Build a Intelligent Document Processor

Build a document processor that extracts structured data from PDFs and images:

**Supported Document Types:**
- Invoices
- Receipts
- Contracts (key terms extraction)
- Forms (any structured form)

**For Invoices, Extract:**
- Vendor name and address
- Invoice number and date
- Due date
- Line items (description, quantity, unit price, total)
- Subtotal, tax, total
- Payment terms

**Processing Flow:**
1. Upload PDF or image
2. OCR if needed (use Claude vision for images)
3. AI identifies document type
4. Extracts relevant fields
5. Validates data (dates make sense, numbers add up)
6. Outputs structured JSON/CSV

**UI Requirements:**
- Drag-and-drop upload area
- Progress indicator during processing
- Side-by-side view: original doc | extracted data
- Edit extracted fields before confirming
- Export to CSV or JSON
- Batch processing for multiple files

**Error Handling:**
- Flag low-confidence extractions
- Highlight fields that need human review
- Log extraction quality metrics

**Tech Stack:**
- Next.js frontend
- Claude API with vision for document understanding
- pdf-parse for PDF text extraction
- Supabase to store processed documents
- React-pdf for document preview

**Start Simple:**
Build for invoices only first. Support just PDF (not images). Add image support and other document types later.