Lethe

Lethe is a local document de-identification app for people who work with confidential contracts, client records or deal documents and want to use AI without sharing the names involved. It replaces names with consistent placeholders, then restores them in AI-generated replies, including summaries and translations.

It runs on Windows, macOS and Linux and is open source under Apache 2.0. Document processing works offline, with no API key or telemetry. Documents, name dictionaries and passphrase-encrypted restoration mappings stay on your computer. Optional detection and OCR models require downloads; any document you later share with ChatGPT or Claude goes to that service.

A dictionary links names and aliases to the same placeholder across a batch of files, so an AI can still distinguish the parties. Lethe also detects email addresses, phone numbers and account numbers. Optional Microsoft Presidio and spaCy models suggest additional names and organisations. These suggestions are disabled by default and require the NLP extra or a bundled installer. You review proposed replacements before Lethe writes the output.

Word, PowerPoint and Excel files retain their formats, with Excel formulas preserved. PDFs and email messages become Word documents; PDF output retains source-page references. Local OCR reads scanned PDF pages and flags uncertain results for review. Lethe can also restore placeholders created outside the app.

Detection has limits: embedded images, metadata and some document objects can still contain names. Email attachments aren't processed. Restoration requires unchanged placeholders and, for encrypted jobs, the original passphrase.