feat(ocr): Add docTR OCR engine with metrics infrastructure

Add docTR as primary OCR engine with 2-tier sequential processing,
OCR metrics tracking, and simplified engine selection.

Features:
- docTR OCR engine with light+medium preprocessing tiers
- doctr_plus mode with early exit optimization (~65% fast path)
- OCR metrics dashboard with per-engine statistics
- User OCR preference persistence
- Parallel worker pool for OCR processing
- Cross-validation for extraction quality

Engine options: tesseract, doctr, doctr_plus (recommended), paddleocr

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

docs/OCR_MEMORY_SOLUTIONS_RESEARCH.md

@@ -0,0 +1,220 @@
+# OCR Memory Leak Solutions Research
+
+**Data**: 2026-01-01
+**Problema**: Worker OCR cade după 10 bonuri din cauza memory leak în WSL2
+**Mediu**: WSL2 pe Windows, 8GB RAM alocat, docTR + PyTorch
+
+---
+
+## Diagnosticul Problemei
+
+- **WSL RAM**: 7.5 GB (din 8GB configurat)
+- **WSL Swap**: 2 GB folosit complet (config 8GB nu se aplică?)
+- **Crash**: După ~10 PDF-uri procesate cu hybrid-doctr
+- **Eroare**: "A process in the process pool was terminated abruptly"
+
+---
+
+## Soluții Găsite (ordonate după prioritate)
+
+### ✅ Nivel 1: Environment Variables (ÎNCERCAT PRIMUL)
+
+```python
+# În ocr_worker_process.py sau la începutul aplicației
+import os
+os.environ["DOCTR_MULTIPROCESSING_DISABLE"] = "TRUE"
+os.environ["ONEDNN_PRIMITIVE_CACHE_CAPACITY"] = "1"
+```
+
+**Sursa**: [docTR Issue #1594](https://github.com/mindee/doctr/issues/1594)
+
+---
+
+### ✅ Nivel 2: maxtasksperchild în ProcessPool
+
+```python
+from concurrent.futures import ProcessPoolExecutor
+
+# Worker-ul repornește la fiecare N task-uri
+executor = ProcessPoolExecutor(max_workers=1, max_tasks_per_child=5)
+```
+
+**Efect**: Memoria se eliberează complet când worker-ul repornește
+**Cost**: +10-15s la fiecare 5 job-uri (reload model docTR)
+
+**Sursa**: [Python Multiprocessing Memory](https://www.pythontutorials.net/blog/memory-usage-keep-growing-with-python-s-multiprocessing-pool/)
+
+---
+
+### ✅ Nivel 3: Manual Memory Cleanup
+
+```python
+import gc
+import torch
+
+def process_image(image):
+    with torch.no_grad():  # Dezactivează gradient tracking
+        result = doctr_engine(image)
+
+    # Cleanup explicit
+    del image
+    gc.collect()
+
+    if torch.cuda.is_available():
+        torch.cuda.empty_cache()
+
+    return result
+```
+
+---
+
+### ✅ Nivel 4: WSL Memory Reclaim
+
+**În .wslconfig** (`C:\Users\<username>\.wslconfig`):
+```ini
+[wsl2]
+memory=8GB
+processors=2
+swap=8GB
+
+[experimental]
+autoMemoryReclaim=gradual
+```
+
+**Manual drop caches** (în WSL):
+```bash
+echo 1 | sudo tee /proc/sys/vm/drop_caches
+echo 1 | sudo tee /proc/sys/vm/compact_memory
+```
+
+**După modificare .wslconfig**:
+```powershell
+wsl --shutdown
+```
+
+**Sursa**: [Microsoft DevBlogs - Memory Reclaim](https://devblogs.microsoft.com/commandline/memory-reclaim-in-the-windows-subsystem-for-linux-2/)
+
+⚠️ **Atenție**: `autoMemoryReclaim` poate afecta Docker!
+
+---
+
+### ✅ Nivel 5: Procesare Secvențială cu Cleanup
+
+Restructurare `_process_hybrid_doctr()`:
+
+```python
+def _process_hybrid_doctr_memory_efficient(image, doctr_engine, preprocessor, extractor):
+    results = []
+
+    for pass_type in ['light', 'medium', 'heavy']:
+        # Preprocess
+        if pass_type == 'light':
+            processed = preprocessor.preprocess_light(image)
+        elif pass_type == 'medium':
+            processed = preprocessor.preprocess_medium(image)
+        else:
+            processed = preprocessor.preprocess_heavy(image)
+
+        # OCR
+        with torch.no_grad():
+            ocr_result = _doctr_recognize(doctr_engine, processed)
+
+        # Cleanup IMEDIAT
+        del processed
+        gc.collect()
+
+        if ocr_result:
+            extraction = extractor.extract(ocr_result.text)
+            results.append(extraction)
+
+            # Early exit dacă e suficient de bun
+            if _is_extraction_complete(extraction) and extraction.overall_confidence > 0.9:
+                break
+
+    return _merge_extractions(results)
+```
+
+---
+
+### ✅ Nivel 6: Alternative la ProcessPoolExecutor
+
+**Ray** (mai bun pentru ML):
+```python
+import ray
+
+@ray.remote
+def process_ocr(image_bytes):
+    # Processing...
+    return result
+
+# Ray gestionează memoria mai bine
+result = ray.get(process_ocr.remote(image_bytes))
+```
+
+**Dask**:
+```python
+from dask import delayed
+
+@delayed
+def process_ocr(image_bytes):
+    return result
+```
+
+**Sursa**: [Managing Memory Issues](https://eyxibnib.biz.id/2024/07/05/managing-memory-issues-with-pythons-threadpoolexecutor-and-processpoolexecutor/)
+
+---
+
+### ✅ Nivel 7: Downscale Imagini Mari
+
+```python
+def preprocess_with_size_limit(image, max_size=2000):
+    h, w = image.shape[:2]
+    if max(h, w) > max_size:
+        scale = max_size / max(h, w)
+        image = cv2.resize(image, None, fx=scale, fy=scale, interpolation=cv2.INTER_AREA)
+    return image
+```
+
+---
+
+## Referințe Complete
+
+### docTR Memory Issues
+- [Discussion #1422 - Memory Leak on inference](https://github.com/mindee/doctr/discussions/1422)
+- [Issue #1594 - CPU memory usage increase](https://github.com/mindee/doctr/issues/1594)
+
+### PyTorch Multiprocessing
+- [PyTorch Forums - Memory leak with multiprocessing](https://discuss.pytorch.org/t/memory-leak-with-multiprocessing/209512)
+- [Issue #44156 - CUDA memory leak](https://github.com/pytorch/pytorch/issues/44156)
+- [Issue #13246 - DataLoader memory replication](https://github.com/pytorch/pytorch/issues/13246)
+
+### Python ProcessPoolExecutor
+- [CPython Issue #90943 - Exception memory leak](https://github.com/python/cpython/issues/90943)
+- [NumPy Issue #12122 - Memory leak with ProcessPoolExecutor](https://github.com/numpy/numpy/issues/12122)
+
+### WSL2 Memory
+- [Microsoft - Memory Reclaim in WSL2](https://devblogs.microsoft.com/commandline/memory-reclaim-in-the-windows-subsystem-for-linux-2/)
+- [WSL Issue #4166 - Massive RAM consumption](https://github.com/microsoft/WSL/issues/4166)
+- [Limiting Memory Usage in WSL2](https://www.aleksandrhovhannisyan.com/blog/limiting-memory-usage-in-wsl-2/)
+
+---
+
+## Plan de Implementare
+
+1. ✅ **Pas 1**: Adaugă env vars (DOCTR_MULTIPROCESSING_DISABLE, ONEDNN)
+2. ✅ **Pas 2**: Adaugă maxtasksperchild=5 în worker pool
+3. ⏳ **Pas 3**: Testează - dacă tot cade, adaugă gc.collect() explicit
+4. ⏳ **Pas 4**: Dacă tot cade, restructurează procesarea secvențială
+5. ⏳ **Pas 5**: Dacă tot cade, crește memoria WSL sau folosește Ray
+
+---
+
+## Notă Importantă
+
+Swap-ul din .wslconfig poate să nu se aplice corect. După modificări:
+```powershell
+wsl --shutdown
+# Apoi redeschide WSL
+```
+
+Verifică cu `free -h` că noile setări sunt aplicate.

docs/OCR_TEST_RESULTS.md

@@ -0,0 +1,106 @@
+# OCR Test Results - docTR+ Engine
+
+**Date:** 2026-01-02 | **Receipts:** 26 | **Test:** Sequential
+
+## Summary Comparison
+
+| Workers | Avg | Total | Mem Used | Mem Avail |
+|---------|-----|-------|----------|-----------|
+| 1       | 6.8s | 176s | 3.2GB | 4.1GB |
+| 2       | 7.2s | 187s | 3.1GB | 4.1GB |
+| 3       | 6.8s | 176s | 3.9GB | 3.3GB |
+
+**Success Rate:** 80.8% (21/26) - same for all configs
+
+**Note:** For sequential tests, 1 worker ≈ 3 workers speed!
+Multiple workers only help with parallel requests.
+
+## Detailed Results (1 Worker)
+
+| # | Receipt | Time | Tier | Result | Notes |
+|---|---------|------|------|--------|-------|
+| 01 | abonament kineterra | 6.8s | T1 | ✓ | 97% |
+| 02 | benzina 14 august | 6.0s | T1 | ✓ | 83% |
+| 03 | benzina 27 octombrie | 5.9s | T1 | ✓ | 83% |
+| 04 | igiena 11 octombrie | 7.7s | T1 | ✓ | 97% |
+| 05 | igiena 14 dec five-holding | 11.5s | T1+T2 | ✗ | TOTAL ±1 |
+| 06 | rechizite 12 dec pictus | 5.9s | T1 | ✓ | 97% |
+| 07 | benzina 10 mai 2025 | 5.1s | T1 | ✓ | 83% |
+| 08 | brick consumabil 604 50% | 4.8s | T1 | ✓ | 97% |
+| 09 | benzina 13 septembrie | 4.9s | T1 | ✓ | 83% |
+| 10 | brick consumabile 604 | 5.3s | T1 | ✓ | 97% |
+| 11 | benzina 20 dec | 5.8s | T1 | ✓ | 79% |
+| 12 | bon fiscal Dedeman | 5.7s | T1 | ✓ | 90% |
+| 13 | factura Dedeman | 6.8s | T1 | ✓ | 97% |
+| 14 | benzina 13 iulie | 5.7s | T1 | ✓ | 95% |
+| 15 | best print stampila | 4.5s | T1 | ✓ | 94% |
+| 16 | electrobering telecomanda | 4.8s | T1 | ✓ | 97% |
+| 17 | brick igiena 8 oct | 11.9s | T1+T2 | ✗ | TOTAL/CUI |
+| 18 | gama ink refill toner | 5.9s | T1 | ✓ | 94% |
+| 19 | kineterra fizioterapie | 4.6s | T1 | ✓ | 97% |
+| 20 | brick igiena 1 sept | 12.5s | T1+T2 | ✗ | ALL None |
+| 21 | kineterra abonament | 5.6s | T1 | ✓ | 97% |
+| 22 | brick igiena electrice | 15.9s | T1+T2 | ✗ | DATE None |
+| 23 | electrobering igiena | 4.4s | T1 | ✓ | 97% |
+| 24 | Lidl papetarie 604 | 5.8s | T1 | ✓ | 87% |
+| 25 | brick igiena 604 | 6.8s | T1 | ✗ | DATE ±1 |
+| 26 | unlimited duplicat | 4.8s | T1 | ✓ | 86% |
+
+## Time Comparison by Receipt
+
+| # | Receipt | 1W | 2W | 3W |
+|---|---------|----|----|-----|
+| 01 | abonament kineterra | 6.8s | 6.7s | 5.8s |
+| 02 | benzina 14 august | 6.0s | 5.5s | 5.8s |
+| 03 | benzina 27 octombrie | 5.9s | 5.9s | 5.7s |
+| 04 | igiena 11 octombrie | 7.7s | 8.9s | 7.4s |
+| 05 | igiena 14 dec (FAIL) | 11.5s | 12.3s | 11.9s |
+| 06 | rechizite pictus | 5.9s | 5.9s | 5.7s |
+| 07 | benzina 10 mai | 5.1s | 6.0s | 5.8s |
+| 08 | brick 50% | 4.8s | 5.9s | 5.5s |
+| 09 | benzina 13 sept | 4.9s | 5.9s | 5.3s |
+| 10 | brick consumabile | 5.3s | 5.7s | 5.7s |
+| 11 | benzina 20 dec | 5.8s | 5.4s | 5.8s |
+| 12 | bon Dedeman | 5.7s | 5.9s | 5.8s |
+| 13 | factura Dedeman | 6.8s | 6.9s | 6.8s |
+| 14 | benzina 13 iulie | 5.7s | 6.1s | 5.4s |
+| 15 | best print | 4.5s | 5.8s | 4.8s |
+| 16 | electrobering | 4.8s | 4.2s | 4.7s |
+| 17 | brick 8 oct (FAIL) | 11.9s | 13.1s | 12.0s |
+| 18 | gama ink | 5.9s | 5.9s | 4.7s |
+| 19 | kineterra fizioterapie | 4.6s | 5.9s | 4.8s |
+| 20 | brick 1 sept (FAIL) | 12.5s | 13.2s | 13.1s |
+| 21 | kineterra abonament | 5.6s | 4.9s | 4.8s |
+| 22 | brick electrice (FAIL) | 15.9s | 17.0s | 15.5s |
+| 23 | electrobering igiena | 4.4s | 5.4s | 5.0s |
+| 24 | Lidl papetarie | 5.8s | 6.9s | 5.8s |
+| 25 | brick 604 (FAIL) | 6.8s | 6.5s | 6.9s |
+| 26 | unlimited duplicat | 4.8s | 5.8s | 5.0s |
+|---|---------|----|----|-----|
+| **AVG** | | **6.8s** | **7.2s** | **6.8s** |
+| **TOTAL** | | **176s** | **187s** | **176s** |
+
+## Tier Analysis
+
+- **T1 only (early exit):** 21 receipts (~5-6s)
+- **T1+T2 (full):** 5 receipts (~12-16s)
+
+## Failures (5)
+
+| Receipt | Issue | Fixable |
+|---------|-------|---------|
+| igiena 14 dec | TOTAL ±1 | No |
+| brick 8 oct | TOTAL/CUI | Maybe |
+| brick 1 sept | ALL None | No (bad doc) |
+| brick electrice | DATE None | Maybe |
+| brick 604 | DATE ±1 | No |
+
+## Recommendation
+
+```
+OCR_WORKERS=1                    # Best for sequential, saves RAM
+OCR_WORKERS=2                    # For parallel requests (production)
+OCR_MAX_TASKS_PER_CHILD=0        # No restart
+```
+
+**For 8GB RAM:** Use 1-2 workers max