Skip to content

🌍 Project Domains:

  • https://GRIMdata.org
  • https://LittleRainbowRights.com

PIPELINE FLOW

This document explains the flow of data through the pipeline.

1. Input Sources

  • Scrapers pull content from:

  • AU policy PDFs (demo implemented)

  • OHCHR Treaty Body Database (planned)

  • UPR, UNICEF, ACERWC, ACHPR (planned)

  • Manual files can be dropped into data/raw/manual/.

2. Processing

  • File type detection Attempt to process based on extension.

  • .pdf → pdf_to_text.py

  • .docx → docx_to_text.py
  • .html → html_to_text.py

  • Fallback → try processors in order until success.

  • Normalization json_normalizer.py cleans country/region names. Preserves _raw fields for provenance.

  • Tagging tagger.py applies regex rules from configs/tags_v1.json. Tags are stored in tags_history in metadata.json.

  • Recommendations (future) Config-driven extraction of recommendations.

3. Metadata

  • metadata.json tracks:
  • Project identity
  • Documents ingested
  • Tags history (with versions)
  • Recommendations history (with versions)
  • Last processed timestamp

4. Exports

  • tags_summary.py → counts tags, outputs CSV with branding footer
  • tags_timeline*.py → (future) timeline exports
  • comparison.py → (future) compare tagging/recommendations across versions

5. Logging

  • logger.py provides unified + per-module logging
  • Controlled via --no-module-logs flag in pipeline_runner.py
  • Logs written to logs/ directory

6. Review

  • Researchers can inspect:
  • Raw files in data/raw/
  • Processed text in data/processed/
  • Metadata in data/metadata/metadata.json
  • Exports in data/exports/
  • Logs in logs/