1)' })
+  x4!: number;
+
+  @ApiProperty({ type: 'number', format: 'double', description: 'Normalized y coordinate of box corner 4 (0-1)' })
+  y4!: number;
+
+  @ApiProperty({ type: 'number', format: 'double', description: 'Confidence score for text detection box' })
+  boxScore!: number;
+
+  @ApiProperty({ type: 'number', format: 'double', description: 'Confidence score for text recognition' })
+  textScore!: number;
+
+  @ApiProperty({ type: 'string', description: 'Recognized text' })
+  text!: string;
+}

server/src/services/asset.service.spec.ts

@@ -700,6 +700,42 @@ describe(AssetService.name, () => {
     });
   });
 
+  describe('getOcr', () => {
+    it('should require asset read permission', async () => {
+      mocks.access.asset.checkOwnerAccess.mockResolvedValue(new Set());
+
+      await expect(sut.getOcr(authStub.admin, 'asset-1')).rejects.toBeInstanceOf(BadRequestException);
+
+      expect(mocks.ocr.getByAssetId).not.toHaveBeenCalled();
+    });
+
+    it('should return OCR data for an asset', async () => {
+      const ocr1 = factory.assetOcr({ text: 'Hello World' });
+      const ocr2 = factory.assetOcr({ text: 'Test Image' });
+
+      mocks.access.asset.checkOwnerAccess.mockResolvedValue(new Set(['asset-1']));
+      mocks.ocr.getByAssetId.mockResolvedValue([ocr1, ocr2]);
+
+      await expect(sut.getOcr(authStub.admin, 'asset-1')).resolves.toEqual([ocr1, ocr2]);
+
+      expect(mocks.access.asset.checkOwnerAccess).toHaveBeenCalledWith(
+        authStub.admin.user.id,
+        new Set(['asset-1']),
+        undefined,
+      );
+      expect(mocks.ocr.getByAssetId).toHaveBeenCalledWith('asset-1');
+    });
+
+    it('should return empty array when no OCR data exists', async () => {
+      mocks.access.asset.checkOwnerAccess.mockResolvedValue(new Set(['asset-1']));
+      mocks.ocr.getByAssetId.mockResolvedValue([]);
+
+      await expect(sut.getOcr(authStub.admin, 'asset-1')).resolves.toEqual([]);
+
+      expect(mocks.ocr.getByAssetId).toHaveBeenCalledWith('asset-1');
+    });
+  });
+
   describe('run', () => {
     it('should run the refresh faces job', async () => {
       mocks.access.asset.checkOwnerAccess.mockResolvedValue(new Set(['asset-1']));

server/src/services/asset.service.ts

@@ -16,6 +16,7 @@ import {
   mapStats,
 } from 'src/dtos/asset.dto';
 import { AuthDto } from 'src/dtos/auth.dto';
+import { AssetOcrResponseDto } from 'src/dtos/ocr.dto';
 import { AssetMetadataKey, AssetStatus, AssetVisibility, JobName, JobStatus, Permission, QueueName } from 'src/enum';
 import { BaseService } from 'src/services/base.service';
 import { ISidecarWriteJob, JobItem, JobOf } from 'src/types';
@@ -289,6 +290,11 @@ export class AssetService extends BaseService {
     return this.assetRepository.getMetadata(id);
   }
 
+  async getOcr(auth: AuthDto, id: string): Promise<AssetOcrResponseDto[]> {
+    await this.requireAccess({ auth, permission: Permission.AssetRead, ids: [id] });
+    return this.ocrRepository.getByAssetId(id);
+  }
+
   async upsertMetadata(auth: AuthDto, id: string, dto: AssetMetadataUpsertDto): Promise<AssetMetadataResponseDto[]> {
     await this.requireAccess({ auth, permission: Permission.AssetUpdate, ids: [id] });
     return this.assetRepository.upsertMetadata(id, dto.items);