Tractor

Welcome to the Tractor documentation.

Tractor is a document redaction application with ML-powered Named Entity Recognition. Users upload documents, the system identifies sensitive information using GLiNER and Presidio, and users can accept/reject redactions before exporting redacted versions.

Key Features

• Document Upload: Support for PDF and docx files.
• Automatic Redaction: Hybrid three-model NER pipeline — SpanCat (trained), GLiNER (zero-shot), and Presidio (pattern-based)
• Review Workflow: Accept, reject, or bulk-action suggested redactions. Adjacent same-type spans are automatically merged for easier review.
• Model Training: Train custom SpanCat models from accepted redactions; GLiNER model managed separately
• Export: Generate disclosure packages with original, redacted, and edited versions

Quick Links

• Getting Started - Set up and start using Tractor
• Redacting Documents - Learn how to redact documents
• Settings - Configure exemption templates and other settings
• Developer Setup - Set up your development environment
• Architecture - Understand how Tractor works

License

Tractor is licensed under the AGPL-3.0 License.