feat(ocr): add /train endpoint to OCR service and OcrClient.trainModel()

- POST /train in ocr-service with ZIP Slip validation, TemporaryDirectory,
  ketos transfer learning, timestamped backups (keep last 3), in-process reload
- X-Training-Token auth (no-op in dev when TRAINING_TOKEN env is empty)
- trainModel() in OcrClient interface + RestClientOcrClient (10-min timeout,
  multipart upload, forwards X-Training-Token when configured)
- TRAINING_TOKEN env var wired in docker-compose; --workers 2 in Dockerfile
  so /health stays responsive during synchronous training

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

backend/src/main/java/org/raddatz/familienarchiv/service/OcrClient.java

@@ -10,6 +10,16 @@ import java.util.function.Consumer;
 public interface OcrClient {
     List<OcrBlockResult> extractBlocks(String pdfUrl, ScriptType scriptType);
 
+    /**
+     * Send a training ZIP to the OCR service for fine-tuning the Kurrent model.
+     *
+     * @param trainingDataZip raw ZIP bytes produced by TrainingDataExportService
+     * @return training result metrics (loss, accuracy, epochs)
+     */
+    TrainingResult trainModel(byte[] trainingDataZip);
+
+    record TrainingResult(Double loss, Double accuracy, Integer epochs) {}
+
     /**
      * Stream OCR results page-by-page via NDJSON. Implementations should override
      * this method. The default exists only for backward compatibility during migration

backend/src/main/java/org/raddatz/familienarchiv/service/RestClientOcrClient.java

@@ -9,9 +9,14 @@ import lombok.extern.slf4j.Slf4j;
 import org.raddatz.familienarchiv.model.ScriptType;
 import org.springframework.beans.factory.annotation.Value;
 import org.springframework.core.ParameterizedTypeReference;
+import org.springframework.core.io.ByteArrayResource;
+import org.springframework.http.HttpEntity;
+import org.springframework.http.HttpHeaders;
 import org.springframework.http.MediaType;
 import org.springframework.http.client.JdkClientHttpRequestFactory;
 import org.springframework.stereotype.Component;
+import org.springframework.util.LinkedMultiValueMap;
+import org.springframework.util.MultiValueMap;
 import org.springframework.web.client.RestClient;
 
 import java.io.BufferedReader;
@@ -36,11 +41,16 @@ public class RestClientOcrClient implements OcrClient, OcrHealthClient {
             .configure(DeserializationFeature.FAIL_ON_UNKNOWN_PROPERTIES, true);
 
     private final RestClient restClient;
+    private final RestClient trainingRestClient;
     private final HttpClient streamingHttpClient;
     private final String baseUrl;
+    private final String trainingToken;
 
-    public RestClientOcrClient(@Value("${app.ocr.base-url:http://ocr-service:8000}") String baseUrl) {
+    public RestClientOcrClient(
+            @Value("${app.ocr.base-url:http://ocr-service:8000}") String baseUrl,
+            @Value("${app.ocr.training-token:}") String trainingToken) {
         this.baseUrl = baseUrl;
+        this.trainingToken = trainingToken;
 
         HttpClient httpClient = HttpClient.newBuilder()
                 .version(HttpClient.Version.HTTP_1_1)
@@ -54,6 +64,17 @@ public class RestClientOcrClient implements OcrClient, OcrHealthClient {
                 .requestFactory(requestFactory)
                 .build();
 
+        HttpClient trainingHttpClient = HttpClient.newBuilder()
+                .version(HttpClient.Version.HTTP_1_1)
+                .connectTimeout(Duration.ofSeconds(10))
+                .build();
+        JdkClientHttpRequestFactory trainingRequestFactory = new JdkClientHttpRequestFactory(trainingHttpClient);
+        trainingRequestFactory.setReadTimeout(Duration.ofMinutes(10));
+        this.trainingRestClient = RestClient.builder()
+                .baseUrl(baseUrl)
+                .requestFactory(trainingRequestFactory)
+                .build();
+
         this.streamingHttpClient = HttpClient.newBuilder()
                 .version(HttpClient.Version.HTTP_1_1)
                 .connectTimeout(Duration.ofSeconds(10))
@@ -81,6 +102,35 @@ public class RestClientOcrClient implements OcrClient, OcrHealthClient {
                 .toList();
     }
 
+    @Override
+    public OcrClient.TrainingResult trainModel(byte[] trainingDataZip) {
+        ByteArrayResource zipResource = new ByteArrayResource(trainingDataZip) {
+            @Override
+            public String getFilename() { return "training-data.zip"; }
+        };
+
+        MultiValueMap<String, Object> body = new LinkedMultiValueMap<>();
+        HttpHeaders partHeaders = new HttpHeaders();
+        partHeaders.setContentType(MediaType.parseMediaType("application/zip"));
+        body.add("file", new HttpEntity<>(zipResource, partHeaders));
+
+        var spec = trainingRestClient.post()
+                .uri("/train")
+                .contentType(MediaType.MULTIPART_FORM_DATA);
+
+        if (trainingToken != null && !trainingToken.isBlank()) {
+            spec = spec.header("X-Training-Token", trainingToken);
+        }
+
+        TrainingResultJson result = spec
+                .body(body)
+                .retrieve()
+                .body(TrainingResultJson.class);
+
+        if (result == null) return new OcrClient.TrainingResult(null, null, null);
+        return new OcrClient.TrainingResult(result.loss(), result.accuracy(), result.epochs());
+    }
+
     @Override
     public boolean isHealthy() {
         try {
@@ -171,6 +221,8 @@ public class RestClientOcrClient implements OcrClient, OcrHealthClient {
         }
     }
 
+    record TrainingResultJson(Double loss, Double accuracy, Integer epochs) {}
+
     record OcrBlockJson(
             @JsonProperty("pageNumber") int pageNumber,
             double x,

backend/src/test/java/org/raddatz/familienarchiv/service/OcrClientDefaultStreamTest.java

@@ -12,10 +12,15 @@ class OcrClientDefaultStreamTest {
 
     @Test
     void defaultStreamBlocksSynthesizesEventsFromExtractBlocks() {
-        OcrClient client = (pdfUrl, scriptType) -> List.of(
+        OcrClient client = new OcrClient() {
-                new OcrBlockResult(0, 0.1, 0.1, 0.8, 0.04, null, "Line 1"),
+            @Override public List<OcrBlockResult> extractBlocks(String pdfUrl, ScriptType scriptType) {
-                new OcrBlockResult(0, 0.1, 0.2, 0.8, 0.04, null, "Line 2"),
+                return List.of(
-                new OcrBlockResult(1, 0.1, 0.1, 0.8, 0.04, null, "Line 3"));
+                    new OcrBlockResult(0, 0.1, 0.1, 0.8, 0.04, null, "Line 1"),
+                    new OcrBlockResult(0, 0.1, 0.2, 0.8, 0.04, null, "Line 2"),
+                    new OcrBlockResult(1, 0.1, 0.1, 0.8, 0.04, null, "Line 3"));
+            }
+            @Override public TrainingResult trainModel(byte[] zip) { return null; }
+        };
 
         List<OcrStreamEvent> events = new ArrayList<>();
         client.streamBlocks("http://test", ScriptType.TYPEWRITER, events::add);
@@ -42,7 +47,10 @@ class OcrClientDefaultStreamTest {
 
     @Test
     void defaultStreamBlocksHandlesEmptyResults() {
-        OcrClient client = (pdfUrl, scriptType) -> List.of();
+        OcrClient client = new OcrClient() {
+            @Override public List<OcrBlockResult> extractBlocks(String pdfUrl, ScriptType scriptType) { return List.of(); }
+            @Override public TrainingResult trainModel(byte[] zip) { return null; }
+        };
 
         List<OcrStreamEvent> events = new ArrayList<>();
         client.streamBlocks("http://test", ScriptType.TYPEWRITER, events::add);

docker-compose.yml

@@ -87,6 +87,7 @@ services:
       - ocr_cache:/root/.cache
     environment:
       KRAKEN_MODEL_PATH: /app/models/german_kurrent.mlmodel
+      TRAINING_TOKEN: "${OCR_TRAINING_TOKEN:-}"
       OCR_CONFIDENCE_THRESHOLD: "0.3"
       OCR_CONFIDENCE_THRESHOLD_KURRENT: "0.5"
       RECOGNITION_BATCH_SIZE: "16"

ocr-service/Dockerfile

@@ -23,4 +23,4 @@ COPY . .
 
 EXPOSE 8000
 
-CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8000"]
+CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8000", "--workers", "2"]

ocr-service/main.py

@@ -1,16 +1,21 @@
 """OCR microservice — FastAPI app with Surya and Kraken engine support."""
 
 import asyncio
+import glob
 import io
 import json
 import logging
 import os
+import shutil
+import tempfile
+import zipfile
 from contextlib import asynccontextmanager
+from datetime import datetime, timezone
 from urllib.parse import urlparse
 
 import httpx
 import pypdfium2 as pdfium
-from fastapi import FastAPI, HTTPException
+from fastapi import FastAPI, Header, HTTPException, UploadFile
 from fastapi.responses import StreamingResponse
 from PIL import Image
 
@@ -19,6 +24,9 @@ from engines import kraken as kraken_engine
 from engines import surya as surya_engine
 from models import OcrBlock, OcrRequest
 
+TRAINING_TOKEN = os.environ.get("TRAINING_TOKEN", "")
+KRAKEN_MODEL_PATH = os.environ.get("KRAKEN_MODEL_PATH", "/app/models/german_kurrent.mlmodel")
+
 logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger(__name__)
 
@@ -169,6 +177,107 @@ async def run_ocr_stream(request: OcrRequest):
     )
 
 
+def _check_training_token(x_training_token: str | None) -> None:
+    """Validate training token if TRAINING_TOKEN env var is set."""
+    if TRAINING_TOKEN and x_training_token != TRAINING_TOKEN:
+        raise HTTPException(status_code=403, detail="Invalid or missing X-Training-Token")
+
+
+def _validate_zip_entry(name: str, extract_dir: str) -> None:
+    """Reject ZIP Slip attacks: path traversal and absolute paths."""
+    if os.path.isabs(name) or name.startswith(".."):
+        raise HTTPException(status_code=400, detail=f"Unsafe ZIP entry: {name}")
+    resolved = os.path.realpath(os.path.join(extract_dir, name))
+    if not resolved.startswith(os.path.realpath(extract_dir)):
+        raise HTTPException(status_code=400, detail=f"ZIP Slip detected: {name}")
+
+
+def _rotate_backups(model_path: str, keep: int = 3) -> None:
+    """Keep only the last `keep` timestamped backups of the model."""
+    pattern = model_path + ".*.bak"
+    backups = sorted(glob.glob(pattern))
+    for old in backups[:-keep]:
+        try:
+            os.remove(old)
+        except OSError:
+            logger.warning("Could not remove old backup: %s", old)
+
+
+@app.post("/train")
+async def train_model(
+    file: UploadFile,
+    x_training_token: str | None = Header(default=None),
+):
+    """Fine-tune the Kurrent recognition model with uploaded training data.
+
+    Accepts a ZIP archive containing .png/.gt.txt training pairs exported
+    by the Java backend. Training mutates in-process model state — not safe
+    if the service is replicated.
+    """
+    _check_training_token(x_training_token)
+
+    if not _models_ready:
+        raise HTTPException(status_code=503, detail="Models not loaded yet")
+
+    zip_bytes = await file.read()
+
+    training_run_id = datetime.now(timezone.utc).strftime("%Y%m%dT%H%M%SZ")
+    log = logging.LoggerAdapter(logger, {"training_run_id": training_run_id})
+    log.info("Starting training run %s", training_run_id)
+
+    def _run_training() -> dict:
+        with tempfile.TemporaryDirectory() as tmp_dir:
+            # Extract ZIP with safety checks
+            with zipfile.ZipFile(io.BytesIO(zip_bytes)) as zf:
+                for entry in zf.namelist():
+                    _validate_zip_entry(entry, tmp_dir)
+                zf.extractall(tmp_dir)
+
+            log.info("Extracted %d ZIP entries to %s", len(os.listdir(tmp_dir)), tmp_dir)
+
+            # Run ketos train (transfer learning from existing model)
+            from kraken import ketos
+            ground_truth = glob.glob(os.path.join(tmp_dir, "*.gt.txt"))
+            if not ground_truth:
+                raise HTTPException(status_code=422, detail="No ground-truth files found in ZIP")
+
+            log.info("Training on %d ground-truth pairs", len(ground_truth))
+            output_model_path = os.path.join(tmp_dir, "fine_tuned.mlmodel")
+
+            result = ketos.train(
+                ground_truth=ground_truth,
+                load=KRAKEN_MODEL_PATH,
+                output=output_model_path,
+                format_type="path",
+            )
+
+            epochs = getattr(result, "epochs", None) or 0
+            loss = getattr(result, "best_loss", None)
+            accuracy = getattr(result, "best_accuracy", None)
+
+            log.info("Training complete — epochs=%s loss=%s accuracy=%s", epochs, loss, accuracy)
+
+            # Backup existing model and replace
+            if os.path.exists(KRAKEN_MODEL_PATH):
+                timestamp = datetime.now(timezone.utc).strftime("%Y%m%dT%H%M%SZ")
+                backup_path = f"{KRAKEN_MODEL_PATH}.{timestamp}.bak"
+                shutil.copy2(KRAKEN_MODEL_PATH, backup_path)
+                log.info("Backed up model to %s", backup_path)
+                _rotate_backups(KRAKEN_MODEL_PATH, keep=3)
+
+            shutil.move(output_model_path, KRAKEN_MODEL_PATH)
+            log.info("Replaced model at %s", KRAKEN_MODEL_PATH)
+
+            # Reload model in-process
+            kraken_engine.load_models()
+            log.info("Reloaded Kraken model in-process")
+
+            return {"loss": loss, "accuracy": accuracy, "epochs": epochs}
+
+    result = await asyncio.to_thread(_run_training)
+    return result
+
+
 async def _download_and_convert_pdf(url: str) -> list[Image.Image]:
     """Download a PDF from a presigned URL and convert each page to a PIL Image."""
     _validate_url(url)