PII GUI: Local Desktop PII Redaction App
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- PII GUI runs entirely on-device, detecting PII via built-in regex rules or optional quantized ONNX models, with the only network access being a one-time model download from Hugging Face.
- The app supports PDF, Markdown, and plain-text input, parsing PDFs with pdf.js to preserve per-character positions so detections are highlighted directly on rendered pages.
- Exported PDFs use pdf-lib to burn opaque rectangles into pages, making redacted text irrecoverable from the output file, distinguishing it from simple text redaction.
- A local SQLite database and on-disk result files persist tabs, custom rules, and filter results across restarts, with a fixed privacy taxonomy covering accounts, addresses, emails, persons, phones, URLs, dates, and secrets.
- The AGPL v3.0-licensed app is built on pdf.js, pdf-lib, ONNX Runtime, and Hugging Face tokenizers, with an English/Korean/Japanese UI and installers for macOS, Windows, and Linux.
- Roadmap items include burned-in PDF export, local persistence, multilingual QA, an accessibility and keyboard-only review pass, and future integration with coding agents like Codex, Claude Code, and Cursor.
Why it matters: PII GUI gives privacy-conscious users and developers a fully offline workflow for sanitizing documents before sharing them with AI tools or third parties, addressing the common risk that sensitive text gets pasted into cloud-based LLMs. The burned-in PDF export and AGPL v3.0 license make it usable in regulated environments where true redaction and source transparency are non-negotiable.
Ask SkimNews
