Complete OCR question image persistence

docker-compose.yml

@@ -88,6 +88,18 @@ services:
       APP_JWT_EXPIRATION_MS: ${APP_JWT_EXPIRATION_MS:-86400000}
       APP_AUTH_GOOGLE_STATE_COOKIE_SECURE: ${GOOGLE_OAUTH_STATE_COOKIE_SECURE:-true}
       OCR_SERVICE_API_KEY: ${OCR_API_KEY:?Set OCR_API_KEY for backend-to-OCR authentication}
+      # Object storage. Keep local for the default development profile;
+      # use STORAGE_PROVIDER=s3 with --profile storage for MinIO.
+      STORAGE_PROVIDER: ${STORAGE_PROVIDER:-local}
+      STORAGE_LOCAL_ROOT: ${STORAGE_LOCAL_ROOT:-/tmp/kixi/uploads}
+      STORAGE_PUBLIC_BASE_URL: ${STORAGE_PUBLIC_BASE_URL:-/uploads}
+      STORAGE_S3_ENDPOINT: ${STORAGE_S3_ENDPOINT:-http://minio:9000}
+      STORAGE_S3_REGION: ${STORAGE_S3_REGION:-us-east-1}
+      STORAGE_S3_BUCKET: ${STORAGE_S3_BUCKET:-kixi-images}
+      STORAGE_S3_ACCESS_KEY: ${STORAGE_S3_ACCESS_KEY:-kixi-minio}
+      STORAGE_S3_SECRET_KEY: ${STORAGE_S3_SECRET_KEY:-kixi-minio-secret}
+      STORAGE_S3_PATH_STYLE_ACCESS: ${STORAGE_S3_PATH_STYLE_ACCESS:-true}
+      STORAGE_MAX_OBJECT_SIZE_BYTES: ${STORAGE_MAX_OBJECT_SIZE_BYTES:-20971520}
     ports:
       - "${BACKEND_PORT:-8080}:8080"
     depends_on:
@@ -255,6 +267,53 @@ services:
     networks:
       - kixi-network
 
+  # ===========================================================================
+  # S3-compatible object storage (Optional Profile)
+  # ===========================================================================
+  # Start with STORAGE_PROVIDER=s3 docker-compose --profile storage up
+  minio:
+    image: minio/minio:RELEASE.2024-06-13T22-53-53Z
+    container_name: kixi-minio
+    profiles:
+      - storage
+    restart: unless-stopped
+    command: server /data --console-address ":9001"
+    environment:
+      MINIO_ROOT_USER: ${STORAGE_S3_ACCESS_KEY:-kixi-minio}
+      MINIO_ROOT_PASSWORD: ${STORAGE_S3_SECRET_KEY:-kixi-minio-secret}
+    ports:
+      - "${MINIO_PORT:-9000}:9000"
+      - "${MINIO_CONSOLE_PORT:-9001}:9001"
+    volumes:
+      - minio_data:/data
+    healthcheck:
+      test: ["CMD", "mc", "ready", "local"]
+      interval: 10s
+      timeout: 5s
+      retries: 5
+    networks:
+      - kixi-network
+
+  minio-init:
+    image: minio/mc:RELEASE.2024-06-13T21-21-32Z
+    container_name: kixi-minio-init
+    profiles:
+      - storage
+    depends_on:
+      minio:
+        condition: service_healthy
+    entrypoint: /bin/sh
+    command:
+      - -c
+      - >-
+        mc alias set local http://minio:9000
+        ${STORAGE_S3_ACCESS_KEY:-kixi-minio}
+        ${STORAGE_S3_SECRET_KEY:-kixi-minio-secret}
+        && mc mb --ignore-existing local/${STORAGE_S3_BUCKET:-kixi-images}
+        && mc anonymous set download local/${STORAGE_S3_BUCKET:-kixi-images}
+    networks:
+      - kixi-network
+
 # =============================================================================
 # Networks
 # =============================================================================
@@ -282,3 +341,7 @@ volumes:
   # Redis data persistence
   redis_data:
     name: kixi-redis-data
+
+  # MinIO object data for local integration runs
+  minio_data:
+    name: kixi-minio-data

docs/IMAGE_STORAGE.md

@@ -0,0 +1,41 @@
+# Image storage
+
+Question images are stored through the `ImageStorage` port. The backend supports:
+
+- `local` (default): writes below `storage.local-root` and preserves the existing
+  `/uploads` URL base for local development;
+- `s3`: uses AWS S3 or any S3-compatible endpoint such as MinIO.
+
+The object key is generated by the backend (`questions/{questionId}/{uuid}.ext`).
+The multipart filename is never used as a path component. Only JPEG, PNG and
+WebP are accepted, and the default maximum object size is 20 MiB. The key is
+stored in `question_images.storage_key`; legacy rows remain purgeable without a
+storage delete because they predate this column.
+
+## MinIO local integration
+
+Run the optional profile with the S3 provider enabled:
+
+```bash
+STORAGE_PROVIDER=s3 \
+STORAGE_PUBLIC_BASE_URL=http://localhost:9000/kixi-images \
+docker compose --profile storage up --build
+```
+
+The profile creates the `kixi-images` bucket and enables anonymous downloads for
+local development only. Do not carry that bucket policy to production. In
+production, configure a private bucket with a controlled CDN or signed delivery
+endpoint through `STORAGE_PUBLIC_BASE_URL`.
+
+## OCR boundary
+
+OCR returns a versioned region contract (`contractVersion: 1`) with page index,
+source-file index, source dimensions and a bounded crop rectangle. The
+persistence endpoint retains the submitted raster bytes, crops `questao_N`
+regions, stores the PNG through the same `ImageStorage` port and associates it
+with the persisted question.
+
+PDF sources and `cabecalho`/`rodape` regions remain metadata-only in this slice:
+the OCR service renders PDF pages internally and `question_images` requires a
+question association. A future statement-asset contract must define those
+cases before they are persisted automatically.

services/backend-api/pom.xml

@@ -27,6 +27,12 @@
     </properties>
 
     <dependencies>
+        <dependency>
+            <groupId>software.amazon.awssdk</groupId>
+            <artifactId>s3</artifactId>
+            <version>2.25.60</version>
+        </dependency>
+
         <dependency>
             <groupId>org.springframework.boot</groupId>
             <artifactId>spring-boot-starter-webflux</artifactId>

services/backend-api/src/main/java/ao/creativemode/kixi/client/OcrServiceClient.java

@@ -139,6 +139,53 @@ public Mono<OcrResponse> extractText(List<FilePart> files) {
                 .doOnError(error -> log.error("OCR request failed: type={}", error.getClass().getSimpleName()));
     }
 
+    /**
+     * Extract text from bytes retained by the persistence flow. This variant
+     * avoids replaying a consumed multipart stream when OCR regions are later
+     * associated with their source page.
+     */
+    public Mono<OcrResponse> extractTextFromUploadedFiles(List<OcrUploadedFile> files) {
+        if (files == null || files.isEmpty()) {
+            return Mono.error(new IllegalArgumentException("At least one file is required"));
+        }
+
+        MultipartBodyBuilder builder = new MultipartBodyBuilder();
+        for (OcrUploadedFile file : files) {
+            builder.part("images", file.content())
+                    .filename(file.filename())
+                    .contentType(file.contentType() != null
+                            ? file.contentType()
+                            : getContentType(file.filename()));
+        }
+
+        log.info("Sending retained OCR request: {} file(s)", files.size());
+        return webClient.post()
+                .uri("/ocr/v1/extract")
+                .headers(this::applyAuthentication)
+                .contentType(MediaType.MULTIPART_FORM_DATA)
+                .body(BodyInserters.fromMultipartData(builder.build()))
+                .retrieve()
+                .onStatus(HttpStatusCode::is4xxClientError, response ->
+                        response.bodyToMono(String.class)
+                                .flatMap(body -> Mono.error(new OcrClientException(
+                                        "OCR request failed: " + body,
+                                        response.statusCode().value()))))
+                .onStatus(HttpStatusCode::is5xxServerError, response ->
+                        response.bodyToMono(String.class)
+                                .flatMap(body -> Mono.error(new OcrServerException(
+                                        "OCR service error: " + body,
+                                        response.statusCode().value()))))
+                .bodyToMono(OcrResponse.class)
+                .timeout(timeout)
+                .retryWhen(Retry.backoff(maxRetries, Duration.ofSeconds(1))
+                        .filter(this::isRetryable))
+                .doOnSuccess(response -> log.info(
+                        "OCR retained request successful: requestId={}, status={}",
+                        response.requestId(), response.status()))
+                .doOnError(error -> log.error("OCR retained request failed: type={}",
+                        error.getClass().getSimpleName()));
+    }
+
     /**
      * Extract text from raw image bytes.
      *

services/backend-api/src/main/java/ao/creativemode/kixi/client/OcrUploadedFile.java

@@ -0,0 +1,22 @@
+package ao.creativemode.kixi.client;
+
+import org.springframework.http.MediaType;
+
+/**
+ * Immutable upload retained for OCR persistence. Keeping the bytes here lets
+ * the backend associate OCR regions with the exact source page it submitted.
+ */
+public record OcrUploadedFile(
+        String filename,
+        MediaType contentType,
+        byte[] content) {
+
+    public OcrUploadedFile {
+        if (filename == null || filename.isBlank()) {
+            throw new IllegalArgumentException("Filename is required");
+        }
+        if (content == null || content.length == 0) {
+            throw new IllegalArgumentException("File content cannot be empty");
+        }
+    }
+}

services/backend-api/src/main/java/ao/creativemode/kixi/config/StorageConfiguration.java

@@ -0,0 +1,38 @@
+package ao.creativemode.kixi.config;
+
+import java.net.URI;
+
+import org.springframework.boot.autoconfigure.condition.ConditionalOnProperty;
+import org.springframework.context.annotation.Bean;
+import org.springframework.context.annotation.Configuration;
+
+import software.amazon.awssdk.auth.credentials.AwsBasicCredentials;
+import software.amazon.awssdk.auth.credentials.StaticCredentialsProvider;
+import software.amazon.awssdk.regions.Region;
+import software.amazon.awssdk.services.s3.S3Client;
+import software.amazon.awssdk.services.s3.S3Configuration;
+
+@Configuration(proxyBeanMethods = false)
+public class StorageConfiguration {
+
+    @Bean
+    @ConditionalOnProperty(prefix = "storage", name = "provider", havingValue = "s3")
+    S3Client s3Client(StorageProperties properties) {
+        if (properties.getS3AccessKey().isBlank() || properties.getS3SecretKey().isBlank()) {
+            throw new IllegalStateException("storage.s3-access-key and storage.s3-secret-key must be configured for S3 storage");
+        }
+
+        var builder = S3Client.builder()
+                .region(Region.of(properties.getS3Region()))
+                .credentialsProvider(StaticCredentialsProvider.create(
+                        AwsBasicCredentials.create(properties.getS3AccessKey(), properties.getS3SecretKey())))
+                .serviceConfiguration(S3Configuration.builder()
+                        .pathStyleAccessEnabled(properties.isS3PathStyleAccess())
+                        .build());
+
+        if (!properties.getS3Endpoint().isBlank()) {
+            builder.endpointOverride(URI.create(properties.getS3Endpoint()));
+        }
+        return builder.build();
+    }
+}

services/backend-api/src/main/java/ao/creativemode/kixi/config/StorageProperties.java

@@ -0,0 +1,104 @@
+package ao.creativemode.kixi.config;
+
+import org.springframework.boot.context.properties.ConfigurationProperties;
+import org.springframework.stereotype.Component;
+
+/**
+ * Object storage configuration. The default keeps local development compatible
+ * with the existing static upload directory; production should use S3.
+ */
+@Component
+@ConfigurationProperties(prefix = "storage")
+public class StorageProperties {
+
+    private String provider = "local";
+    private String localRoot = "services/backend-api/src/main/resources/static/uploads";
+    private String publicBaseUrl = "/uploads";
+    private String s3Endpoint = "";
+    private String s3Region = "us-east-1";
+    private String s3Bucket = "kixi-images";
+    private String s3AccessKey = "";
+    private String s3SecretKey = "";
+    private boolean s3PathStyleAccess = true;
+    private long maxObjectSizeBytes = 20L * 1024 * 1024;
+
+    public String getProvider() {
+        return provider;
+    }
+
+    public void setProvider(String provider) {
+        this.provider = provider;
+    }
+
+    public String getLocalRoot() {
+        return localRoot;
+    }
+
+    public void setLocalRoot(String localRoot) {
+        this.localRoot = localRoot;
+    }
+
+    public String getPublicBaseUrl() {
+        return publicBaseUrl;
+    }
+
+    public void setPublicBaseUrl(String publicBaseUrl) {
+        this.publicBaseUrl = publicBaseUrl;
+    }
+
+    public String getS3Endpoint() {
+        return s3Endpoint;
+    }
+
+    public void setS3Endpoint(String s3Endpoint) {
+        this.s3Endpoint = s3Endpoint;
+    }
+
+    public String getS3Region() {
+        return s3Region;
+    }
+
+    public void setS3Region(String s3Region) {
+        this.s3Region = s3Region;
+    }
+
+    public String getS3Bucket() {
+        return s3Bucket;
+    }
+
+    public void setS3Bucket(String s3Bucket) {
+        this.s3Bucket = s3Bucket;
+    }
+
+    public String getS3AccessKey() {
+        return s3AccessKey;
+    }
+
+    public void setS3AccessKey(String s3AccessKey) {
+        this.s3AccessKey = s3AccessKey;
+    }
+
+    public String getS3SecretKey() {
+        return s3SecretKey;
+    }
+
+    public void setS3SecretKey(String s3SecretKey) {
+        this.s3SecretKey = s3SecretKey;
+    }
+
+    public boolean isS3PathStyleAccess() {
+        return s3PathStyleAccess;
+    }
+
+    public void setS3PathStyleAccess(boolean s3PathStyleAccess) {
+        this.s3PathStyleAccess = s3PathStyleAccess;
+    }
+
+    public long getMaxObjectSizeBytes() {
+        return maxObjectSizeBytes;
+    }
+
+    public void setMaxObjectSizeBytes(long maxObjectSizeBytes) {
+        this.maxObjectSizeBytes = maxObjectSizeBytes;
+    }
+}