OCR-to-RAG pipeline for life science screenshots

PaddleOCR with preprocessing, scispaCy NER, figure/table detection, citation extraction, and chunked Markdown output with frontmatter. Includes watch mode and notebook reprocessing.
2026-07-18 19:34:22 +00:00
parent 012549b4bc
commit a32b7508c7
68 changed files with 8708 additions and 0 deletions
--- a/.env.example
+++ b/.env.example
@@ -0,0 +1,26 @@
+# OCR Pipeline Configuration
+# Copy to .env and modify as needed
+
+# Logging
+LOG_LEVEL=INFO
+LOG_FILE=./data/logs/ocr_pipeline.log
+
+# Input paths (override config.yaml)
+INPUT_PATHS=~/Pictures,/mnt/storage3/aman/screenshots
+
+# Processing
+WORKERS=4
+OCR_ENGINE=paddleocr
+USE_GPU=false
+
+# Output
+OUTPUT_DIR=./data/ocr_output
+ORGANIZE_BY=date_run
+
+# Watch mode
+WATCH_ENABLED=true
+WATCH_DB_PATH=./data/processed_files.db
+
+# Optional: GROBID for citation parsing
+GROBID_ENABLED=false
+GROBID_URL=http://localhost:8070
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@@ -0,0 +1,27 @@
+repos:
+  - repo: https://github.com/astral-sh/ruff-pre-commit
+    rev: v0.2.0
+    hooks:
+      - id: ruff
+        args: [--fix]
+      - id: ruff-format
+
+  - repo: https://github.com/pre-commit/mirrors-mypy
+    rev: v1.8.0
+    hooks:
+      - id: mypy
+        additional_dependencies: [pydantic, pyyaml, typer, rich, structlog, numpy, opencv-python-headless]
+
+  - repo: https://github.com/pre-commit/pre-commit-hooks
+    rev: v4.5.0
+    hooks:
+      - id: trailing-whitespace
+      - id: end-of-file-fixer
+      - id: check-yaml
+      - id: check-toml
+      - id: check-added-large-files
+
+  - repo: https://github.com/pycqa/isort
+    rev: 5.13.2
+    hooks:
+      - id: isort
--- a/26
+++ b/26
@@ -0,0 +1,26 @@
+FROM python:3.11-slim
+
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    tesseract-ocr \
+    libtesseract-dev \
+    poppler-utils \
+    libgl1 \
+    libglib2.0-0 \
+    libsm6 \
+    libxext6 \
+    libxrender-dev \
+    libgomp1 \
+    && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /app
+
+COPY pyproject.toml ./
+RUN pip install --no-cache-dir uv && uv sync --frozen
+
+COPY src ./src
+COPY config.yaml ./
+
+ENV PYTHONPATH=/app/src
+ENV OMP_NUM_THREADS=4
+
+ENTRYPOINT ["ocr-pipeline"]
--- a/README.md
+++ b/README.md
@@ -0,0 +1,175 @@
+# OCR Pipeline for Life Science Screenshots
+
+Turns scientific screenshots into RAG-ready Markdown, with metadata attached.
+
+- OCR via PaddleOCR, falls back to Tesseract
+- Preprocessing: deskew, denoise, CLAHE, line removal
+- Figure/table detection
+- Entity extraction with scispaCy
+- Citation matching
+- Chunking for embedding
+
+## Quick start
+
+```bash
+uv sync           # or: pip install -e .
+pre-commit install
+
+ocr-pipeline run                  # process screenshots
+ocr-pipeline watch                # watch a folder
+ocr-pipeline reprocess ocr_sc.ipynb   # pull screenshots out of a notebook
+ocr-pipeline status
+```
+
+## Docker
+
+```bash
+docker build -t ocr-pipeline .
+
+docker run -v ~/Pictures:/data/screenshots -v ./data:/app/data ocr-pipeline run
+docker run -v ~/Pictures:/data/screenshots -v ./data:/app/data ocr-pipeline watch
+```
+
+## Config
+
+<details>
+<summary><code>config.yaml</code></summary>
+
+```yaml
+input:
+  paths: ["~/Pictures", "/mnt/storage3/aman/screenshots"]
+  patterns: ["SCR-*.png", "*.jpg", "*.jpeg", "*.tiff"]
+  recursive: true
+
+ocr:
+  engine: "paddleocr"  # paddleocr | tesseract | auto
+  languages: ["en", "latin"]
+  use_gpu: false
+  preprocess:
+    deskew: true
+    denoise: true
+    clahe: true
+    adaptive_threshold: true
+    remove_lines: true
+
+processing:
+  workers: 4
+  batch_size: 10
+  retry_attempts: 2
+
+detectors:
+  figures:
+    enabled: true
+    confidence_threshold: 0.7
+  tables:
+    enabled: true
+    confidence_threshold: 0.7
+
+entities:
+  enabled: true
+  model: "en_core_sci_lg"
+
+citations:
+  regex_enabled: true
+  grobid_enabled: false  # set true if you're running a GROBID server
+
+chunking:
+  chunk_size: 1000
+  chunk_overlap: 200
+
+output:
+  base_directory: "./data/ocr_output"
+  organize_by: "date_run"  # date_run | source_dir | flat
+  frontmatter:
+    - source_path
+    - source_hash
+    - timestamp
+    - ocr_engine
+    - ocr_confidence_mean
+    - language
+    - detected_entities
+    - has_figures
+    - has_tables
+    - citations_found
+    - chunk_index
+    - total_chunks
+
+watch:
+  enabled: true
+  debounce_seconds: 5
+  db_path: "./data/processed_files.db"
+```
+
+</details>
+
+## Output
+
+Each chunk is a Markdown file with YAML frontmatter:
+
+```markdown
+---
+source_path: "/Users/Aman/Pictures/SCR-20250115-gel.png"
+source_hash: "a1b2c3d4e5f6..."
+timestamp: "2025-01-15T10:30:00Z"
+ocr_engine: "paddleocr"
+ocr_confidence_mean: 0.91
+language: "en"
+detected_entities: ["GENE", "PROTEIN", "CHEMICAL"]
+has_figures: true
+has_tables: false
+citations_found: ["DOI:10.1038/nature12345", "PMID:12345678"]
+chunk_index: 0
+total_chunks: 2
+---
+
+# Screenshot: SCR-20250115-gel.png
+
+## Figures
+
+### Figure 1
+- BBox: [100, 200, 800, 600]
+- Confidence: 0.92
+- Caption: "Western blot showing BRCA1 expression..."
+
+## Detected Entities
+
+- BRCA1
+- CRISPR
+- β-actin
+
+## Citations
+
+- DOI: 10.1038/nature12345
+- PMID: 12345678
+
+## Extracted Text
+
+**Western Blot Analysis of BRCA1 Expression**
+
+Lane 1: WT control
+Lane 2: BRCA1 KO (CRISPR)
+Lane 3: BRCA1 KO + pBRCA1-WT rescue
+Lane 4: BRCA1 KO + pBRCA1-C61G mutant
+
+Anti-BRCA1 (1:1000), Anti-β-actin (1:5000)
+```
+
+Files land in `data/ocr_output/<date>/run_NNN/`.
+
+## Life science specifics
+
+Gene/protein names go through scispaCy's `en_core_sci_lg`. Chemical formulas and units (µM, ng/mL, kb/Mb/Gb, °C, ×g) get normalized, scientific notation gets cleaned up (`1.5×10⁻³` → `1.5×10^-3`), and gel/blot figures get their captions pulled out separately. Citations are matched by regex for DOI, PMID, arXiv, PMC, and ISBN.
+
+## Models
+
+First run downloads what it needs, cached in `~/.cache/ocr_pipeline/`:
+
+- PaddleOCR models (~200MB)
+- scispaCy `en_core_sci_lg` (~800MB)
+- Table Transformer (~500MB)
+- LayoutParser PubLayNet (~300MB)
+
+## TODO
+
+- [ ] Build a knowledge graph from extracted entities/citations
+
--- a/config.yaml
+++ b/config.yaml
@@ -0,0 +1,105 @@
+input:
+  paths:
+    - "~/Pictures"
+    - "/mnt/storage3/aman/screenshots"
+  patterns:
+    - "SCR-*.png"
+    - "*.jpg"
+    - "*.jpeg"
+    - "*.tiff"
+    - "*.bmp"
+  recursive: true
+
+ocr:
+  engine: "paddleocr"  # "paddleocr" | "tesseract" | "auto"
+  languages: ["en"]  # For tesseract: eng, lat, deu, fra, spa, ita, por, chi_sim, jpn, kor
+  use_gpu: false
+  use_angle_cls: true
+  det_db_thresh: 0.3
+  det_db_box_thresh: 0.6
+  det_db_unclip_ratio: 1.5
+  rec_batch_num: 6
+  cpu_threads: 4
+  preprocess:
+    deskew: true
+    denoise: true
+    clahe: true
+    adaptive_threshold: true
+    remove_lines: true
+    max_dimension: 4096
+
+processing:
+  workers: 4
+  batch_size: 10
+  retry_attempts: 2
+  timeout_per_image: 120
+  skip_existing: true
+
+detectors:
+  figures:
+    enabled: true
+    model: "lp://PubLayNet/faster_rcnn_R_50_FPN_3x/config"
+    confidence_threshold: 0.7
+  tables:
+    enabled: true
+    model: "microsoft/table-transformer-detection"
+    confidence_threshold: 0.7
+  captions:
+    enabled: true
+    model: "lp://PubLayNet/faster_rcnn_R_50_FPN_3x/config"
+    confidence_threshold: 0.5
+
+entities:
+  enabled: true
+  model: "en_core_sci_lg"
+  types:
+    - "GENE"
+    - "PROTEIN"
+    - "CHEMICAL"
+    - "SPECIES"
+    - "DISEASE"
+    - "CELL_LINE"
+    - "ORGANISM"
+    - "CELL_TYPE"
+  merge_entities: true
+
+citations:
+  regex_enabled: true
+  grobid_enabled: false
+  grobid_url: "http://localhost:8070"
+  grobid_timeout: 30
+
+chunking:
+  chunk_size: 1000
+  chunk_overlap: 200
+  separators: ["\n\n", "\n", ". ", " ", ""]
+  keep_separator: true
+
+output:
+  base_directory: "./data/ocr_output"
+  format: "markdown"
+  organize_by: "date_run"  # "date_run" | "source_dir" | "flat"
+  frontmatter:
+    - "source_path"
+    - "source_hash"
+    - "timestamp"
+    - "ocr_engine"
+    - "ocr_confidence_mean"
+    - "language"
+    - "detected_entities"
+    - "has_figures"
+    - "has_tables"
+    - "citations_found"
+    - "chunk_index"
+    - "total_chunks"
+
+watch:
+  enabled: true
+  debounce_seconds: 5
+  ignore_patterns:
+    - ".DS_Store"
+    - "*.tmp"
+    - "*.partial"
+    - "*.crdownload"
+  db_path: "./data/processed_files.db"
+  poll_interval: 1.0
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -0,0 +1,134 @@
+[project]
+name = "ocr-pipeline"
+version = "0.1.0"
+description = "OCR pipeline for life science screenshots - RAG ready"
+readme = "README.md"
+requires-python = ">=3.11"
+dependencies = [
+    # Core
+    "typer[all]>=0.9.0",
+    "pydantic>=2.6.0",
+    "pydantic-settings>=2.2.0",
+    "pyyaml>=6.0.1",
+    "structlog>=24.1.0",
+    "rich>=13.7.0",
+    "tqdm>=4.66.0",
+
+    # Image processing
+    "opencv-python-headless>=4.9.0",
+    "pillow>=10.2.0",
+    "numpy>=1.26.0",
+    "scikit-image>=0.22.0",
+
+    # OCR engines
+    "paddleocr>=2.7.0",
+    "pytesseract>=0.3.10",
+
+    # Table detection
+    "torch>=2.2.0",
+    "torchvision>=0.17.0",
+    "transformers>=4.38.0",
+    "layoutparser>=0.3.0",
+
+    # NLP
+    "langchain-text-splitters>=0.0.2",
+    "spacy>=3.7.0",
+
+    # Database
+    "sqlite-utils>=3.37.0",
+
+    # File watching
+    "watchdog>=3.0.0",
+
+    # Utilities
+    "python-slugify>=8.0.0",
+    "xxhash>=3.4.0",
+    "python-magic>=0.4.27",
+]
+
+[project.optional-dependencies]
+dev = [
+    "pytest>=8.0.0",
+    "pytest-cov>=4.1.0",
+    "pytest-mock>=3.12.0",
+    "ruff>=0.2.0",
+    "mypy>=1.8.0",
+    "pre-commit>=3.6.0",
+]
+grobid = [
+    "grobid-client>=0.8.0",
+]
+full = [
+    "ocr-pipeline[dev]",
+    "ocr-pipeline[grobid]",
+]
+
+[project.entry-points.console_scripts]
+ocr-pipeline = "ocr_pipeline.cli:app"
+
+[tool.uv]
+dev-dependencies = [
+    "pytest>=8.0.0",
+    "pytest-cov>=4.1.0",
+    "pytest-mock>=3.12.0",
+    "ruff>=0.2.0",
+    "mypy>=1.8.0",
+    "pre-commit>=3.6.0",
+]
+
+[tool.ruff]
+line-length = 100
+target-version = "py311"
+select = [
+    "E",   # pycodestyle
+    "F",   # pyflakes
+    "I",   # isort
+    "N",   # pep8-naming
+    "W",   # pycodestyle warnings
+    "UP",  # pyupgrade
+    "B",   # flake8-bugbear
+    "C4",  # flake8-comprehensions
+    "T20", # flake8-print
+]
+ignore = [
+    "E501",  # line too long (handled by formatter)
+    "B008",  # function calls in default argument
+]
+fixable = ["I", "E", "W", "F"]
+unfixable = ["N", "B", "C4", "T20", "UP"]
+
+[tool.ruff.format]
+quote-style = "double"
+indent-style = "space"
+skip-magic-trailing-comma = false
+line-ending = "lf"
+
+[tool.mypy]
+python_version = "3.11"
+warn_return_any = true
+warn_unused_configs = true
+disallow_untyped_defs = true
+disallow_incomplete_defs = true
+check_untyped_defs = true
+no_implicit_optional = true
+strict_equality = true
+show_error_codes = true
+pretty = true
+
+[tool.pytest.ini_options]
+testpaths = ["tests"]
+python_files = ["test_*.py"]
+python_classes = ["Test*"]
+python_functions = ["test_*"]
+addopts = "-v --tb=short"
+filterwarnings = [
+    "ignore::DeprecationWarning",
+    "ignore::UserWarning:torch",
+    "ignore::UserWarning:transformers",
+]
+
+[tool.pre-commit-hooks]
+ruff = { id = "ruff", args = ["check", "--fix"] }
+ruff-format = { id = "ruff", args = ["format"] }
+mypy = { id = "mypy" }
+pytest = { id = "pytest", args = ["-x"] }
--- a/src/ocr_pipeline.egg-info/PKG-INFO
+++ b/src/ocr_pipeline.egg-info/PKG-INFO
@@ -0,0 +1,276 @@
+Metadata-Version: 2.4
+Name: ocr-pipeline
+Version: 0.1.0
+Summary: OCR pipeline for life science screenshots - RAG ready
+Requires-Python: >=3.11
+Description-Content-Type: text/markdown
+Requires-Dist: typer[all]>=0.9.0
+Requires-Dist: pydantic>=2.6.0
+Requires-Dist: pydantic-settings>=2.2.0
+Requires-Dist: pyyaml>=6.0.1
+Requires-Dist: structlog>=24.1.0
+Requires-Dist: rich>=13.7.0
+Requires-Dist: tqdm>=4.66.0
+Requires-Dist: opencv-python-headless>=4.9.0
+Requires-Dist: pillow>=10.2.0
+Requires-Dist: numpy>=1.26.0
+Requires-Dist: scikit-image>=0.22.0
+Requires-Dist: paddleocr>=2.7.0
+Requires-Dist: pytesseract>=0.3.10
+Requires-Dist: torch>=2.2.0
+Requires-Dist: torchvision>=0.17.0
+Requires-Dist: transformers>=4.38.0
+Requires-Dist: layoutparser>=0.3.0
+Requires-Dist: langchain-text-splitters>=0.0.2
+Requires-Dist: spacy>=3.7.0
+Requires-Dist: sqlite-utils>=3.37.0
+Requires-Dist: watchdog>=3.0.0
+Requires-Dist: python-slugify>=8.0.0
+Requires-Dist: xxhash>=3.4.0
+Requires-Dist: python-magic>=0.4.27
+Provides-Extra: dev
+Requires-Dist: pytest>=8.0.0; extra == "dev"
+Requires-Dist: pytest-cov>=4.1.0; extra == "dev"
+Requires-Dist: pytest-mock>=3.12.0; extra == "dev"
+Requires-Dist: ruff>=0.2.0; extra == "dev"
+Requires-Dist: mypy>=1.8.0; extra == "dev"
+Requires-Dist: pre-commit>=3.6.0; extra == "dev"
+Provides-Extra: grobid
+Requires-Dist: grobid-client>=0.8.0; extra == "grobid"
+Provides-Extra: full
+Requires-Dist: ocr-pipeline[dev]; extra == "full"
+Requires-Dist: ocr-pipeline[grobid]; extra == "full"
+
+# OCR Pipeline for Life Science Screenshots
+
+A production-ready OCR pipeline that extracts text from scientific screenshots and converts them into RAG-ready Markdown files with rich metadata.
+
+## Features
+
+- **Multi-engine OCR**: PaddleOCR (primary) with Tesseract fallback
+- **Image preprocessing**: Deskewing, denoising, CLAHE contrast enhancement, line removal
+- **Parallel processing**: Multi-process worker pool for throughput
+- **Scientific entity recognition**: Genes, proteins, chemicals, species, diseases, cell lines via scispaCy
+- **Figure & table detection**: LayoutParser + Table Transformer
+- **Citation extraction**: DOI, PMID, arXiv, PMC, ISBN via regex + optional GROBID
+- **Semantic chunking**: LangChain recursive splitter optimized for scientific text
+- **RAG-ready output**: Markdown with YAML frontmatter (source, hash, timestamp, entities, confidence)
+- **Watch mode**: File system monitoring with SQLite persistence for incremental processing
+- **Notebook migration**: Reprocess screenshots from existing Jupyter notebooks
+- **Life-science focused**: Handles scientific notation, units, gene symbols, chemical formulas
+
+## Quick Start
+
+```bash
+# Install with uv (recommended)
+uv sync
+
+# Or with pip
+pip install -e .
+
+# Install pre-commit hooks
+pre-commit install
+
+# Run on your screenshots
+ocr-pipeline run
+
+# Watch for new screenshots
+ocr-pipeline watch
+
+# Reprocess from notebook
+ocr-pipeline reprocess ocr_sc.ipynb
+
+# Check status
+ocr-pipeline status
+```
+
+## Configuration
+
+Edit `config.yaml`:
+
+```yaml
+input:
+  paths: ["~/Pictures", "/mnt/storage3/aman/screenshots"]
+  patterns: ["SCR-*.png", "*.jpg", "*.jpeg", "*.tiff"]
+  recursive: true
+
+ocr:
+  engine: "paddleocr"  # paddleocr | tesseract | auto
+  languages: ["en", "latin"]
+  use_gpu: false
+  preprocess:
+    deskew: true
+    denoise: true
+    clahe: true
+    adaptive_threshold: true
+    remove_lines: true
+
+processing:
+  workers: 4
+  batch_size: 10
+  retry_attempts: 2
+
+detectors:
+  figures:
+    enabled: true
+    confidence_threshold: 0.7
+  tables:
+    enabled: true
+    confidence_threshold: 0.7
+
+entities:
+  enabled: true
+  model: "en_core_sci_lg"  # scispaCy large model
+
+citations:
+  regex_enabled: true
+  grobid_enabled: false  # Set true if running GROBID server
+
+chunking:
+  chunk_size: 1000
+  chunk_overlap: 200
+
+output:
+  base_directory: "./data/ocr_output"
+  organize_by: "date_run"  # date_run | source_dir | flat
+  frontmatter:
+    - source_path
+    - source_hash
+    - timestamp
+    - ocr_engine
+    - ocr_confidence_mean
+    - language
+    - detected_entities
+    - has_figures
+    - has_tables
+    - citations_found
+    - chunk_index
+    - total_chunks
+
+watch:
+  enabled: true
+  debounce_seconds: 5
+  db_path: "./data/processed_files.db"
+```
+
+## Output Format
+
+Each chunk produces a Markdown file with YAML frontmatter:
+
+```markdown
+---
+source_path: "/Users/Aman/Pictures/SCR-20250115-gel.png"
+source_hash: "a1b2c3d4e5f6..."
+timestamp: "2025-01-15T10:30:00Z"
+ocr_engine: "paddleocr"
+ocr_confidence_mean: 0.91
+language: "en"
+detected_entities: ["GENE", "PROTEIN", "CHEMICAL"]
+has_figures: true
+has_tables: false
+citations_found: ["DOI:10.1038/nature12345", "PMID:12345678"]
+chunk_index: 0
+total_chunks: 2
+---
+
+# Screenshot: SCR-20250115-gel.png
+
+## Figures
+
+### Figure 1
+- **BBox**: [100, 200, 800, 600]
+- **Confidence**: 0.92
+- **Caption**: "Western blot showing BRCA1 expression..."
+
+## Detected Entities
+
+- BRCA1
+- CRISPR
+- β-actin
+
+## Citations
+
+- DOI: 10.1038/nature12345
+- PMID: 12345678
+
+## Extracted Text
+
+**Western Blot Analysis of BRCA1 Expression**
+
+Lane 1: WT control
+Lane 2: BRCA1 KO (CRISPR)
+Lane 3: BRCA1 KO + pBRCA1-WT rescue
+Lane 4: BRCA1 KO + pBRCA1-C61G mutant
+
+Anti-BRCA1 (1:1000), Anti-β-actin (1:5000)
+```
+
+## Directory Structure
+
+```
+data/ocr_output/
+├── 2025-01-15/
+│   ├── run_001/
+│   │   ├── SCR-20250115-gel_chunk_000.md
+│   │   └── SCR-20250115-gel_chunk_001.md
+│   └── run_002/
+│       └── ...
+└── 2025-01-16/
+    └── run_001/
+        └── ...
+```
+
+## Docker
+
+```bash
+# Build
+docker build -t ocr-pipeline .
+
+# Run once
+docker run -v ~/Pictures:/data/screenshots -v ./data:/app/data ocr-pipeline run
+
+# Watch mode
+docker run -v ~/Pictures:/data/screenshots -v ./data:/app/data ocr-pipeline watch
+```
+
+## Life Science Optimizations
+
+| Feature | Implementation |
+|---------|----------------|
+| Gene/Protein names | scispaCy `en_core_sci_lg` NER |
+| Chemical formulas | Regex + unit normalization (µM, ng/mL, kb, etc.) |
+| Scientific notation | `1.5×10⁻³` → `1.5×10^-3` |
+| Gel/blot lanes | Figure detection + caption extraction |
+| Citations | DOI, PMID, arXiv, PMC, ISBN patterns |
+| Units | µM, ng/mL, kb/Mb/Gb, °C, ×g, etc. |
+
+## Development
+
+```bash
+# Install dev dependencies
+uv sync --dev
+
+# Run tests
+pytest
+
+# Lint
+ruff check .
+ruff format .
+
+# Type check
+mypy src/
+```
+
+## Model Downloads
+
+First run downloads models automatically:
+- PaddleOCR detection/recognition models (~200MB)
+- scispaCy `en_core_sci_lg` (~800MB)
+- Table Transformer (~500MB)
+- LayoutParser PubLayNet (~300MB)
+
+Cache location: `~/.cache/ocr_pipeline/`
+
+## License
+
+MIT
--- a/src/ocr_pipeline.egg-info/SOURCES.txt
+++ b/src/ocr_pipeline.egg-info/SOURCES.txt
@@ -0,0 +1,35 @@
+README.md
+pyproject.toml
+src/ocr_pipeline/__init__.py
+src/ocr_pipeline/cli.py
+src/ocr_pipeline/config.py
+src/ocr_pipeline/pipeline.py
+src/ocr_pipeline/watch.py
+src/ocr_pipeline.egg-info/PKG-INFO
+src/ocr_pipeline.egg-info/SOURCES.txt
+src/ocr_pipeline.egg-info/dependency_links.txt
+src/ocr_pipeline.egg-info/entry_points.txt
+src/ocr_pipeline.egg-info/requires.txt
+src/ocr_pipeline.egg-info/top_level.txt
+src/ocr_pipeline/citations/__init__.py
+src/ocr_pipeline/citations/extractor.py
+src/ocr_pipeline/citations/regex_extractor.py
+src/ocr_pipeline/detectors/__init__.py
+src/ocr_pipeline/detectors/figures.py
+src/ocr_pipeline/detectors/tables.py
+src/ocr_pipeline/ocr/__init__.py
+src/ocr_pipeline/ocr/engine.py
+src/ocr_pipeline/ocr/parallel.py
+src/ocr_pipeline/output/__init__.py
+src/ocr_pipeline/output/markdown.py
+src/ocr_pipeline/postprocess/__init__.py
+src/ocr_pipeline/postprocess/chunk.py
+src/ocr_pipeline/postprocess/clean.py
+src/ocr_pipeline/postprocess/entities.py
+src/ocr_pipeline/utils/__init__.py
+src/ocr_pipeline/utils/db.py
+src/ocr_pipeline/utils/logging.py
+src/ocr_pipeline/utils/migrate.py
+src/ocr_pipeline/watch/__init__.py
+src/ocr_pipeline/watch/watcher.py
+tests/test_postprocess.py
--- a/src/ocr_pipeline.egg-info/dependency_links.txt
+++ b/src/ocr_pipeline.egg-info/dependency_links.txt
@@ -0,0 +1 @@
+
--- a/src/ocr_pipeline.egg-info/entry_points.txt
+++ b/src/ocr_pipeline.egg-info/entry_points.txt
@@ -0,0 +1,2 @@
+[console_scripts]
+ocr-pipeline = ocr_pipeline.cli:app
--- a/src/ocr_pipeline.egg-info/requires.txt
+++ b/src/ocr_pipeline.egg-info/requires.txt
@@ -0,0 +1,39 @@
+typer[all]>=0.9.0
+pydantic>=2.6.0
+pydantic-settings>=2.2.0
+pyyaml>=6.0.1
+structlog>=24.1.0
+rich>=13.7.0
+tqdm>=4.66.0
+opencv-python-headless>=4.9.0
+pillow>=10.2.0
+numpy>=1.26.0
+scikit-image>=0.22.0
+paddleocr>=2.7.0
+pytesseract>=0.3.10
+torch>=2.2.0
+torchvision>=0.17.0
+transformers>=4.38.0
+layoutparser>=0.3.0
+langchain-text-splitters>=0.0.2
+spacy>=3.7.0
+sqlite-utils>=3.37.0
+watchdog>=3.0.0
+python-slugify>=8.0.0
+xxhash>=3.4.0
+python-magic>=0.4.27
+
+[dev]
+pytest>=8.0.0
+pytest-cov>=4.1.0
+pytest-mock>=3.12.0
+ruff>=0.2.0
+mypy>=1.8.0
+pre-commit>=3.6.0
+
+[full]
+ocr-pipeline[dev]
+ocr-pipeline[grobid]
+
+[grobid]
+grobid-client>=0.8.0
--- a/src/ocr_pipeline.egg-info/top_level.txt
+++ b/src/ocr_pipeline.egg-info/top_level.txt
@@ -0,0 +1 @@
+ocr_pipeline
--- a/src/ocr_pipeline/init.py
+++ b/src/ocr_pipeline/init.py
@@ -0,0 +1,12 @@
+from __future__ import annotations
+
+__version__ = "0.1.0"
+
+from .cli import main
+from .pipeline import OCRPipeline, PipelineResult
+
+__all__ = [
+    "OCRPipeline",
+    "PipelineResult",
+    "main",
+]
--- a/src/ocr_pipeline/pycache/init.cpython-313.pyc
+++ b/src/ocr_pipeline/pycache/init.cpython-313.pyc
--- a/src/ocr_pipeline/pycache/cli.cpython-313.pyc
+++ b/src/ocr_pipeline/pycache/cli.cpython-313.pyc
--- a/src/ocr_pipeline/pycache/config.cpython-313.pyc
+++ b/src/ocr_pipeline/pycache/config.cpython-313.pyc
--- a/src/ocr_pipeline/pycache/pipeline.cpython-313.pyc
+++ b/src/ocr_pipeline/pycache/pipeline.cpython-313.pyc
--- a/src/ocr_pipeline/pycache/watch.cpython-313.pyc
+++ b/src/ocr_pipeline/pycache/watch.cpython-313.pyc
--- a/src/ocr_pipeline/citations/init.py
+++ b/src/ocr_pipeline/citations/init.py
@@ -0,0 +1,7 @@
+from .regex_extractor import Citation, extract_citations, format_citations_for_markdown
+
+__all__ = [
+    "extract_citations",
+    "format_citations_for_markdown",
+    "Citation",
+]
--- a/src/ocr_pipeline/citations/pycache/init.cpython-313.pyc
+++ b/src/ocr_pipeline/citations/pycache/init.cpython-313.pyc
--- a/src/ocr_pipeline/citations/pycache/extractor.cpython-313.pyc
+++ b/src/ocr_pipeline/citations/pycache/extractor.cpython-313.pyc
--- a/src/ocr_pipeline/citations/pycache/regex_extractor.cpython-313.pyc
+++ b/src/ocr_pipeline/citations/pycache/regex_extractor.cpython-313.pyc
--- a/src/ocr_pipeline/citations/extractor.py
+++ b/src/ocr_pipeline/citations/extractor.py
@@ -0,0 +1,191 @@
+from __future__ import annotations
+
+import re
+from dataclasses import dataclass
+from typing import Any
+
+import requests
+
+from ocr_pipeline.config import settings
+from ocr_pipeline.utils.logging import get_logger
+
+logger = get_logger(__name__)
+
+
+@dataclass
+class Citation:
+    type: str
+    value: str
+    start: int
+    end: int
+    confidence: float
+    metadata: dict[str, Any] | None = None
+
+
+class RegexCitationExtractor:
+    DOI_PATTERN = re.compile(r"\b(?:10\.\d{4,9}/[-._;()/:A-Z0-9]+)\b", re.IGNORECASE)
+    PMID_PATTERN = re.compile(r"\b(?:PMID|PubMed|pubmed)[:/\s]?(\d{1,8})\b", re.IGNORECASE)
+    ARXIV_PATTERN = re.compile(
+        r"\b(?:arXiv|arxiv)[:/\s]?(\d{4}\.\d{4,5}(?:v\d+)?)\b", re.IGNORECASE
+    )
+    PMC_PATTERN = re.compile(r"\b(?:PMC|pmc)[:/\s]?(\d{1,8})\b", re.IGNORECASE)
+    ISBN_PATTERN = re.compile(
+        r"\b(?:ISBN[-:\s]?)?(?:97[89][-\s]?)?\d{1,5}[-\s]?\d{1,7}[-\s]?\d{1,7}[-\s]?[\dX]\b",
+        re.IGNORECASE,
+    )
+
+    @classmethod
+    def extract(cls, text: str) -> list[Citation]:
+        citations = []
+
+        for match in cls.DOI_PATTERN.finditer(text):
+            citations.append(
+                Citation(
+                    type="DOI",
+                    value=match.group(),
+                    start=match.start(),
+                    end=match.end(),
+                    confidence=0.95,
+                    metadata={"url": f"https://doi.org/{match.group()}"},
+                )
+            )
+
+        for match in cls.PMID_PATTERN.finditer(text):
+            citations.append(
+                Citation(
+                    type="PMID",
+                    value=f"PMID:{match.group(1)}",
+                    start=match.start(),
+                    end=match.end(),
+                    confidence=0.9,
+                    metadata={"url": f"https://pubmed.ncbi.nlm.nih.gov/{match.group(1)}/"},
+                )
+            )
+
+        for match in cls.ARXIV_PATTERN.finditer(text):
+            citations.append(
+                Citation(
+                    type="arXiv",
+                    value=f"arXiv:{match.group(1)}",
+                    start=match.start(),
+                    end=match.end(),
+                    confidence=0.9,
+                    metadata={"url": f"https://arxiv.org/abs/{match.group(1)}"},
+                )
+            )
+
+        for match in cls.PMC_PATTERN.finditer(text):
+            citations.append(
+                Citation(
+                    type="PMC",
+                    value=f"PMC:{match.group(1)}",
+                    start=match.start(),
+                    end=match.end(),
+                    confidence=0.85,
+                    metadata={
+                        "url": f"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC{match.group(1)}/"
+                    },
+                )
+            )
+
+        for match in cls.ISBN_PATTERN.finditer(text):
+            isbn = re.sub(r"[^\dX]", "", match.group(), flags=re.IGNORECASE)
+            if len(isbn) in (10, 13):
+                citations.append(
+                    Citation(
+                        type="ISBN",
+                        value=isbn,
+                        start=match.start(),
+                        end=match.end(),
+                        confidence=0.8,
+                        metadata={"url": f"https://isbnsearch.org/isbn/{isbn}"},
+                    )
+                )
+
+        return citations
+
+
+class GrobidClient:
+    def __init__(self):
+        self.url = settings.citations.grobid_url
+        self.timeout = settings.citations.grobid_timeout
+
+    def extract_references(self, text: str) -> list[Citation]:
+        if not settings.citations.grobid_enabled:
+            return []
+
+        try:
+            response = requests.post(
+                f"{self.url}/api/processCitationList",
+                data={"citations": text},
+                headers={"Accept": "application/xml"},
+                timeout=self.timeout,
+            )
+            if response.status_code != 200:
+                logger.warning("grobid_request_failed", status=response.status_code)
+                return []
+
+            return self._parse_grobid_xml(response.text)
+        except Exception as e:
+            logger.warning("grobid_extraction_failed", error=str(e))
+            return []
+
+    def _parse_grobid_xml(self, xml_text: str) -> list[Citation]:
+        citations = []
+        try:
+            import xml.etree.ElementTree as ET
+
+            root = ET.fromstring(xml_text)
+            for cit in root.findall(".//{*}citation"):
+                citation_text = ET.tostring(cit, encoding="unicode")
+                doi_match = self.DOI_PATTERN.search(citation_text)
+                if doi_match:
+                    citations.append(
+                        Citation(
+                            type="DOI",
+                            value=doi_match.group(),
+                            start=0,
+                            end=len(citation_text),
+                            confidence=0.99,
+                            metadata={"source": "grobid", "full_citation": citation_text},
+                        )
+                    )
+        except Exception:
+            pass
+        return citations
+
+
+def extract_citations(text: str) -> list[Citation]:
+    citations = []
+
+    if settings.citations.regex_enabled:
+        citations.extend(RegexCitationExtractor.extract(text))
+
+    if settings.citations.grobid_enabled:
+        grobid = GrobidClient()
+        citations.extend(grobid.extract_references(text))
+
+    seen = set()
+    unique = []
+    for cit in citations:
+        key = (cit.type, cit.value)
+        if key not in seen:
+            seen.add(key)
+            unique.append(cit)
+
+    return unique
+
+
+def format_citations_for_markdown(citations: list[Citation]) -> str:
+    if not citations:
+        return "No citations found."
+
+    lines = ["## Citations", ""]
+    for cit in citations:
+        url = cit.metadata.get("url", "") if cit.metadata else ""
+        if url:
+            lines.append(f"- [{cit.type}: {cit.value}]({url})")
+        else:
+            lines.append(f"- {cit.type}: {cit.value}")
+
+    return "\n".join(lines)
--- a/src/ocr_pipeline/citations/regex_extractor.py
+++ b/src/ocr_pipeline/citations/regex_extractor.py
@@ -0,0 +1,196 @@
+from __future__ import annotations
+
+import re
+from dataclasses import dataclass
+
+import httpx
+
+from ocr_pipeline.config import settings
+from ocr_pipeline.utils.logging import get_logger
+
+logger = get_logger(__name__)
+
+
+@dataclass
+class Citation:
+    raw: str
+    type: str
+    identifier: str
+    confidence: float
+    context: str = ""
+
+
+class RegexCitationExtractor:
+    DOI_PATTERN = re.compile(
+        r"\b(?:10\.\d{4,9}/[-._;()/:A-Z0-9]+)\b",
+        re.IGNORECASE,
+    )
+    PMID_PATTERN = re.compile(
+        r"\b(?:PMID|PubMed\s*ID|PubMed)\s*[:\-]?\s*(\d{8})\b",
+        re.IGNORECASE,
+    )
+    ARXIV_PATTERN = re.compile(
+        r"\b(?:arXiv|arxiv)\s*[:\-]?\s*(\d{4}\.\d{4,5}(?:v\d+)?)\b",
+        re.IGNORECASE,
+    )
+    ISBN_PATTERN = re.compile(
+        r"\b(?:ISBN[-\s]*(?:13|10)?[-\s]*)?(?:97[89][-\s]*)?\d{1,5}[-\s]*\d{1,7}[-\s]*\d{1,7}[-\s]*[\dX]\b",
+        re.IGNORECASE,
+    )
+
+    @classmethod
+    def extract(cls, text: str) -> list[Citation]:
+        citations = []
+
+        for match in cls.DOI_PATTERN.finditer(text):
+            ctx_start = max(0, match.start() - 80)
+            ctx_end = min(len(text), match.end() + 80)
+            citations.append(
+                Citation(
+                    raw=match.group(),
+                    type="doi",
+                    identifier=match.group().lower(),
+                    confidence=0.95,
+                    context=text[ctx_start:ctx_end].strip(),
+                )
+            )
+
+        for match in cls.PMID_PATTERN.finditer(text):
+            ctx_start = max(0, match.start() - 80)
+            ctx_end = min(len(text), match.end() + 80)
+            citations.append(
+                Citation(
+                    raw=match.group(),
+                    type="pmid",
+                    identifier=f"PMID:{match.group(1)}",
+                    confidence=0.9,
+                    context=text[ctx_start:ctx_end].strip(),
+                )
+            )
+
+        for match in cls.ARXIV_PATTERN.finditer(text):
+            ctx_start = max(0, match.start() - 80)
+            ctx_end = min(len(text), match.end() + 80)
+            citations.append(
+                Citation(
+                    raw=match.group(),
+                    type="arxiv",
+                    identifier=f"arXiv:{match.group(1)}",
+                    confidence=0.9,
+                    context=text[ctx_start:ctx_end].strip(),
+                )
+            )
+
+        for match in cls.ISBN_PATTERN.finditer(text):
+            ctx_start = max(0, match.start() - 80)
+            ctx_end = min(len(text), match.end() + 80)
+            isbn = re.sub(r"[-\s]", "", match.group())
+            citations.append(
+                Citation(
+                    raw=match.group(),
+                    type="isbn",
+                    identifier=f"ISBN:{isbn}",
+                    confidence=0.7,
+                    context=text[ctx_start:ctx_end].strip(),
+                )
+            )
+
+        return citations
+
+
+class GrobidClient:
+    def __init__(self):
+        self.url = settings.citations.grobid_url
+        self.timeout = settings.citations.grobid_timeout
+
+    def process_text(self, text: str) -> list[Citation]:
+        if not settings.citations.grobid_enabled:
+            return []
+
+        try:
+            with httpx.Client(timeout=self.timeout) as client:
+                response = client.post(
+                    f"{self.url}/api/processCitationText",
+                    data={"text": text},
+                )
+                response.raise_for_status()
+                return self._parse_grobid_xml(response.text)
+        except Exception as e:
+            logger.warning("grobid_request_failed", error=str(e))
+            return []
+
+    def _parse_grobid_xml(self, xml_text: str) -> list[Citation]:
+        try:
+            import xml.etree.ElementTree as ET
+
+            root = ET.fromstring(xml_text)
+            citations = []
+
+            ns = {"tei": "http://www.tei-c.org/ns/1.0"}
+            for cit in root.findall(".//tei:biblStruct", ns):
+                doi_elem = cit.find(".//tei:idno[@type='DOI']", ns)
+                pmid_elem = cit.find(".//tei:idno[@type='PMID']", ns)
+
+                if doi_elem is not None and doi_elem.text:
+                    citations.append(
+                        Citation(
+                            raw=doi_elem.text,
+                            type="doi",
+                            identifier=doi_elem.text.lower(),
+                            confidence=0.99,
+                        )
+                    )
+                if pmid_elem is not None and pmid_elem.text:
+                    citations.append(
+                        Citation(
+                            raw=pmid_elem.text,
+                            type="pmid",
+                            identifier=f"PMID:{pmid_elem.text}",
+                            confidence=0.99,
+                        )
+                    )
+
+            return citations
+        except Exception as e:
+            logger.warning("grobid_xml_parse_failed", error=str(e))
+            return []
+
+
+def extract_citations(text: str) -> list[Citation]:
+    citations = []
+
+    if settings.citations.regex_enabled:
+        citations.extend(RegexCitationExtractor.extract(text))
+
+    if settings.citations.grobid_enabled:
+        client = GrobidClient()
+        citations.extend(client.process_text(text))
+
+    seen = set()
+    unique = []
+    for cit in citations:
+        key = (cit.type, cit.identifier)
+        if key not in seen:
+            seen.add(key)
+            unique.append(cit)
+
+    return unique
+
+
+def format_citations_for_markdown(citations: list[Citation]) -> str:
+    if not citations:
+        return "No citations found."
+
+    lines = [
+        "## Citations",
+        "",
+        "| Type | Identifier | Context |",
+        "|------|------------|---------|",
+    ]
+    for cit in citations:
+        context = (
+            cit.context[:80].replace("|", "\\|") + "..." if len(cit.context) > 80 else cit.context
+        )
+        lines.append(f"| {cit.type.upper()} | `{cit.identifier}` | {context} |")
+
+    return "\n".join(lines)
--- a/src/ocr_pipeline/cli.py
+++ b/src/ocr_pipeline/cli.py
@@ -0,0 +1,163 @@
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+
+import typer
+from rich.console import Console
+from rich.table import Table
+
+from ocr_pipeline.config import settings
+from ocr_pipeline.pipeline import OCRPipeline
+from ocr_pipeline.utils.logging import get_logger, setup_logging
+from ocr_pipeline.utils.migrate import migrate_notebook
+
+app = typer.Typer(
+    name="ocr-pipeline",
+    help="OCR pipeline for screenshots -> RAG-ready Markdown",
+    add_completion=False,
+)
+console = Console()
+logger = get_logger(__name__)
+
+
+@app.callback()
+def callback(
+    config: Path | None = typer.Option(None, "--config", "-c", help="Config file path"),
+    log_level: str = typer.Option("INFO", "--log-level", "-l", help="Log level"),
+    log_file: Path | None = typer.Option(None, "--log-file", help="Log file path"),
+):
+    """OCR Pipeline - Extract text from screenshots for RAG"""
+    setup_logging(log_level, log_file)
+
+    if config:
+        settings.update_from_yaml(config)
+
+
+@app.command()
+def run(
+    input_dir: Path | None = typer.Option(None, "--input-dir", "-i", help="Input directory"),
+    output_dir: Path | None = typer.Option(None, "--output-dir", "-o", help="Output directory"),
+    workers: int | None = typer.Option(
+        None, "--workers", "-w", help="Number of worker processes"
+    ),
+    engine: str | None = typer.Option(
+        None, "--engine", "-e", help="OCR engine (paddleocr/tesseract/auto)"
+    ),
+):
+    """Run OCR pipeline on all screenshots"""
+    if input_dir:
+        settings.input.paths = [str(input_dir)]
+    if output_dir:
+        settings.output.base_directory = str(output_dir)
+    if workers:
+        settings.processing.workers = workers
+    if engine:
+        settings.ocr.engine = engine
+
+    pipeline = OCRPipeline()
+    result = pipeline.process_all()
+
+    table = Table(title="Pipeline Results")
+    table.add_column("Metric", style="cyan")
+    table.add_column("Value", style="green")
+
+    table.add_row("Total Images", str(result.total_images))
+    table.add_row("Successful", str(result.successful))
+    table.add_row("Failed", str(result.failed))
+    table.add_row("Chunks Created", str(result.chunks_created))
+    table.add_row("Output Files", str(len(result.output_files)))
+
+    console.print(table)
+
+    if result.errors:
+        console.print("\n[red]Errors:[/red]")
+        for err in result.errors:
+            console.print(f"  {err['path']}: {err['error']}")
+
+    if result.failed > 0:
+        sys.exit(1)
+
+
+@app.command()
+def watch():
+    """Watch for new screenshots and process automatically"""
+    pipeline = OCRPipeline()
+    console.print("[green]Starting watch mode...[/green]")
+    console.print("Press Ctrl+C to stop")
+    try:
+        pipeline.watch()
+    except KeyboardInterrupt:
+        console.print("\n[yellow]Stopping watcher...[/yellow]")
+
+
+@app.command()
+def reprocess(
+    notebook: Path = typer.Argument(..., help="Path to Jupyter notebook"),
+):
+    """Reprocess screenshots from a Jupyter notebook"""
+    pipeline = OCRPipeline()
+    console.print(f"[green]Migrating from notebook: {notebook}[/green]")
+    result = migrate_notebook(notebook, pipeline)
+
+    table = Table(title="Migration Results")
+    table.add_column("Metric", style="cyan")
+    table.add_column("Value", style="green")
+
+    for key, value in result.items():
+        if key != "errors":
+            table.add_row(key.capitalize(), str(value))
+
+    console.print(table)
+
+    if result.get("errors"):
+        console.print("\n[red]Errors:[/red]")
+        for err in result["errors"]:
+            console.print(f"  {err['path']}: {err['error']}")
+
+
+@app.command()
+def status():
+    """Show pipeline status and statistics"""
+    pipeline = OCRPipeline()
+    stats = pipeline.get_stats()
+
+    table = Table(title="Pipeline Status")
+    table.add_column("Metric", style="cyan")
+    table.add_column("Value", style="green")
+
+    for key, value in stats.items():
+        table.add_row(key.replace("_", " ").title(), str(value))
+
+    console.print(table)
+
+
+@app.command()
+def stats(days: int = typer.Option(30, "--days", "-d", help="Number of days to analyze")):
+    """Show processing statistics"""
+    pipeline = OCRPipeline()
+    stats = pipeline.get_stats()
+
+    table = Table(title=f"Statistics (last {days} days)")
+    table.add_column("Metric", style="cyan")
+    table.add_column("Value", style="green")
+
+    for key, value in stats.items():
+        table.add_row(key.replace("_", " ").title(), str(value))
+
+    console.print(table)
+
+
+@app.command()
+def config():
+    """Show current configuration"""
+    console.print("[cyan]Current Configuration:[/cyan]")
+    console.print(settings.model_dump_json(indent=2))
+
+
+def main():
+    app()
+
+
+if __name__ == "__main__":
+    main()
--- a/src/ocr_pipeline/config.py
+++ b/src/ocr_pipeline/config.py
@@ -0,0 +1,246 @@
+from __future__ import annotations
+
+import os
+from pathlib import Path
+from typing import Any, ClassVar
+
+import yaml
+from pydantic import Field, field_validator
+from pydantic_settings import BaseSettings, SettingsConfigDict
+
+
+class InputConfig(BaseSettings):
+    paths: list[str] = Field(default_factory=lambda: ["~/Pictures"])
+    patterns: list[str] = Field(
+        default_factory=lambda: ["SCR-*.png", "*.jpg", "*.jpeg", "*.tiff", "*.bmp"]
+    )
+    recursive: bool = True
+
+    @field_validator("paths", mode="before")
+    @classmethod
+    def expand_paths(cls, v: list[str]) -> list[str]:
+        return [os.path.expanduser(p) for p in v]
+
+
+class PreprocessConfig(BaseSettings):
+    deskew: bool = True
+    denoise: bool = True
+    clahe: bool = True
+    adaptive_threshold: bool = True
+    remove_lines: bool = True
+    max_dimension: int = 4096
+
+
+class OCRConfig(BaseSettings):
+    engine: str = "paddleocr"  # paddleocr | tesseract | auto
+    languages: list[str] = Field(default_factory=lambda: ["en", "latin"])
+    use_gpu: bool = False
+    use_angle_cls: bool = True
+    det_db_thresh: float = 0.3
+    det_db_box_thresh: float = 0.6
+    det_db_unclip_ratio: float = 1.5
+    rec_batch_num: int = 6
+    cpu_threads: int = 4
+    preprocess: PreprocessConfig = Field(default_factory=PreprocessConfig)
+
+
+class ProcessingConfig(BaseSettings):
+    workers: int = 4
+    batch_size: int = 10
+    retry_attempts: int = 2
+    timeout_per_image: int = 120
+    skip_existing: bool = True
+
+
+class DetectorConfig(BaseSettings):
+    enabled: bool = True
+    model: str = ""
+    confidence_threshold: float = 0.7
+
+
+class FigureDetectorConfig(DetectorConfig):
+    model: str = "lp://PubLayNet/faster_rcnn_R_50_FPN_3x/config"
+    confidence_threshold: float = 0.7
+
+
+class TableDetectorConfig(DetectorConfig):
+    model: str = "microsoft/table-transformer-detection"
+    confidence_threshold: float = 0.7
+
+
+class CaptionDetectorConfig(DetectorConfig):
+    model: str = "lp://PubLayNet/faster_rcnn_R_50_FPN_3x/config"
+    confidence_threshold: float = 0.5
+
+
+class DetectorsConfig(BaseSettings):
+    figures: FigureDetectorConfig = Field(default_factory=FigureDetectorConfig)
+    tables: TableDetectorConfig = Field(default_factory=TableDetectorConfig)
+    captions: CaptionDetectorConfig = Field(default_factory=CaptionDetectorConfig)
+
+
+class EntitiesConfig(BaseSettings):
+    enabled: bool = True
+    model: str = "en_core_sci_lg"
+    types: list[str] = Field(
+        default_factory=lambda: [
+            "GENE",
+            "PROTEIN",
+            "CHEMICAL",
+            "SPECIES",
+            "DISEASE",
+            "CELL_LINE",
+            "ORGANISM",
+            "CELL_TYPE",
+        ]
+    )
+    merge_entities: bool = True
+
+
+class CitationsConfig(BaseSettings):
+    regex_enabled: bool = True
+    grobid_enabled: bool = False
+    grobid_url: str = "http://localhost:8070"
+    grobid_timeout: int = 30
+
+
+class ChunkingConfig(BaseSettings):
+    chunk_size: int = 1000
+    chunk_overlap: int = 200
+    separators: list[str] = Field(default_factory=lambda: ["\n\n", "\n", ". ", " ", ""])
+    keep_separator: bool = True
+
+
+class FrontmatterConfig(BaseSettings):
+    fields: list[str] = Field(
+        default_factory=lambda: [
+            "source_path",
+            "source_hash",
+            "timestamp",
+            "ocr_engine",
+            "ocr_confidence_mean",
+            "language",
+            "detected_entities",
+            "has_figures",
+            "has_tables",
+            "citations_found",
+            "chunk_index",
+            "total_chunks",
+        ]
+    )
+    include_raw_text: bool = False
+
+
+class OutputConfig(BaseSettings):
+    base_directory: str = "./data/ocr_output"
+    format: str = "markdown"
+    organize_by: str = "date_run"  # date_run | source_dir | flat
+    frontmatter: FrontmatterConfig = Field(default_factory=FrontmatterConfig)
+
+
+class WatchConfig(BaseSettings):
+    enabled: bool = True
+    debounce_seconds: int = 5
+    ignore_patterns: list[str] = Field(
+        default_factory=lambda: [
+            ".DS_Store",
+            "*.tmp",
+            "*.partial",
+            "*.crdownload",
+        ]
+    )
+    db_path: str = "./data/processed_files.db"
+    poll_interval: float = 1.0
+
+
+class _BaseSettings(BaseSettings):
+    model_config = SettingsConfigDict(
+        env_file=".env",
+        env_nested_delimiter="__",
+        extra="ignore",
+    )
+
+
+def _load_yaml_config(path: str = "config.yaml") -> dict[str, Any]:
+    """Load YAML config file if it exists."""
+    config_path = Path(path)
+    if not config_path.is_absolute():
+        config_path = Path.cwd() / config_path
+    if config_path.exists():
+        with open(config_path) as f:
+            return yaml.safe_load(f) or {}
+    return {}
+
+
+def _build_settings_from_yaml(yaml_dict: dict[str, Any]) -> dict[str, Any]:
+    """Convert YAML dict to properly typed nested models for Settings."""
+    if not yaml_dict:
+        return {}
+
+    result = {}
+
+    # Map YAML keys to model classes
+    model_map = {
+        "input": InputConfig,
+        "ocr": OCRConfig,
+        "processing": ProcessingConfig,
+        "detectors": DetectorsConfig,
+        "entities": EntitiesConfig,
+        "citations": CitationsConfig,
+        "chunking": ChunkingConfig,
+        "output": OutputConfig,
+        "watch": WatchConfig,
+    }
+
+    for key, model_class in model_map.items():
+        if key in yaml_dict:
+            data = yaml_dict[key]
+            # Special handling for output config to transform frontmatter list -> dict
+            if key == "output" and "frontmatter" in data and isinstance(data["frontmatter"], list):
+                data = {**data, "frontmatter": {"fields": data["frontmatter"]}}
+            result[key] = model_class(**data)
+
+    return result
+
+
+class Settings(_BaseSettings):
+    input: InputConfig = Field(default_factory=InputConfig)
+    ocr: OCRConfig = Field(default_factory=OCRConfig)
+    processing: ProcessingConfig = Field(default_factory=ProcessingConfig)
+    detectors: DetectorsConfig = Field(default_factory=DetectorsConfig)
+    entities: EntitiesConfig = Field(default_factory=EntitiesConfig)
+    citations: CitationsConfig = Field(default_factory=CitationsConfig)
+    chunking: ChunkingConfig = Field(default_factory=ChunkingConfig)
+    output: OutputConfig = Field(default_factory=OutputConfig)
+    watch: WatchConfig = Field(default_factory=WatchConfig)
+
+    # Store YAML config for manual override
+    _yaml_config: ClassVar[dict[str, Any]] = {}
+
+    def __init__(self, **kwargs):
+        if not Settings._yaml_config:
+            Settings._yaml_config = _load_yaml_config()
+
+        # Build nested models from YAML
+        yaml_models = _build_settings_from_yaml(Settings._yaml_config)
+        super().__init__(**yaml_models, **kwargs)
+
+    def update_from_yaml(self, path: str) -> None:
+        """Reload settings from a YAML file."""
+        Settings._yaml_config = _load_yaml_config(path)
+        # Re-initialize with new config
+        yaml_models = _build_settings_from_yaml(Settings._yaml_config)
+        for key, value in yaml_models.items():
+            if hasattr(self, key):
+                setattr(self, key, value)
+
+    @property
+    def output_dir(self) -> Path:
+        return Path(self.output.base_directory).expanduser().resolve()
+
+    @property
+    def watch_db_path(self) -> Path:
+        return Path(self.watch.db_path).expanduser().resolve()
+
+
+settings = Settings()
--- a/src/ocr_pipeline/detectors/init.py
+++ b/src/ocr_pipeline/detectors/init.py
@@ -0,0 +1,15 @@
+from __future__ import annotations
+
+from .figures import CaptionExtractor, DetectedFigure, FigureDetector, detect_figures
+from .tables import DetectedTable, TableDetector, TableStructureRecognizer, detect_tables
+
+__all__ = [
+    "detect_figures",
+    "FigureDetector",
+    "CaptionExtractor",
+    "DetectedFigure",
+    "detect_tables",
+    "TableDetector",
+    "TableStructureRecognizer",
+    "DetectedTable",
+]
--- a/src/ocr_pipeline/detectors/pycache/init.cpython-313.pyc
+++ b/src/ocr_pipeline/detectors/pycache/init.cpython-313.pyc
--- a/src/ocr_pipeline/detectors/pycache/figures.cpython-313.pyc
+++ b/src/ocr_pipeline/detectors/pycache/figures.cpython-313.pyc
--- a/src/ocr_pipeline/detectors/pycache/tables.cpython-313.pyc
+++ b/src/ocr_pipeline/detectors/pycache/tables.cpython-313.pyc
--- a/src/ocr_pipeline/detectors/figures.py
+++ b/src/ocr_pipeline/detectors/figures.py
@@ -0,0 +1,166 @@
+from __future__ import annotations
+
+from dataclasses import dataclass
+
+import numpy as np
+
+from ocr_pipeline.config import settings
+from ocr_pipeline.utils.logging import get_logger
+
+logger = get_logger(__name__)
+
+
+@dataclass
+class DetectedFigure:
+    bbox: list[float]
+    confidence: float
+    label: str
+    caption: str | None = None
+    image_crop: np.ndarray | None = None
+
+
+class FigureDetector:
+    def __init__(self):
+        self._predictor = None
+        self._initialized = False
+
+    def _init(self):
+        if self._initialized:
+            return
+        if not settings.detectors.figures.enabled:
+            self._initialized = True
+            return
+        try:
+            import layoutparser as lp
+
+            self._predictor = lp.Detectron2LayoutModel(
+                config_path=settings.detectors.figures.model,
+                label_map={0: "Text", 1: "Title", 2: "List", 3: "Table", 4: "Figure"},
+                extra_config=[
+                    "MODEL.ROI_HEADS.SCORE_THRESH_TEST",
+                    settings.detectors.figures.confidence_threshold,
+                ],
+            )
+            self._initialized = True
+            logger.info("figure_detector_initialized", model=settings.detectors.figures.model)
+        except Exception as e:
+            logger.warning("figure_detector_init_failed", error=str(e))
+            self._initialized = True
+
+    def detect(self, image: np.ndarray) -> list[DetectedFigure]:
+        if not settings.detectors.figures.enabled:
+            return []
+
+        self._init()
+        if self._predictor is None:
+            return []
+
+        try:
+            layout = self._predictor.detect(image)
+            figures = []
+
+            for block in layout:
+                if block.type == "Figure":
+                    x1, y1, x2, y2 = (
+                        block.block.x_1,
+                        block.block.y_1,
+                        block.block.x_2,
+                        block.block.y_2,
+                    )
+                    crop = image[int(y1) : int(y2), int(x1) : int(x2)]
+                    figures.append(
+                        DetectedFigure(
+                            bbox=[float(x1), float(y1), float(x2), float(y2)],
+                            confidence=float(block.score),
+                            label="Figure",
+                            image_crop=crop if crop.size > 0 else None,
+                        )
+                    )
+
+            logger.debug("figures_detected", count=len(figures))
+            return figures
+        except Exception as e:
+            logger.warning("figure_detection_failed", error=str(e))
+            return []
+
+
+class CaptionExtractor:
+    def __init__(self):
+        self._predictor = None
+        self._initialized = False
+
+    def _init(self):
+        if self._initialized:
+            return
+        if not settings.detectors.captions.enabled:
+            self._initialized = True
+            return
+        try:
+            import layoutparser as lp
+
+            self._predictor = lp.Detectron2LayoutModel(
+                config_path=settings.detectors.captions.model,
+                label_map={0: "Text", 1: "Title", 2: "List", 3: "Table", 4: "Figure"},
+                extra_config=[
+                    "MODEL.ROI_HEADS.SCORE_THRESH_TEST",
+                    settings.detectors.captions.confidence_threshold,
+                ],
+            )
+            self._initialized = True
+        except Exception as e:
+            logger.warning("caption_detector_init_failed", error=str(e))
+            self._initialized = True
+
+    def extract_captions(
+        self, image: np.ndarray, figures: list[DetectedFigure]
+    ) -> list[DetectedFigure]:
+        if not settings.detectors.captions.enabled or not figures:
+            return figures
+
+        self._init()
+        if self._predictor is None:
+            return figures
+
+        try:
+            layout = self._predictor.detect(image)
+            captions = [
+                b
+                for b in layout
+                if b.type in ("Title", "Text")
+                and b.score > settings.detectors.captions.confidence_threshold
+            ]
+
+            for fig in figures:
+                fig_x1, fig_y1, fig_x2, fig_y2 = fig.bbox
+                fig_center_y = (fig_y1 + fig_y2) / 2
+
+                best_caption = None
+                min_dist = float("inf")
+
+                for cap in captions:
+                    cap_y1, cap_y2 = cap.block.y_1, cap.block.y_2
+                    cap_center_y = (cap_y1 + cap_y2) / 2
+                    dist = abs(cap_center_y - fig_center_y)
+
+                    if cap_y2 < fig_y1 or cap_y1 > fig_y2:
+                        if dist < min_dist:
+                            min_dist = dist
+                            best_caption = cap
+
+                if best_caption and min_dist < 200:
+                    fig.caption = best_caption.block.text
+
+            return figures
+        except Exception as e:
+            logger.warning("caption_extraction_failed", error=str(e))
+            return figures
+
+
+def detect_figures(image: np.ndarray) -> list[DetectedFigure]:
+    detector = FigureDetector()
+    figures = detector.detect(image)
+
+    captioner = CaptionExtractor()
+    figures = captioner.extract_captions(image, figures)
+
+    return figures
--- a/src/ocr_pipeline/detectors/tables.py
+++ b/src/ocr_pipeline/detectors/tables.py
@@ -0,0 +1,202 @@
+from __future__ import annotations
+
+from dataclasses import dataclass
+from typing import Any
+
+import cv2
+import numpy as np
+
+from ocr_pipeline.config import settings
+from ocr_pipeline.utils.logging import get_logger
+
+logger = get_logger(__name__)
+
+
+@dataclass
+class DetectedTable:
+    bbox: list[float]
+    confidence: float
+    label: str
+    structure: list[dict[str, Any]] | None = None
+    markdown: str | None = None
+    image_crop: np.ndarray | None = None
+
+
+class TableDetector:
+    def __init__(self):
+        self._model = None
+        self._processor = None
+        self._initialized = False
+
+    def _init(self):
+        if self._initialized:
+            return
+        if not settings.detectors.tables.enabled:
+            self._initialized = True
+            return
+        try:
+            from transformers import AutoImageProcessor, AutoModelForObjectDetection
+
+            self._processor = AutoImageProcessor.from_pretrained(settings.detectors.tables.model)
+            self._model = AutoModelForObjectDetection.from_pretrained(
+                settings.detectors.tables.model
+            )
+            self._model.eval()
+            self._initialized = True
+            logger.info("table_detector_initialized", model=settings.detectors.tables.model)
+        except Exception as e:
+            logger.warning("table_detector_init_failed", error=str(e))
+            self._initialized = True
+
+    def detect(self, image: np.ndarray) -> list[DetectedTable]:
+        if not settings.detectors.tables.enabled:
+            return []
+
+        self._init()
+        if self._model is None or self._processor is None:
+            return []
+
+        try:
+            import torch
+
+            pil_image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
+            pil_image = cv2.resize(pil_image, (800, 800))
+
+            inputs = self._processor(images=pil_image, return_tensors="pt")
+            with torch.no_grad():
+                outputs = self._model(**inputs)
+
+            target_sizes = torch.tensor([pil_image.shape[:2]])
+            results = self._processor.post_process_object_detection(
+                outputs,
+                threshold=settings.detectors.tables.confidence_threshold,
+                target_sizes=target_sizes,
+            )[0]
+
+            tables = []
+            for score, _label, box in zip(
+                results["scores"], results["labels"], results["boxes"], strict=True
+            ):
+                box = [float(x) for x in box]
+                x1, y1, x2, y2 = map(int, box)
+                crop = image[y1:y2, x1:x2]
+                tables.append(
+                    DetectedTable(
+                        bbox=box,
+                        confidence=float(score),
+                        label="Table",
+                        image_crop=crop if crop.size > 0 else None,
+                    )
+                )
+
+            logger.debug("tables_detected", count=len(tables))
+            return tables
+        except Exception as e:
+            logger.warning("table_detection_failed", error=str(e))
+            return []
+
+
+class TableStructureRecognizer:
+    def __init__(self):
+        self._model = None
+        self._processor = None
+        self._initialized = False
+
+    def _init(self):
+        if self._initialized:
+            return
+        try:
+            from transformers import AutoImageProcessor, AutoModelForObjectDetection
+
+            self._processor = AutoImageProcessor.from_pretrained(
+                "microsoft/table-transformer-structure-recognition"
+            )
+            self._model = AutoModelForObjectDetection.from_pretrained(
+                "microsoft/table-transformer-structure-recognition"
+            )
+            self._model.eval()
+            self._initialized = True
+            logger.info("table_structure_recognizer_initialized")
+        except Exception as e:
+            logger.warning("table_structure_recognizer_init_failed", error=str(e))
+            self._initialized = True
+
+    def recognize(self, table_crop: np.ndarray) -> tuple[list[dict[str, Any]] | None, str | None]:
+        if self._model is None or self._processor is None:
+            return None, None
+
+        try:
+            import torch
+
+            pil_image = cv2.cvtColor(table_crop, cv2.COLOR_BGR2RGB)
+            inputs = self._processor(images=pil_image, return_tensors="pt")
+            with torch.no_grad():
+                outputs = self._model(**inputs)
+
+            target_sizes = torch.tensor([pil_image.shape[:2]])
+            results = self._processor.post_process_object_detection(
+                outputs, threshold=0.5, target_sizes=target_sizes
+            )[0]
+
+            structure = []
+            for score, label, box in zip(
+                results["scores"], results["labels"], results["boxes"], strict=True
+            ):
+                structure.append(
+                    {
+                        "label": self._model.config.id2label[int(label)],
+                        "confidence": float(score),
+                        "bbox": [float(x) for x in box],
+                    }
+                )
+
+            markdown = self._structure_to_markdown(structure, table_crop.shape)
+            return structure, markdown
+        except Exception as e:
+            logger.warning("table_structure_recognition_failed", error=str(e))
+            return None, None
+
+    def _structure_to_markdown(self, structure: list[dict[str, Any]], shape: tuple) -> str | None:
+        if not structure:
+            return None
+
+        rows = {}
+        for cell in structure:
+            if cell["label"] in ("table row", "table cell"):
+                y_center = (cell["bbox"][1] + cell["bbox"][3]) / 2
+                x_center = (cell["bbox"][0] + cell["bbox"][2]) / 2
+                row_idx = int(y_center / (shape[0] / 20))
+                if row_idx not in rows:
+                    rows[row_idx] = []
+                rows[row_idx].append((x_center, cell["label"]))
+
+        markdown_lines = []
+        for row_idx in sorted(rows.keys()):
+            cells = sorted(rows[row_idx], key=lambda x: x[0])
+            row_cells = [c[1] for c in cells]
+            markdown_lines.append("| " + " | ".join(row_cells) + " |")
+
+        if markdown_lines:
+            header = markdown_lines[0]
+            separator = "| " + " | ".join(["---"] * len(header.split("|")[1:-1])) + " |"
+            markdown_lines.insert(1, separator)
+            return "\n".join(markdown_lines)
+
+        return None
+
+
+def detect_tables(image: np.ndarray) -> list[DetectedTable]:
+    detector = TableDetector()
+    tables = detector.detect(image)
+
+    if not tables:
+        return tables
+
+    recognizer = TableStructureRecognizer()
+    for table in tables:
+        if table.image_crop is not None:
+            structure, markdown = recognizer.recognize(table.image_crop)
+            table.structure = structure
+            table.markdown = markdown
+
+    return tables
--- a/src/ocr_pipeline/ocr/init.py
+++ b/src/ocr_pipeline/ocr/init.py
@@ -0,0 +1,22 @@
+from __future__ import annotations
+
+from .engine import (
+    ImagePreprocessor,
+    OCREngine,
+    OCRResult,
+    PaddleOCREngine,
+    TesseractEngine,
+    compute_image_hash,
+)
+from .parallel import ParallelProcessor, ProcessedImage
+
+__all__ = [
+    "OCREngine",
+    "OCRResult",
+    "ImagePreprocessor",
+    "compute_image_hash",
+    "PaddleOCREngine",
+    "TesseractEngine",
+    "ParallelProcessor",
+    "ProcessedImage",
+]
--- a/src/ocr_pipeline/ocr/pycache/init.cpython-313.pyc
+++ b/src/ocr_pipeline/ocr/pycache/init.cpython-313.pyc
--- a/src/ocr_pipeline/ocr/pycache/engine.cpython-313.pyc
+++ b/src/ocr_pipeline/ocr/pycache/engine.cpython-313.pyc
--- a/src/ocr_pipeline/ocr/pycache/parallel.cpython-313.pyc
+++ b/src/ocr_pipeline/ocr/pycache/parallel.cpython-313.pyc
--- a/src/ocr_pipeline/ocr/engine.py
+++ b/src/ocr_pipeline/ocr/engine.py
@@ -0,0 +1,236 @@
+from __future__ import annotations
+
+import time
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Any
+
+import cv2
+import numpy as np
+import pytesseract
+from PIL import Image
+
+from ocr_pipeline.config import settings
+from ocr_pipeline.utils.logging import get_logger
+
+logger = get_logger(__name__)
+
+# Language code mapping: config names -> engine-specific codes
+TESSERACT_LANG_MAP = {
+    "en": "eng",
+    "latin": "lat",
+    "de": "deu",
+    "fr": "fra",
+    "es": "spa",
+    "it": "ita",
+    "pt": "por",
+    "zh": "chi_sim",
+    "ja": "jpn",
+    "ko": "kor",
+}
+
+PADDLEOCR_LANG_MAP = {
+    "en": "en",
+    "latin": "latin",
+    "de": "german",
+    "fr": "french",
+    "es": "spanish",
+    "it": "italian",
+    "pt": "portuguese",
+    "zh": "ch",
+    "ja": "japan",
+    "ko": "korean",
+}
+
+
+def map_languages(languages: list[str], engine: str) -> str:
+    """Map config language names to engine-specific codes."""
+    if engine == "tesseract":
+        mapped = [TESSERACT_LANG_MAP.get(lang, lang) for lang in languages]
+        return "+".join(mapped)
+    else:  # paddleocr
+        # PaddleOCR uses primary language only for detection
+        primary = languages[0] if languages else "en"
+        return PADDLEOCR_LANG_MAP.get(primary, primary)
+
+
+@dataclass
+class OCRResult:
+    text: str
+    confidence: float
+    language: str
+    engine: str
+    boxes: list[dict[str, Any]]
+    raw_result: Any
+    processing_time: float
+
+
+class PaddleOCREngine:
+    def __init__(self):
+        self._ocr = None
+        self._initialized = False
+
+    def _init(self):
+        if self._initialized:
+            return
+        try:
+            from paddleocr import PaddleOCR
+
+            paddle_lang = map_languages(settings.ocr.languages, "paddleocr")
+
+            self._ocr = PaddleOCR(
+                use_angle_cls=settings.ocr.use_angle_cls,
+                lang=paddle_lang,
+                use_gpu=settings.ocr.use_gpu,
+                show_log=False,
+                det_db_thresh=settings.ocr.det_db_thresh,
+                det_db_box_thresh=settings.ocr.det_db_box_thresh,
+                det_db_unclip_ratio=settings.ocr.det_db_unclip_ratio,
+                rec_batch_num=settings.ocr.rec_batch_num,
+                cpu_threads=settings.ocr.cpu_threads,
+            )
+            self._initialized = True
+            logger.info("paddleocr_initialized", language=paddle_lang)
+        except Exception as e:
+            logger.error("paddleocr_init_failed", error=str(e))
+            raise
+
+    def process(self, image: np.ndarray) -> OCRResult:
+        self._init()
+        start = time.time()
+
+        result = self._ocr.ocr(image, cls=True)
+        processing_time = time.time() - start
+
+        if not result or not result[0]:
+            return OCRResult(
+                text="",
+                confidence=0.0,
+                language=map_languages(settings.ocr.languages, "paddleocr"),
+                engine="paddleocr",
+                boxes=[],
+                raw_result=result,
+                processing_time=processing_time,
+            )
+
+        texts = []
+        confidences = []
+        boxes = []
+
+        for line in result[0]:
+            box = line[0]
+            text = line[1][0]
+            conf = line[1][1]
+
+            texts.append(text)
+            confidences.append(conf)
+            boxes.append(
+                {
+                    "bbox": box,
+                    "text": text,
+                    "confidence": conf,
+                }
+            )
+
+        full_text = "\n".join(texts)
+        mean_conf = sum(confidences) / len(confidences) if confidences else 0.0
+
+        return OCRResult(
+            text=full_text,
+            confidence=mean_conf,
+            language=map_languages(settings.ocr.languages, "paddleocr"),
+            engine="paddleocr",
+            boxes=boxes,
+            raw_result=result,
+            processing_time=processing_time,
+        )
+
+
+class TesseractEngine:
+    def process(self, image: np.ndarray) -> OCRResult:
+        start = time.time()
+
+        pil_image = Image.fromarray(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))
+
+        tess_lang = map_languages(settings.ocr.languages, "tesseract")
+        config = f"--oem 3 --psm 6 -l {tess_lang}"
+        data = pytesseract.image_to_data(
+            pil_image, config=config, output_type=pytesseract.Output.DICT
+        )
+
+        texts = []
+        confidences = []
+        boxes = []
+
+        for i in range(len(data["text"])):
+            text = data["text"][i].strip()
+            conf = data["conf"][i]
+            if text and conf > 0:
+                texts.append(text)
+                confidences.append(conf / 100.0)
+                boxes.append(
+                    {
+                        "bbox": [
+                            [data["left"][i], data["top"][i]],
+                            [data["left"][i] + data["width"][i], data["top"][i]],
+                            [
+                                data["left"][i] + data["width"][i],
+                                data["top"][i] + data["height"][i],
+                            ],
+                            [data["left"][i], data["top"][i] + data["height"][i]],
+                        ],
+                        "text": text,
+                        "confidence": conf / 100.0,
+                    }
+                )
+
+        full_text = " ".join(texts)
+        mean_conf = sum(confidences) / len(confidences) if confidences else 0.0
+
+        return OCRResult(
+            text=full_text,
+            confidence=mean_conf,
+            language=map_languages(settings.ocr.languages, "tesseract"),
+            engine="tesseract",
+            boxes=boxes,
+            raw_result=data,
+            processing_time=time.time() - start,
+        )
+
+
+class OCREngine:
+    def __init__(self):
+        self.paddle = PaddleOCREngine()
+        self.tesseract = TesseractEngine()
+        self.engine_preference = settings.ocr.engine
+
+    def process(self, image: np.ndarray) -> OCRResult:
+        """Process a pre-loaded image array."""
+        if self.engine_preference == "paddleocr":
+            try:
+                return self.paddle.process(image)
+            except Exception as e:
+                logger.warning("paddleocr_failed_fallback", error=str(e))
+                return self.tesseract.process(image)
+
+        elif self.engine_preference == "tesseract":
+            return self.tesseract.process(image)
+
+        else:
+            try:
+                return self.paddle.process(image)
+            except Exception as e:
+                logger.warning("paddleocr_failed_fallback", error=str(e))
+                return self.tesseract.process(image)
+
+    def process_file(self, image_path: Path) -> OCRResult:
+        """Load and process an image file."""
+        logger.info("processing_image", path=str(image_path), engine=self.engine_preference)
+
+        image = cv2.imread(str(image_path))
+        if image is None:
+            raise ValueError(f"Failed to load image: {image_path}")
+        return self.process(image)
+
+    # Alias for backward compatibility
+    recognize = process
--- a/src/ocr_pipeline/ocr/parallel.py
+++ b/src/ocr_pipeline/ocr/parallel.py
@@ -0,0 +1,147 @@
+from __future__ import annotations
+
+import time
+from concurrent.futures import ProcessPoolExecutor, as_completed
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Any
+
+from ocr_pipeline.config import settings
+from ocr_pipeline.utils.logging import ProgressTracker, get_logger
+
+logger = get_logger(__name__)
+
+
+@dataclass
+class ProcessedImage:
+    path: Path
+    image_hash: str
+    ocr_result: Any
+    preprocessing_time: float
+    ocr_time: float
+    success: bool
+    error: str | None = None
+
+
+def _process_single_image(args: tuple[Path, dict]) -> ProcessedImage:
+    """Worker function for multiprocessing - must be at module level."""
+    path, config_dict = args
+    from ocr_pipeline.ocr.engine import ImagePreprocessor, OCREngine, compute_image_hash
+
+    preprocessor = ImagePreprocessor()
+    ocr_engine = OCREngine()
+
+    try:
+        image_hash = compute_image_hash(path)
+        prep_start = time.time()
+        image = preprocessor.preprocess(path)
+        preprocessing_time = time.time() - prep_start
+
+        ocr_start = time.time()
+        result = ocr_engine.process(image)
+        ocr_time = time.time() - ocr_start
+
+        return ProcessedImage(
+            path=path,
+            image_hash=image_hash,
+            ocr_result=result,
+            preprocessing_time=preprocessing_time,
+            ocr_time=ocr_time,
+            success=True,
+        )
+    except Exception as e:
+        return ProcessedImage(
+            path=path,
+            image_hash="",
+            ocr_result=None,
+            preprocessing_time=0,
+            ocr_time=0,
+            success=False,
+            error=str(e),
+        )
+
+
+class ParallelProcessor:
+    def __init__(self, workers: int | None = None):
+        self.workers = workers or settings.processing.workers
+        self.batch_size = settings.processing.batch_size
+        self.retry_attempts = settings.processing.retry_attempts
+        self.timeout = settings.processing.timeout_per_image
+
+    def find_images(self) -> list[Path]:
+        all_images = []
+        for base_path_str in settings.input.paths:
+            base_path = Path(base_path_str).expanduser()
+            if not base_path.exists():
+                logger.warning("input_path_not_found", path=str(base_path))
+                continue
+
+            for pattern in settings.input.patterns:
+                if settings.input.recursive:
+                    images = list(base_path.rglob(pattern))
+                else:
+                    images = list(base_path.glob(pattern))
+                all_images.extend(images)
+
+        unique_images = []
+        seen = set()
+        for img in all_images:
+            if img.is_file() and img.suffix.lower() in (".png", ".jpg", ".jpeg", ".tiff", ".bmp"):
+                if img not in seen:
+                    seen.add(img)
+                    unique_images.append(img)
+
+        logger.info("found_images", count=len(unique_images))
+        return sorted(unique_images)
+
+    def process(self, image_paths: list[Path] | None = None) -> list[ProcessedImage]:
+        if image_paths is None:
+            image_paths = self.find_images()
+
+        if not image_paths:
+            logger.warning("no_images_to_process")
+            return []
+
+        results = []
+        failed = []
+
+        with ProgressTracker(len(image_paths), "OCR Processing") as tracker:
+            with ProcessPoolExecutor(max_workers=self.workers) as executor:
+                future_to_path = {
+                    executor.submit(_process_single_image, (path, {})): path for path in image_paths
+                }
+
+                for future in as_completed(future_to_path):
+                    path = future_to_path[future]
+                    try:
+                        result = future.result(timeout=self.timeout)
+                        if not result.success and self.retry_attempts > 0:
+                            for attempt in range(self.retry_attempts):
+                                logger.warning("retrying", path=str(path), attempt=attempt + 1)
+                                retry_result = _process_single_image((path, {}))
+                                if retry_result.success:
+                                    result = retry_result
+                                    break
+                        results.append(result)
+                        tracker.update(success=result.success)
+                        tracker.log_result(path, result.success, result.error)
+                    except Exception as e:
+                        logger.error("processing_failed", path=str(path), error=str(e))
+                        results.append(
+                            ProcessedImage(
+                                path=path,
+                                image_hash="",
+                                ocr_result=None,
+                                preprocessing_time=0,
+                                ocr_time=0,
+                                success=False,
+                                error=str(e),
+                            )
+                        )
+                        tracker.update(success=False)
+                        tracker.log_result(path, False, str(e))
+
+        successful = [r for r in results if r.success]
+        failed = [r for r in results if not r.success]
+        logger.info("processing_complete", successful=len(successful), failed=len(failed))
+        return results
--- a/src/ocr_pipeline/output/init.py
+++ b/src/ocr_pipeline/output/init.py
@@ -0,0 +1,8 @@
+from __future__ import annotations
+
+from .markdown import MarkdownWriter, OutputMetadata
+
+__all__ = [
+    "MarkdownWriter",
+    "OutputMetadata",
+]
--- a/src/ocr_pipeline/output/pycache/init.cpython-313.pyc
+++ b/src/ocr_pipeline/output/pycache/init.cpython-313.pyc
--- a/src/ocr_pipeline/output/pycache/markdown.cpython-313.pyc
+++ b/src/ocr_pipeline/output/pycache/markdown.cpython-313.pyc
--- a/src/ocr_pipeline/output/markdown.py
+++ b/src/ocr_pipeline/output/markdown.py
@@ -0,0 +1,143 @@
+from __future__ import annotations
+
+import datetime
+from pathlib import Path
+from typing import Any
+
+import yaml
+
+from ocr_pipeline.config import settings
+from ocr_pipeline.utils.logging import get_logger
+
+logger = get_logger(__name__)
+
+
+class MarkdownWriter:
+    def __init__(self):
+        self.base_dir = Path(settings.output.base_directory).expanduser().resolve()
+        self.organize_by = settings.output.organize_by
+        self.frontmatter_fields = settings.output.frontmatter.fields
+
+    def _get_output_dir(self, timestamp: str) -> Path:
+        dt = datetime.datetime.fromisoformat(timestamp.replace("Z", "+00:00"))
+
+        if self.organize_by == "date_run":
+            date_dir = self.base_dir / dt.strftime("%Y-%m-%d")
+            run_dirs = sorted(date_dir.glob("run_*"))
+            run_num = len(run_dirs) + 1
+            return date_dir / f"run_{run_num:03d}"
+        elif self.organize_by == "source_dir":
+            return self.base_dir / "by_source"
+        else:
+            return self.base_dir
+
+    def write_chunk(
+        self,
+        content: str,
+        metadata: dict[str, Any],
+        figures: list[Any] = None,
+        tables: list[Any] = None,
+        entities: Any = None,
+        citations: list[Any] = None,
+    ) -> Path:
+        output_dir = self._get_output_dir(metadata.get("timestamp", ""))
+        output_dir.mkdir(parents=True, exist_ok=True)
+
+        source_path = Path(metadata.get("source_path", "unknown"))
+        safe_name = "".join(c for c in source_path.stem if c.isalnum() or c in "-_")[:80]
+        chunk_idx = metadata.get("chunk_index", 0)
+        total_chunks = metadata.get("total_chunks", 1)
+
+        if total_chunks > 1:
+            filename = f"{safe_name}_chunk_{chunk_idx:03d}.md"
+        else:
+            filename = f"{safe_name}.md"
+
+        output_path = output_dir / filename
+
+        frontmatter = self._build_frontmatter(metadata)
+        body = self._build_body(content, figures, tables, entities, citations)
+
+        with open(output_path, "w") as f:
+            f.write("---\n")
+            yaml.dump(frontmatter, f, sort_keys=False, allow_unicode=True)
+            f.write("---\n\n")
+            f.write(body)
+
+        logger.debug("markdown_written", path=str(output_path))
+        return output_path
+
+    def _build_frontmatter(self, metadata: dict[str, Any]) -> dict[str, Any]:
+        fm = {}
+        for field in self.frontmatter_fields:
+            if field in metadata:
+                fm[field] = metadata[field]
+        return fm
+
+    def _build_body(
+        self,
+        content: str,
+        figures: list[Any] = None,
+        tables: list[Any] = None,
+        entities: Any = None,
+        citations: list[Any] = None,
+    ) -> str:
+        parts = []
+
+        source_name = Path(content.split("\n")[0]).name if content else "Unknown"
+        parts.append(f"# Screenshot: {source_name}")
+        parts.append("")
+
+        if content:
+            parts.append("## Extracted Text")
+            parts.append("")
+            parts.append(content.strip())
+            parts.append("")
+
+        if figures:
+            parts.append("## Figures Detected")
+            parts.append("")
+            for i, fig in enumerate(figures):
+                parts.append(f"### Figure {i + 1}")
+                parts.append(f"- **Bounding Box**: {fig.bbox}")
+                parts.append(f"- **Confidence**: {fig.confidence:.2f}")
+                if fig.caption:
+                    parts.append(f"- **Caption**: {fig.caption}")
+                parts.append("")
+
+        if tables:
+            parts.append("## Tables Detected")
+            parts.append("")
+            for i, table in enumerate(tables):
+                parts.append(f"### Table {i + 1}")
+                parts.append(f"- **Bounding Box**: {table.bbox}")
+                parts.append(f"- **Confidence**: {table.confidence:.2f}")
+                if table.markdown:
+                    parts.append("")
+                    parts.append("#### Table Content (Markdown)")
+                    parts.append("")
+                    parts.append(table.markdown)
+                parts.append("")
+
+        if entities and entities.entities:
+            parts.append("## Detected Entities")
+            parts.append("")
+            parts.append("| Entity | Type | Context |")
+            parts.append("|--------|------|---------|")
+            for ent in entities.entities[:50]:
+                context = ent.context[:80].replace("|", "\\|") + (
+                    "..." if len(ent.context) > 80 else ""
+                )
+                parts.append(f"| {ent.text} | {ent.label} | {context} |")
+            parts.append("")
+
+        if citations:
+            parts.append("## Citations Found")
+            parts.append("")
+            parts.append("| Type | Identifier |")
+            parts.append("|------|------------|")
+            for cit in citations:
+                parts.append(f"| {cit.type.upper()} | `{cit.identifier}` |")
+            parts.append("")
+
+        return "\n".join(parts)
--- a/src/ocr_pipeline/pipeline.py
+++ b/src/ocr_pipeline/pipeline.py
@@ -0,0 +1,188 @@
+from __future__ import annotations
+
+import time
+from concurrent.futures import ProcessPoolExecutor, as_completed
+from dataclasses import dataclass, field
+from pathlib import Path
+from typing import Any
+
+from ocr_pipeline.citations import extract_citations
+from ocr_pipeline.config import settings
+from ocr_pipeline.detectors import detect_figures, detect_tables
+from ocr_pipeline.ocr import ImagePreprocessor, OCREngine, compute_image_hash
+from ocr_pipeline.ocr.parallel import ParallelProcessor
+from ocr_pipeline.output import MarkdownWriter
+from ocr_pipeline.postprocess import chunk_text, clean_text, extract_entities
+from ocr_pipeline.utils.db import get_db
+from ocr_pipeline.utils.logging import get_logger
+from ocr_pipeline.watch import Watcher
+
+try:
+    from rich.console import Console
+
+    _console = Console()
+except ImportError:
+    _console = None
+
+logger = get_logger(__name__)
+
+
+@dataclass
+class PipelineResult:
+    total_images: int = 0
+    successful: int = 0
+    failed: int = 0
+    chunks_created: int = 0
+    output_files: list[Path] = field(default_factory=list)
+    errors: list[dict[str, Any]] = field(default_factory=list)
+
+
+class OCRPipeline:
+    def __init__(self):
+        self.engine = OCREngine()
+        self.preprocessor = ImagePreprocessor()
+        self.writer = MarkdownWriter()
+        self.db = get_db()
+
+    def _process_single(self, image_path: Path) -> list[dict[str, Any]]:
+        image = self.preprocessor.preprocess(image_path)
+        result = self.engine.process(image)
+
+        image_hash = compute_image_hash(image_path)
+        if self.db.is_processed(image_hash, image_path):
+            logger.debug("already_processed", path=str(image_path))
+            return []
+
+        cleaned = clean_text(result.text)
+        chunks = chunk_text(cleaned.text)
+        entities = extract_entities(cleaned.text)
+        figures = detect_figures(image)
+        tables = detect_tables(image)
+        citations = extract_citations(cleaned.text)
+
+        output_files = []
+        for i, chunk in enumerate(chunks):
+            output_path = self.writer.write_chunk(
+                source_path=image_path,
+                source_hash=image_hash,
+                text=chunk.content,
+                entities=[e.text for e in entities.entities],
+                figures=figures,
+                tables=tables,
+                citations=[f"{c.type}:{c.identifier}" for c in citations],
+                chunk_index=i,
+                total_chunks=len(chunks),
+                ocr_engine=result.engine,
+                ocr_confidence=result.confidence,
+                language=result.language,
+            )
+            output_files.append(output_path)
+
+        self.db.mark_processed(
+            file_hash=image_hash,
+            path=str(image_path),
+            timestamp=time.time(),
+            engine=result.engine,
+            confidence=result.confidence,
+            chunks=len(chunks),
+            run_id=0,
+        )
+
+        return [{"path": p, "chunks": len(chunks)} for p in output_files]
+
+    def process_all(self, image_paths: list[Path] | None = None) -> PipelineResult:
+        """Process images through the full pipeline."""
+        if image_paths is None:
+            processor = ParallelProcessor()
+            results = processor.process()
+            image_paths = [r.path for r in results if r.success]
+        else:
+            # Validate paths exist
+            image_paths = [p for p in image_paths if p.exists()]
+
+        result = PipelineResult(total_images=len(image_paths))
+
+        if not image_paths:
+            return result
+
+        with ProcessPoolExecutor(max_workers=settings.processing.workers) as executor:
+            future_to_path = {executor.submit(self._process_single, p): p for p in image_paths}
+
+            for future in as_completed(future_to_path):
+                path = future_to_path[future]
+                try:
+                    output_info = future.result(timeout=settings.processing.timeout_per_image)
+                    result.successful += 1
+                    result.chunks_created += sum(info["chunks"] for info in output_info)
+                    result.output_files.extend([info["path"] for info in output_info])
+                except Exception as e:
+                    result.failed += 1
+                    result.errors.append({"path": str(path), "error": str(e)})
+                    logger.error("pipeline_failed", path=str(path), error=str(e))
+
+        return result
+
+    def watch(self):
+        def process_callback(path: Path):
+            self._process_single(path)
+
+        watcher = Watcher(process_callback)
+        watcher.start()
+
+        try:
+            while True:
+                time.sleep(1)
+        except KeyboardInterrupt:
+            watcher.stop()
+
+    def get_stats(self) -> dict[str, Any]:
+        db_stats = self.db.get_stats()
+        return {
+            "total_files_processed": db_stats.get("total_processed", 0),
+            "total_runs": db_stats.get("total_runs", 0),
+            "latest_run": db_stats.get("latest_run"),
+        }
+
+
+def migrate_notebook(notebook_path: Path, pipeline: OCRPipeline) -> dict[str, Any]:
+    import json
+
+    with open(notebook_path) as f:
+        nb = json.load(f)
+
+    image_paths = set()
+    for cell in nb.get("cells", []):
+        if cell.get("cell_type") == "code":
+            for output in cell.get("outputs", []):
+                if output.get("output_type") == "stream":
+                    for line in output.get("text", []):
+                        parts = line.split()
+                        for p in parts:
+                            if "SCR-" in p and (
+                                p.endswith(".png") or p.endswith(".jpg") or p.endswith(".jpeg")
+                            ):
+                                image_paths.add(p)
+
+    unique_paths = [Path(p) for p in image_paths if Path(p).exists()]
+
+    try:
+        from rich.console import Console
+
+        console = Console()
+        console.print(f"Found {len(unique_paths)} images in notebook")
+    except ImportError:
+        logger.info("Found %d images in notebook", len(unique_paths))
+
+    if not unique_paths:
+        return {"total": 0, "successful": 0, "failed": 0, "errors": []}
+
+    result = pipeline.process_all(unique_paths)
+
+    return {
+        "total": result.total_images,
+        "successful": result.successful,
+        "failed": result.failed,
+        "chunks_created": result.chunks_created,
+        "output_files": len(result.output_files),
+        "errors": result.errors,
+    }
--- a/src/ocr_pipeline/postprocess/init.py
+++ b/src/ocr_pipeline/postprocess/init.py
@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+from .chunk import SemanticChunker, TextChunk, chunk_text
+from .clean import CleanResult, clean_text
+from .entities import Entity, SciSpacyNER, extract_entities, summarize_entities
+
+__all__ = [
+    "clean_text",
+    "CleanResult",
+    "chunk_text",
+    "TextChunk",
+    "SemanticChunker",
+    "extract_entities",
+    "summarize_entities",
+    "Entity",
+    "SciSpacyNER",
+]
--- a/src/ocr_pipeline/postprocess/pycache/init.cpython-313.pyc
+++ b/src/ocr_pipeline/postprocess/pycache/init.cpython-313.pyc
--- a/src/ocr_pipeline/postprocess/pycache/chunk.cpython-313.pyc
+++ b/src/ocr_pipeline/postprocess/pycache/chunk.cpython-313.pyc
--- a/src/ocr_pipeline/postprocess/pycache/clean.cpython-313.pyc
+++ b/src/ocr_pipeline/postprocess/pycache/clean.cpython-313.pyc
--- a/src/ocr_pipeline/postprocess/pycache/entities.cpython-313.pyc
+++ b/src/ocr_pipeline/postprocess/pycache/entities.cpython-313.pyc
--- a/src/ocr_pipeline/postprocess/chunk.py
+++ b/src/ocr_pipeline/postprocess/chunk.py
@@ -0,0 +1,61 @@
+from __future__ import annotations
+
+from dataclasses import dataclass
+from typing import Any
+
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+
+from ocr_pipeline.config import settings
+from ocr_pipeline.utils.logging import get_logger
+
+logger = get_logger(__name__)
+
+
+@dataclass
+class TextChunk:
+    content: str
+    chunk_index: int
+    start_char: int
+    end_char: int
+    metadata: dict[str, Any]
+
+
+class SemanticChunker:
+    def __init__(self):
+        self.splitter = RecursiveCharacterTextSplitter(
+            chunk_size=settings.chunking.chunk_size,
+            chunk_overlap=settings.chunking.chunk_overlap,
+            separators=settings.chunking.separators,
+            keep_separator=settings.chunking.keep_separator,
+            length_function=len,
+        )
+
+    def chunk(self, text: str, base_metadata: dict[str, Any] | None = None) -> list[TextChunk]:
+        if not text or not text.strip():
+            return []
+
+        base_metadata = base_metadata or {}
+        docs = self.splitter.create_documents([text], metadatas=[base_metadata])
+
+        chunks = []
+        for i, doc in enumerate(docs):
+            start = text.find(doc.page_content)
+            if start == -1:
+                start = 0
+            chunks.append(
+                TextChunk(
+                    content=doc.page_content,
+                    chunk_index=i,
+                    start_char=start,
+                    end_char=start + len(doc.page_content),
+                    metadata={**doc.metadata, "chunk_index": i, "total_chunks": len(docs)},
+                )
+            )
+
+        logger.debug("chunked_text", chunks=len(chunks), original_length=len(text))
+        return chunks
+
+
+def chunk_text(text: str, metadata: dict[str, Any] | None = None) -> list[TextChunk]:
+    chunker = SemanticChunker()
+    return chunker.chunk(text, metadata)
--- a/src/ocr_pipeline/postprocess/clean.py
+++ b/src/ocr_pipeline/postprocess/clean.py
@@ -0,0 +1,163 @@
+from __future__ import annotations
+
+import re
+import unicodedata
+from dataclasses import dataclass
+
+from ocr_pipeline.utils.logging import get_logger
+
+logger = get_logger(__name__)
+
+
+@dataclass
+class CleanResult:
+    text: str
+    original_length: int
+    cleaned_length: int
+    operations: list[str]
+
+
+def fix_encoding(text: str) -> str:
+    try:
+        return text.encode("latin-1").decode("utf-8")
+    except UnicodeError:
+        return text
+
+
+def normalize_unicode(text: str) -> str:
+    return unicodedata.normalize("NFC", text)
+
+
+def fix_ligatures(text: str) -> str:
+    ligatures = {
+        "ﬁ": "fi",
+        "ﬂ": "fl",
+        "ﬀ": "ff",
+        "ﬃ": "ffi",
+        "ﬄ": "ffl",
+        "ﬅ": "ft",
+        "ﬆ": "st",
+        "Ꜳ": "AA",
+        "ꜳ": "aa",
+        "Ꜵ": "AO",
+        "ꜵ": "ao",
+        "Ꜷ": "AU",
+        "ꜷ": "au",
+        "Ꜹ": "AV",
+        "ꜹ": "av",
+        "Ꜻ": "AV",
+        "ꜻ": "av",
+        "Ꜽ": "AY",
+        "ꜽ": "ay",
+    }
+    for lig, repl in ligatures.items():
+        text = text.replace(lig, repl)
+    return text
+
+
+def remove_control_chars(text: str) -> str:
+    return "".join(ch for ch in text if unicodedata.category(ch)[0] != "C" or ch in "\n\t")
+
+
+def fix_hyphenation(text: str) -> str:
+    return re.sub(r"(\w+)-\n(\w+)", r"\1\2", text)
+
+
+def normalize_whitespace(text: str) -> str:
+    text = re.sub(r"[ \t]+", " ", text)
+    text = re.sub(r"\n{3,}", "\n\n", text)
+    return text.strip()
+
+
+def remove_ocr_artifacts(text: str) -> str:
+    patterns = [
+        r"^\s*[\|┃║]\s*$",
+        r"^\s*[─━]\s*$",
+        r"^\s*[┌┐└┘├┤┬┴┼]\s*$",
+        r"(.)\1{10,}",
+        r"^\s*[•·▪▫◦‣⁃]\s*$",
+    ]
+    lines = text.split("\n")
+    cleaned = []
+    for line in lines:
+        if any(re.match(p, line.strip()) for p in patterns):
+            continue
+        cleaned.append(line)
+    return "\n".join(cleaned)
+
+
+def fix_scientific_notation(text: str) -> str:
+    text = re.sub(r"(\d)\s*[×xX]\s*10\s*[\^]?\s*([+-]?\d+)", r"\1×10^\2", text)
+    text = re.sub(r"(\d)\s*[×xX]\s*10\s*([+-]?\d+)", r"\1×10^\2", text)
+    return text
+
+
+def normalize_units(text: str) -> str:
+    unit_fixes = {
+        r"(\d)\s*u[mM]\b": r"\1 µm",
+        r"(\d)\s*[µu]g\b": r"\1 µg",
+        r"(\d)\s*[mn]g\b": r"\1 mg",
+        r"(\d)\s*[kK]b\b": r"\1 kb",
+        r"(\d)\s*[mM]b\b": r"\1 Mb",
+        r"(\d)\s*[gG]b\b": r"\1 Gb",
+        r"(\d)\s*[pP]?[mM]\b": r"\1 pm",
+        r"(\d)\s*[nN][mM]\b": r"\1 nm",
+        r"(\d)\s*[cC][mM]\b": r"\1 cm",
+        r"(\d)\s*[mM][lL]\b": r"\1 mL",
+        r"(\d)\s*[uU][lL]\b": r"\1 µL",
+        r"(\d)\s*°\s*[cC]\b": r"\1 °C",
+    }
+    for pattern, repl in unit_fixes.items():
+        text = re.sub(pattern, repl, text)
+    return text
+
+
+def fix_line_breaks(text: str) -> str:
+    text = re.sub(r"(?<!\n)\n(?!\n)", " ", text)
+    text = re.sub(r"\n{3,}", "\n\n", text)
+    return text
+
+
+def clean_text(text: str) -> CleanResult:
+    original_len = len(text)
+    operations = []
+
+    text = fix_encoding(text)
+    operations.append("fix_encoding")
+
+    text = normalize_unicode(text)
+    operations.append("normalize_unicode")
+
+    text = fix_ligatures(text)
+    operations.append("fix_ligatures")
+
+    text = remove_control_chars(text)
+    operations.append("remove_control_chars")
+
+    text = fix_hyphenation(text)
+    operations.append("fix_hyphenation")
+
+    text = normalize_whitespace(text)
+    operations.append("normalize_whitespace")
+
+    text = remove_ocr_artifacts(text)
+    operations.append("remove_ocr_artifacts")
+
+    text = fix_scientific_notation(text)
+    operations.append("fix_scientific_notation")
+
+    text = normalize_units(text)
+    operations.append("normalize_units")
+
+    text = fix_line_breaks(text)
+    operations.append("fix_line_breaks")
+
+    cleaned_len = len(text)
+    logger.debug("text_cleaned", original=original_len, cleaned=cleaned_len, ops=operations)
+
+    return CleanResult(
+        text=text,
+        original_length=original_len,
+        cleaned_length=cleaned_len,
+        operations=operations,
+    )
--- a/src/ocr_pipeline/postprocess/entities.py
+++ b/src/ocr_pipeline/postprocess/entities.py
@@ -0,0 +1,227 @@
+from __future__ import annotations
+
+import re
+from dataclasses import dataclass
+
+import spacy
+
+from ocr_pipeline.config import settings
+from ocr_pipeline.utils.logging import get_logger
+
+logger = get_logger(__name__)
+
+
+@dataclass
+class Entity:
+    text: str
+    label: str
+    start: int
+    end: int
+    confidence: float
+    context: str = ""
+
+
+@dataclass
+class EntityResult:
+    entities: list[Entity]
+    text: str
+    total_entities: int
+    entities_by_type: dict[str, int]
+
+
+class ScientificEntityRecognizer:
+    def __init__(self):
+        self._nlp = None
+        self._initialized = False
+
+    def _init(self):
+        if self._initialized:
+            return
+        try:
+
+            self._nlp = spacy.load(settings.entities.model)
+            self._nlp.max_length = 2_000_000
+            self._initialized = True
+            logger.info("scispacy_model_loaded", model=settings.entities.model)
+        except Exception as e:
+            logger.warning("scispacy_load_failed", error=str(e), model=settings.entities.model)
+            self._fallback_init()
+            self._initialized = True
+
+    def _fallback_init(self):
+        try:
+            self._nlp = spacy.blank("en")
+            self._nlp.add_pipe("sentencizer")
+            logger.info("using_fallback_blank_model")
+        except Exception:
+            self._nlp = None
+
+    def extract(self, text: str) -> EntityResult:
+        if not settings.entities.enabled:
+            return EntityResult(entities=[], text=text, total_entities=0, entities_by_type={})
+
+        self._init()
+        if self._nlp is None:
+            return EntityResult(entities=[], text=text, total_entities=0, entities_by_type={})
+
+        try:
+            doc = self._nlp(text)
+            entities = []
+
+            for ent in doc.ents:
+                if ent.label_ in settings.entities.types:
+                    context_start = max(0, ent.start_char - 100)
+                    context_end = min(len(text), ent.end_char + 100)
+                    context = text[context_start:context_end].strip()
+
+                    entities.append(
+                        Entity(
+                            text=ent.text,
+                            label=ent.label_,
+                            start=ent.start_char,
+                            end=ent.end_char,
+                            confidence=0.9,
+                            context=context,
+                        )
+                    )
+
+            entities = (
+                self._merge_entities(entities, text) if settings.entities.merge_entities else entities
+            )
+            entities = self._deduplicate_entities(entities)
+
+            by_type = {}
+            for e in entities:
+                by_type[e.label] = by_type.get(e.label, 0) + 1
+
+            return EntityResult(
+                entities=entities,
+                text=text,
+                total_entities=len(entities),
+                entities_by_type=by_type,
+            )
+        except Exception as e:
+            logger.warning("entity_extraction_failed", error=str(e))
+            return EntityResult(entities=[], text=text, total_entities=0, entities_by_type={})
+
+    def _merge_entities(self, entities: list[Entity], text: str) -> list[Entity]:
+        if not entities:
+            return entities
+
+        entities.sort(key=lambda e: (e.start, -e.end))
+        merged = [entities[0]]
+
+        for ent in entities[1:]:
+            last = merged[-1]
+            if ent.start <= last.end + 5 and ent.label == last.label:
+                merged[-1] = Entity(
+                    text=last.text + text[last.end : ent.start] + ent.text,
+                    label=last.label,
+                    start=last.start,
+                    end=ent.end,
+                    confidence=max(last.confidence, ent.confidence),
+                    context=last.context,
+                )
+            else:
+                merged.append(ent)
+
+        return merged
+
+    def _deduplicate_entities(self, entities: list[Entity]) -> list[Entity]:
+        seen = set()
+        unique = []
+        for ent in entities:
+            key = (ent.text.lower(), ent.label)
+            if key not in seen:
+                seen.add(key)
+                unique.append(ent)
+        return unique
+
+
+class RegexEntityRecognizer:
+    PATTERNS = {
+        "GENE": [
+            r"\b(?:[A-Z]{2,5}\d*[A-Z]?|[A-Z]{3,10})\b",
+            r"\b(?:BRCA[12]|TP53|EGFR|KRAS|MYC|PIK3CA|PTEN|APC|MLH1|MSH2|MSH6|PMS2|ATM|CHEK2|PALB2|RAD51|BARD1|BRIP1|CDH1|STK11|VHL|WT1|NF1|NF2|RB1|MEN1|RET|MET|ALK|ROS1|NTRK[123]|BRAF|NRAS|HRAS|IDH[12]|TERT|FGFR[1234]|KIT|PDGFRA|FLT3|JAK2|MPL|CALR|ASXL1|DNMT3A|TET2|SRSF2|U2AF1|SF3B1|ZRSR2|CBL|CSF3R|RUNX1|CEBPA|NPM1|FLT3|DNMT3A|IDH[12]|TET2|ASXL1|SRSF2|U2AF1|SF3B1|ZRSR2|CBL|CSF3R|RUNX1|CEBPA|NPM1)\b",
+        ],
+        "PROTEIN": [
+            r"\b(?:p53|p21|p16|CDK[1246]|cyclin\s+[ADEB]|Rb|E2F|MDM2|BAX|BCL[2XL]|CASP[389]|PARP|H2AX|ATM|ATR|CHK[12]|BRCA[12]|RAD51|FANCD2|FANCI|PCNA|RPA|MSH[26]|MLH1|PMS2|EXO1|MRE11|RAD50|NBN|CTIP|BRCA1|BRCA2|PALB2|RAD51C|RAD51D|BRIP1|BARD1)\b",
+        ],
+        "CHEMICAL": [
+            r"\b(?:DMSO|PBS|EDTA|Tris|HEPES|SDS|DTT|BME|NaCl|KCl|MgCl2|CaCl2|NaOH|HCl|H2SO4|HNO3|EtOH|MeOH|IPA|DMSO|DMF|DMA|THF|DCM|CHCl3|CH2Cl2|EtOAc|hexane|pentane|acetone|acetonitrile|water|H2O|buffer|media|serum|FBS|BSA|pen/strep|penicillin|streptomycin|trypsin|EDTA|collagenase|dispase|accutase)\b",
+            r"\b(?:[A-Z][a-z]?\d*(?:\([^)]+\))?(?:\s*[+\-]\s*[A-Z][a-z]?\d*(?:\([^)]+\))?)*)\b",
+        ],
+        "CONCENTRATION": [
+            r"\b\d+(?:\.\d+)?\s*(?:[µumMpnc]?[Mm]|[µumMpnc]?[Mm]/[Ll]|[µumMpnc]?[Mm]\s*[Ll]?|[µumMpnc]?[gG]\s*/\s*[Ll]|[µumMpnc]?[gG]\s*/\s*[mM][lL]|[µumMpnc]?[gG]\s*/\s*[dL]|[µumMpnc]?[gG]\s*/\s*[mM][lL]|[µumMpnc]?[gG]\s*/\s*[dD][lL]|[µumMpnc]?[gG]\s*/\s*100\s*[mM][lL]|[µumMpnc]?[gG]\s*/\s*[kK][gG])\b",
+        ],
+    }
+
+    @classmethod
+    def extract(cls, text: str) -> list[Entity]:
+        entities = []
+        for label, patterns in cls.PATTERNS.items():
+            for pattern in patterns:
+                for match in re.finditer(pattern, text, re.IGNORECASE):
+                    context_start = max(0, match.start() - 50)
+                    context_end = min(len(text), match.end() + 50)
+                    entities.append(
+                        Entity(
+                            text=match.group(),
+                            label=label,
+                            start=match.start(),
+                            end=match.end(),
+                            confidence=0.7,
+                            context=text[context_start:context_end].strip(),
+                        )
+                    )
+        return entities
+
+
+def extract_entities(text: str) -> EntityResult:
+    recognizer = ScientificEntityRecognizer()
+    result = recognizer.extract(text)
+
+    regex_entities = RegexEntityRecognizer.extract(text)
+    all_entities = result.entities + regex_entities
+    all_entities = _deduplicate(all_entities)
+
+    by_type = {}
+    for e in all_entities:
+        by_type[e.label] = by_type.get(e.label, 0) + 1
+
+    return EntityResult(
+        entities=all_entities,
+        text=text,
+        total_entities=len(all_entities),
+        entities_by_type=by_type,
+    )
+
+
+def _deduplicate(entities: list[Entity]) -> list[Entity]:
+    seen = set()
+    unique = []
+    for ent in sorted(entities, key=lambda e: (e.start, -e.end)):
+        key = (ent.text.lower(), ent.label)
+        if key not in seen:
+            seen.add(key)
+            unique.append(ent)
+    return unique
+
+
+def format_entities_for_markdown(result: EntityResult) -> str:
+    if not result.entities:
+        return "No entities detected."
+
+    lines = [
+        "## Detected Entities",
+        "",
+        "| Entity | Type | Count | Contexts |",
+        "|--------|------|-------|----------|",
+    ]
+
+    for label in sorted(result.entities_by_type.keys()):
+        ents = [e for e in result.entities if e.label == label]
+        contexts = {e.context[:80] for e in ents[:3]}
+        lines.append(f"| {ents[0].text} | {label} | {len(ents)} | {'; '.join(contexts)} |")
+
+    return "\n".join(lines)
--- a/src/ocr_pipeline/utils/init.py
+++ b/src/ocr_pipeline/utils/init.py
@@ -0,0 +1,18 @@
+from __future__ import annotations
+
+from .db import Database, ProcessedFile, get_db
+from .logging import ProgressTracker, create_progress, get_logger, setup_logging
+from .migrate import compute_file_hash, migrate_notebook, parse_notebook_for_images
+
+__all__ = [
+    "setup_logging",
+    "get_logger",
+    "create_progress",
+    "ProgressTracker",
+    "parse_notebook_for_images",
+    "migrate_notebook",
+    "compute_file_hash",
+    "get_db",
+    "Database",
+    "ProcessedFile",
+]
--- a/src/ocr_pipeline/utils/pycache/init.cpython-313.pyc
+++ b/src/ocr_pipeline/utils/pycache/init.cpython-313.pyc
--- a/src/ocr_pipeline/utils/pycache/db.cpython-313.pyc
+++ b/src/ocr_pipeline/utils/pycache/db.cpython-313.pyc
--- a/src/ocr_pipeline/utils/pycache/logging.cpython-313.pyc
+++ b/src/ocr_pipeline/utils/pycache/logging.cpython-313.pyc
--- a/src/ocr_pipeline/utils/pycache/migrate.cpython-313.pyc
+++ b/src/ocr_pipeline/utils/pycache/migrate.cpython-313.pyc
--- a/src/ocr_pipeline/utils/db.py
+++ b/src/ocr_pipeline/utils/db.py
@@ -0,0 +1,166 @@
+from __future__ import annotations
+
+import sqlite3
+import time
+from contextlib import contextmanager
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Any
+
+from ocr_pipeline.utils.logging import get_logger
+
+logger = get_logger(__name__)
+
+
+@dataclass
+class ProcessedFile:
+    path: str
+    hash: str
+    timestamp: float
+    run_id: int
+    size: int
+    mtime: float
+
+
+class Database:
+    def __init__(self, db_path: Path):
+        self.db_path = Path(db_path).expanduser().resolve()
+        self.db_path.parent.mkdir(parents=True, exist_ok=True)
+        self._init_db()
+
+    @contextmanager
+    def _conn(self):
+        conn = sqlite3.connect(self.db_path)
+        conn.row_factory = sqlite3.Row
+        try:
+            yield conn
+        finally:
+            conn.close()
+
+    def _init_db(self):
+        with self._conn() as conn:
+            conn.executescript("""
+                CREATE TABLE IF NOT EXISTS processed_files (
+                    hash TEXT PRIMARY KEY,
+                    path TEXT NOT NULL,
+                    timestamp TEXT NOT NULL,
+                    ocr_engine TEXT,
+                    confidence REAL,
+                    chunks_created INTEGER DEFAULT 0,
+                    created_at REAL DEFAULT (strftime('%s', 'now'))
+                );
+
+                CREATE TABLE IF NOT EXISTS runs (
+                    id INTEGER PRIMARY KEY AUTOINCREMENT,
+                    date TEXT NOT NULL,
+                    run_number INTEGER NOT NULL,
+                    total_images INTEGER DEFAULT 0,
+                    successful INTEGER DEFAULT 0,
+                    failed INTEGER DEFAULT 0,
+                    chunks_created INTEGER DEFAULT 0,
+                    started_at REAL DEFAULT (strftime('%s', 'now')),
+                    completed_at REAL,
+                    UNIQUE(date, run_number)
+                );
+
+                CREATE INDEX IF NOT EXISTS idx_processed_hash ON processed_files(hash);
+                CREATE INDEX IF NOT EXISTS idx_runs_date ON runs(date);
+            """)
+
+    def is_processed(self, file_hash: str, path: Path | str | None = None) -> bool:
+        """Check if a file is already processed by hash (and optionally path)."""
+        with self._conn() as conn:
+            if path:
+                cursor = conn.execute(
+                    "SELECT 1 FROM processed_files WHERE hash = ? AND path = ?",
+                    (file_hash, str(path)),
+                )
+            else:
+                cursor = conn.execute(
+                    "SELECT 1 FROM processed_files WHERE hash = ?",
+                    (file_hash,),
+                )
+            return cursor.fetchone() is not None
+
+    def get_processed(self, file_hash: str) -> dict[str, Any] | None:
+        with self._conn() as conn:
+            cursor = conn.execute("SELECT * FROM processed_files WHERE hash = ?", (file_hash,))
+            row = cursor.fetchone()
+            return dict(row) if row else None
+
+    def mark_processed(
+        self,
+        file_hash: str,
+        path: str,
+        timestamp: float | str,
+        engine: str = "",
+        confidence: float = 0.0,
+        chunks: int = 0,
+        run_id: int = 0,
+    ):
+        with self._conn() as conn:
+            conn.execute(
+                """INSERT OR REPLACE INTO processed_files
+                   (hash, path, timestamp, ocr_engine, confidence, chunks_created)
+                   VALUES (?, ?, ?, ?, ?, ?)""",
+                (file_hash, path, str(timestamp), engine, confidence, chunks),
+            )
+
+    def get_stats(self, days: int = 30) -> dict[str, Any]:
+        with self._conn() as conn:
+            cursor = conn.execute("""
+                SELECT
+                    COUNT(*) as total_processed,
+                    SUM(chunks_created) as total_chunks,
+                    AVG(confidence) as avg_confidence
+                FROM processed_files
+            """)
+            row = cursor.fetchone()
+            return dict(row) if row else {}
+
+    def start_run(self, date: str | None = None) -> int:
+        if date is None:
+            date = time.strftime("%Y-%m-%d")
+        with self._conn() as conn:
+            cursor = conn.execute(
+                "SELECT COALESCE(MAX(run_number), 0) + 1 FROM runs WHERE date = ?", (date,)
+            )
+            run_number = cursor.fetchone()[0]
+            conn.execute(
+                "INSERT INTO runs (date, run_number, total_images) VALUES (?, ?, 0)",
+                (date, run_number),
+            )
+            return run_number
+
+    def complete_run(self, date: str, run_number: int, stats: dict[str, int]):
+        with self._conn() as conn:
+            conn.execute(
+                """UPDATE runs SET
+                    total_images = ?,
+                    successful = ?,
+                    failed = ?,
+                    chunks_created = ?,
+                    completed_at = ?
+                WHERE date = ? AND run_number = ?""",
+                (
+                    stats.get("total", 0),
+                    stats.get("successful", 0),
+                    stats.get("failed", 0),
+                    stats.get("chunks", 0),
+                    time.time(),
+                    date,
+                    run_number,
+                ),
+            )
+
+
+_db_instance: Database | None = None
+
+
+def get_db() -> Database:
+    global _db_instance
+    if _db_instance is None:
+        from ocr_pipeline.config import settings
+
+        _db_instance = Database(settings.watch_db_path)
+    return _db_instance
--- a/src/ocr_pipeline/utils/logging.py
+++ b/src/ocr_pipeline/utils/logging.py
@@ -0,0 +1,132 @@
+from __future__ import annotations
+
+import logging
+from pathlib import Path
+
+import structlog
+from rich.logging import RichHandler
+from rich.progress import (
+    BarColumn,
+    Progress,
+    SpinnerColumn,
+    TaskProgressColumn,
+    TextColumn,
+    TimeElapsedColumn,
+)
+
+
+def setup_logging(level: str = "INFO", log_file: Path | None = None) -> None:
+    """Configure structlog with rich console output and optional file logging."""
+    log_level = getattr(logging, level.upper(), logging.INFO)
+
+    shared_processors = [
+        structlog.contextvars.merge_contextvars,
+        structlog.stdlib.add_logger_name,
+        structlog.stdlib.add_log_level,
+        structlog.processors.TimeStamper(fmt="ISO"),
+        structlog.processors.StackInfoRenderer(),
+        structlog.dev.set_exc_info,
+    ]
+
+    structlog.configure(
+        processors=shared_processors
+        + [
+            structlog.stdlib.ProcessorFormatter.wrap_for_formatter,
+        ],
+        logger_factory=structlog.stdlib.LoggerFactory(),
+        wrapper_class=structlog.stdlib.BoundLogger,
+        context_class=dict,
+    )
+
+    formatter = structlog.stdlib.ProcessorFormatter(
+        foreign_pre_chain=shared_processors,
+        processors=[
+            structlog.stdlib.ProcessorFormatter.remove_processors_meta,
+            structlog.dev.ConsoleRenderer(colors=True),
+        ],
+    )
+
+    console_handler = RichHandler(
+        rich_tracebacks=True,
+        tracebacks_show_locals=False,
+        markup=True,
+        show_time=False,
+        show_path=False,
+    )
+    console_handler.setFormatter(formatter)
+
+    handlers: list[logging.Handler] = [console_handler]
+
+    if log_file:
+        log_file.parent.mkdir(parents=True, exist_ok=True)
+        file_handler = logging.FileHandler(log_file)
+        file_formatter = structlog.stdlib.ProcessorFormatter(
+            foreign_pre_chain=shared_processors,
+            processors=[
+                structlog.stdlib.ProcessorFormatter.remove_processors_meta,
+                structlog.processors.JSONRenderer(),
+            ],
+        )
+        file_handler.setFormatter(file_formatter)
+        handlers.append(file_handler)
+
+    root_logger = logging.getLogger()
+    root_logger.handlers = []
+    root_logger.setLevel(log_level)
+    for handler in handlers:
+        root_logger.addHandler(handler)
+
+    logging.getLogger("PIL").setLevel(logging.WARNING)
+    logging.getLogger("paddle").setLevel(logging.WARNING)
+    logging.getLogger("torch").setLevel(logging.WARNING)
+    logging.getLogger("transformers").setLevel(logging.WARNING)
+    logging.getLogger("watchdog").setLevel(logging.WARNING)
+
+
+def get_logger(name: str) -> structlog.BoundLogger:
+    return structlog.get_logger(name)
+
+
+def create_progress() -> Progress:
+    return Progress(
+        SpinnerColumn(),
+        TextColumn("[bold blue]{task.description}"),
+        BarColumn(),
+        TaskProgressColumn(),
+        TimeElapsedColumn(),
+        transient=True,
+    )
+
+
+class ProgressTracker:
+    def __init__(self, total: int, description: str = "Processing"):
+        self.total = total
+        self.description = description
+        self.progress: Progress | None = None
+        self.task_id: int | None = None
+        self.completed = 0
+        self.failed = 0
+        self.logger = get_logger(__name__)
+
+    def __enter__(self) -> ProgressTracker:
+        self.progress = create_progress()
+        self.progress.start()
+        self.task_id = self.progress.add_task(self.description, total=self.total)
+        return self
+
+    def __exit__(self, exc_type, exc_val, exc_tb) -> None:
+        if self.progress:
+            self.progress.stop()
+
+    def update(self, advance: int = 1, success: bool = True) -> None:
+        if self.progress and self.task_id is not None:
+            self.progress.update(self.task_id, advance=advance)
+        self.completed += 1
+        if not success:
+            self.failed += 1
+
+    def log_result(self, path: Path, success: bool, error: str | None = None) -> None:
+        if success:
+            self.logger.info("processed", path=str(path))
+        else:
+            self.logger.error("failed", path=str(path), error=error)
--- a/src/ocr_pipeline/utils/migrate.py
+++ b/src/ocr_pipeline/utils/migrate.py
@@ -0,0 +1,92 @@
+from __future__ import annotations
+
+import hashlib
+from pathlib import Path
+from typing import Any
+
+import nbformat
+
+from ocr_pipeline.utils.logging import get_logger
+
+logger = get_logger(__name__)
+
+
+def parse_notebook_for_images(notebook_path: Path) -> list[Path]:
+    image_paths = []
+    try:
+        with open(notebook_path) as f:
+            nb = nbformat.read(f, as_version=4)
+
+        for cell in nb.cells:
+            if cell.cell_type == "code":
+                for output in cell.outputs:
+                    if output.output_type == "stream" and output.name == "stdout":
+                        text = output.text
+                        for line in text.split("\n"):
+                            line = line.strip()
+                            if line.startswith("/") and any(
+                                line.endswith(ext)
+                                for ext in (".png", ".jpg", ".jpeg", ".tiff", ".bmp")
+                            ):
+                                path = Path(line)
+                                if path.exists():
+                                    image_paths.append(path)
+                            elif "SCR-" in line and any(
+                                line.endswith(ext) for ext in (".png", ".jpg", ".jpeg")
+                            ):
+                                for part in line.split():
+                                    if part.startswith("SCR-") and any(
+                                        part.endswith(ext) for ext in (".png", ".jpg", ".jpeg")
+                                    ):
+                                        path = Path("/Users/Aman/Pictures") / part
+                                        if path.exists():
+                                            image_paths.append(path)
+
+    except Exception as e:
+        logger.warning("notebook_parse_failed", path=str(notebook_path), error=str(e))
+
+    return list(set(image_paths))
+
+
+def compute_file_hash(path: Path) -> str:
+    hasher = hashlib.sha256()
+    with open(path, "rb") as f:
+        for chunk in iter(lambda: f.read(8192), b""):
+            hasher.update(chunk)
+    return hasher.hexdigest()
+
+
+def migrate_notebook(notebook_path: Path, pipeline: Any) -> dict[str, Any]:
+    logger.info("starting_notebook_migration", notebook=str(notebook_path))
+
+    image_paths = parse_notebook_for_images(notebook_path)
+    logger.info("found_images_in_notebook", count=len(image_paths))
+
+    results = {
+        "total": len(image_paths),
+        "processed": 0,
+        "skipped": 0,
+        "failed": 0,
+        "errors": [],
+    }
+
+    for img_path in image_paths:
+        try:
+            img_hash = compute_file_hash(img_path)
+            existing = pipeline.db.get_processed(img_hash)
+            if existing:
+                logger.info("skipping_already_processed", path=str(img_path), hash=img_hash[:16])
+                results["skipped"] += 1
+                continue
+
+            logger.info("migrating_image", path=str(img_path))
+            pipeline.process_single(img_path)
+            results["processed"] += 1
+
+        except Exception as e:
+            logger.error("migration_failed", path=str(img_path), error=str(e))
+            results["failed"] += 1
+            results["errors"].append({"path": str(img_path), "error": str(e)})
+
+    logger.info("migration_complete", **results)
+    return results
--- a/src/ocr_pipeline/watch.py
+++ b/src/ocr_pipeline/watch.py
@@ -0,0 +1,144 @@
+from __future__ import annotations
+
+import threading
+import time
+from collections.abc import Callable
+from pathlib import Path
+
+from watchdog.events import FileSystemEventHandler
+from watchdog.observers import Observer
+
+from ocr_pipeline.config import settings
+from ocr_pipeline.utils.db import ProcessedFile, get_db
+from ocr_pipeline.utils.logging import get_logger
+
+logger = get_logger(__name__)
+
+
+class ScreenshotHandler(FileSystemEventHandler):
+    def __init__(self, process_callback: Callable[[Path], None]):
+        self.process_callback = process_callback
+        self.pending: dict[str, float] = {}
+        self.lock = threading.Lock()
+        self.debounce = settings.watch.debounce_seconds
+
+    def _should_process(self, path: Path) -> bool:
+        if not path.is_file():
+            return False
+        if path.suffix.lower() not in (".png", ".jpg", ".jpeg", ".tiff", ".bmp"):
+            return False
+        for pattern in settings.watch.ignore_patterns:
+            if path.match(pattern):
+                return False
+        return True
+
+    def on_created(self, event):
+        if not event.is_directory:
+            path = Path(event.src_path)
+            if self._should_process(path):
+                with self.lock:
+                    self.pending[str(path)] = time.time()
+
+    def on_modified(self, event):
+        if not event.is_directory:
+            path = Path(event.src_path)
+            if self._should_process(path):
+                with self.lock:
+                    self.pending[str(path)] = time.time()
+
+    def check_pending(self) -> list[Path]:
+        now = time.time()
+        to_process = []
+
+        with self.lock:
+            for path_str, timestamp in list(self.pending.items()):
+                if now - timestamp >= self.debounce:
+                    path = Path(path_str)
+                    if path.exists():
+                        to_process.append(path)
+                    del self.pending[path_str]
+
+        return to_process
+
+
+class Watcher:
+    def __init__(self, process_callback: Callable[[Path], None]):
+        self.process_callback = process_callback
+        self.observer: Observer | None = None
+        self.handler: ScreenshotHandler | None = None
+        self.running = False
+        self.db = get_db()
+        self.run_id: int | None = None
+
+    def start(self) -> None:
+        if self.running:
+            return
+
+        self.run_id = self.db.start_run()
+        self.running = True
+
+        self.handler = ScreenshotHandler(self._process_file)
+        self.observer = Observer()
+
+        for base_path_str in settings.input.paths:
+            base_path = Path(base_path_str).expanduser()
+            if base_path.exists():
+                self.observer.schedule(
+                    self.handler, str(base_path), recursive=settings.input.recursive
+                )
+                logger.info("watching_directory", path=str(base_path))
+
+        self.observer.start()
+
+        self._watch_thread = threading.Thread(target=self._watch_loop, daemon=True)
+        self._watch_thread.start()
+
+        logger.info("watcher_started", run_id=self.run_id)
+
+    def _watch_loop(self) -> None:
+        poll_interval = settings.watch.poll_interval
+        processed = 0
+        failed = 0
+
+        while self.running:
+            time.sleep(poll_interval)
+
+            if self.handler:
+                for path in self.handler.check_pending():
+                    try:
+                        self._process_file(path)
+                        processed += 1
+                    except Exception as e:
+                        logger.error("watch_process_failed", path=str(path), error=str(e))
+                        failed += 1
+
+        self.db.end_run(self.run_id, processed, failed)
+        logger.info("watcher_stopped", run_id=self.run_id, processed=processed, failed=failed)
+
+    def _process_file(self, path: Path) -> None:
+        from ocr_pipeline.ocr.engine import compute_image_hash
+
+        file_hash = compute_image_hash(path)
+        if self.db.is_processed(path, file_hash):
+            logger.debug("file_already_processed", path=str(path))
+            return
+
+        logger.info("processing_new_file", path=str(path))
+        self.process_callback(path)
+        self.db.mark_processed(
+            ProcessedFile(
+                path=str(path),
+                hash=file_hash,
+                timestamp=time.time(),
+                run_id=self.run_id,
+                size=path.stat().st_size,
+                mtime=path.stat().st_mtime,
+            )
+        )
+
+    def stop(self) -> None:
+        self.running = False
+        if self.observer:
+            self.observer.stop()
+            self.observer.join(timeout=5)
+        logger.info("watcher_stopped")
--- a/src/ocr_pipeline/watch/init.py
+++ b/src/ocr_pipeline/watch/init.py
@@ -0,0 +1,5 @@
+from __future__ import annotations
+
+from .watcher import Watcher
+
+__all__ = ["Watcher"]
--- a/src/ocr_pipeline/watch/pycache/init.cpython-313.pyc
+++ b/src/ocr_pipeline/watch/pycache/init.cpython-313.pyc
--- a/src/ocr_pipeline/watch/pycache/watcher.cpython-313.pyc
+++ b/src/ocr_pipeline/watch/pycache/watcher.cpython-313.pyc
--- a/src/ocr_pipeline/watch/watcher.py
+++ b/src/ocr_pipeline/watch/watcher.py
@@ -0,0 +1,144 @@
+from __future__ import annotations
+
+import threading
+import time
+from collections.abc import Callable
+from pathlib import Path
+
+from watchdog.events import FileSystemEventHandler
+from watchdog.observers import Observer
+
+from ocr_pipeline.config import settings
+from ocr_pipeline.utils.db import get_db
+from ocr_pipeline.utils.logging import get_logger
+
+logger = get_logger(__name__)
+
+
+class ScreenshotHandler(FileSystemEventHandler):
+    def __init__(self, process_callback: Callable[[Path], None]):
+        self.process_callback = process_callback
+        self.pending: dict[str, float] = {}
+        self.lock = threading.Lock()
+        self.debounce = settings.watch.debounce_seconds
+
+    def _should_process(self, path: Path) -> bool:
+        if not path.is_file():
+            return False
+        if path.suffix.lower() not in (".png", ".jpg", ".jpeg", ".tiff", ".bmp"):
+            return False
+        for pattern in settings.watch.ignore_patterns:
+            if path.match(pattern):
+                return False
+        return True
+
+    def on_created(self, event):
+        if not event.is_directory:
+            path = Path(event.src_path)
+            if self._should_process(path):
+                with self.lock:
+                    self.pending[str(path)] = time.time()
+
+    def on_modified(self, event):
+        if not event.is_directory:
+            path = Path(event.src_path)
+            if self._should_process(path):
+                with self.lock:
+                    self.pending[str(path)] = time.time()
+
+    def check_pending(self) -> list[Path]:
+        now = time.time()
+        to_process = []
+
+        with self.lock:
+            for path_str, timestamp in list(self.pending.items()):
+                if now - timestamp >= self.debounce:
+                    path = Path(path_str)
+                    if path.exists():
+                        to_process.append(path)
+                    del self.pending[path_str]
+
+        return to_process
+
+
+class Watcher:
+    def __init__(self, process_callback: Callable[[Path], None]):
+        self.process_callback = process_callback
+        self.observer: Observer | None = None
+        self.handler: ScreenshotHandler | None = None
+        self.running = False
+        self.db = get_db()
+        self.run_id: int | None = None
+
+    def start(self) -> None:
+        if self.running:
+            return
+
+        self.run_id = self.db.start_run()
+        self.running = True
+
+        self.handler = ScreenshotHandler(self._process_file)
+        self.observer = Observer()
+
+        for base_path_str in settings.input.paths:
+            base_path = Path(base_path_str).expanduser()
+            if base_path.exists():
+                self.observer.schedule(
+                    self.handler, str(base_path), recursive=settings.input.recursive
+                )
+                logger.info("watching_directory", path=str(base_path))
+
+        self.observer.start()
+
+        self._watch_thread = threading.Thread(target=self._watch_loop, daemon=True)
+        self._watch_thread.start()
+
+        logger.info("watcher_started", run_id=self.run_id)
+
+    def _watch_loop(self) -> None:
+        poll_interval = settings.watch.poll_interval
+        processed = 0
+        failed = 0
+
+        while self.running:
+            time.sleep(poll_interval)
+
+            if self.handler:
+                for path in self.handler.check_pending():
+                    try:
+                        self._process_file(path)
+                        processed += 1
+                    except Exception as e:
+                        logger.error("watch_process_failed", path=str(path), error=str(e))
+                        failed += 1
+
+        self.db.complete_run(
+            time.strftime("%Y-%m-%d"),
+            self.run_id,
+            {"total": processed, "successful": processed, "failed": failed, "chunks": 0},
+        )
+        logger.info("watcher_stopped", run_id=self.run_id, processed=processed, failed=failed)
+
+    def _process_file(self, path: Path) -> None:
+        from ocr_pipeline.ocr.engine import compute_image_hash
+
+        file_hash = compute_image_hash(path)
+        if self.db.is_processed(file_hash, path):
+            logger.debug("file_already_processed", path=str(path))
+            return
+
+        logger.info("processing_new_file", path=str(path))
+        self.process_callback(path)
+        self.db.mark_processed(
+            file_hash=file_hash,
+            path=str(path),
+            timestamp=time.time(),
+            run_id=self.run_id,
+        )
+
+    def stop(self) -> None:
+        self.running = False
+        if self.observer:
+            self.observer.stop()
+            self.observer.join(timeout=5)
+        logger.info("watcher_stop_requested")
--- a/tests/init.py
+++ b/tests/init.py
--- a/tests/conftest.py
+++ b/tests/conftest.py
@@ -0,0 +1,4 @@
+import sys
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).parent.parent / "src"))
--- a/tests/test_postprocess.py
+++ b/tests/test_postprocess.py
@@ -0,0 +1,73 @@
+import pytest
+from pathlib import Path
+
+from ocr_pipeline.postprocess.clean import clean_text, CleanResult
+from ocr_pipeline.postprocess.chunk import chunk_text, TextChunk
+from ocr_pipeline.config import settings
+
+
+class TestTextCleaning:
+    def test_fix_encoding(self):
+        result = clean_text("café")
+        assert "café" in result.text or "cafe" in result.text
+
+    def test_normalize_whitespace(self):
+        result = clean_text("hello    world\n\n\n\nfoo")
+        assert "hello world" in result.text
+        assert "foo" in result.text
+
+    def test_fix_hyphenation(self):
+        result = clean_text("hyphen-\nated")
+        assert "hyphenated" in result.text
+
+    def test_scientific_notation(self):
+        result = clean_text("1.5 x 10 ^ -3")
+        assert "×10^-3" in result.text or "x10^-3" in result.text
+
+    def test_normalize_units(self):
+        result = clean_text("50 uM and 100 ug/mL")
+        assert "µM" in result.text
+        assert "µg/mL" in result.text or "µg/mL" in result.text
+
+    def test_remove_ocr_artifacts(self):
+        result = clean_text("normal text\n|||\n───\nmore text")
+        assert "normal text" in result.text
+        assert "more text" in result.text
+
+
+class TestChunking:
+    def test_basic_chunking(self):
+        text = "A" * 500 + "\n\n" + "B" * 500 + "\n\n" + "C" * 500
+        chunks = chunk_text(text)
+        assert len(chunks) >= 2
+        assert all(isinstance(c, TextChunk) for c in chunks)
+
+    def test_empty_text(self):
+        chunks = chunk_text("")
+        assert chunks == []
+
+    def test_single_chunk(self):
+        text = "Short text that fits in one chunk"
+        chunks = chunk_text(text)
+        assert len(chunks) == 1
+
+    def test_metadata_propagation(self):
+        text = "A" * 1500
+        meta = {"source": "test.png", "custom": "value"}
+        chunks = chunk_text(text, meta)
+        assert all(c.metadata.get("source") == "test.png" for c in chunks)
+        assert all(c.metadata.get("custom") == "value" for c in chunks)
+
+
+class TestConfig:
+    def test_settings_load(self):
+        assert settings.input.paths
+        assert settings.ocr.engine in ("paddleocr", "tesseract", "auto")
+        assert settings.output.base_directory
+
+    def test_output_dir_property(self):
+        assert settings.output_dir.exists() or settings.output_dir.parent.exists()
+
+
+if __name__ == "__main__":
+    pytest.main([__file__, "-v"])
--- a/uv.lock
+++ b/uv.lock