docparse

Document layout parser and structured extraction.

docparse extracts structured data from PDFs, scanned forms, and mixed-format documents — tables, multi-column layouts, checkboxes — ready to pipe into LLMs or your own pipelines.

from docparse import parse_document, extract

doc = parse_document("loan_application.pdf")

# Use a schema to extract exactly the fields you need
data = extract(doc, schema={
    "applicant_name": "string",
    "annual_income": "number",
    "employment_status": "string",
    "loan_amount_requested": "number",
})

# {'applicant_name': 'Jane Doe', 'annual_income': 95000, ...}

Status

🚧 Early development. Star to follow progress.

What it does

• Layout parsing — tables, multi-column, forms, checkboxes from any PDF or scanned image
• Schema-based extraction — define the fields you want, get structured JSON back
• Confidence scores — per-field confidence, route low-confidence fields to human review
• Async + webhooks — batch processing for high-volume pipelines
• Vertical packs — pre-built schemas for mortgage docs, insurance forms, medical claims, invoices

Roadmap

• Python library
• Table detection benchmark
• Schema-based extraction API
• Vertical pack: mortgage documents
• Vertical pack: insurance (ACORD forms)
• Hosted service (mawlaia.com)
• HIPAA BAA

License

MIT