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.

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