# Extract scanned PDFs and images into Apache Iceberg with self-hosted OCR

**Scanned invoices, faxes, and forms become searchable, structured
rows in Iceberg — OCR runs on your own GPU with your own vision
model, so documents with PHI or PII never leave your environment.**
Lakewright detects which PDFs need OCR (no text layer) and routes
them through a vision model you configure.

## How does the pipeline decide what needs OCR?
PDF profiling reads page count and text-layer presence. Born-digital
PDFs extract text directly; scanned ones render page images for the
vision model. Images (TIFF/JPEG/PNG) route straight to OCR. Document
metadata (kind, pages, dimensions) lands as inventory tables either
way.

## The steps
```bash
export LAKEWRIGHT_LLM_BASE_URL=http://your-gpu:11434/v1  # your model
lakewright scan  s3://document-dump      # flags needs_ocr per file
lakewright approve plans/claim_scans.plan.yaml
lakewright load  plans/claim_scans.plan.yaml --rest $CATALOG
```

## FAQ
**Which models?** Any OpenAI-compatible vision endpoint — reference
setup is qwen3-vl on a single local GPU box.
**Email attachments?** .eml/mbox files expand automatically: message
inventory plus every attachment routed through the same pipeline.