# `data/source_docs/` — drop your policy PDFs / Word docs here

Put your travel-insurance **policy knowledge** documents here as `.pdf` or `.docx`:
policy wordings, benefit schedules, coverage tables, FAQs, claim guides.

Then convert them to citable markdown (headings + tables preserved) with:

```bash
python scripts/ingest_docs.py --input data/source_docs --config config/config.yaml
```

This writes one `.md` per document into `data/raw/`, where `prepare_data.py` chunks and indexes it.

Notes:
- **Legacy `.doc`** isn't supported — re-save as `.docx` first.
- **Scanned/image-only PDFs** have no extractable text — OCR them first (the `pdf` skill can help).
- Benefit/coverage **tables** are converted to markdown tables so limits (e.g. "$100,000") stay
  attached to their labels instead of being flattened.
- This folder is for **policy knowledge only**. Do not place customer/policyholder PII here — that
  would require a different, access-controlled path (not fine-tuning or a shared RAG index).