grime 0.1.0

Archival document management as a reusable Django app

grime

Grime is an archival document management platform — a reusable Django app for ingesting scanned documents, running OCR / NER pipelines, and annotating pages with tagged regions.

Models

Model	Role
Document	A standalone archival document (PDF, scanned booklet, ledger).
DocumentPage	A single page of a Document — the unit of OCR processing.
Word	One OCR-extracted word on a DocumentPage, with bbox, confidence, optional human correction, and optional BIO NER label.
Tag	A manually drawn region on a Document or DocumentPage with a label and word-level subcomponents.
OCRPass	Audit record for one OCR run on a DocumentPage.
NERPass	Audit record for one NER run on a DocumentPage.

Management commands

python manage.py ocr        --document 42 [--page N] [--textract] [--force] [--dry-run]
python manage.py ner        --document 42 [--page N] [--threshold 0.85] [--force] [--dry-run]
python manage.py match_tags --label "member entry" [--source-document 3] [--target-document 5] \
                            [--create-tags] [--force] [--min-score 0.5] [--tolerance 0.08]

Quick start

pip install -e ".[dev]"
python manage.py migrate
python manage.py createsuperuser
python manage.py runserver
# then visit http://127.0.0.1:8000/admin/

Optional dependencies

Extra	Adds
ocr	Tesseract (pytesseract, opencv-python, numpy)
textract	AWS Textract via boto3
hf	HuggingFace historical NER (transformers, torch)
viz	match_tags --video rendering (imageio[ffmpeg])
dev	All of the above

System prerequisites for the ocr extra:

sudo apt install tesseract-ocr poppler-utils

Status

This is an initial scaffold. The admin loads and the management commands run end-to-end, but the embedded document viewer (templates/admin/grime/_document_viewer.html) is read-only: bboxes and tags render on the page image, but interactive editing (OCR correction, tag CRUD, NER label correction) needs AJAX endpoints that have not been implemented yet.