platform agnosticity

config.yaml

@@ -1,7 +1,6 @@
 input:
   paths:
     - "~/Pictures"
-    - "/mnt/storage3/aman/screenshots"
   patterns:
     - "SCR-*.png"
     - "*.jpg"
@@ -11,8 +10,8 @@ input:
   recursive: true
 
 ocr:
-  engine: "paddleocr"  # "paddleocr" | "tesseract" | "auto"
-  languages: ["en"]  # For tesseract: eng, lat, deu, fra, spa, ita, por, chi_sim, jpn, kor
+  engine: "paddleocr"
+  languages: ["en", "latin"]
   use_gpu: false
   use_angle_cls: true
   det_db_thresh: 0.3
@@ -20,13 +19,6 @@ ocr:
   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
@@ -76,30 +68,32 @@ chunking:
   keep_separator: true
 
 output:
-  base_directory: "./data/ocr_output"
+  # Uses platformdirs default: ~/.local/share/ocr-pipeline (Linux), ~/Library/Application Support/ocr-pipeline (macOS), %LOCALAPPDATA%\ocr-pipeline (Windows)
   format: "markdown"
-  organize_by: "date_run"  # "date_run" | "source_dir" | "flat"
+  organize_by: "date_run"
   frontmatter:
-    - "source_path"
-    - "source_hash"
-    - "timestamp"
-    - "ocr_engine"
-    - "ocr_confidence_mean"
-    - "language"
-    - "detected_entities"
-    - "has_figures"
-    - "has_tables"
-    - "citations_found"
-    - "chunk_index"
-    - "total_chunks"
+    fields:
+      - "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: false
 
 watch:
   enabled: true
   debounce_seconds: 5
-  ignore_patterns:
-    - ".DS_Store"
-    - "*.tmp"
-    - "*.partial"
-    - "*.crdownload"
-  db_path: "./data/processed_files.db"
+  # Uses platform-aware defaults from config.py
+  # ignore_patterns:
+  #   - ".DS_Store"
+  #   - "*.tmp"
+  #   - "*.partial"
+  #   - "*.crdownload"
   poll_interval: 1.0