Docuparse

AI-powered bill parser that extracts structured data from PDF receipts using Claude and optionally uploads them to Paperless-ngx.

Requirements

• Python 3.12+
• Anthropic API Key
• (Optional) Paperless-ngx instance

Installation

pip install -e .

Configuration

Create a .env file in the project root:

ANTHROPIC_API_KEY=your-api-key

# Optional: Paperless-ngx integration
PAPERLESS_URL=https://your-paperless-instance.com
PAPERLESS_API_TOKEN=your-paperless-token

To disable Paperless upload, set PAPERLESS_UPLOAD_ENABLE = False in docuparse/config.py.

Usage

docuparse

Or run as module:

python -m docuparse

A file dialog opens to select PDF bills. The extracted data is saved to ~/Downloads/bills-YYYY-MM-DD.json.

Development

pip install -e ".[dev]"

Lint and format:

ruff check docuparse/
ruff format docuparse/

Type checking:

mypy docuparse/

Run tests:

pytest